{ "best_metric": null, "best_model_checkpoint": null, "epoch": 4.999698213423467, "global_step": 82835, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 0.0, "loss": 8.9372, "step": 2 }, { "epoch": 0.0, "learning_rate": 3.0180479266010743e-09, "loss": 9.7629, "step": 4 }, { "epoch": 0.0, "learning_rate": 9.054143779803223e-09, "loss": 9.8613, "step": 6 }, { "epoch": 0.0, "learning_rate": 1.2072191706404297e-08, "loss": 9.7387, "step": 8 }, { "epoch": 0.0, "learning_rate": 1.8108287559606446e-08, "loss": 9.5245, "step": 10 }, { "epoch": 0.0, "learning_rate": 2.4144383412808595e-08, "loss": 9.721, "step": 12 }, { "epoch": 0.0, "learning_rate": 3.0180479266010743e-08, "loss": 9.8041, "step": 14 }, { "epoch": 0.0, "learning_rate": 3.621657511921289e-08, "loss": 9.854, "step": 16 }, { "epoch": 0.0, "learning_rate": 4.225267097241505e-08, "loss": 9.0043, "step": 18 }, { "epoch": 0.0, "learning_rate": 4.828876682561719e-08, "loss": 10.0382, "step": 20 }, { "epoch": 0.0, "learning_rate": 5.432486267881934e-08, "loss": 9.255, "step": 22 }, { "epoch": 0.0, "learning_rate": 6.036095853202149e-08, "loss": 9.8957, "step": 24 }, { "epoch": 0.0, "learning_rate": 6.639705438522364e-08, "loss": 9.874, "step": 26 }, { "epoch": 0.0, "learning_rate": 7.243315023842578e-08, "loss": 9.5943, "step": 28 }, { "epoch": 0.0, "learning_rate": 7.846924609162794e-08, "loss": 9.8768, "step": 30 }, { "epoch": 0.0, "learning_rate": 8.45053419448301e-08, "loss": 9.589, "step": 32 }, { "epoch": 0.0, "learning_rate": 8.752338987143117e-08, "loss": 10.3253, "step": 34 }, { "epoch": 0.0, "learning_rate": 9.355948572463331e-08, "loss": 9.5414, "step": 36 }, { "epoch": 0.0, "learning_rate": 9.959558157783546e-08, "loss": 9.7475, "step": 38 }, { "epoch": 0.0, "learning_rate": 1.0563167743103762e-07, "loss": 9.8865, "step": 40 }, { "epoch": 0.0, "learning_rate": 1.1166777328423976e-07, "loss": 9.8662, "step": 42 }, { "epoch": 0.0, "learning_rate": 1.1770386913744192e-07, "loss": 9.3723, "step": 44 }, { "epoch": 0.0, "learning_rate": 1.2373996499064405e-07, "loss": 9.7019, "step": 46 }, { "epoch": 0.0, "learning_rate": 1.2977606084384621e-07, "loss": 9.7947, "step": 48 }, { "epoch": 0.0, "learning_rate": 1.3581215669704836e-07, "loss": 9.8253, "step": 50 }, { "epoch": 0.0, "learning_rate": 1.418482525502505e-07, "loss": 10.2321, "step": 52 }, { "epoch": 0.0, "learning_rate": 1.4788434840345267e-07, "loss": 9.8124, "step": 54 }, { "epoch": 0.0, "learning_rate": 1.539204442566548e-07, "loss": 9.2466, "step": 56 }, { "epoch": 0.0, "learning_rate": 1.5995654010985695e-07, "loss": 8.4751, "step": 58 }, { "epoch": 0.0, "learning_rate": 1.659926359630591e-07, "loss": 9.4499, "step": 60 }, { "epoch": 0.0, "learning_rate": 1.7202873181626123e-07, "loss": 9.4864, "step": 62 }, { "epoch": 0.0, "learning_rate": 1.780648276694634e-07, "loss": 10.0747, "step": 64 }, { "epoch": 0.0, "learning_rate": 1.8410092352266555e-07, "loss": 9.8141, "step": 66 }, { "epoch": 0.0, "learning_rate": 1.9013701937586771e-07, "loss": 9.6327, "step": 68 }, { "epoch": 0.0, "learning_rate": 1.9617311522906983e-07, "loss": 9.8094, "step": 70 }, { "epoch": 0.0, "learning_rate": 2.02209211082272e-07, "loss": 9.3862, "step": 72 }, { "epoch": 0.0, "learning_rate": 2.0824530693547414e-07, "loss": 9.5979, "step": 74 }, { "epoch": 0.0, "learning_rate": 2.142814027886763e-07, "loss": 9.4632, "step": 76 }, { "epoch": 0.0, "learning_rate": 2.2031749864187845e-07, "loss": 9.6623, "step": 78 }, { "epoch": 0.0, "learning_rate": 2.2635359449508057e-07, "loss": 9.7787, "step": 80 }, { "epoch": 0.0, "learning_rate": 2.3238969034828276e-07, "loss": 9.7191, "step": 82 }, { "epoch": 0.01, "learning_rate": 2.384257862014849e-07, "loss": 8.976, "step": 84 }, { "epoch": 0.01, "learning_rate": 2.4446188205468705e-07, "loss": 8.8727, "step": 86 }, { "epoch": 0.01, "learning_rate": 2.504979779078892e-07, "loss": 8.8634, "step": 88 }, { "epoch": 0.01, "learning_rate": 2.565340737610914e-07, "loss": 9.3587, "step": 90 }, { "epoch": 0.01, "learning_rate": 2.6257016961429347e-07, "loss": 8.2269, "step": 92 }, { "epoch": 0.01, "learning_rate": 2.686062654674956e-07, "loss": 8.7696, "step": 94 }, { "epoch": 0.01, "learning_rate": 2.746423613206978e-07, "loss": 8.7275, "step": 96 }, { "epoch": 0.01, "learning_rate": 2.8067845717389995e-07, "loss": 8.6478, "step": 98 }, { "epoch": 0.01, "learning_rate": 2.867145530271021e-07, "loss": 8.8562, "step": 100 }, { "epoch": 0.01, "learning_rate": 2.9275064888030424e-07, "loss": 9.1851, "step": 102 }, { "epoch": 0.01, "learning_rate": 2.987867447335064e-07, "loss": 8.0504, "step": 104 }, { "epoch": 0.01, "learning_rate": 3.048228405867085e-07, "loss": 8.0933, "step": 106 }, { "epoch": 0.01, "learning_rate": 3.1085893643991066e-07, "loss": 8.0142, "step": 108 }, { "epoch": 0.01, "learning_rate": 3.1689503229311286e-07, "loss": 8.2026, "step": 110 }, { "epoch": 0.01, "learning_rate": 3.2293112814631495e-07, "loss": 8.1533, "step": 112 }, { "epoch": 0.01, "learning_rate": 3.2896722399951714e-07, "loss": 8.1662, "step": 114 }, { "epoch": 0.01, "learning_rate": 3.350033198527193e-07, "loss": 7.8035, "step": 116 }, { "epoch": 0.01, "learning_rate": 3.410394157059214e-07, "loss": 8.674, "step": 118 }, { "epoch": 0.01, "learning_rate": 3.4707551155912357e-07, "loss": 7.9901, "step": 120 }, { "epoch": 0.01, "learning_rate": 3.531116074123257e-07, "loss": 7.6294, "step": 122 }, { "epoch": 0.01, "learning_rate": 3.591477032655279e-07, "loss": 8.0513, "step": 124 }, { "epoch": 0.01, "learning_rate": 3.6518379911873e-07, "loss": 8.4402, "step": 126 }, { "epoch": 0.01, "learning_rate": 3.712198949719322e-07, "loss": 8.0391, "step": 128 }, { "epoch": 0.01, "learning_rate": 3.7725599082513433e-07, "loss": 7.9661, "step": 130 }, { "epoch": 0.01, "learning_rate": 3.8329208667833647e-07, "loss": 8.022, "step": 132 }, { "epoch": 0.01, "learning_rate": 3.893281825315386e-07, "loss": 7.8235, "step": 134 }, { "epoch": 0.01, "learning_rate": 3.9536427838474076e-07, "loss": 7.534, "step": 136 }, { "epoch": 0.01, "learning_rate": 4.0140037423794295e-07, "loss": 7.3391, "step": 138 }, { "epoch": 0.01, "learning_rate": 4.074364700911451e-07, "loss": 7.5548, "step": 140 }, { "epoch": 0.01, "learning_rate": 4.134725659443472e-07, "loss": 7.7969, "step": 142 }, { "epoch": 0.01, "learning_rate": 4.195086617975494e-07, "loss": 7.2439, "step": 144 }, { "epoch": 0.01, "learning_rate": 4.255447576507515e-07, "loss": 6.9898, "step": 146 }, { "epoch": 0.01, "learning_rate": 4.315808535039536e-07, "loss": 7.1331, "step": 148 }, { "epoch": 0.01, "learning_rate": 4.376169493571558e-07, "loss": 8.0848, "step": 150 }, { "epoch": 0.01, "learning_rate": 4.43653045210358e-07, "loss": 7.0698, "step": 152 }, { "epoch": 0.01, "learning_rate": 4.4968914106356014e-07, "loss": 6.5403, "step": 154 }, { "epoch": 0.01, "learning_rate": 4.5572523691676223e-07, "loss": 6.671, "step": 156 }, { "epoch": 0.01, "learning_rate": 4.617613327699644e-07, "loss": 7.3358, "step": 158 }, { "epoch": 0.01, "learning_rate": 4.6779742862316657e-07, "loss": 7.0013, "step": 160 }, { "epoch": 0.01, "learning_rate": 4.7383352447636866e-07, "loss": 6.608, "step": 162 }, { "epoch": 0.01, "learning_rate": 4.798696203295709e-07, "loss": 6.6057, "step": 164 }, { "epoch": 0.01, "learning_rate": 4.85905716182773e-07, "loss": 6.4483, "step": 166 }, { "epoch": 0.01, "learning_rate": 4.919418120359752e-07, "loss": 6.7205, "step": 168 }, { "epoch": 0.01, "learning_rate": 4.979779078891773e-07, "loss": 5.9447, "step": 170 }, { "epoch": 0.01, "learning_rate": 5.040140037423794e-07, "loss": 6.9293, "step": 172 }, { "epoch": 0.01, "learning_rate": 5.100500995955816e-07, "loss": 6.764, "step": 174 }, { "epoch": 0.01, "learning_rate": 5.160861954487837e-07, "loss": 7.206, "step": 176 }, { "epoch": 0.01, "learning_rate": 5.221222913019859e-07, "loss": 6.3251, "step": 178 }, { "epoch": 0.01, "learning_rate": 5.281583871551881e-07, "loss": 6.1435, "step": 180 }, { "epoch": 0.01, "learning_rate": 5.341944830083902e-07, "loss": 6.487, "step": 182 }, { "epoch": 0.01, "learning_rate": 5.402305788615924e-07, "loss": 6.7475, "step": 184 }, { "epoch": 0.01, "learning_rate": 5.462666747147945e-07, "loss": 6.2869, "step": 186 }, { "epoch": 0.01, "learning_rate": 5.523027705679967e-07, "loss": 6.4313, "step": 188 }, { "epoch": 0.01, "learning_rate": 5.583388664211988e-07, "loss": 6.1132, "step": 190 }, { "epoch": 0.01, "learning_rate": 5.643749622744009e-07, "loss": 6.6992, "step": 192 }, { "epoch": 0.01, "learning_rate": 5.704110581276031e-07, "loss": 6.5636, "step": 194 }, { "epoch": 0.01, "learning_rate": 5.764471539808052e-07, "loss": 5.9689, "step": 196 }, { "epoch": 0.01, "learning_rate": 5.824832498340074e-07, "loss": 6.1412, "step": 198 }, { "epoch": 0.01, "learning_rate": 5.885193456872095e-07, "loss": 5.7085, "step": 200 }, { "epoch": 0.01, "learning_rate": 5.945554415404117e-07, "loss": 5.8916, "step": 202 }, { "epoch": 0.01, "learning_rate": 6.005915373936138e-07, "loss": 5.669, "step": 204 }, { "epoch": 0.01, "learning_rate": 6.06627633246816e-07, "loss": 6.0997, "step": 206 }, { "epoch": 0.01, "learning_rate": 6.126637291000182e-07, "loss": 6.2007, "step": 208 }, { "epoch": 0.01, "learning_rate": 6.186998249532203e-07, "loss": 6.0631, "step": 210 }, { "epoch": 0.01, "learning_rate": 6.247359208064225e-07, "loss": 5.6989, "step": 212 }, { "epoch": 0.01, "learning_rate": 6.307720166596246e-07, "loss": 5.9532, "step": 214 }, { "epoch": 0.01, "learning_rate": 6.368081125128268e-07, "loss": 5.8602, "step": 216 }, { "epoch": 0.01, "learning_rate": 6.428442083660288e-07, "loss": 6.0182, "step": 218 }, { "epoch": 0.01, "learning_rate": 6.48880304219231e-07, "loss": 5.2881, "step": 220 }, { "epoch": 0.01, "learning_rate": 6.549164000724332e-07, "loss": 5.6634, "step": 222 }, { "epoch": 0.01, "learning_rate": 6.609524959256353e-07, "loss": 5.3456, "step": 224 }, { "epoch": 0.01, "learning_rate": 6.669885917788374e-07, "loss": 5.3702, "step": 226 }, { "epoch": 0.01, "learning_rate": 6.730246876320396e-07, "loss": 5.1404, "step": 228 }, { "epoch": 0.01, "learning_rate": 6.790607834852418e-07, "loss": 5.9002, "step": 230 }, { "epoch": 0.01, "learning_rate": 6.850968793384439e-07, "loss": 5.7089, "step": 232 }, { "epoch": 0.01, "learning_rate": 6.911329751916461e-07, "loss": 4.6848, "step": 234 }, { "epoch": 0.01, "learning_rate": 6.971690710448483e-07, "loss": 5.5505, "step": 236 }, { "epoch": 0.01, "learning_rate": 7.032051668980504e-07, "loss": 5.759, "step": 238 }, { "epoch": 0.01, "learning_rate": 7.092412627512525e-07, "loss": 5.9438, "step": 240 }, { "epoch": 0.01, "learning_rate": 7.152773586044547e-07, "loss": 5.163, "step": 242 }, { "epoch": 0.01, "learning_rate": 7.213134544576569e-07, "loss": 5.0442, "step": 244 }, { "epoch": 0.01, "learning_rate": 7.273495503108589e-07, "loss": 4.9266, "step": 246 }, { "epoch": 0.01, "learning_rate": 7.333856461640611e-07, "loss": 5.0532, "step": 248 }, { "epoch": 0.02, "learning_rate": 7.394217420172633e-07, "loss": 5.5874, "step": 250 }, { "epoch": 0.02, "learning_rate": 7.454578378704654e-07, "loss": 5.2224, "step": 252 }, { "epoch": 0.02, "learning_rate": 7.514939337236675e-07, "loss": 4.9778, "step": 254 }, { "epoch": 0.02, "learning_rate": 7.575300295768697e-07, "loss": 5.0015, "step": 256 }, { "epoch": 0.02, "learning_rate": 7.635661254300719e-07, "loss": 5.0586, "step": 258 }, { "epoch": 0.02, "learning_rate": 7.69602221283274e-07, "loss": 4.8217, "step": 260 }, { "epoch": 0.02, "learning_rate": 7.756383171364762e-07, "loss": 5.3707, "step": 262 }, { "epoch": 0.02, "learning_rate": 7.816744129896784e-07, "loss": 4.8991, "step": 264 }, { "epoch": 0.02, "learning_rate": 7.877105088428805e-07, "loss": 5.2239, "step": 266 }, { "epoch": 0.02, "learning_rate": 7.937466046960827e-07, "loss": 4.7818, "step": 268 }, { "epoch": 0.02, "learning_rate": 7.997827005492846e-07, "loss": 4.5426, "step": 270 }, { "epoch": 0.02, "learning_rate": 8.058187964024868e-07, "loss": 4.9632, "step": 272 }, { "epoch": 0.02, "learning_rate": 8.11854892255689e-07, "loss": 4.6608, "step": 274 }, { "epoch": 0.02, "learning_rate": 8.178909881088912e-07, "loss": 4.7291, "step": 276 }, { "epoch": 0.02, "learning_rate": 8.239270839620934e-07, "loss": 4.7637, "step": 278 }, { "epoch": 0.02, "learning_rate": 8.299631798152955e-07, "loss": 5.013, "step": 280 }, { "epoch": 0.02, "learning_rate": 8.359992756684977e-07, "loss": 4.3658, "step": 282 }, { "epoch": 0.02, "learning_rate": 8.420353715216997e-07, "loss": 4.3738, "step": 284 }, { "epoch": 0.02, "learning_rate": 8.480714673749019e-07, "loss": 4.9247, "step": 286 }, { "epoch": 0.02, "learning_rate": 8.541075632281041e-07, "loss": 4.7354, "step": 288 }, { "epoch": 0.02, "learning_rate": 8.601436590813063e-07, "loss": 4.1141, "step": 290 }, { "epoch": 0.02, "learning_rate": 8.661797549345085e-07, "loss": 4.6962, "step": 292 }, { "epoch": 0.02, "learning_rate": 8.722158507877106e-07, "loss": 5.0078, "step": 294 }, { "epoch": 0.02, "learning_rate": 8.782519466409128e-07, "loss": 5.0416, "step": 296 }, { "epoch": 0.02, "learning_rate": 8.842880424941147e-07, "loss": 3.7375, "step": 298 }, { "epoch": 0.02, "learning_rate": 8.903241383473169e-07, "loss": 3.8638, "step": 300 }, { "epoch": 0.02, "learning_rate": 8.963602342005191e-07, "loss": 4.3227, "step": 302 }, { "epoch": 0.02, "learning_rate": 9.023963300537213e-07, "loss": 4.4707, "step": 304 }, { "epoch": 0.02, "learning_rate": 9.084324259069235e-07, "loss": 4.5495, "step": 306 }, { "epoch": 0.02, "learning_rate": 9.144685217601256e-07, "loss": 4.2036, "step": 308 }, { "epoch": 0.02, "learning_rate": 9.205046176133278e-07, "loss": 3.9846, "step": 310 }, { "epoch": 0.02, "learning_rate": 9.265407134665298e-07, "loss": 4.8941, "step": 312 }, { "epoch": 0.02, "learning_rate": 9.32576809319732e-07, "loss": 4.4499, "step": 314 }, { "epoch": 0.02, "learning_rate": 9.386129051729342e-07, "loss": 4.4182, "step": 316 }, { "epoch": 0.02, "learning_rate": 9.446490010261364e-07, "loss": 4.5527, "step": 318 }, { "epoch": 0.02, "learning_rate": 9.506850968793386e-07, "loss": 4.2955, "step": 320 }, { "epoch": 0.02, "learning_rate": 9.567211927325407e-07, "loss": 3.7982, "step": 322 }, { "epoch": 0.02, "learning_rate": 9.62757288585743e-07, "loss": 4.2134, "step": 324 }, { "epoch": 0.02, "learning_rate": 9.687933844389448e-07, "loss": 4.7128, "step": 326 }, { "epoch": 0.02, "learning_rate": 9.71811432365546e-07, "loss": 4.9122, "step": 328 }, { "epoch": 0.02, "learning_rate": 9.778475282187482e-07, "loss": 3.9668, "step": 330 }, { "epoch": 0.02, "learning_rate": 9.838836240719505e-07, "loss": 4.1216, "step": 332 }, { "epoch": 0.02, "learning_rate": 9.899197199251524e-07, "loss": 5.1625, "step": 334 }, { "epoch": 0.02, "learning_rate": 9.959558157783547e-07, "loss": 4.3852, "step": 336 }, { "epoch": 0.02, "learning_rate": 1.0019919116315568e-06, "loss": 4.3818, "step": 338 }, { "epoch": 0.02, "learning_rate": 1.0080280074847588e-06, "loss": 4.3734, "step": 340 }, { "epoch": 0.02, "learning_rate": 1.0140641033379611e-06, "loss": 4.0326, "step": 342 }, { "epoch": 0.02, "learning_rate": 1.0201001991911632e-06, "loss": 4.185, "step": 344 }, { "epoch": 0.02, "learning_rate": 1.0261362950443655e-06, "loss": 3.9322, "step": 346 }, { "epoch": 0.02, "learning_rate": 1.0321723908975674e-06, "loss": 4.5106, "step": 348 }, { "epoch": 0.02, "learning_rate": 1.0382084867507697e-06, "loss": 4.4923, "step": 350 }, { "epoch": 0.02, "learning_rate": 1.0442445826039718e-06, "loss": 4.3362, "step": 352 }, { "epoch": 0.02, "learning_rate": 1.0502806784571739e-06, "loss": 4.1472, "step": 354 }, { "epoch": 0.02, "learning_rate": 1.0563167743103762e-06, "loss": 4.8752, "step": 356 }, { "epoch": 0.02, "learning_rate": 1.0623528701635783e-06, "loss": 4.7386, "step": 358 }, { "epoch": 0.02, "learning_rate": 1.0683889660167804e-06, "loss": 4.4184, "step": 360 }, { "epoch": 0.02, "learning_rate": 1.0744250618699825e-06, "loss": 4.5112, "step": 362 }, { "epoch": 0.02, "learning_rate": 1.0804611577231848e-06, "loss": 3.6309, "step": 364 }, { "epoch": 0.02, "learning_rate": 1.0864972535763868e-06, "loss": 4.4038, "step": 366 }, { "epoch": 0.02, "learning_rate": 1.092533349429589e-06, "loss": 4.7573, "step": 368 }, { "epoch": 0.02, "learning_rate": 1.0985694452827912e-06, "loss": 4.1415, "step": 370 }, { "epoch": 0.02, "learning_rate": 1.1046055411359933e-06, "loss": 4.1076, "step": 372 }, { "epoch": 0.02, "learning_rate": 1.1106416369891954e-06, "loss": 3.6889, "step": 374 }, { "epoch": 0.02, "learning_rate": 1.1166777328423975e-06, "loss": 4.4522, "step": 376 }, { "epoch": 0.02, "learning_rate": 1.1227138286955998e-06, "loss": 4.0264, "step": 378 }, { "epoch": 0.02, "learning_rate": 1.1287499245488019e-06, "loss": 3.8275, "step": 380 }, { "epoch": 0.02, "learning_rate": 1.134786020402004e-06, "loss": 4.5662, "step": 382 }, { "epoch": 0.02, "learning_rate": 1.1408221162552063e-06, "loss": 4.3301, "step": 384 }, { "epoch": 0.02, "learning_rate": 1.1468582121084084e-06, "loss": 4.1985, "step": 386 }, { "epoch": 0.02, "learning_rate": 1.1528943079616105e-06, "loss": 3.4489, "step": 388 }, { "epoch": 0.02, "learning_rate": 1.1589304038148126e-06, "loss": 3.8898, "step": 390 }, { "epoch": 0.02, "learning_rate": 1.1649664996680149e-06, "loss": 3.8185, "step": 392 }, { "epoch": 0.02, "learning_rate": 1.171002595521217e-06, "loss": 4.1466, "step": 394 }, { "epoch": 0.02, "learning_rate": 1.177038691374419e-06, "loss": 4.5973, "step": 396 }, { "epoch": 0.02, "learning_rate": 1.1830747872276213e-06, "loss": 3.4598, "step": 398 }, { "epoch": 0.02, "learning_rate": 1.1891108830808234e-06, "loss": 4.5472, "step": 400 }, { "epoch": 0.02, "learning_rate": 1.1921289310074245e-06, "loss": 4.3111, "step": 402 }, { "epoch": 0.02, "learning_rate": 1.1981650268606266e-06, "loss": 3.7894, "step": 404 }, { "epoch": 0.02, "learning_rate": 1.2042011227138289e-06, "loss": 3.7301, "step": 406 }, { "epoch": 0.02, "learning_rate": 1.210237218567031e-06, "loss": 3.7228, "step": 408 }, { "epoch": 0.02, "learning_rate": 1.216273314420233e-06, "loss": 4.3969, "step": 410 }, { "epoch": 0.02, "learning_rate": 1.2223094102734351e-06, "loss": 3.9891, "step": 412 }, { "epoch": 0.02, "learning_rate": 1.2283455061266374e-06, "loss": 3.8524, "step": 414 }, { "epoch": 0.03, "learning_rate": 1.2343816019798395e-06, "loss": 4.0453, "step": 416 }, { "epoch": 0.03, "learning_rate": 1.2404176978330416e-06, "loss": 3.7115, "step": 418 }, { "epoch": 0.03, "learning_rate": 1.246453793686244e-06, "loss": 4.2122, "step": 420 }, { "epoch": 0.03, "learning_rate": 1.252489889539446e-06, "loss": 3.8081, "step": 422 }, { "epoch": 0.03, "learning_rate": 1.258525985392648e-06, "loss": 3.9204, "step": 424 }, { "epoch": 0.03, "learning_rate": 1.2645620812458502e-06, "loss": 4.1756, "step": 426 }, { "epoch": 0.03, "learning_rate": 1.2705981770990525e-06, "loss": 4.0314, "step": 428 }, { "epoch": 0.03, "learning_rate": 1.2766342729522546e-06, "loss": 3.959, "step": 430 }, { "epoch": 0.03, "learning_rate": 1.2826703688054566e-06, "loss": 3.5939, "step": 432 }, { "epoch": 0.03, "learning_rate": 1.288706464658659e-06, "loss": 3.8975, "step": 434 }, { "epoch": 0.03, "learning_rate": 1.294742560511861e-06, "loss": 3.5231, "step": 436 }, { "epoch": 0.03, "learning_rate": 1.3007786563650631e-06, "loss": 4.4205, "step": 438 }, { "epoch": 0.03, "learning_rate": 1.3068147522182652e-06, "loss": 3.6687, "step": 440 }, { "epoch": 0.03, "learning_rate": 1.3128508480714673e-06, "loss": 3.5962, "step": 442 }, { "epoch": 0.03, "learning_rate": 1.3188869439246696e-06, "loss": 3.6106, "step": 444 }, { "epoch": 0.03, "learning_rate": 1.3249230397778717e-06, "loss": 3.6641, "step": 446 }, { "epoch": 0.03, "learning_rate": 1.330959135631074e-06, "loss": 3.8254, "step": 448 }, { "epoch": 0.03, "learning_rate": 1.336995231484276e-06, "loss": 4.2167, "step": 450 }, { "epoch": 0.03, "learning_rate": 1.3430313273374782e-06, "loss": 4.2577, "step": 452 }, { "epoch": 0.03, "learning_rate": 1.3490674231906803e-06, "loss": 3.4431, "step": 454 }, { "epoch": 0.03, "learning_rate": 1.3551035190438824e-06, "loss": 3.6207, "step": 456 }, { "epoch": 0.03, "learning_rate": 1.3611396148970847e-06, "loss": 3.7758, "step": 458 }, { "epoch": 0.03, "learning_rate": 1.3671757107502867e-06, "loss": 3.6439, "step": 460 }, { "epoch": 0.03, "learning_rate": 1.373211806603489e-06, "loss": 4.228, "step": 462 }, { "epoch": 0.03, "learning_rate": 1.3792479024566911e-06, "loss": 3.472, "step": 464 }, { "epoch": 0.03, "learning_rate": 1.3852839983098932e-06, "loss": 3.5919, "step": 466 }, { "epoch": 0.03, "learning_rate": 1.3913200941630953e-06, "loss": 3.2747, "step": 468 }, { "epoch": 0.03, "learning_rate": 1.3973561900162974e-06, "loss": 3.4521, "step": 470 }, { "epoch": 0.03, "learning_rate": 1.4033922858694997e-06, "loss": 3.3162, "step": 472 }, { "epoch": 0.03, "learning_rate": 1.4094283817227018e-06, "loss": 3.4543, "step": 474 }, { "epoch": 0.03, "learning_rate": 1.415464477575904e-06, "loss": 3.6578, "step": 476 }, { "epoch": 0.03, "learning_rate": 1.4215005734291062e-06, "loss": 3.7214, "step": 478 }, { "epoch": 0.03, "learning_rate": 1.4275366692823083e-06, "loss": 3.5865, "step": 480 }, { "epoch": 0.03, "learning_rate": 1.4335727651355104e-06, "loss": 3.6149, "step": 482 }, { "epoch": 0.03, "learning_rate": 1.4396088609887124e-06, "loss": 3.9983, "step": 484 }, { "epoch": 0.03, "learning_rate": 1.4456449568419147e-06, "loss": 4.0547, "step": 486 }, { "epoch": 0.03, "learning_rate": 1.4516810526951168e-06, "loss": 3.3879, "step": 488 }, { "epoch": 0.03, "learning_rate": 1.4577171485483191e-06, "loss": 3.3529, "step": 490 }, { "epoch": 0.03, "learning_rate": 1.4637532444015212e-06, "loss": 3.5543, "step": 492 }, { "epoch": 0.03, "learning_rate": 1.4697893402547233e-06, "loss": 3.1971, "step": 494 }, { "epoch": 0.03, "learning_rate": 1.4758254361079254e-06, "loss": 3.3137, "step": 496 }, { "epoch": 0.03, "learning_rate": 1.4818615319611275e-06, "loss": 3.8202, "step": 498 }, { "epoch": 0.03, "learning_rate": 1.4878976278143298e-06, "loss": 3.6116, "step": 500 }, { "epoch": 0.03, "learning_rate": 1.4939337236675319e-06, "loss": 3.8588, "step": 502 }, { "epoch": 0.03, "learning_rate": 1.4999698195207342e-06, "loss": 3.5506, "step": 504 }, { "epoch": 0.03, "learning_rate": 1.5060059153739363e-06, "loss": 3.3879, "step": 506 }, { "epoch": 0.03, "learning_rate": 1.5120420112271384e-06, "loss": 3.4886, "step": 508 }, { "epoch": 0.03, "learning_rate": 1.5180781070803405e-06, "loss": 4.2108, "step": 510 }, { "epoch": 0.03, "learning_rate": 1.5241142029335425e-06, "loss": 3.4955, "step": 512 }, { "epoch": 0.03, "learning_rate": 1.5301502987867448e-06, "loss": 3.9766, "step": 514 }, { "epoch": 0.03, "learning_rate": 1.536186394639947e-06, "loss": 3.8097, "step": 516 }, { "epoch": 0.03, "learning_rate": 1.5422224904931492e-06, "loss": 3.397, "step": 518 }, { "epoch": 0.03, "learning_rate": 1.5482585863463513e-06, "loss": 3.5808, "step": 520 }, { "epoch": 0.03, "learning_rate": 1.5542946821995534e-06, "loss": 3.3537, "step": 522 }, { "epoch": 0.03, "learning_rate": 1.5603307780527555e-06, "loss": 3.2779, "step": 524 }, { "epoch": 0.03, "learning_rate": 1.5663668739059576e-06, "loss": 3.2188, "step": 526 }, { "epoch": 0.03, "learning_rate": 1.5724029697591597e-06, "loss": 3.8721, "step": 528 }, { "epoch": 0.03, "learning_rate": 1.578439065612362e-06, "loss": 3.2544, "step": 530 }, { "epoch": 0.03, "learning_rate": 1.584475161465564e-06, "loss": 3.581, "step": 532 }, { "epoch": 0.03, "learning_rate": 1.5905112573187664e-06, "loss": 3.6326, "step": 534 }, { "epoch": 0.03, "learning_rate": 1.5965473531719685e-06, "loss": 3.3451, "step": 536 }, { "epoch": 0.03, "learning_rate": 1.6025834490251708e-06, "loss": 3.3818, "step": 538 }, { "epoch": 0.03, "learning_rate": 1.6086195448783726e-06, "loss": 3.7094, "step": 540 }, { "epoch": 0.03, "learning_rate": 1.6146556407315747e-06, "loss": 4.0553, "step": 542 }, { "epoch": 0.03, "learning_rate": 1.620691736584777e-06, "loss": 3.3644, "step": 544 }, { "epoch": 0.03, "learning_rate": 1.6267278324379791e-06, "loss": 3.7227, "step": 546 }, { "epoch": 0.03, "learning_rate": 1.6327639282911814e-06, "loss": 2.5999, "step": 548 }, { "epoch": 0.03, "learning_rate": 1.6388000241443835e-06, "loss": 3.7067, "step": 550 }, { "epoch": 0.03, "learning_rate": 1.6448361199975858e-06, "loss": 2.9748, "step": 552 }, { "epoch": 0.03, "learning_rate": 1.6508722158507877e-06, "loss": 3.1797, "step": 554 }, { "epoch": 0.03, "learning_rate": 1.6569083117039898e-06, "loss": 3.1848, "step": 556 }, { "epoch": 0.03, "learning_rate": 1.662944407557192e-06, "loss": 3.1106, "step": 558 }, { "epoch": 0.03, "learning_rate": 1.6689805034103942e-06, "loss": 3.1589, "step": 560 }, { "epoch": 0.03, "learning_rate": 1.6750165992635965e-06, "loss": 3.5205, "step": 562 }, { "epoch": 0.03, "learning_rate": 1.6810526951167986e-06, "loss": 4.0288, "step": 564 }, { "epoch": 0.03, "learning_rate": 1.6870887909700009e-06, "loss": 3.6733, "step": 566 }, { "epoch": 0.03, "learning_rate": 1.6931248868232027e-06, "loss": 3.1477, "step": 568 }, { "epoch": 0.03, "learning_rate": 1.6991609826764048e-06, "loss": 3.3632, "step": 570 }, { "epoch": 0.03, "learning_rate": 1.7051970785296071e-06, "loss": 3.2987, "step": 572 }, { "epoch": 0.03, "learning_rate": 1.7112331743828092e-06, "loss": 3.2361, "step": 574 }, { "epoch": 0.03, "learning_rate": 1.7172692702360115e-06, "loss": 3.1541, "step": 576 }, { "epoch": 0.03, "learning_rate": 1.7233053660892136e-06, "loss": 3.3276, "step": 578 }, { "epoch": 0.04, "learning_rate": 1.729341461942416e-06, "loss": 2.8208, "step": 580 }, { "epoch": 0.04, "learning_rate": 1.7353775577956178e-06, "loss": 2.2377, "step": 582 }, { "epoch": 0.04, "learning_rate": 1.7414136536488199e-06, "loss": 3.2501, "step": 584 }, { "epoch": 0.04, "learning_rate": 1.7474497495020222e-06, "loss": 3.5237, "step": 586 }, { "epoch": 0.04, "learning_rate": 1.7534858453552243e-06, "loss": 3.7244, "step": 588 }, { "epoch": 0.04, "learning_rate": 1.7595219412084266e-06, "loss": 3.4353, "step": 590 }, { "epoch": 0.04, "learning_rate": 1.7655580370616286e-06, "loss": 3.5764, "step": 592 }, { "epoch": 0.04, "learning_rate": 1.771594132914831e-06, "loss": 3.3612, "step": 594 }, { "epoch": 0.04, "learning_rate": 1.7776302287680328e-06, "loss": 3.4932, "step": 596 }, { "epoch": 0.04, "learning_rate": 1.783666324621235e-06, "loss": 3.2602, "step": 598 }, { "epoch": 0.04, "learning_rate": 1.7897024204744372e-06, "loss": 3.128, "step": 600 }, { "epoch": 0.04, "learning_rate": 1.7957385163276393e-06, "loss": 3.1888, "step": 602 }, { "epoch": 0.04, "learning_rate": 1.8017746121808416e-06, "loss": 2.9256, "step": 604 }, { "epoch": 0.04, "learning_rate": 1.8078107080340437e-06, "loss": 3.8179, "step": 606 }, { "epoch": 0.04, "learning_rate": 1.813846803887246e-06, "loss": 3.3194, "step": 608 }, { "epoch": 0.04, "learning_rate": 1.8198828997404479e-06, "loss": 3.0185, "step": 610 }, { "epoch": 0.04, "learning_rate": 1.82591899559365e-06, "loss": 3.6034, "step": 612 }, { "epoch": 0.04, "learning_rate": 1.8319550914468523e-06, "loss": 2.9802, "step": 614 }, { "epoch": 0.04, "learning_rate": 1.8379911873000544e-06, "loss": 3.3134, "step": 616 }, { "epoch": 0.04, "learning_rate": 1.8440272831532567e-06, "loss": 3.5019, "step": 618 }, { "epoch": 0.04, "learning_rate": 1.8500633790064587e-06, "loss": 2.794, "step": 620 }, { "epoch": 0.04, "learning_rate": 1.856099474859661e-06, "loss": 3.0259, "step": 622 }, { "epoch": 0.04, "learning_rate": 1.862135570712863e-06, "loss": 3.39, "step": 624 }, { "epoch": 0.04, "learning_rate": 1.868171666566065e-06, "loss": 2.6833, "step": 626 }, { "epoch": 0.04, "learning_rate": 1.8742077624192673e-06, "loss": 3.0818, "step": 628 }, { "epoch": 0.04, "learning_rate": 1.8802438582724694e-06, "loss": 3.1044, "step": 630 }, { "epoch": 0.04, "learning_rate": 1.8862799541256717e-06, "loss": 3.1241, "step": 632 }, { "epoch": 0.04, "learning_rate": 1.8923160499788738e-06, "loss": 3.2367, "step": 634 }, { "epoch": 0.04, "learning_rate": 1.898352145832076e-06, "loss": 3.1184, "step": 636 }, { "epoch": 0.04, "learning_rate": 1.904388241685278e-06, "loss": 3.8658, "step": 638 }, { "epoch": 0.04, "learning_rate": 1.91042433753848e-06, "loss": 3.5826, "step": 640 }, { "epoch": 0.04, "learning_rate": 1.9164604333916826e-06, "loss": 2.8831, "step": 642 }, { "epoch": 0.04, "learning_rate": 1.9224965292448842e-06, "loss": 2.7834, "step": 644 }, { "epoch": 0.04, "learning_rate": 1.9285326250980868e-06, "loss": 2.8425, "step": 646 }, { "epoch": 0.04, "learning_rate": 1.934568720951289e-06, "loss": 2.7064, "step": 648 }, { "epoch": 0.04, "learning_rate": 1.940604816804491e-06, "loss": 2.7312, "step": 650 }, { "epoch": 0.04, "learning_rate": 1.946640912657693e-06, "loss": 3.3009, "step": 652 }, { "epoch": 0.04, "learning_rate": 1.952677008510895e-06, "loss": 3.378, "step": 654 }, { "epoch": 0.04, "learning_rate": 1.9587131043640976e-06, "loss": 3.2142, "step": 656 }, { "epoch": 0.04, "learning_rate": 1.9647492002172993e-06, "loss": 2.7, "step": 658 }, { "epoch": 0.04, "learning_rate": 1.970785296070502e-06, "loss": 3.3326, "step": 660 }, { "epoch": 0.04, "learning_rate": 1.976821391923704e-06, "loss": 2.7705, "step": 662 }, { "epoch": 0.04, "learning_rate": 1.982857487776906e-06, "loss": 4.0469, "step": 664 }, { "epoch": 0.04, "learning_rate": 1.988893583630108e-06, "loss": 3.5934, "step": 666 }, { "epoch": 0.04, "learning_rate": 1.99492967948331e-06, "loss": 2.6605, "step": 668 }, { "epoch": 0.04, "learning_rate": 2.0009657753365127e-06, "loss": 2.8015, "step": 670 }, { "epoch": 0.04, "learning_rate": 2.0070018711897143e-06, "loss": 2.891, "step": 672 }, { "epoch": 0.04, "learning_rate": 2.013037967042917e-06, "loss": 3.063, "step": 674 }, { "epoch": 0.04, "learning_rate": 2.019074062896119e-06, "loss": 3.1426, "step": 676 }, { "epoch": 0.04, "learning_rate": 2.025110158749321e-06, "loss": 3.0866, "step": 678 }, { "epoch": 0.04, "learning_rate": 2.031146254602523e-06, "loss": 2.5907, "step": 680 }, { "epoch": 0.04, "learning_rate": 2.037182350455725e-06, "loss": 2.4373, "step": 682 }, { "epoch": 0.04, "learning_rate": 2.0432184463089277e-06, "loss": 3.1531, "step": 684 }, { "epoch": 0.04, "learning_rate": 2.0492545421621294e-06, "loss": 3.3823, "step": 686 }, { "epoch": 0.04, "learning_rate": 2.055290638015332e-06, "loss": 2.8776, "step": 688 }, { "epoch": 0.04, "learning_rate": 2.061326733868534e-06, "loss": 2.8164, "step": 690 }, { "epoch": 0.04, "learning_rate": 2.067362829721736e-06, "loss": 3.1098, "step": 692 }, { "epoch": 0.04, "learning_rate": 2.073398925574938e-06, "loss": 3.2972, "step": 694 }, { "epoch": 0.04, "learning_rate": 2.0794350214281402e-06, "loss": 2.8249, "step": 696 }, { "epoch": 0.04, "learning_rate": 2.0854711172813428e-06, "loss": 3.1051, "step": 698 }, { "epoch": 0.04, "learning_rate": 2.0915072131345444e-06, "loss": 2.9447, "step": 700 }, { "epoch": 0.04, "learning_rate": 2.097543308987747e-06, "loss": 3.1943, "step": 702 }, { "epoch": 0.04, "learning_rate": 2.103579404840949e-06, "loss": 3.1506, "step": 704 }, { "epoch": 0.04, "learning_rate": 2.109615500694151e-06, "loss": 2.4647, "step": 706 }, { "epoch": 0.04, "learning_rate": 2.115651596547353e-06, "loss": 2.964, "step": 708 }, { "epoch": 0.04, "learning_rate": 2.1216876924005553e-06, "loss": 2.3707, "step": 710 }, { "epoch": 0.04, "learning_rate": 2.127723788253758e-06, "loss": 2.938, "step": 712 }, { "epoch": 0.04, "learning_rate": 2.1337598841069595e-06, "loss": 3.002, "step": 714 }, { "epoch": 0.04, "learning_rate": 2.139795979960162e-06, "loss": 2.967, "step": 716 }, { "epoch": 0.04, "learning_rate": 2.145832075813364e-06, "loss": 3.1926, "step": 718 }, { "epoch": 0.04, "learning_rate": 2.151868171666566e-06, "loss": 2.3176, "step": 720 }, { "epoch": 0.04, "learning_rate": 2.1579042675197683e-06, "loss": 2.3749, "step": 722 }, { "epoch": 0.04, "learning_rate": 2.1639403633729703e-06, "loss": 2.787, "step": 724 }, { "epoch": 0.04, "learning_rate": 2.169976459226173e-06, "loss": 2.9403, "step": 726 }, { "epoch": 0.04, "learning_rate": 2.1760125550793745e-06, "loss": 3.1704, "step": 728 }, { "epoch": 0.04, "learning_rate": 2.182048650932577e-06, "loss": 3.1137, "step": 730 }, { "epoch": 0.04, "learning_rate": 2.188084746785779e-06, "loss": 2.8336, "step": 732 }, { "epoch": 0.04, "learning_rate": 2.1941208426389812e-06, "loss": 3.761, "step": 734 }, { "epoch": 0.04, "learning_rate": 2.2001569384921833e-06, "loss": 3.3272, "step": 736 }, { "epoch": 0.04, "learning_rate": 2.2061930343453854e-06, "loss": 3.1208, "step": 738 }, { "epoch": 0.04, "learning_rate": 2.212229130198588e-06, "loss": 3.0743, "step": 740 }, { "epoch": 0.04, "learning_rate": 2.2182652260517896e-06, "loss": 3.045, "step": 742 }, { "epoch": 0.04, "learning_rate": 2.224301321904992e-06, "loss": 2.9993, "step": 744 }, { "epoch": 0.05, "learning_rate": 2.230337417758194e-06, "loss": 2.9975, "step": 746 }, { "epoch": 0.05, "learning_rate": 2.2363735136113963e-06, "loss": 3.1561, "step": 748 }, { "epoch": 0.05, "learning_rate": 2.2424096094645984e-06, "loss": 2.2974, "step": 750 }, { "epoch": 0.05, "learning_rate": 2.2484457053178004e-06, "loss": 2.4949, "step": 752 }, { "epoch": 0.05, "learning_rate": 2.254481801171003e-06, "loss": 2.6208, "step": 754 }, { "epoch": 0.05, "learning_rate": 2.2605178970242046e-06, "loss": 2.7056, "step": 756 }, { "epoch": 0.05, "learning_rate": 2.266553992877407e-06, "loss": 3.1087, "step": 758 }, { "epoch": 0.05, "learning_rate": 2.2725900887306092e-06, "loss": 3.1075, "step": 760 }, { "epoch": 0.05, "learning_rate": 2.2786261845838113e-06, "loss": 3.0956, "step": 762 }, { "epoch": 0.05, "learning_rate": 2.2846622804370134e-06, "loss": 3.0782, "step": 764 }, { "epoch": 0.05, "learning_rate": 2.2906983762902155e-06, "loss": 2.8229, "step": 766 }, { "epoch": 0.05, "learning_rate": 2.296734472143418e-06, "loss": 3.0271, "step": 768 }, { "epoch": 0.05, "learning_rate": 2.3027705679966197e-06, "loss": 2.4753, "step": 770 }, { "epoch": 0.05, "learning_rate": 2.308806663849822e-06, "loss": 2.7933, "step": 772 }, { "epoch": 0.05, "learning_rate": 2.3148427597030243e-06, "loss": 2.4667, "step": 774 }, { "epoch": 0.05, "learning_rate": 2.3208788555562264e-06, "loss": 2.8174, "step": 776 }, { "epoch": 0.05, "learning_rate": 2.3269149514094284e-06, "loss": 2.8554, "step": 778 }, { "epoch": 0.05, "learning_rate": 2.3329510472626305e-06, "loss": 2.8909, "step": 780 }, { "epoch": 0.05, "learning_rate": 2.338987143115833e-06, "loss": 2.4176, "step": 782 }, { "epoch": 0.05, "learning_rate": 2.3450232389690347e-06, "loss": 3.0384, "step": 784 }, { "epoch": 0.05, "learning_rate": 2.3510593348222372e-06, "loss": 3.0306, "step": 786 }, { "epoch": 0.05, "learning_rate": 2.3570954306754393e-06, "loss": 2.73, "step": 788 }, { "epoch": 0.05, "learning_rate": 2.3631315265286414e-06, "loss": 2.8732, "step": 790 }, { "epoch": 0.05, "learning_rate": 2.3691676223818435e-06, "loss": 2.8171, "step": 792 }, { "epoch": 0.05, "learning_rate": 2.3752037182350456e-06, "loss": 2.6983, "step": 794 }, { "epoch": 0.05, "learning_rate": 2.381239814088248e-06, "loss": 2.6043, "step": 796 }, { "epoch": 0.05, "learning_rate": 2.3872759099414498e-06, "loss": 3.1073, "step": 798 }, { "epoch": 0.05, "learning_rate": 2.3933120057946523e-06, "loss": 2.9928, "step": 800 }, { "epoch": 0.05, "learning_rate": 2.3993481016478544e-06, "loss": 2.3456, "step": 802 }, { "epoch": 0.05, "learning_rate": 2.4053841975010565e-06, "loss": 3.0018, "step": 804 }, { "epoch": 0.05, "learning_rate": 2.4114202933542585e-06, "loss": 2.9814, "step": 806 }, { "epoch": 0.05, "learning_rate": 2.4174563892074606e-06, "loss": 2.4818, "step": 808 }, { "epoch": 0.05, "learning_rate": 2.423492485060663e-06, "loss": 3.4806, "step": 810 }, { "epoch": 0.05, "learning_rate": 2.429528580913865e-06, "loss": 2.6425, "step": 812 }, { "epoch": 0.05, "learning_rate": 2.4355646767670673e-06, "loss": 2.8857, "step": 814 }, { "epoch": 0.05, "learning_rate": 2.4416007726202694e-06, "loss": 2.685, "step": 816 }, { "epoch": 0.05, "learning_rate": 2.4476368684734715e-06, "loss": 3.0317, "step": 818 }, { "epoch": 0.05, "learning_rate": 2.4536729643266736e-06, "loss": 2.7044, "step": 820 }, { "epoch": 0.05, "learning_rate": 2.4597090601798757e-06, "loss": 2.7928, "step": 822 }, { "epoch": 0.05, "learning_rate": 2.465745156033078e-06, "loss": 2.3706, "step": 824 }, { "epoch": 0.05, "learning_rate": 2.47178125188628e-06, "loss": 2.5899, "step": 826 }, { "epoch": 0.05, "learning_rate": 2.4778173477394824e-06, "loss": 2.4988, "step": 828 }, { "epoch": 0.05, "learning_rate": 2.4838534435926845e-06, "loss": 3.1348, "step": 830 }, { "epoch": 0.05, "learning_rate": 2.4898895394458865e-06, "loss": 3.2034, "step": 832 }, { "epoch": 0.05, "learning_rate": 2.4959256352990886e-06, "loss": 2.5677, "step": 834 }, { "epoch": 0.05, "learning_rate": 2.5019617311522907e-06, "loss": 2.8781, "step": 836 }, { "epoch": 0.05, "learning_rate": 2.5079978270054932e-06, "loss": 2.9319, "step": 838 }, { "epoch": 0.05, "learning_rate": 2.514033922858695e-06, "loss": 2.8165, "step": 840 }, { "epoch": 0.05, "learning_rate": 2.5200700187118974e-06, "loss": 2.8147, "step": 842 }, { "epoch": 0.05, "learning_rate": 2.5261061145650995e-06, "loss": 2.6757, "step": 844 }, { "epoch": 0.05, "learning_rate": 2.5321422104183016e-06, "loss": 2.8099, "step": 846 }, { "epoch": 0.05, "learning_rate": 2.5381783062715037e-06, "loss": 3.0371, "step": 848 }, { "epoch": 0.05, "learning_rate": 2.5442144021247058e-06, "loss": 2.6826, "step": 850 }, { "epoch": 0.05, "learning_rate": 2.5502504979779083e-06, "loss": 2.5326, "step": 852 }, { "epoch": 0.05, "learning_rate": 2.55628659383111e-06, "loss": 2.7641, "step": 854 }, { "epoch": 0.05, "learning_rate": 2.5623226896843125e-06, "loss": 2.941, "step": 856 }, { "epoch": 0.05, "learning_rate": 2.5683587855375146e-06, "loss": 3.001, "step": 858 }, { "epoch": 0.05, "learning_rate": 2.5743948813907166e-06, "loss": 2.6757, "step": 860 }, { "epoch": 0.05, "learning_rate": 2.5804309772439187e-06, "loss": 2.4132, "step": 862 }, { "epoch": 0.05, "learning_rate": 2.586467073097121e-06, "loss": 3.1929, "step": 864 }, { "epoch": 0.05, "learning_rate": 2.5925031689503233e-06, "loss": 2.7021, "step": 866 }, { "epoch": 0.05, "learning_rate": 2.598539264803525e-06, "loss": 2.7515, "step": 868 }, { "epoch": 0.05, "learning_rate": 2.6045753606567275e-06, "loss": 2.2607, "step": 870 }, { "epoch": 0.05, "learning_rate": 2.6106114565099296e-06, "loss": 2.5672, "step": 872 }, { "epoch": 0.05, "learning_rate": 2.6166475523631317e-06, "loss": 2.7251, "step": 874 }, { "epoch": 0.05, "learning_rate": 2.6226836482163338e-06, "loss": 2.4853, "step": 876 }, { "epoch": 0.05, "learning_rate": 2.628719744069536e-06, "loss": 2.4278, "step": 878 }, { "epoch": 0.05, "learning_rate": 2.634755839922738e-06, "loss": 2.7445, "step": 880 }, { "epoch": 0.05, "learning_rate": 2.64079193577594e-06, "loss": 2.4881, "step": 882 }, { "epoch": 0.05, "learning_rate": 2.6468280316291426e-06, "loss": 2.1576, "step": 884 }, { "epoch": 0.05, "learning_rate": 2.6528641274823446e-06, "loss": 2.8573, "step": 886 }, { "epoch": 0.05, "learning_rate": 2.6589002233355467e-06, "loss": 2.9177, "step": 888 }, { "epoch": 0.05, "learning_rate": 2.664936319188749e-06, "loss": 2.9603, "step": 890 }, { "epoch": 0.05, "learning_rate": 2.670972415041951e-06, "loss": 2.888, "step": 892 }, { "epoch": 0.05, "learning_rate": 2.677008510895153e-06, "loss": 2.2032, "step": 894 }, { "epoch": 0.05, "learning_rate": 2.683044606748355e-06, "loss": 2.6634, "step": 896 }, { "epoch": 0.05, "learning_rate": 2.6890807026015576e-06, "loss": 2.5843, "step": 898 }, { "epoch": 0.05, "learning_rate": 2.6951167984547597e-06, "loss": 2.9446, "step": 900 }, { "epoch": 0.05, "learning_rate": 2.7011528943079618e-06, "loss": 2.4518, "step": 902 }, { "epoch": 0.05, "learning_rate": 2.707188990161164e-06, "loss": 2.6606, "step": 904 }, { "epoch": 0.05, "learning_rate": 2.713225086014366e-06, "loss": 2.6793, "step": 906 }, { "epoch": 0.05, "learning_rate": 2.719261181867568e-06, "loss": 2.5038, "step": 908 }, { "epoch": 0.05, "learning_rate": 2.72529727772077e-06, "loss": 2.8233, "step": 910 }, { "epoch": 0.06, "learning_rate": 2.7313333735739727e-06, "loss": 2.3882, "step": 912 }, { "epoch": 0.06, "learning_rate": 2.7373694694271747e-06, "loss": 2.4259, "step": 914 }, { "epoch": 0.06, "learning_rate": 2.743405565280377e-06, "loss": 2.7738, "step": 916 }, { "epoch": 0.06, "learning_rate": 2.749441661133579e-06, "loss": 2.2671, "step": 918 }, { "epoch": 0.06, "learning_rate": 2.755477756986781e-06, "loss": 2.27, "step": 920 }, { "epoch": 0.06, "learning_rate": 2.761513852839983e-06, "loss": 2.9642, "step": 922 }, { "epoch": 0.06, "learning_rate": 2.767549948693185e-06, "loss": 2.7758, "step": 924 }, { "epoch": 0.06, "learning_rate": 2.7735860445463877e-06, "loss": 2.4951, "step": 926 }, { "epoch": 0.06, "learning_rate": 2.77962214039959e-06, "loss": 2.6907, "step": 928 }, { "epoch": 0.06, "learning_rate": 2.785658236252792e-06, "loss": 2.8699, "step": 930 }, { "epoch": 0.06, "learning_rate": 2.791694332105994e-06, "loss": 2.891, "step": 932 }, { "epoch": 0.06, "learning_rate": 2.797730427959196e-06, "loss": 2.6109, "step": 934 }, { "epoch": 0.06, "learning_rate": 2.803766523812398e-06, "loss": 2.2325, "step": 936 }, { "epoch": 0.06, "learning_rate": 2.8098026196656002e-06, "loss": 2.5436, "step": 938 }, { "epoch": 0.06, "learning_rate": 2.8158387155188027e-06, "loss": 2.8025, "step": 940 }, { "epoch": 0.06, "learning_rate": 2.821874811372005e-06, "loss": 2.5036, "step": 942 }, { "epoch": 0.06, "learning_rate": 2.8279109072252065e-06, "loss": 2.5834, "step": 944 }, { "epoch": 0.06, "learning_rate": 2.833947003078409e-06, "loss": 2.5402, "step": 946 }, { "epoch": 0.06, "learning_rate": 2.839983098931611e-06, "loss": 2.954, "step": 948 }, { "epoch": 0.06, "learning_rate": 2.846019194784813e-06, "loss": 2.2881, "step": 950 }, { "epoch": 0.06, "learning_rate": 2.8520552906380153e-06, "loss": 2.6567, "step": 952 }, { "epoch": 0.06, "learning_rate": 2.858091386491218e-06, "loss": 2.0207, "step": 954 }, { "epoch": 0.06, "learning_rate": 2.86412748234442e-06, "loss": 2.4373, "step": 956 }, { "epoch": 0.06, "learning_rate": 2.8701635781976216e-06, "loss": 2.5319, "step": 958 }, { "epoch": 0.06, "learning_rate": 2.876199674050824e-06, "loss": 2.3624, "step": 960 }, { "epoch": 0.06, "learning_rate": 2.882235769904026e-06, "loss": 2.2062, "step": 962 }, { "epoch": 0.06, "learning_rate": 2.8882718657572282e-06, "loss": 3.0351, "step": 964 }, { "epoch": 0.06, "learning_rate": 2.8943079616104303e-06, "loss": 2.856, "step": 966 }, { "epoch": 0.06, "learning_rate": 2.900344057463633e-06, "loss": 2.7833, "step": 968 }, { "epoch": 0.06, "learning_rate": 2.906380153316835e-06, "loss": 2.5611, "step": 970 }, { "epoch": 0.06, "learning_rate": 2.9124162491700366e-06, "loss": 2.622, "step": 972 }, { "epoch": 0.06, "learning_rate": 2.918452345023239e-06, "loss": 2.543, "step": 974 }, { "epoch": 0.06, "learning_rate": 2.924488440876441e-06, "loss": 2.3011, "step": 976 }, { "epoch": 0.06, "learning_rate": 2.9305245367296433e-06, "loss": 2.1305, "step": 978 }, { "epoch": 0.06, "learning_rate": 2.9365606325828454e-06, "loss": 2.2576, "step": 980 }, { "epoch": 0.06, "learning_rate": 2.942596728436048e-06, "loss": 2.4068, "step": 982 }, { "epoch": 0.06, "learning_rate": 2.94863282428925e-06, "loss": 2.2406, "step": 984 }, { "epoch": 0.06, "learning_rate": 2.9546689201424516e-06, "loss": 2.9537, "step": 986 }, { "epoch": 0.06, "learning_rate": 2.960705015995654e-06, "loss": 2.7028, "step": 988 }, { "epoch": 0.06, "learning_rate": 2.9667411118488562e-06, "loss": 2.3991, "step": 990 }, { "epoch": 0.06, "learning_rate": 2.9727772077020583e-06, "loss": 2.3192, "step": 992 }, { "epoch": 0.06, "learning_rate": 2.9788133035552604e-06, "loss": 2.6921, "step": 994 }, { "epoch": 0.06, "learning_rate": 2.984849399408463e-06, "loss": 2.6509, "step": 996 }, { "epoch": 0.06, "learning_rate": 2.990885495261665e-06, "loss": 3.1303, "step": 998 }, { "epoch": 0.06, "learning_rate": 2.9969215911148667e-06, "loss": 2.3591, "step": 1000 }, { "epoch": 0.06, "learning_rate": 3.002957686968069e-06, "loss": 2.5376, "step": 1002 }, { "epoch": 0.06, "learning_rate": 3.0089937828212713e-06, "loss": 2.5465, "step": 1004 }, { "epoch": 0.06, "learning_rate": 3.0150298786744734e-06, "loss": 2.3886, "step": 1006 }, { "epoch": 0.06, "learning_rate": 3.0210659745276755e-06, "loss": 2.1726, "step": 1008 }, { "epoch": 0.06, "learning_rate": 3.027102070380878e-06, "loss": 2.5477, "step": 1010 }, { "epoch": 0.06, "learning_rate": 3.03313816623408e-06, "loss": 2.6417, "step": 1012 }, { "epoch": 0.06, "learning_rate": 3.0391742620872817e-06, "loss": 2.7609, "step": 1014 }, { "epoch": 0.06, "learning_rate": 3.0452103579404843e-06, "loss": 2.3949, "step": 1016 }, { "epoch": 0.06, "learning_rate": 3.0512464537936863e-06, "loss": 2.4817, "step": 1018 }, { "epoch": 0.06, "learning_rate": 3.0572825496468884e-06, "loss": 2.0452, "step": 1020 }, { "epoch": 0.06, "learning_rate": 3.0633186455000905e-06, "loss": 2.6061, "step": 1022 }, { "epoch": 0.06, "learning_rate": 3.069354741353293e-06, "loss": 2.94, "step": 1024 }, { "epoch": 0.06, "learning_rate": 3.075390837206495e-06, "loss": 2.7724, "step": 1026 }, { "epoch": 0.06, "learning_rate": 3.0814269330596968e-06, "loss": 2.1715, "step": 1028 }, { "epoch": 0.06, "learning_rate": 3.0874630289128993e-06, "loss": 2.0949, "step": 1030 }, { "epoch": 0.06, "learning_rate": 3.0934991247661014e-06, "loss": 2.5225, "step": 1032 }, { "epoch": 0.06, "learning_rate": 3.0995352206193035e-06, "loss": 2.189, "step": 1034 }, { "epoch": 0.06, "learning_rate": 3.1055713164725056e-06, "loss": 2.8, "step": 1036 }, { "epoch": 0.06, "learning_rate": 3.111607412325708e-06, "loss": 2.7138, "step": 1038 }, { "epoch": 0.06, "learning_rate": 3.11764350817891e-06, "loss": 2.3621, "step": 1040 }, { "epoch": 0.06, "learning_rate": 3.123679604032112e-06, "loss": 2.726, "step": 1042 }, { "epoch": 0.06, "learning_rate": 3.129715699885314e-06, "loss": 2.4293, "step": 1044 }, { "epoch": 0.06, "learning_rate": 3.135751795738517e-06, "loss": 2.8536, "step": 1046 }, { "epoch": 0.06, "learning_rate": 3.1417878915917185e-06, "loss": 2.4997, "step": 1048 }, { "epoch": 0.06, "learning_rate": 3.1478239874449206e-06, "loss": 2.7168, "step": 1050 }, { "epoch": 0.06, "learning_rate": 3.1538600832981227e-06, "loss": 2.3717, "step": 1052 }, { "epoch": 0.06, "learning_rate": 3.1598961791513252e-06, "loss": 2.2821, "step": 1054 }, { "epoch": 0.06, "learning_rate": 3.1659322750045273e-06, "loss": 2.3718, "step": 1056 }, { "epoch": 0.06, "learning_rate": 3.1719683708577294e-06, "loss": 2.2403, "step": 1058 }, { "epoch": 0.06, "learning_rate": 3.1780044667109315e-06, "loss": 2.0922, "step": 1060 }, { "epoch": 0.06, "learning_rate": 3.1840405625641336e-06, "loss": 2.3835, "step": 1062 }, { "epoch": 0.06, "learning_rate": 3.190076658417336e-06, "loss": 2.2675, "step": 1064 }, { "epoch": 0.06, "learning_rate": 3.196112754270538e-06, "loss": 2.7488, "step": 1066 }, { "epoch": 0.06, "learning_rate": 3.2021488501237403e-06, "loss": 2.1118, "step": 1068 }, { "epoch": 0.06, "learning_rate": 3.208184945976942e-06, "loss": 2.4171, "step": 1070 }, { "epoch": 0.06, "learning_rate": 3.214221041830144e-06, "loss": 2.5757, "step": 1072 }, { "epoch": 0.06, "learning_rate": 3.220257137683347e-06, "loss": 2.4707, "step": 1074 }, { "epoch": 0.06, "learning_rate": 3.2262932335365486e-06, "loss": 2.1329, "step": 1076 }, { "epoch": 0.07, "learning_rate": 3.2323293293897507e-06, "loss": 2.8618, "step": 1078 }, { "epoch": 0.07, "learning_rate": 3.238365425242953e-06, "loss": 2.4211, "step": 1080 }, { "epoch": 0.07, "learning_rate": 3.2444015210961553e-06, "loss": 1.8236, "step": 1082 }, { "epoch": 0.07, "learning_rate": 3.2504376169493574e-06, "loss": 2.2477, "step": 1084 }, { "epoch": 0.07, "learning_rate": 3.2564737128025595e-06, "loss": 2.1637, "step": 1086 }, { "epoch": 0.07, "learning_rate": 3.2625098086557616e-06, "loss": 2.6983, "step": 1088 }, { "epoch": 0.07, "learning_rate": 3.2685459045089637e-06, "loss": 2.1878, "step": 1090 }, { "epoch": 0.07, "learning_rate": 3.274582000362166e-06, "loss": 1.6509, "step": 1092 }, { "epoch": 0.07, "learning_rate": 3.2806180962153683e-06, "loss": 2.0083, "step": 1094 }, { "epoch": 0.07, "learning_rate": 3.2866541920685704e-06, "loss": 2.6962, "step": 1096 }, { "epoch": 0.07, "learning_rate": 3.292690287921772e-06, "loss": 2.523, "step": 1098 }, { "epoch": 0.07, "learning_rate": 3.298726383774974e-06, "loss": 2.3391, "step": 1100 }, { "epoch": 0.07, "learning_rate": 3.304762479628177e-06, "loss": 2.658, "step": 1102 }, { "epoch": 0.07, "learning_rate": 3.3107985754813787e-06, "loss": 2.1884, "step": 1104 }, { "epoch": 0.07, "learning_rate": 3.316834671334581e-06, "loss": 2.2637, "step": 1106 }, { "epoch": 0.07, "learning_rate": 3.322870767187783e-06, "loss": 2.5305, "step": 1108 }, { "epoch": 0.07, "learning_rate": 3.3289068630409854e-06, "loss": 2.5952, "step": 1110 }, { "epoch": 0.07, "learning_rate": 3.3349429588941875e-06, "loss": 2.0384, "step": 1112 }, { "epoch": 0.07, "learning_rate": 3.3409790547473896e-06, "loss": 2.4144, "step": 1114 }, { "epoch": 0.07, "learning_rate": 3.3470151506005917e-06, "loss": 2.2756, "step": 1116 }, { "epoch": 0.07, "learning_rate": 3.3530512464537938e-06, "loss": 2.5856, "step": 1118 }, { "epoch": 0.07, "learning_rate": 3.3590873423069963e-06, "loss": 2.5791, "step": 1120 }, { "epoch": 0.07, "learning_rate": 3.3651234381601984e-06, "loss": 2.6113, "step": 1122 }, { "epoch": 0.07, "learning_rate": 3.3711595340134005e-06, "loss": 2.5179, "step": 1124 }, { "epoch": 0.07, "learning_rate": 3.377195629866602e-06, "loss": 2.3817, "step": 1126 }, { "epoch": 0.07, "learning_rate": 3.383231725719804e-06, "loss": 2.254, "step": 1128 }, { "epoch": 0.07, "learning_rate": 3.389267821573007e-06, "loss": 2.5646, "step": 1130 }, { "epoch": 0.07, "learning_rate": 3.395303917426209e-06, "loss": 2.3408, "step": 1132 }, { "epoch": 0.07, "learning_rate": 3.401340013279411e-06, "loss": 2.0616, "step": 1134 }, { "epoch": 0.07, "learning_rate": 3.407376109132613e-06, "loss": 2.1062, "step": 1136 }, { "epoch": 0.07, "learning_rate": 3.4134122049858155e-06, "loss": 1.8076, "step": 1138 }, { "epoch": 0.07, "learning_rate": 3.4194483008390176e-06, "loss": 2.1056, "step": 1140 }, { "epoch": 0.07, "learning_rate": 3.4254843966922197e-06, "loss": 2.3638, "step": 1142 }, { "epoch": 0.07, "learning_rate": 3.4315204925454218e-06, "loss": 1.7532, "step": 1144 }, { "epoch": 0.07, "learning_rate": 3.437556588398624e-06, "loss": 2.1699, "step": 1146 }, { "epoch": 0.07, "learning_rate": 3.4435926842518264e-06, "loss": 1.8287, "step": 1148 }, { "epoch": 0.07, "learning_rate": 3.4496287801050285e-06, "loss": 2.5079, "step": 1150 }, { "epoch": 0.07, "learning_rate": 3.4556648759582305e-06, "loss": 2.0705, "step": 1152 }, { "epoch": 0.07, "learning_rate": 3.4617009718114322e-06, "loss": 2.222, "step": 1154 }, { "epoch": 0.07, "learning_rate": 3.4677370676646343e-06, "loss": 2.4302, "step": 1156 }, { "epoch": 0.07, "learning_rate": 3.4737731635178372e-06, "loss": 1.7067, "step": 1158 }, { "epoch": 0.07, "learning_rate": 3.479809259371039e-06, "loss": 2.1507, "step": 1160 }, { "epoch": 0.07, "learning_rate": 3.485845355224241e-06, "loss": 1.9669, "step": 1162 }, { "epoch": 0.07, "learning_rate": 3.491881451077443e-06, "loss": 1.9887, "step": 1164 }, { "epoch": 0.07, "learning_rate": 3.4979175469306456e-06, "loss": 2.2431, "step": 1166 }, { "epoch": 0.07, "learning_rate": 3.5039536427838477e-06, "loss": 2.2312, "step": 1168 }, { "epoch": 0.07, "learning_rate": 3.5099897386370498e-06, "loss": 2.3152, "step": 1170 }, { "epoch": 0.07, "learning_rate": 3.516025834490252e-06, "loss": 2.0005, "step": 1172 }, { "epoch": 0.07, "learning_rate": 3.522061930343454e-06, "loss": 2.4826, "step": 1174 }, { "epoch": 0.07, "learning_rate": 3.5280980261966565e-06, "loss": 2.1861, "step": 1176 }, { "epoch": 0.07, "learning_rate": 3.5341341220498586e-06, "loss": 1.863, "step": 1178 }, { "epoch": 0.07, "learning_rate": 3.5401702179030606e-06, "loss": 2.4304, "step": 1180 }, { "epoch": 0.07, "learning_rate": 3.5462063137562623e-06, "loss": 1.9352, "step": 1182 }, { "epoch": 0.07, "learning_rate": 3.5522424096094644e-06, "loss": 2.4623, "step": 1184 }, { "epoch": 0.07, "learning_rate": 3.5582785054626673e-06, "loss": 2.5849, "step": 1186 }, { "epoch": 0.07, "learning_rate": 3.564314601315869e-06, "loss": 2.6356, "step": 1188 }, { "epoch": 0.07, "learning_rate": 3.570350697169071e-06, "loss": 2.1213, "step": 1190 }, { "epoch": 0.07, "learning_rate": 3.576386793022273e-06, "loss": 2.3881, "step": 1192 }, { "epoch": 0.07, "learning_rate": 3.5824228888754753e-06, "loss": 2.3887, "step": 1194 }, { "epoch": 0.07, "learning_rate": 3.5884589847286778e-06, "loss": 2.3674, "step": 1196 }, { "epoch": 0.07, "learning_rate": 3.59449508058188e-06, "loss": 2.7366, "step": 1198 }, { "epoch": 0.07, "learning_rate": 3.600531176435082e-06, "loss": 1.8281, "step": 1200 }, { "epoch": 0.07, "learning_rate": 3.606567272288284e-06, "loss": 2.6025, "step": 1202 }, { "epoch": 0.07, "learning_rate": 3.6126033681414866e-06, "loss": 2.309, "step": 1204 }, { "epoch": 0.07, "learning_rate": 3.6186394639946887e-06, "loss": 2.121, "step": 1206 }, { "epoch": 0.07, "learning_rate": 3.6246755598478907e-06, "loss": 2.0244, "step": 1208 }, { "epoch": 0.07, "learning_rate": 3.6307116557010924e-06, "loss": 1.927, "step": 1210 }, { "epoch": 0.07, "learning_rate": 3.6367477515542945e-06, "loss": 2.0422, "step": 1212 }, { "epoch": 0.07, "learning_rate": 3.6427838474074974e-06, "loss": 2.1372, "step": 1214 }, { "epoch": 0.07, "learning_rate": 3.648819943260699e-06, "loss": 1.9762, "step": 1216 }, { "epoch": 0.07, "learning_rate": 3.654856039113901e-06, "loss": 1.9158, "step": 1218 }, { "epoch": 0.07, "learning_rate": 3.6608921349671033e-06, "loss": 2.0641, "step": 1220 }, { "epoch": 0.07, "learning_rate": 3.6669282308203054e-06, "loss": 2.5005, "step": 1222 }, { "epoch": 0.07, "learning_rate": 3.672964326673508e-06, "loss": 2.3558, "step": 1224 }, { "epoch": 0.07, "learning_rate": 3.67900042252671e-06, "loss": 1.8034, "step": 1226 }, { "epoch": 0.07, "learning_rate": 3.685036518379912e-06, "loss": 2.1213, "step": 1228 }, { "epoch": 0.07, "learning_rate": 3.691072614233114e-06, "loss": 2.1486, "step": 1230 }, { "epoch": 0.07, "learning_rate": 3.6971087100863167e-06, "loss": 2.0399, "step": 1232 }, { "epoch": 0.07, "learning_rate": 3.7031448059395187e-06, "loss": 2.4597, "step": 1234 }, { "epoch": 0.07, "learning_rate": 3.709180901792721e-06, "loss": 2.0423, "step": 1236 }, { "epoch": 0.07, "learning_rate": 3.7152169976459225e-06, "loss": 2.4713, "step": 1238 }, { "epoch": 0.07, "learning_rate": 3.7212530934991246e-06, "loss": 1.9475, "step": 1240 }, { "epoch": 0.07, "learning_rate": 3.7272891893523275e-06, "loss": 2.0874, "step": 1242 }, { "epoch": 0.08, "learning_rate": 3.733325285205529e-06, "loss": 2.6202, "step": 1244 }, { "epoch": 0.08, "learning_rate": 3.7393613810587313e-06, "loss": 2.3321, "step": 1246 }, { "epoch": 0.08, "learning_rate": 3.7453974769119334e-06, "loss": 2.7036, "step": 1248 }, { "epoch": 0.08, "learning_rate": 3.7514335727651355e-06, "loss": 2.622, "step": 1250 }, { "epoch": 0.08, "learning_rate": 3.757469668618338e-06, "loss": 2.05, "step": 1252 }, { "epoch": 0.08, "learning_rate": 3.76350576447154e-06, "loss": 1.9706, "step": 1254 }, { "epoch": 0.08, "learning_rate": 3.769541860324742e-06, "loss": 2.2418, "step": 1256 }, { "epoch": 0.08, "learning_rate": 3.7755779561779442e-06, "loss": 2.1195, "step": 1258 }, { "epoch": 0.08, "learning_rate": 3.7816140520311468e-06, "loss": 2.4261, "step": 1260 }, { "epoch": 0.08, "learning_rate": 3.787650147884349e-06, "loss": 2.1147, "step": 1262 }, { "epoch": 0.08, "learning_rate": 3.793686243737551e-06, "loss": 3.0482, "step": 1264 }, { "epoch": 0.08, "learning_rate": 3.7997223395907526e-06, "loss": 2.4639, "step": 1266 }, { "epoch": 0.08, "learning_rate": 3.8057584354439547e-06, "loss": 2.2623, "step": 1268 }, { "epoch": 0.08, "learning_rate": 3.8117945312971576e-06, "loss": 1.9502, "step": 1270 }, { "epoch": 0.08, "learning_rate": 3.81783062715036e-06, "loss": 1.9393, "step": 1272 }, { "epoch": 0.08, "learning_rate": 3.823866723003561e-06, "loss": 2.0054, "step": 1274 }, { "epoch": 0.08, "learning_rate": 3.829902818856763e-06, "loss": 2.747, "step": 1276 }, { "epoch": 0.08, "learning_rate": 3.8359389147099656e-06, "loss": 2.2113, "step": 1278 }, { "epoch": 0.08, "learning_rate": 3.841975010563168e-06, "loss": 1.9945, "step": 1280 }, { "epoch": 0.08, "learning_rate": 3.84801110641637e-06, "loss": 1.6801, "step": 1282 }, { "epoch": 0.08, "learning_rate": 3.854047202269572e-06, "loss": 2.1488, "step": 1284 }, { "epoch": 0.08, "learning_rate": 3.860083298122774e-06, "loss": 2.0644, "step": 1286 }, { "epoch": 0.08, "learning_rate": 3.866119393975976e-06, "loss": 2.2608, "step": 1288 }, { "epoch": 0.08, "learning_rate": 3.872155489829179e-06, "loss": 2.2877, "step": 1290 }, { "epoch": 0.08, "learning_rate": 3.878191585682381e-06, "loss": 1.9623, "step": 1292 }, { "epoch": 0.08, "learning_rate": 3.884227681535583e-06, "loss": 2.3395, "step": 1294 }, { "epoch": 0.08, "learning_rate": 3.890263777388785e-06, "loss": 1.8777, "step": 1296 }, { "epoch": 0.08, "learning_rate": 3.896299873241987e-06, "loss": 2.2753, "step": 1298 }, { "epoch": 0.08, "learning_rate": 3.90233596909519e-06, "loss": 2.3395, "step": 1300 }, { "epoch": 0.08, "learning_rate": 3.9083720649483915e-06, "loss": 2.1487, "step": 1302 }, { "epoch": 0.08, "learning_rate": 3.914408160801593e-06, "loss": 2.0298, "step": 1304 }, { "epoch": 0.08, "learning_rate": 3.920444256654796e-06, "loss": 2.1007, "step": 1306 }, { "epoch": 0.08, "learning_rate": 3.926480352507998e-06, "loss": 2.4285, "step": 1308 }, { "epoch": 0.08, "learning_rate": 3.9325164483612e-06, "loss": 2.0977, "step": 1310 }, { "epoch": 0.08, "learning_rate": 3.938552544214402e-06, "loss": 1.9189, "step": 1312 }, { "epoch": 0.08, "learning_rate": 3.944588640067604e-06, "loss": 2.176, "step": 1314 }, { "epoch": 0.08, "learning_rate": 3.9506247359208065e-06, "loss": 2.3566, "step": 1316 }, { "epoch": 0.08, "learning_rate": 3.956660831774009e-06, "loss": 2.1042, "step": 1318 }, { "epoch": 0.08, "learning_rate": 3.962696927627211e-06, "loss": 2.165, "step": 1320 }, { "epoch": 0.08, "learning_rate": 3.968733023480413e-06, "loss": 2.2178, "step": 1322 }, { "epoch": 0.08, "learning_rate": 3.974769119333615e-06, "loss": 1.9688, "step": 1324 }, { "epoch": 0.08, "learning_rate": 3.980805215186817e-06, "loss": 2.3867, "step": 1326 }, { "epoch": 0.08, "learning_rate": 3.98684131104002e-06, "loss": 1.9631, "step": 1328 }, { "epoch": 0.08, "learning_rate": 3.9928774068932216e-06, "loss": 2.0472, "step": 1330 }, { "epoch": 0.08, "learning_rate": 3.998913502746423e-06, "loss": 2.3393, "step": 1332 }, { "epoch": 0.08, "learning_rate": 4.004949598599626e-06, "loss": 2.0667, "step": 1334 }, { "epoch": 0.08, "learning_rate": 4.010985694452828e-06, "loss": 1.8774, "step": 1336 }, { "epoch": 0.08, "learning_rate": 4.01702179030603e-06, "loss": 2.1721, "step": 1338 }, { "epoch": 0.08, "learning_rate": 4.0230578861592324e-06, "loss": 1.7749, "step": 1340 }, { "epoch": 0.08, "learning_rate": 4.029093982012434e-06, "loss": 2.1946, "step": 1342 }, { "epoch": 0.08, "learning_rate": 4.035130077865637e-06, "loss": 2.0909, "step": 1344 }, { "epoch": 0.08, "learning_rate": 4.041166173718839e-06, "loss": 2.3998, "step": 1346 }, { "epoch": 0.08, "learning_rate": 4.047202269572041e-06, "loss": 2.2121, "step": 1348 }, { "epoch": 0.08, "learning_rate": 4.053238365425243e-06, "loss": 2.0676, "step": 1350 }, { "epoch": 0.08, "learning_rate": 4.059274461278445e-06, "loss": 1.9915, "step": 1352 }, { "epoch": 0.08, "learning_rate": 4.0653105571316475e-06, "loss": 2.2493, "step": 1354 }, { "epoch": 0.08, "learning_rate": 4.07134665298485e-06, "loss": 1.8073, "step": 1356 }, { "epoch": 0.08, "learning_rate": 4.077382748838052e-06, "loss": 2.3829, "step": 1358 }, { "epoch": 0.08, "learning_rate": 4.083418844691253e-06, "loss": 2.0302, "step": 1360 }, { "epoch": 0.08, "learning_rate": 4.089454940544456e-06, "loss": 2.1802, "step": 1362 }, { "epoch": 0.08, "learning_rate": 4.095491036397658e-06, "loss": 2.1847, "step": 1364 }, { "epoch": 0.08, "learning_rate": 4.10152713225086e-06, "loss": 1.8642, "step": 1366 }, { "epoch": 0.08, "learning_rate": 4.1075632281040625e-06, "loss": 2.0391, "step": 1368 }, { "epoch": 0.08, "learning_rate": 4.113599323957264e-06, "loss": 1.7441, "step": 1370 }, { "epoch": 0.08, "learning_rate": 4.119635419810467e-06, "loss": 1.7279, "step": 1372 }, { "epoch": 0.08, "learning_rate": 4.125671515663669e-06, "loss": 2.2362, "step": 1374 }, { "epoch": 0.08, "learning_rate": 4.131707611516871e-06, "loss": 1.9671, "step": 1376 }, { "epoch": 0.08, "learning_rate": 4.137743707370073e-06, "loss": 1.7161, "step": 1378 }, { "epoch": 0.08, "learning_rate": 4.143779803223275e-06, "loss": 1.9398, "step": 1380 }, { "epoch": 0.08, "learning_rate": 4.149815899076478e-06, "loss": 2.3631, "step": 1382 }, { "epoch": 0.08, "learning_rate": 4.15585199492968e-06, "loss": 1.8303, "step": 1384 }, { "epoch": 0.08, "learning_rate": 4.161888090782882e-06, "loss": 2.1285, "step": 1386 }, { "epoch": 0.08, "learning_rate": 4.167924186636083e-06, "loss": 1.8706, "step": 1388 }, { "epoch": 0.08, "learning_rate": 4.173960282489286e-06, "loss": 1.8374, "step": 1390 }, { "epoch": 0.08, "learning_rate": 4.1799963783424884e-06, "loss": 1.7686, "step": 1392 }, { "epoch": 0.08, "learning_rate": 4.18603247419569e-06, "loss": 2.0689, "step": 1394 }, { "epoch": 0.08, "learning_rate": 4.192068570048893e-06, "loss": 2.3006, "step": 1396 }, { "epoch": 0.08, "learning_rate": 4.198104665902094e-06, "loss": 2.2192, "step": 1398 }, { "epoch": 0.08, "learning_rate": 4.204140761755297e-06, "loss": 1.8676, "step": 1400 }, { "epoch": 0.08, "learning_rate": 4.210176857608499e-06, "loss": 2.1769, "step": 1402 }, { "epoch": 0.08, "learning_rate": 4.216212953461701e-06, "loss": 2.1608, "step": 1404 }, { "epoch": 0.08, "learning_rate": 4.2222490493149035e-06, "loss": 2.483, "step": 1406 }, { "epoch": 0.08, "learning_rate": 4.228285145168105e-06, "loss": 2.0881, "step": 1408 }, { "epoch": 0.09, "learning_rate": 4.234321241021308e-06, "loss": 1.929, "step": 1410 }, { "epoch": 0.09, "learning_rate": 4.24035733687451e-06, "loss": 2.3047, "step": 1412 }, { "epoch": 0.09, "learning_rate": 4.246393432727712e-06, "loss": 1.997, "step": 1414 }, { "epoch": 0.09, "learning_rate": 4.2524295285809135e-06, "loss": 2.0981, "step": 1416 }, { "epoch": 0.09, "learning_rate": 4.258465624434116e-06, "loss": 2.0301, "step": 1418 }, { "epoch": 0.09, "learning_rate": 4.2645017202873185e-06, "loss": 1.8369, "step": 1420 }, { "epoch": 0.09, "learning_rate": 4.27053781614052e-06, "loss": 2.6446, "step": 1422 }, { "epoch": 0.09, "learning_rate": 4.276573911993723e-06, "loss": 2.2927, "step": 1424 }, { "epoch": 0.09, "learning_rate": 4.282610007846924e-06, "loss": 1.8813, "step": 1426 }, { "epoch": 0.09, "learning_rate": 4.288646103700127e-06, "loss": 1.8738, "step": 1428 }, { "epoch": 0.09, "learning_rate": 4.294682199553329e-06, "loss": 2.3059, "step": 1430 }, { "epoch": 0.09, "learning_rate": 4.300718295406531e-06, "loss": 2.2097, "step": 1432 }, { "epoch": 0.09, "learning_rate": 4.306754391259734e-06, "loss": 1.9767, "step": 1434 }, { "epoch": 0.09, "learning_rate": 4.312790487112935e-06, "loss": 2.0023, "step": 1436 }, { "epoch": 0.09, "learning_rate": 4.318826582966138e-06, "loss": 2.0216, "step": 1438 }, { "epoch": 0.09, "learning_rate": 4.32486267881934e-06, "loss": 2.2074, "step": 1440 }, { "epoch": 0.09, "learning_rate": 4.330898774672542e-06, "loss": 2.2926, "step": 1442 }, { "epoch": 0.09, "learning_rate": 4.336934870525744e-06, "loss": 2.0177, "step": 1444 }, { "epoch": 0.09, "learning_rate": 4.342970966378946e-06, "loss": 1.934, "step": 1446 }, { "epoch": 0.09, "learning_rate": 4.349007062232149e-06, "loss": 2.1793, "step": 1448 }, { "epoch": 0.09, "learning_rate": 4.35504315808535e-06, "loss": 2.424, "step": 1450 }, { "epoch": 0.09, "learning_rate": 4.361079253938553e-06, "loss": 1.5621, "step": 1452 }, { "epoch": 0.09, "learning_rate": 4.3671153497917545e-06, "loss": 2.0931, "step": 1454 }, { "epoch": 0.09, "learning_rate": 4.373151445644957e-06, "loss": 2.4004, "step": 1456 }, { "epoch": 0.09, "learning_rate": 4.3791875414981595e-06, "loss": 1.6691, "step": 1458 }, { "epoch": 0.09, "learning_rate": 4.385223637351361e-06, "loss": 1.6597, "step": 1460 }, { "epoch": 0.09, "learning_rate": 4.391259733204564e-06, "loss": 2.2091, "step": 1462 }, { "epoch": 0.09, "learning_rate": 4.397295829057765e-06, "loss": 1.7431, "step": 1464 }, { "epoch": 0.09, "learning_rate": 4.403331924910968e-06, "loss": 1.8576, "step": 1466 }, { "epoch": 0.09, "learning_rate": 4.40936802076417e-06, "loss": 2.2262, "step": 1468 }, { "epoch": 0.09, "learning_rate": 4.415404116617372e-06, "loss": 2.0001, "step": 1470 }, { "epoch": 0.09, "learning_rate": 4.421440212470574e-06, "loss": 2.388, "step": 1472 }, { "epoch": 0.09, "learning_rate": 4.427476308323776e-06, "loss": 1.7493, "step": 1474 }, { "epoch": 0.09, "learning_rate": 4.433512404176979e-06, "loss": 1.9935, "step": 1476 }, { "epoch": 0.09, "learning_rate": 4.43954850003018e-06, "loss": 1.7015, "step": 1478 }, { "epoch": 0.09, "learning_rate": 4.445584595883383e-06, "loss": 1.63, "step": 1480 }, { "epoch": 0.09, "learning_rate": 4.4516206917365846e-06, "loss": 2.0909, "step": 1482 }, { "epoch": 0.09, "learning_rate": 4.457656787589787e-06, "loss": 2.2059, "step": 1484 }, { "epoch": 0.09, "learning_rate": 4.46369288344299e-06, "loss": 1.9331, "step": 1486 }, { "epoch": 0.09, "learning_rate": 4.469728979296191e-06, "loss": 1.8934, "step": 1488 }, { "epoch": 0.09, "learning_rate": 4.475765075149394e-06, "loss": 1.9759, "step": 1490 }, { "epoch": 0.09, "learning_rate": 4.4818011710025954e-06, "loss": 2.3743, "step": 1492 }, { "epoch": 0.09, "learning_rate": 4.487837266855798e-06, "loss": 2.0094, "step": 1494 }, { "epoch": 0.09, "learning_rate": 4.4938733627090005e-06, "loss": 1.9575, "step": 1496 }, { "epoch": 0.09, "learning_rate": 4.499909458562202e-06, "loss": 1.8623, "step": 1498 }, { "epoch": 0.09, "learning_rate": 4.505945554415404e-06, "loss": 2.1978, "step": 1500 }, { "epoch": 0.09, "learning_rate": 4.511981650268606e-06, "loss": 2.3611, "step": 1502 }, { "epoch": 0.09, "learning_rate": 4.518017746121809e-06, "loss": 1.6948, "step": 1504 }, { "epoch": 0.09, "learning_rate": 4.5240538419750105e-06, "loss": 1.9698, "step": 1506 }, { "epoch": 0.09, "learning_rate": 4.530089937828213e-06, "loss": 1.8178, "step": 1508 }, { "epoch": 0.09, "learning_rate": 4.536126033681415e-06, "loss": 2.2389, "step": 1510 }, { "epoch": 0.09, "learning_rate": 4.542162129534617e-06, "loss": 1.7853, "step": 1512 }, { "epoch": 0.09, "learning_rate": 4.54819822538782e-06, "loss": 1.9741, "step": 1514 }, { "epoch": 0.09, "learning_rate": 4.554234321241021e-06, "loss": 1.6623, "step": 1516 }, { "epoch": 0.09, "learning_rate": 4.560270417094224e-06, "loss": 2.1464, "step": 1518 }, { "epoch": 0.09, "learning_rate": 4.5663065129474255e-06, "loss": 1.9378, "step": 1520 }, { "epoch": 0.09, "learning_rate": 4.572342608800628e-06, "loss": 1.3998, "step": 1522 }, { "epoch": 0.09, "learning_rate": 4.5783787046538306e-06, "loss": 2.1053, "step": 1524 }, { "epoch": 0.09, "learning_rate": 4.584414800507032e-06, "loss": 2.3882, "step": 1526 }, { "epoch": 0.09, "learning_rate": 4.590450896360234e-06, "loss": 2.224, "step": 1528 }, { "epoch": 0.09, "learning_rate": 4.596486992213436e-06, "loss": 1.9731, "step": 1530 }, { "epoch": 0.09, "learning_rate": 4.602523088066639e-06, "loss": 1.9813, "step": 1532 }, { "epoch": 0.09, "learning_rate": 4.608559183919841e-06, "loss": 2.1683, "step": 1534 }, { "epoch": 0.09, "learning_rate": 4.614595279773043e-06, "loss": 1.8934, "step": 1536 }, { "epoch": 0.09, "learning_rate": 4.620631375626245e-06, "loss": 1.8823, "step": 1538 }, { "epoch": 0.09, "learning_rate": 4.626667471479447e-06, "loss": 1.8883, "step": 1540 }, { "epoch": 0.09, "learning_rate": 4.63270356733265e-06, "loss": 1.9718, "step": 1542 }, { "epoch": 0.09, "learning_rate": 4.6387396631858515e-06, "loss": 2.2811, "step": 1544 }, { "epoch": 0.09, "learning_rate": 4.644775759039054e-06, "loss": 2.0018, "step": 1546 }, { "epoch": 0.09, "learning_rate": 4.650811854892256e-06, "loss": 1.9539, "step": 1548 }, { "epoch": 0.09, "learning_rate": 4.656847950745458e-06, "loss": 1.7945, "step": 1550 }, { "epoch": 0.09, "learning_rate": 4.662884046598661e-06, "loss": 2.0272, "step": 1552 }, { "epoch": 0.09, "learning_rate": 4.668920142451862e-06, "loss": 1.6376, "step": 1554 }, { "epoch": 0.09, "learning_rate": 4.674956238305064e-06, "loss": 2.0278, "step": 1556 }, { "epoch": 0.09, "learning_rate": 4.6809923341582665e-06, "loss": 2.0079, "step": 1558 }, { "epoch": 0.09, "learning_rate": 4.687028430011469e-06, "loss": 1.9201, "step": 1560 }, { "epoch": 0.09, "learning_rate": 4.693064525864671e-06, "loss": 1.8545, "step": 1562 }, { "epoch": 0.09, "learning_rate": 4.699100621717873e-06, "loss": 2.0584, "step": 1564 }, { "epoch": 0.09, "learning_rate": 4.705136717571075e-06, "loss": 1.7133, "step": 1566 }, { "epoch": 0.09, "learning_rate": 4.711172813424277e-06, "loss": 2.0776, "step": 1568 }, { "epoch": 0.09, "learning_rate": 4.71720890927748e-06, "loss": 1.6575, "step": 1570 }, { "epoch": 0.09, "learning_rate": 4.7232450051306816e-06, "loss": 1.6523, "step": 1572 }, { "epoch": 0.1, "learning_rate": 4.729281100983884e-06, "loss": 1.9355, "step": 1574 }, { "epoch": 0.1, "learning_rate": 4.735317196837086e-06, "loss": 1.8711, "step": 1576 }, { "epoch": 0.1, "learning_rate": 4.741353292690288e-06, "loss": 1.885, "step": 1578 }, { "epoch": 0.1, "learning_rate": 4.747389388543491e-06, "loss": 2.0599, "step": 1580 }, { "epoch": 0.1, "learning_rate": 4.753425484396692e-06, "loss": 1.5303, "step": 1582 }, { "epoch": 0.1, "learning_rate": 4.759461580249894e-06, "loss": 1.7966, "step": 1584 }, { "epoch": 0.1, "learning_rate": 4.765497676103097e-06, "loss": 1.878, "step": 1586 }, { "epoch": 0.1, "learning_rate": 4.771533771956299e-06, "loss": 1.7692, "step": 1588 }, { "epoch": 0.1, "learning_rate": 4.777569867809501e-06, "loss": 1.9787, "step": 1590 }, { "epoch": 0.1, "learning_rate": 4.783605963662703e-06, "loss": 1.6868, "step": 1592 }, { "epoch": 0.1, "learning_rate": 4.789642059515905e-06, "loss": 1.8588, "step": 1594 }, { "epoch": 0.1, "learning_rate": 4.7956781553691075e-06, "loss": 2.1473, "step": 1596 }, { "epoch": 0.1, "learning_rate": 4.80171425122231e-06, "loss": 1.6, "step": 1598 }, { "epoch": 0.1, "learning_rate": 4.807750347075512e-06, "loss": 2.0502, "step": 1600 }, { "epoch": 0.1, "learning_rate": 4.813786442928714e-06, "loss": 2.1347, "step": 1602 }, { "epoch": 0.1, "learning_rate": 4.819822538781916e-06, "loss": 1.8167, "step": 1604 }, { "epoch": 0.1, "learning_rate": 4.825858634635118e-06, "loss": 2.0158, "step": 1606 }, { "epoch": 0.1, "learning_rate": 4.831894730488321e-06, "loss": 1.766, "step": 1608 }, { "epoch": 0.1, "learning_rate": 4.8379308263415225e-06, "loss": 1.8321, "step": 1610 }, { "epoch": 0.1, "learning_rate": 4.843966922194724e-06, "loss": 2.3845, "step": 1612 }, { "epoch": 0.1, "learning_rate": 4.850003018047927e-06, "loss": 1.7642, "step": 1614 }, { "epoch": 0.1, "learning_rate": 4.856039113901129e-06, "loss": 1.7696, "step": 1616 }, { "epoch": 0.1, "learning_rate": 4.862075209754331e-06, "loss": 2.1376, "step": 1618 }, { "epoch": 0.1, "learning_rate": 4.868111305607533e-06, "loss": 1.9776, "step": 1620 }, { "epoch": 0.1, "learning_rate": 4.874147401460735e-06, "loss": 1.7406, "step": 1622 }, { "epoch": 0.1, "learning_rate": 4.8801834973139376e-06, "loss": 2.0122, "step": 1624 }, { "epoch": 0.1, "learning_rate": 4.88621959316714e-06, "loss": 1.9774, "step": 1626 }, { "epoch": 0.1, "learning_rate": 4.892255689020342e-06, "loss": 2.2756, "step": 1628 }, { "epoch": 0.1, "learning_rate": 4.898291784873544e-06, "loss": 1.863, "step": 1630 }, { "epoch": 0.1, "learning_rate": 4.904327880726746e-06, "loss": 2.4344, "step": 1632 }, { "epoch": 0.1, "learning_rate": 4.9103639765799484e-06, "loss": 2.0121, "step": 1634 }, { "epoch": 0.1, "learning_rate": 4.916400072433151e-06, "loss": 1.8208, "step": 1636 }, { "epoch": 0.1, "learning_rate": 4.922436168286353e-06, "loss": 1.9523, "step": 1638 }, { "epoch": 0.1, "learning_rate": 4.928472264139554e-06, "loss": 1.6836, "step": 1640 }, { "epoch": 0.1, "learning_rate": 4.934508359992757e-06, "loss": 1.6685, "step": 1642 }, { "epoch": 0.1, "learning_rate": 4.940544455845959e-06, "loss": 2.1782, "step": 1644 }, { "epoch": 0.1, "learning_rate": 4.946580551699161e-06, "loss": 1.9122, "step": 1646 }, { "epoch": 0.1, "learning_rate": 4.9526166475523635e-06, "loss": 2.0758, "step": 1648 }, { "epoch": 0.1, "learning_rate": 4.958652743405565e-06, "loss": 1.9782, "step": 1650 }, { "epoch": 0.1, "learning_rate": 4.964688839258768e-06, "loss": 1.9154, "step": 1652 }, { "epoch": 0.1, "learning_rate": 4.97072493511197e-06, "loss": 1.9811, "step": 1654 }, { "epoch": 0.1, "learning_rate": 4.976761030965172e-06, "loss": 1.919, "step": 1656 }, { "epoch": 0.1, "learning_rate": 4.982797126818374e-06, "loss": 1.8792, "step": 1658 }, { "epoch": 0.1, "learning_rate": 4.988833222671576e-06, "loss": 2.4134, "step": 1660 }, { "epoch": 0.1, "learning_rate": 4.9948693185247785e-06, "loss": 1.9918, "step": 1662 }, { "epoch": 0.1, "learning_rate": 5.000905414377981e-06, "loss": 1.5943, "step": 1664 }, { "epoch": 0.1, "learning_rate": 5.006941510231183e-06, "loss": 1.508, "step": 1666 }, { "epoch": 0.1, "learning_rate": 5.012977606084384e-06, "loss": 1.5667, "step": 1668 }, { "epoch": 0.1, "learning_rate": 5.019013701937587e-06, "loss": 1.6652, "step": 1670 }, { "epoch": 0.1, "learning_rate": 5.025049797790789e-06, "loss": 2.1967, "step": 1672 }, { "epoch": 0.1, "learning_rate": 5.031085893643991e-06, "loss": 1.8951, "step": 1674 }, { "epoch": 0.1, "learning_rate": 5.0371219894971936e-06, "loss": 1.9727, "step": 1676 }, { "epoch": 0.1, "learning_rate": 5.043158085350395e-06, "loss": 1.869, "step": 1678 }, { "epoch": 0.1, "learning_rate": 5.049194181203598e-06, "loss": 1.7271, "step": 1680 }, { "epoch": 0.1, "learning_rate": 5.0552302770568e-06, "loss": 1.8827, "step": 1682 }, { "epoch": 0.1, "learning_rate": 5.061266372910002e-06, "loss": 2.073, "step": 1684 }, { "epoch": 0.1, "learning_rate": 5.0673024687632044e-06, "loss": 1.639, "step": 1686 }, { "epoch": 0.1, "learning_rate": 5.073338564616406e-06, "loss": 1.684, "step": 1688 }, { "epoch": 0.1, "learning_rate": 5.079374660469609e-06, "loss": 1.7013, "step": 1690 }, { "epoch": 0.1, "learning_rate": 5.085410756322811e-06, "loss": 1.8463, "step": 1692 }, { "epoch": 0.1, "learning_rate": 5.091446852176013e-06, "loss": 2.0369, "step": 1694 }, { "epoch": 0.1, "learning_rate": 5.0974829480292145e-06, "loss": 2.0457, "step": 1696 }, { "epoch": 0.1, "learning_rate": 5.103519043882417e-06, "loss": 1.9152, "step": 1698 }, { "epoch": 0.1, "learning_rate": 5.1095551397356195e-06, "loss": 1.6193, "step": 1700 }, { "epoch": 0.1, "learning_rate": 5.115591235588821e-06, "loss": 1.8288, "step": 1702 }, { "epoch": 0.1, "learning_rate": 5.121627331442024e-06, "loss": 1.9383, "step": 1704 }, { "epoch": 0.1, "learning_rate": 5.127663427295225e-06, "loss": 1.982, "step": 1706 }, { "epoch": 0.1, "learning_rate": 5.133699523148428e-06, "loss": 1.7188, "step": 1708 }, { "epoch": 0.1, "learning_rate": 5.13973561900163e-06, "loss": 1.615, "step": 1710 }, { "epoch": 0.1, "learning_rate": 5.145771714854832e-06, "loss": 1.9668, "step": 1712 }, { "epoch": 0.1, "learning_rate": 5.1518078107080345e-06, "loss": 1.5326, "step": 1714 }, { "epoch": 0.1, "learning_rate": 5.157843906561236e-06, "loss": 1.7383, "step": 1716 }, { "epoch": 0.1, "learning_rate": 5.163880002414439e-06, "loss": 2.1414, "step": 1718 }, { "epoch": 0.1, "learning_rate": 5.169916098267641e-06, "loss": 1.7895, "step": 1720 }, { "epoch": 0.1, "learning_rate": 5.175952194120843e-06, "loss": 1.648, "step": 1722 }, { "epoch": 0.1, "learning_rate": 5.1819882899740446e-06, "loss": 2.2053, "step": 1724 }, { "epoch": 0.1, "learning_rate": 5.188024385827247e-06, "loss": 1.6918, "step": 1726 }, { "epoch": 0.1, "learning_rate": 5.19406048168045e-06, "loss": 2.111, "step": 1728 }, { "epoch": 0.1, "learning_rate": 5.200096577533651e-06, "loss": 1.9407, "step": 1730 }, { "epoch": 0.1, "learning_rate": 5.206132673386854e-06, "loss": 2.0757, "step": 1732 }, { "epoch": 0.1, "learning_rate": 5.2121687692400554e-06, "loss": 1.6571, "step": 1734 }, { "epoch": 0.1, "learning_rate": 5.218204865093258e-06, "loss": 1.9857, "step": 1736 }, { "epoch": 0.1, "learning_rate": 5.2242409609464605e-06, "loss": 2.0395, "step": 1738 }, { "epoch": 0.11, "learning_rate": 5.230277056799662e-06, "loss": 1.7851, "step": 1740 }, { "epoch": 0.11, "learning_rate": 5.236313152652865e-06, "loss": 1.989, "step": 1742 }, { "epoch": 0.11, "learning_rate": 5.242349248506066e-06, "loss": 2.3709, "step": 1744 }, { "epoch": 0.11, "learning_rate": 5.248385344359269e-06, "loss": 1.6758, "step": 1746 }, { "epoch": 0.11, "learning_rate": 5.254421440212471e-06, "loss": 2.287, "step": 1748 }, { "epoch": 0.11, "learning_rate": 5.260457536065673e-06, "loss": 2.5342, "step": 1750 }, { "epoch": 0.11, "learning_rate": 5.266493631918875e-06, "loss": 1.9809, "step": 1752 }, { "epoch": 0.11, "learning_rate": 5.272529727772077e-06, "loss": 1.9784, "step": 1754 }, { "epoch": 0.11, "learning_rate": 5.27856582362528e-06, "loss": 1.9496, "step": 1756 }, { "epoch": 0.11, "learning_rate": 5.284601919478481e-06, "loss": 1.5767, "step": 1758 }, { "epoch": 0.11, "learning_rate": 5.290638015331684e-06, "loss": 2.0544, "step": 1760 }, { "epoch": 0.11, "learning_rate": 5.2966741111848855e-06, "loss": 1.951, "step": 1762 }, { "epoch": 0.11, "learning_rate": 5.302710207038088e-06, "loss": 1.7452, "step": 1764 }, { "epoch": 0.11, "learning_rate": 5.3087463028912906e-06, "loss": 1.9667, "step": 1766 }, { "epoch": 0.11, "learning_rate": 5.314782398744492e-06, "loss": 1.5471, "step": 1768 }, { "epoch": 0.11, "learning_rate": 5.320818494597695e-06, "loss": 2.0163, "step": 1770 }, { "epoch": 0.11, "learning_rate": 5.326854590450896e-06, "loss": 1.9589, "step": 1772 }, { "epoch": 0.11, "learning_rate": 5.332890686304099e-06, "loss": 1.6582, "step": 1774 }, { "epoch": 0.11, "learning_rate": 5.338926782157301e-06, "loss": 1.5788, "step": 1776 }, { "epoch": 0.11, "learning_rate": 5.344962878010503e-06, "loss": 2.036, "step": 1778 }, { "epoch": 0.11, "learning_rate": 5.350998973863705e-06, "loss": 2.2587, "step": 1780 }, { "epoch": 0.11, "learning_rate": 5.357035069716907e-06, "loss": 1.9223, "step": 1782 }, { "epoch": 0.11, "learning_rate": 5.36307116557011e-06, "loss": 1.7554, "step": 1784 }, { "epoch": 0.11, "learning_rate": 5.3691072614233114e-06, "loss": 1.5783, "step": 1786 }, { "epoch": 0.11, "learning_rate": 5.375143357276514e-06, "loss": 1.6535, "step": 1788 }, { "epoch": 0.11, "learning_rate": 5.381179453129716e-06, "loss": 1.7671, "step": 1790 }, { "epoch": 0.11, "learning_rate": 5.387215548982918e-06, "loss": 1.5343, "step": 1792 }, { "epoch": 0.11, "learning_rate": 5.393251644836121e-06, "loss": 1.748, "step": 1794 }, { "epoch": 0.11, "learning_rate": 5.399287740689322e-06, "loss": 2.2145, "step": 1796 }, { "epoch": 0.11, "learning_rate": 5.405323836542525e-06, "loss": 1.5615, "step": 1798 }, { "epoch": 0.11, "learning_rate": 5.4113599323957265e-06, "loss": 1.7798, "step": 1800 }, { "epoch": 0.11, "learning_rate": 5.417396028248929e-06, "loss": 1.8196, "step": 1802 }, { "epoch": 0.11, "learning_rate": 5.4234321241021315e-06, "loss": 1.8351, "step": 1804 }, { "epoch": 0.11, "learning_rate": 5.429468219955333e-06, "loss": 1.4983, "step": 1806 }, { "epoch": 0.11, "learning_rate": 5.435504315808535e-06, "loss": 1.3984, "step": 1808 }, { "epoch": 0.11, "learning_rate": 5.441540411661737e-06, "loss": 1.7774, "step": 1810 }, { "epoch": 0.11, "learning_rate": 5.44757650751494e-06, "loss": 2.0398, "step": 1812 }, { "epoch": 0.11, "learning_rate": 5.4536126033681415e-06, "loss": 1.8237, "step": 1814 }, { "epoch": 0.11, "learning_rate": 5.459648699221344e-06, "loss": 2.0527, "step": 1816 }, { "epoch": 0.11, "learning_rate": 5.465684795074546e-06, "loss": 1.8679, "step": 1818 }, { "epoch": 0.11, "learning_rate": 5.471720890927748e-06, "loss": 1.6323, "step": 1820 }, { "epoch": 0.11, "learning_rate": 5.477756986780951e-06, "loss": 1.5492, "step": 1822 }, { "epoch": 0.11, "learning_rate": 5.483793082634152e-06, "loss": 1.5295, "step": 1824 }, { "epoch": 0.11, "learning_rate": 5.489829178487355e-06, "loss": 1.6683, "step": 1826 }, { "epoch": 0.11, "learning_rate": 5.495865274340557e-06, "loss": 1.6957, "step": 1828 }, { "epoch": 0.11, "learning_rate": 5.501901370193759e-06, "loss": 2.6405, "step": 1830 }, { "epoch": 0.11, "learning_rate": 5.507937466046962e-06, "loss": 1.8225, "step": 1832 }, { "epoch": 0.11, "learning_rate": 5.513973561900163e-06, "loss": 1.7279, "step": 1834 }, { "epoch": 0.11, "learning_rate": 5.520009657753365e-06, "loss": 1.6329, "step": 1836 }, { "epoch": 0.11, "learning_rate": 5.5260457536065675e-06, "loss": 1.4026, "step": 1838 }, { "epoch": 0.11, "learning_rate": 5.53208184945977e-06, "loss": 2.0066, "step": 1840 }, { "epoch": 0.11, "learning_rate": 5.538117945312972e-06, "loss": 1.8047, "step": 1842 }, { "epoch": 0.11, "learning_rate": 5.544154041166174e-06, "loss": 1.7672, "step": 1844 }, { "epoch": 0.11, "learning_rate": 5.550190137019376e-06, "loss": 2.0947, "step": 1846 }, { "epoch": 0.11, "learning_rate": 5.556226232872578e-06, "loss": 1.7141, "step": 1848 }, { "epoch": 0.11, "learning_rate": 5.562262328725781e-06, "loss": 1.7475, "step": 1850 }, { "epoch": 0.11, "learning_rate": 5.5682984245789825e-06, "loss": 1.4427, "step": 1852 }, { "epoch": 0.11, "learning_rate": 5.574334520432185e-06, "loss": 1.9458, "step": 1854 }, { "epoch": 0.11, "learning_rate": 5.580370616285387e-06, "loss": 1.7097, "step": 1856 }, { "epoch": 0.11, "learning_rate": 5.586406712138589e-06, "loss": 1.6405, "step": 1858 }, { "epoch": 0.11, "learning_rate": 5.592442807991792e-06, "loss": 1.6957, "step": 1860 }, { "epoch": 0.11, "learning_rate": 5.598478903844993e-06, "loss": 1.734, "step": 1862 }, { "epoch": 0.11, "learning_rate": 5.604514999698195e-06, "loss": 2.0297, "step": 1864 }, { "epoch": 0.11, "learning_rate": 5.6105510955513975e-06, "loss": 1.9453, "step": 1866 }, { "epoch": 0.11, "learning_rate": 5.6165871914046e-06, "loss": 1.4764, "step": 1868 }, { "epoch": 0.11, "learning_rate": 5.622623287257802e-06, "loss": 2.0552, "step": 1870 }, { "epoch": 0.11, "learning_rate": 5.628659383111004e-06, "loss": 1.948, "step": 1872 }, { "epoch": 0.11, "learning_rate": 5.634695478964206e-06, "loss": 2.2322, "step": 1874 }, { "epoch": 0.11, "learning_rate": 5.640731574817408e-06, "loss": 1.8247, "step": 1876 }, { "epoch": 0.11, "learning_rate": 5.646767670670611e-06, "loss": 1.4478, "step": 1878 }, { "epoch": 0.11, "learning_rate": 5.652803766523813e-06, "loss": 1.1879, "step": 1880 }, { "epoch": 0.11, "learning_rate": 5.658839862377015e-06, "loss": 2.0237, "step": 1882 }, { "epoch": 0.11, "learning_rate": 5.664875958230217e-06, "loss": 2.0018, "step": 1884 }, { "epoch": 0.11, "learning_rate": 5.670912054083419e-06, "loss": 1.8476, "step": 1886 }, { "epoch": 0.11, "learning_rate": 5.676948149936622e-06, "loss": 2.0201, "step": 1888 }, { "epoch": 0.11, "learning_rate": 5.6829842457898235e-06, "loss": 1.6666, "step": 1890 }, { "epoch": 0.11, "learning_rate": 5.689020341643025e-06, "loss": 2.1309, "step": 1892 }, { "epoch": 0.11, "learning_rate": 5.695056437496228e-06, "loss": 1.8863, "step": 1894 }, { "epoch": 0.11, "learning_rate": 5.70109253334943e-06, "loss": 1.8639, "step": 1896 }, { "epoch": 0.11, "learning_rate": 5.707128629202632e-06, "loss": 2.0198, "step": 1898 }, { "epoch": 0.11, "learning_rate": 5.713164725055834e-06, "loss": 1.7553, "step": 1900 }, { "epoch": 0.11, "learning_rate": 5.719200820909036e-06, "loss": 1.589, "step": 1902 }, { "epoch": 0.11, "learning_rate": 5.7252369167622385e-06, "loss": 2.1281, "step": 1904 }, { "epoch": 0.12, "learning_rate": 5.731273012615441e-06, "loss": 2.1904, "step": 1906 }, { "epoch": 0.12, "learning_rate": 5.737309108468643e-06, "loss": 2.1766, "step": 1908 }, { "epoch": 0.12, "learning_rate": 5.743345204321845e-06, "loss": 1.708, "step": 1910 }, { "epoch": 0.12, "learning_rate": 5.749381300175047e-06, "loss": 1.8505, "step": 1912 }, { "epoch": 0.12, "learning_rate": 5.755417396028249e-06, "loss": 1.7225, "step": 1914 }, { "epoch": 0.12, "learning_rate": 5.761453491881452e-06, "loss": 1.6575, "step": 1916 }, { "epoch": 0.12, "learning_rate": 5.7674895877346536e-06, "loss": 1.5659, "step": 1918 }, { "epoch": 0.12, "learning_rate": 5.773525683587855e-06, "loss": 1.9733, "step": 1920 }, { "epoch": 0.12, "learning_rate": 5.779561779441058e-06, "loss": 1.8489, "step": 1922 }, { "epoch": 0.12, "learning_rate": 5.78559787529426e-06, "loss": 1.2323, "step": 1924 }, { "epoch": 0.12, "learning_rate": 5.791633971147462e-06, "loss": 1.5814, "step": 1926 }, { "epoch": 0.12, "learning_rate": 5.7976700670006644e-06, "loss": 1.5951, "step": 1928 }, { "epoch": 0.12, "learning_rate": 5.803706162853866e-06, "loss": 1.6238, "step": 1930 }, { "epoch": 0.12, "learning_rate": 5.809742258707069e-06, "loss": 1.941, "step": 1932 }, { "epoch": 0.12, "learning_rate": 5.815778354560271e-06, "loss": 1.7244, "step": 1934 }, { "epoch": 0.12, "learning_rate": 5.821814450413473e-06, "loss": 1.6151, "step": 1936 }, { "epoch": 0.12, "learning_rate": 5.827850546266675e-06, "loss": 1.5589, "step": 1938 }, { "epoch": 0.12, "learning_rate": 5.833886642119877e-06, "loss": 2.5437, "step": 1940 }, { "epoch": 0.12, "learning_rate": 5.8399227379730795e-06, "loss": 1.9369, "step": 1942 }, { "epoch": 0.12, "learning_rate": 5.845958833826282e-06, "loss": 1.7239, "step": 1944 }, { "epoch": 0.12, "learning_rate": 5.851994929679484e-06, "loss": 2.1428, "step": 1946 }, { "epoch": 0.12, "learning_rate": 5.858031025532685e-06, "loss": 1.9912, "step": 1948 }, { "epoch": 0.12, "learning_rate": 5.864067121385888e-06, "loss": 1.8766, "step": 1950 }, { "epoch": 0.12, "learning_rate": 5.87010321723909e-06, "loss": 1.4359, "step": 1952 }, { "epoch": 0.12, "learning_rate": 5.876139313092292e-06, "loss": 1.5417, "step": 1954 }, { "epoch": 0.12, "learning_rate": 5.8821754089454945e-06, "loss": 2.1995, "step": 1956 }, { "epoch": 0.12, "learning_rate": 5.888211504798696e-06, "loss": 1.8074, "step": 1958 }, { "epoch": 0.12, "learning_rate": 5.894247600651899e-06, "loss": 1.8526, "step": 1960 }, { "epoch": 0.12, "learning_rate": 5.900283696505101e-06, "loss": 1.8699, "step": 1962 }, { "epoch": 0.12, "learning_rate": 5.906319792358303e-06, "loss": 1.8615, "step": 1964 }, { "epoch": 0.12, "learning_rate": 5.912355888211505e-06, "loss": 2.1859, "step": 1966 }, { "epoch": 0.12, "learning_rate": 5.918391984064707e-06, "loss": 1.6305, "step": 1968 }, { "epoch": 0.12, "learning_rate": 5.9244280799179096e-06, "loss": 1.8578, "step": 1970 }, { "epoch": 0.12, "learning_rate": 5.930464175771112e-06, "loss": 1.5954, "step": 1972 }, { "epoch": 0.12, "learning_rate": 5.936500271624314e-06, "loss": 1.5363, "step": 1974 }, { "epoch": 0.12, "learning_rate": 5.942536367477515e-06, "loss": 1.6034, "step": 1976 }, { "epoch": 0.12, "learning_rate": 5.948572463330718e-06, "loss": 1.9764, "step": 1978 }, { "epoch": 0.12, "learning_rate": 5.9546085591839204e-06, "loss": 2.3412, "step": 1980 }, { "epoch": 0.12, "learning_rate": 5.960644655037122e-06, "loss": 1.8276, "step": 1982 }, { "epoch": 0.12, "learning_rate": 5.966680750890325e-06, "loss": 1.873, "step": 1984 }, { "epoch": 0.12, "learning_rate": 5.972716846743526e-06, "loss": 1.5384, "step": 1986 }, { "epoch": 0.12, "learning_rate": 5.978752942596729e-06, "loss": 1.7937, "step": 1988 }, { "epoch": 0.12, "learning_rate": 5.984789038449931e-06, "loss": 1.4818, "step": 1990 }, { "epoch": 0.12, "learning_rate": 5.990825134303133e-06, "loss": 1.4648, "step": 1992 }, { "epoch": 0.12, "learning_rate": 5.9968612301563355e-06, "loss": 1.3047, "step": 1994 }, { "epoch": 0.12, "learning_rate": 6.002897326009537e-06, "loss": 1.8854, "step": 1996 }, { "epoch": 0.12, "learning_rate": 6.00893342186274e-06, "loss": 1.7578, "step": 1998 }, { "epoch": 0.12, "learning_rate": 6.014969517715942e-06, "loss": 1.9089, "step": 2000 }, { "epoch": 0.12, "learning_rate": 6.021005613569144e-06, "loss": 2.1451, "step": 2002 }, { "epoch": 0.12, "learning_rate": 6.0270417094223455e-06, "loss": 1.8581, "step": 2004 }, { "epoch": 0.12, "learning_rate": 6.033077805275548e-06, "loss": 1.9368, "step": 2006 }, { "epoch": 0.12, "learning_rate": 6.0391139011287505e-06, "loss": 1.8187, "step": 2008 }, { "epoch": 0.12, "learning_rate": 6.045149996981952e-06, "loss": 1.7762, "step": 2010 }, { "epoch": 0.12, "learning_rate": 6.051186092835155e-06, "loss": 1.5389, "step": 2012 }, { "epoch": 0.12, "learning_rate": 6.057222188688356e-06, "loss": 2.001, "step": 2014 }, { "epoch": 0.12, "learning_rate": 6.063258284541559e-06, "loss": 1.7436, "step": 2016 }, { "epoch": 0.12, "learning_rate": 6.069294380394761e-06, "loss": 1.8823, "step": 2018 }, { "epoch": 0.12, "learning_rate": 6.075330476247963e-06, "loss": 1.8817, "step": 2020 }, { "epoch": 0.12, "learning_rate": 6.081366572101166e-06, "loss": 2.2152, "step": 2022 }, { "epoch": 0.12, "learning_rate": 6.087402667954367e-06, "loss": 1.6715, "step": 2024 }, { "epoch": 0.12, "learning_rate": 6.09343876380757e-06, "loss": 1.5787, "step": 2026 }, { "epoch": 0.12, "learning_rate": 6.099474859660772e-06, "loss": 1.4552, "step": 2028 }, { "epoch": 0.12, "learning_rate": 6.105510955513974e-06, "loss": 1.7489, "step": 2030 }, { "epoch": 0.12, "learning_rate": 6.111547051367176e-06, "loss": 2.0648, "step": 2032 }, { "epoch": 0.12, "learning_rate": 6.117583147220378e-06, "loss": 1.7187, "step": 2034 }, { "epoch": 0.12, "learning_rate": 6.123619243073581e-06, "loss": 1.7569, "step": 2036 }, { "epoch": 0.12, "learning_rate": 6.129655338926782e-06, "loss": 1.6265, "step": 2038 }, { "epoch": 0.12, "learning_rate": 6.135691434779985e-06, "loss": 1.5355, "step": 2040 }, { "epoch": 0.12, "learning_rate": 6.1417275306331865e-06, "loss": 1.9243, "step": 2042 }, { "epoch": 0.12, "learning_rate": 6.147763626486389e-06, "loss": 1.9813, "step": 2044 }, { "epoch": 0.12, "learning_rate": 6.1537997223395915e-06, "loss": 1.8922, "step": 2046 }, { "epoch": 0.12, "learning_rate": 6.159835818192793e-06, "loss": 1.7566, "step": 2048 }, { "epoch": 0.12, "learning_rate": 6.165871914045996e-06, "loss": 1.2412, "step": 2050 }, { "epoch": 0.12, "learning_rate": 6.171908009899197e-06, "loss": 1.4867, "step": 2052 }, { "epoch": 0.12, "learning_rate": 6.1779441057524e-06, "loss": 1.696, "step": 2054 }, { "epoch": 0.12, "learning_rate": 6.183980201605602e-06, "loss": 1.2743, "step": 2056 }, { "epoch": 0.12, "learning_rate": 6.190016297458804e-06, "loss": 2.1668, "step": 2058 }, { "epoch": 0.12, "learning_rate": 6.196052393312006e-06, "loss": 1.7143, "step": 2060 }, { "epoch": 0.12, "learning_rate": 6.202088489165208e-06, "loss": 1.5902, "step": 2062 }, { "epoch": 0.12, "learning_rate": 6.208124585018411e-06, "loss": 1.6959, "step": 2064 }, { "epoch": 0.12, "learning_rate": 6.214160680871612e-06, "loss": 1.8445, "step": 2066 }, { "epoch": 0.12, "learning_rate": 6.220196776724815e-06, "loss": 1.6793, "step": 2068 }, { "epoch": 0.12, "learning_rate": 6.2262328725780166e-06, "loss": 2.2188, "step": 2070 }, { "epoch": 0.13, "learning_rate": 6.232268968431219e-06, "loss": 1.858, "step": 2072 }, { "epoch": 0.13, "learning_rate": 6.238305064284422e-06, "loss": 1.5909, "step": 2074 }, { "epoch": 0.13, "learning_rate": 6.244341160137623e-06, "loss": 1.5712, "step": 2076 }, { "epoch": 0.13, "learning_rate": 6.250377255990826e-06, "loss": 1.6714, "step": 2078 }, { "epoch": 0.13, "learning_rate": 6.2564133518440274e-06, "loss": 2.0624, "step": 2080 }, { "epoch": 0.13, "learning_rate": 6.26244944769723e-06, "loss": 1.7478, "step": 2082 }, { "epoch": 0.13, "learning_rate": 6.268485543550432e-06, "loss": 1.3942, "step": 2084 }, { "epoch": 0.13, "learning_rate": 6.274521639403634e-06, "loss": 1.5311, "step": 2086 }, { "epoch": 0.13, "learning_rate": 6.280557735256837e-06, "loss": 1.4429, "step": 2088 }, { "epoch": 0.13, "learning_rate": 6.286593831110038e-06, "loss": 1.7223, "step": 2090 }, { "epoch": 0.13, "learning_rate": 6.292629926963241e-06, "loss": 1.5319, "step": 2092 }, { "epoch": 0.13, "learning_rate": 6.2986660228164425e-06, "loss": 1.6046, "step": 2094 }, { "epoch": 0.13, "learning_rate": 6.304702118669645e-06, "loss": 1.4239, "step": 2096 }, { "epoch": 0.13, "learning_rate": 6.3107382145228475e-06, "loss": 1.9597, "step": 2098 }, { "epoch": 0.13, "learning_rate": 6.316774310376049e-06, "loss": 1.5127, "step": 2100 }, { "epoch": 0.13, "learning_rate": 6.322810406229252e-06, "loss": 1.3974, "step": 2102 }, { "epoch": 0.13, "learning_rate": 6.3288465020824525e-06, "loss": 1.4984, "step": 2104 }, { "epoch": 0.13, "learning_rate": 6.334882597935656e-06, "loss": 1.7753, "step": 2106 }, { "epoch": 0.13, "learning_rate": 6.340918693788858e-06, "loss": 1.5701, "step": 2108 }, { "epoch": 0.13, "learning_rate": 6.346954789642059e-06, "loss": 1.3936, "step": 2110 }, { "epoch": 0.13, "learning_rate": 6.3529908854952626e-06, "loss": 1.8284, "step": 2112 }, { "epoch": 0.13, "learning_rate": 6.359026981348463e-06, "loss": 1.3807, "step": 2114 }, { "epoch": 0.13, "learning_rate": 6.365063077201666e-06, "loss": 1.956, "step": 2116 }, { "epoch": 0.13, "learning_rate": 6.371099173054869e-06, "loss": 1.7159, "step": 2118 }, { "epoch": 0.13, "learning_rate": 6.37713526890807e-06, "loss": 1.6526, "step": 2120 }, { "epoch": 0.13, "learning_rate": 6.383171364761273e-06, "loss": 1.5465, "step": 2122 }, { "epoch": 0.13, "learning_rate": 6.389207460614474e-06, "loss": 1.8565, "step": 2124 }, { "epoch": 0.13, "learning_rate": 6.395243556467677e-06, "loss": 1.926, "step": 2126 }, { "epoch": 0.13, "learning_rate": 6.401279652320879e-06, "loss": 1.4444, "step": 2128 }, { "epoch": 0.13, "learning_rate": 6.407315748174081e-06, "loss": 1.6707, "step": 2130 }, { "epoch": 0.13, "learning_rate": 6.4133518440272835e-06, "loss": 1.7096, "step": 2132 }, { "epoch": 0.13, "learning_rate": 6.419387939880486e-06, "loss": 1.6767, "step": 2134 }, { "epoch": 0.13, "learning_rate": 6.425424035733688e-06, "loss": 1.7671, "step": 2136 }, { "epoch": 0.13, "learning_rate": 6.43146013158689e-06, "loss": 1.6423, "step": 2138 }, { "epoch": 0.13, "learning_rate": 6.437496227440092e-06, "loss": 1.6324, "step": 2140 }, { "epoch": 0.13, "learning_rate": 6.443532323293294e-06, "loss": 1.8013, "step": 2142 }, { "epoch": 0.13, "learning_rate": 6.449568419146497e-06, "loss": 1.5862, "step": 2144 }, { "epoch": 0.13, "learning_rate": 6.4556045149996985e-06, "loss": 1.7911, "step": 2146 }, { "epoch": 0.13, "learning_rate": 6.461640610852901e-06, "loss": 1.5625, "step": 2148 }, { "epoch": 0.13, "learning_rate": 6.467676706706103e-06, "loss": 1.6028, "step": 2150 }, { "epoch": 0.13, "learning_rate": 6.473712802559305e-06, "loss": 1.5765, "step": 2152 }, { "epoch": 0.13, "learning_rate": 6.479748898412508e-06, "loss": 1.7739, "step": 2154 }, { "epoch": 0.13, "learning_rate": 6.485784994265709e-06, "loss": 1.5894, "step": 2156 }, { "epoch": 0.13, "learning_rate": 6.491821090118912e-06, "loss": 2.028, "step": 2158 }, { "epoch": 0.13, "learning_rate": 6.497857185972113e-06, "loss": 1.7907, "step": 2160 }, { "epoch": 0.13, "learning_rate": 6.503893281825316e-06, "loss": 1.7793, "step": 2162 }, { "epoch": 0.13, "learning_rate": 6.5099293776785186e-06, "loss": 1.7081, "step": 2164 }, { "epoch": 0.13, "learning_rate": 6.515965473531719e-06, "loss": 1.713, "step": 2166 }, { "epoch": 0.13, "learning_rate": 6.522001569384923e-06, "loss": 1.5607, "step": 2168 }, { "epoch": 0.13, "learning_rate": 6.5280376652381236e-06, "loss": 1.8247, "step": 2170 }, { "epoch": 0.13, "learning_rate": 6.534073761091326e-06, "loss": 1.6909, "step": 2172 }, { "epoch": 0.13, "learning_rate": 6.5401098569445294e-06, "loss": 1.8046, "step": 2174 }, { "epoch": 0.13, "learning_rate": 6.54614595279773e-06, "loss": 1.9332, "step": 2176 }, { "epoch": 0.13, "learning_rate": 6.552182048650933e-06, "loss": 1.8343, "step": 2178 }, { "epoch": 0.13, "learning_rate": 6.5582181445041344e-06, "loss": 1.5162, "step": 2180 }, { "epoch": 0.13, "learning_rate": 6.564254240357337e-06, "loss": 1.6222, "step": 2182 }, { "epoch": 0.13, "learning_rate": 6.5702903362105395e-06, "loss": 2.0345, "step": 2184 }, { "epoch": 0.13, "learning_rate": 6.576326432063741e-06, "loss": 1.4167, "step": 2186 }, { "epoch": 0.13, "learning_rate": 6.582362527916944e-06, "loss": 1.6108, "step": 2188 }, { "epoch": 0.13, "learning_rate": 6.588398623770146e-06, "loss": 1.9628, "step": 2190 }, { "epoch": 0.13, "learning_rate": 6.594434719623348e-06, "loss": 1.7132, "step": 2192 }, { "epoch": 0.13, "learning_rate": 6.60047081547655e-06, "loss": 1.3748, "step": 2194 }, { "epoch": 0.13, "learning_rate": 6.606506911329752e-06, "loss": 1.6876, "step": 2196 }, { "epoch": 0.13, "learning_rate": 6.6125430071829545e-06, "loss": 1.7953, "step": 2198 }, { "epoch": 0.13, "learning_rate": 6.618579103036157e-06, "loss": 1.7828, "step": 2200 }, { "epoch": 0.13, "learning_rate": 6.624615198889359e-06, "loss": 1.8187, "step": 2202 }, { "epoch": 0.13, "learning_rate": 6.630651294742561e-06, "loss": 1.722, "step": 2204 }, { "epoch": 0.13, "learning_rate": 6.636687390595763e-06, "loss": 1.827, "step": 2206 }, { "epoch": 0.13, "learning_rate": 6.642723486448965e-06, "loss": 1.6168, "step": 2208 }, { "epoch": 0.13, "learning_rate": 6.648759582302168e-06, "loss": 1.2514, "step": 2210 }, { "epoch": 0.13, "learning_rate": 6.6547956781553696e-06, "loss": 1.695, "step": 2212 }, { "epoch": 0.13, "learning_rate": 6.660831774008572e-06, "loss": 1.384, "step": 2214 }, { "epoch": 0.13, "learning_rate": 6.666867869861773e-06, "loss": 1.7531, "step": 2216 }, { "epoch": 0.13, "learning_rate": 6.672903965714976e-06, "loss": 1.6514, "step": 2218 }, { "epoch": 0.13, "learning_rate": 6.678940061568179e-06, "loss": 1.9705, "step": 2220 }, { "epoch": 0.13, "learning_rate": 6.68497615742138e-06, "loss": 1.4548, "step": 2222 }, { "epoch": 0.13, "learning_rate": 6.691012253274583e-06, "loss": 1.4008, "step": 2224 }, { "epoch": 0.13, "learning_rate": 6.697048349127784e-06, "loss": 1.6436, "step": 2226 }, { "epoch": 0.13, "learning_rate": 6.703084444980986e-06, "loss": 1.8617, "step": 2228 }, { "epoch": 0.13, "learning_rate": 6.70912054083419e-06, "loss": 2.0886, "step": 2230 }, { "epoch": 0.13, "learning_rate": 6.7151566366873904e-06, "loss": 2.1532, "step": 2232 }, { "epoch": 0.13, "learning_rate": 6.721192732540593e-06, "loss": 1.8985, "step": 2234 }, { "epoch": 0.13, "learning_rate": 6.727228828393795e-06, "loss": 1.662, "step": 2236 }, { "epoch": 0.14, "learning_rate": 6.733264924246997e-06, "loss": 1.9525, "step": 2238 }, { "epoch": 0.14, "learning_rate": 6.7393010201002e-06, "loss": 1.7259, "step": 2240 }, { "epoch": 0.14, "learning_rate": 6.745337115953401e-06, "loss": 1.8483, "step": 2242 }, { "epoch": 0.14, "learning_rate": 6.751373211806604e-06, "loss": 1.5949, "step": 2244 }, { "epoch": 0.14, "learning_rate": 6.757409307659806e-06, "loss": 1.6362, "step": 2246 }, { "epoch": 0.14, "learning_rate": 6.763445403513008e-06, "loss": 1.6322, "step": 2248 }, { "epoch": 0.14, "learning_rate": 6.7694814993662105e-06, "loss": 1.6056, "step": 2250 }, { "epoch": 0.14, "learning_rate": 6.775517595219412e-06, "loss": 1.6117, "step": 2252 }, { "epoch": 0.14, "learning_rate": 6.781553691072615e-06, "loss": 1.6363, "step": 2254 }, { "epoch": 0.14, "learning_rate": 6.787589786925817e-06, "loss": 2.1058, "step": 2256 }, { "epoch": 0.14, "learning_rate": 6.793625882779019e-06, "loss": 2.0608, "step": 2258 }, { "epoch": 0.14, "learning_rate": 6.799661978632221e-06, "loss": 1.6539, "step": 2260 }, { "epoch": 0.14, "learning_rate": 6.805698074485422e-06, "loss": 1.5488, "step": 2262 }, { "epoch": 0.14, "learning_rate": 6.8117341703386256e-06, "loss": 1.5381, "step": 2264 }, { "epoch": 0.14, "learning_rate": 6.817770266191828e-06, "loss": 1.7584, "step": 2266 }, { "epoch": 0.14, "learning_rate": 6.823806362045029e-06, "loss": 1.7318, "step": 2268 }, { "epoch": 0.14, "learning_rate": 6.829842457898232e-06, "loss": 1.6147, "step": 2270 }, { "epoch": 0.14, "learning_rate": 6.835878553751433e-06, "loss": 1.6937, "step": 2272 }, { "epoch": 0.14, "learning_rate": 6.8419146496046364e-06, "loss": 1.3633, "step": 2274 }, { "epoch": 0.14, "learning_rate": 6.847950745457839e-06, "loss": 2.061, "step": 2276 }, { "epoch": 0.14, "learning_rate": 6.85398684131104e-06, "loss": 2.0743, "step": 2278 }, { "epoch": 0.14, "learning_rate": 6.860022937164243e-06, "loss": 1.6671, "step": 2280 }, { "epoch": 0.14, "learning_rate": 6.866059033017444e-06, "loss": 1.8648, "step": 2282 }, { "epoch": 0.14, "learning_rate": 6.8720951288706465e-06, "loss": 1.5546, "step": 2284 }, { "epoch": 0.14, "learning_rate": 6.87813122472385e-06, "loss": 1.6101, "step": 2286 }, { "epoch": 0.14, "learning_rate": 6.884167320577051e-06, "loss": 1.6863, "step": 2288 }, { "epoch": 0.14, "learning_rate": 6.890203416430253e-06, "loss": 1.4213, "step": 2290 }, { "epoch": 0.14, "learning_rate": 6.896239512283455e-06, "loss": 1.661, "step": 2292 }, { "epoch": 0.14, "learning_rate": 6.902275608136657e-06, "loss": 1.4053, "step": 2294 }, { "epoch": 0.14, "learning_rate": 6.90831170398986e-06, "loss": 1.7382, "step": 2296 }, { "epoch": 0.14, "learning_rate": 6.9143477998430615e-06, "loss": 1.3657, "step": 2298 }, { "epoch": 0.14, "learning_rate": 6.920383895696264e-06, "loss": 1.5505, "step": 2300 }, { "epoch": 0.14, "learning_rate": 6.9264199915494665e-06, "loss": 1.9649, "step": 2302 }, { "epoch": 0.14, "learning_rate": 6.932456087402668e-06, "loss": 1.6173, "step": 2304 }, { "epoch": 0.14, "learning_rate": 6.938492183255871e-06, "loss": 2.0728, "step": 2306 }, { "epoch": 0.14, "learning_rate": 6.944528279109072e-06, "loss": 1.6398, "step": 2308 }, { "epoch": 0.14, "learning_rate": 6.950564374962275e-06, "loss": 1.4436, "step": 2310 }, { "epoch": 0.14, "learning_rate": 6.956600470815477e-06, "loss": 1.6626, "step": 2312 }, { "epoch": 0.14, "learning_rate": 6.962636566668679e-06, "loss": 1.6777, "step": 2314 }, { "epoch": 0.14, "learning_rate": 6.968672662521882e-06, "loss": 2.0725, "step": 2316 }, { "epoch": 0.14, "learning_rate": 6.974708758375082e-06, "loss": 1.2266, "step": 2318 }, { "epoch": 0.14, "learning_rate": 6.980744854228286e-06, "loss": 1.4574, "step": 2320 }, { "epoch": 0.14, "learning_rate": 6.986780950081488e-06, "loss": 1.5044, "step": 2322 }, { "epoch": 0.14, "learning_rate": 6.992817045934689e-06, "loss": 1.6963, "step": 2324 }, { "epoch": 0.14, "learning_rate": 6.9988531417878925e-06, "loss": 1.6287, "step": 2326 }, { "epoch": 0.14, "learning_rate": 7.004889237641093e-06, "loss": 1.7782, "step": 2328 }, { "epoch": 0.14, "learning_rate": 7.010925333494296e-06, "loss": 1.4049, "step": 2330 }, { "epoch": 0.14, "learning_rate": 7.016961429347499e-06, "loss": 1.5565, "step": 2332 }, { "epoch": 0.14, "learning_rate": 7.0229975252007e-06, "loss": 1.5479, "step": 2334 }, { "epoch": 0.14, "learning_rate": 7.0290336210539025e-06, "loss": 1.7669, "step": 2336 }, { "epoch": 0.14, "learning_rate": 7.035069716907104e-06, "loss": 1.3539, "step": 2338 }, { "epoch": 0.14, "learning_rate": 7.041105812760307e-06, "loss": 1.5433, "step": 2340 }, { "epoch": 0.14, "learning_rate": 7.04714190861351e-06, "loss": 1.7897, "step": 2342 }, { "epoch": 0.14, "learning_rate": 7.053178004466711e-06, "loss": 1.4345, "step": 2344 }, { "epoch": 0.14, "learning_rate": 7.059214100319913e-06, "loss": 1.8089, "step": 2346 }, { "epoch": 0.14, "learning_rate": 7.065250196173115e-06, "loss": 1.5646, "step": 2348 }, { "epoch": 0.14, "learning_rate": 7.0712862920263175e-06, "loss": 1.4152, "step": 2350 }, { "epoch": 0.14, "learning_rate": 7.07732238787952e-06, "loss": 1.6879, "step": 2352 }, { "epoch": 0.14, "learning_rate": 7.083358483732722e-06, "loss": 1.741, "step": 2354 }, { "epoch": 0.14, "learning_rate": 7.089394579585924e-06, "loss": 1.9651, "step": 2356 }, { "epoch": 0.14, "learning_rate": 7.095430675439127e-06, "loss": 0.992, "step": 2358 }, { "epoch": 0.14, "learning_rate": 7.101466771292328e-06, "loss": 1.6255, "step": 2360 }, { "epoch": 0.14, "learning_rate": 7.107502867145531e-06, "loss": 2.004, "step": 2362 }, { "epoch": 0.14, "learning_rate": 7.1135389629987326e-06, "loss": 1.4532, "step": 2364 }, { "epoch": 0.14, "learning_rate": 7.119575058851935e-06, "loss": 1.6149, "step": 2366 }, { "epoch": 0.14, "learning_rate": 7.125611154705138e-06, "loss": 1.5182, "step": 2368 }, { "epoch": 0.14, "learning_rate": 7.131647250558339e-06, "loss": 1.8795, "step": 2370 }, { "epoch": 0.14, "learning_rate": 7.137683346411542e-06, "loss": 1.7307, "step": 2372 }, { "epoch": 0.14, "learning_rate": 7.143719442264743e-06, "loss": 1.7456, "step": 2374 }, { "epoch": 0.14, "learning_rate": 7.149755538117946e-06, "loss": 1.7935, "step": 2376 }, { "epoch": 0.14, "learning_rate": 7.1557916339711485e-06, "loss": 1.3238, "step": 2378 }, { "epoch": 0.14, "learning_rate": 7.161827729824349e-06, "loss": 1.4187, "step": 2380 }, { "epoch": 0.14, "learning_rate": 7.167863825677553e-06, "loss": 1.7347, "step": 2382 }, { "epoch": 0.14, "learning_rate": 7.1738999215307535e-06, "loss": 1.4341, "step": 2384 }, { "epoch": 0.14, "learning_rate": 7.179936017383956e-06, "loss": 1.4763, "step": 2386 }, { "epoch": 0.14, "learning_rate": 7.185972113237159e-06, "loss": 1.6746, "step": 2388 }, { "epoch": 0.14, "learning_rate": 7.19200820909036e-06, "loss": 1.8172, "step": 2390 }, { "epoch": 0.14, "learning_rate": 7.198044304943563e-06, "loss": 1.3079, "step": 2392 }, { "epoch": 0.14, "learning_rate": 7.204080400796764e-06, "loss": 1.5942, "step": 2394 }, { "epoch": 0.14, "learning_rate": 7.210116496649967e-06, "loss": 1.9864, "step": 2396 }, { "epoch": 0.14, "learning_rate": 7.216152592503169e-06, "loss": 1.8514, "step": 2398 }, { "epoch": 0.14, "learning_rate": 7.222188688356371e-06, "loss": 1.7632, "step": 2400 }, { "epoch": 0.14, "learning_rate": 7.2282247842095735e-06, "loss": 1.5598, "step": 2402 }, { "epoch": 0.15, "learning_rate": 7.234260880062775e-06, "loss": 1.3742, "step": 2404 }, { "epoch": 0.15, "learning_rate": 7.240296975915978e-06, "loss": 1.8272, "step": 2406 }, { "epoch": 0.15, "learning_rate": 7.24633307176918e-06, "loss": 1.9276, "step": 2408 }, { "epoch": 0.15, "learning_rate": 7.252369167622382e-06, "loss": 1.7094, "step": 2410 }, { "epoch": 0.15, "learning_rate": 7.258405263475584e-06, "loss": 1.7564, "step": 2412 }, { "epoch": 0.15, "learning_rate": 7.264441359328787e-06, "loss": 1.441, "step": 2414 }, { "epoch": 0.15, "learning_rate": 7.270477455181989e-06, "loss": 1.479, "step": 2416 }, { "epoch": 0.15, "learning_rate": 7.276513551035191e-06, "loss": 1.6512, "step": 2418 }, { "epoch": 0.15, "learning_rate": 7.282549646888393e-06, "loss": 1.2986, "step": 2420 }, { "epoch": 0.15, "learning_rate": 7.288585742741595e-06, "loss": 1.3352, "step": 2422 }, { "epoch": 0.15, "learning_rate": 7.294621838594798e-06, "loss": 1.4684, "step": 2424 }, { "epoch": 0.15, "learning_rate": 7.3006579344479994e-06, "loss": 1.2159, "step": 2426 }, { "epoch": 0.15, "learning_rate": 7.306694030301202e-06, "loss": 1.8813, "step": 2428 }, { "epoch": 0.15, "learning_rate": 7.312730126154403e-06, "loss": 1.7273, "step": 2430 }, { "epoch": 0.15, "learning_rate": 7.318766222007606e-06, "loss": 1.5823, "step": 2432 }, { "epoch": 0.15, "learning_rate": 7.324802317860809e-06, "loss": 1.8531, "step": 2434 }, { "epoch": 0.15, "learning_rate": 7.3308384137140095e-06, "loss": 1.701, "step": 2436 }, { "epoch": 0.15, "learning_rate": 7.336874509567213e-06, "loss": 1.333, "step": 2438 }, { "epoch": 0.15, "learning_rate": 7.342910605420414e-06, "loss": 2.004, "step": 2440 }, { "epoch": 0.15, "learning_rate": 7.348946701273616e-06, "loss": 1.9782, "step": 2442 }, { "epoch": 0.15, "learning_rate": 7.3549827971268195e-06, "loss": 1.5695, "step": 2444 }, { "epoch": 0.15, "learning_rate": 7.36101889298002e-06, "loss": 1.3869, "step": 2446 }, { "epoch": 0.15, "learning_rate": 7.367054988833223e-06, "loss": 1.6954, "step": 2448 }, { "epoch": 0.15, "learning_rate": 7.3730910846864245e-06, "loss": 1.5166, "step": 2450 }, { "epoch": 0.15, "learning_rate": 7.379127180539627e-06, "loss": 1.376, "step": 2452 }, { "epoch": 0.15, "learning_rate": 7.3851632763928295e-06, "loss": 1.5985, "step": 2454 }, { "epoch": 0.15, "learning_rate": 7.391199372246031e-06, "loss": 1.9097, "step": 2456 }, { "epoch": 0.15, "learning_rate": 7.397235468099234e-06, "loss": 1.5903, "step": 2458 }, { "epoch": 0.15, "learning_rate": 7.403271563952435e-06, "loss": 1.673, "step": 2460 }, { "epoch": 0.15, "learning_rate": 7.409307659805638e-06, "loss": 1.493, "step": 2462 }, { "epoch": 0.15, "learning_rate": 7.41534375565884e-06, "loss": 1.6114, "step": 2464 }, { "epoch": 0.15, "learning_rate": 7.421379851512042e-06, "loss": 1.3599, "step": 2466 }, { "epoch": 0.15, "learning_rate": 7.427415947365245e-06, "loss": 1.5775, "step": 2468 }, { "epoch": 0.15, "learning_rate": 7.433452043218447e-06, "loss": 1.39, "step": 2470 }, { "epoch": 0.15, "learning_rate": 7.439488139071649e-06, "loss": 1.4306, "step": 2472 }, { "epoch": 0.15, "learning_rate": 7.445524234924851e-06, "loss": 1.4714, "step": 2474 }, { "epoch": 0.15, "learning_rate": 7.451560330778053e-06, "loss": 1.5354, "step": 2476 }, { "epoch": 0.15, "learning_rate": 7.4575964266312555e-06, "loss": 1.5925, "step": 2478 }, { "epoch": 0.15, "learning_rate": 7.463632522484458e-06, "loss": 1.5765, "step": 2480 }, { "epoch": 0.15, "learning_rate": 7.46966861833766e-06, "loss": 1.3869, "step": 2482 }, { "epoch": 0.15, "learning_rate": 7.475704714190862e-06, "loss": 1.9677, "step": 2484 }, { "epoch": 0.15, "learning_rate": 7.481740810044063e-06, "loss": 1.3437, "step": 2486 }, { "epoch": 0.15, "learning_rate": 7.487776905897266e-06, "loss": 1.4899, "step": 2488 }, { "epoch": 0.15, "learning_rate": 7.493813001750469e-06, "loss": 1.7444, "step": 2490 }, { "epoch": 0.15, "learning_rate": 7.49984909760367e-06, "loss": 1.3257, "step": 2492 }, { "epoch": 0.15, "learning_rate": 7.505885193456873e-06, "loss": 1.4084, "step": 2494 }, { "epoch": 0.15, "learning_rate": 7.511921289310074e-06, "loss": 1.8421, "step": 2496 }, { "epoch": 0.15, "learning_rate": 7.517957385163276e-06, "loss": 1.3825, "step": 2498 }, { "epoch": 0.15, "learning_rate": 7.52399348101648e-06, "loss": 1.7207, "step": 2500 }, { "epoch": 0.15, "learning_rate": 7.5300295768696805e-06, "loss": 1.6119, "step": 2502 }, { "epoch": 0.15, "learning_rate": 7.536065672722883e-06, "loss": 1.5399, "step": 2504 }, { "epoch": 0.15, "learning_rate": 7.542101768576085e-06, "loss": 1.2805, "step": 2506 }, { "epoch": 0.15, "learning_rate": 7.548137864429287e-06, "loss": 1.8174, "step": 2508 }, { "epoch": 0.15, "learning_rate": 7.55417396028249e-06, "loss": 1.6885, "step": 2510 }, { "epoch": 0.15, "learning_rate": 7.560210056135691e-06, "loss": 1.4421, "step": 2512 }, { "epoch": 0.15, "learning_rate": 7.566246151988894e-06, "loss": 1.6468, "step": 2514 }, { "epoch": 0.15, "learning_rate": 7.572282247842096e-06, "loss": 1.6918, "step": 2516 }, { "epoch": 0.15, "learning_rate": 7.578318343695298e-06, "loss": 1.5914, "step": 2518 }, { "epoch": 0.15, "learning_rate": 7.584354439548501e-06, "loss": 1.5326, "step": 2520 }, { "epoch": 0.15, "learning_rate": 7.590390535401702e-06, "loss": 1.3195, "step": 2522 }, { "epoch": 0.15, "learning_rate": 7.596426631254905e-06, "loss": 1.5768, "step": 2524 }, { "epoch": 0.15, "learning_rate": 7.602462727108107e-06, "loss": 1.6253, "step": 2526 }, { "epoch": 0.15, "learning_rate": 7.608498822961309e-06, "loss": 1.4239, "step": 2528 }, { "epoch": 0.15, "learning_rate": 7.6145349188145115e-06, "loss": 1.9165, "step": 2530 }, { "epoch": 0.15, "learning_rate": 7.620571014667713e-06, "loss": 1.3442, "step": 2532 }, { "epoch": 0.15, "learning_rate": 7.626607110520916e-06, "loss": 1.8811, "step": 2534 }, { "epoch": 0.15, "learning_rate": 7.632643206374119e-06, "loss": 1.7096, "step": 2536 }, { "epoch": 0.15, "learning_rate": 7.63867930222732e-06, "loss": 1.663, "step": 2538 }, { "epoch": 0.15, "learning_rate": 7.644715398080522e-06, "loss": 1.8709, "step": 2540 }, { "epoch": 0.15, "learning_rate": 7.650751493933723e-06, "loss": 1.4963, "step": 2542 }, { "epoch": 0.15, "learning_rate": 7.656787589786926e-06, "loss": 1.6781, "step": 2544 }, { "epoch": 0.15, "learning_rate": 7.662823685640128e-06, "loss": 2.039, "step": 2546 }, { "epoch": 0.15, "learning_rate": 7.66885978149333e-06, "loss": 1.4358, "step": 2548 }, { "epoch": 0.15, "learning_rate": 7.674895877346533e-06, "loss": 1.4495, "step": 2550 }, { "epoch": 0.15, "learning_rate": 7.680931973199734e-06, "loss": 1.5469, "step": 2552 }, { "epoch": 0.15, "learning_rate": 7.686968069052937e-06, "loss": 1.7924, "step": 2554 }, { "epoch": 0.15, "learning_rate": 7.693004164906139e-06, "loss": 1.3431, "step": 2556 }, { "epoch": 0.15, "learning_rate": 7.699040260759342e-06, "loss": 1.5654, "step": 2558 }, { "epoch": 0.15, "learning_rate": 7.705076356612544e-06, "loss": 1.8037, "step": 2560 }, { "epoch": 0.15, "learning_rate": 7.711112452465745e-06, "loss": 1.9034, "step": 2562 }, { "epoch": 0.15, "learning_rate": 7.717148548318947e-06, "loss": 2.0662, "step": 2564 }, { "epoch": 0.15, "learning_rate": 7.72318464417215e-06, "loss": 1.496, "step": 2566 }, { "epoch": 0.15, "learning_rate": 7.729220740025352e-06, "loss": 1.3893, "step": 2568 }, { "epoch": 0.16, "learning_rate": 7.735256835878555e-06, "loss": 1.7935, "step": 2570 }, { "epoch": 0.16, "learning_rate": 7.741292931731756e-06, "loss": 1.436, "step": 2572 }, { "epoch": 0.16, "learning_rate": 7.747329027584958e-06, "loss": 1.5655, "step": 2574 }, { "epoch": 0.16, "learning_rate": 7.75336512343816e-06, "loss": 1.6535, "step": 2576 }, { "epoch": 0.16, "learning_rate": 7.759401219291362e-06, "loss": 1.2197, "step": 2578 }, { "epoch": 0.16, "learning_rate": 7.765437315144566e-06, "loss": 1.73, "step": 2580 }, { "epoch": 0.16, "learning_rate": 7.771473410997767e-06, "loss": 1.2372, "step": 2582 }, { "epoch": 0.16, "learning_rate": 7.777509506850969e-06, "loss": 1.8371, "step": 2584 }, { "epoch": 0.16, "learning_rate": 7.783545602704172e-06, "loss": 1.769, "step": 2586 }, { "epoch": 0.16, "learning_rate": 7.789581698557372e-06, "loss": 1.7701, "step": 2588 }, { "epoch": 0.16, "learning_rate": 7.795617794410575e-06, "loss": 1.7911, "step": 2590 }, { "epoch": 0.16, "learning_rate": 7.80165389026378e-06, "loss": 1.6756, "step": 2592 }, { "epoch": 0.16, "learning_rate": 7.80768998611698e-06, "loss": 1.5372, "step": 2594 }, { "epoch": 0.16, "learning_rate": 7.813726081970183e-06, "loss": 1.4583, "step": 2596 }, { "epoch": 0.16, "learning_rate": 7.819762177823383e-06, "loss": 1.3078, "step": 2598 }, { "epoch": 0.16, "learning_rate": 7.825798273676586e-06, "loss": 1.6537, "step": 2600 }, { "epoch": 0.16, "learning_rate": 7.831834369529788e-06, "loss": 1.5393, "step": 2602 }, { "epoch": 0.16, "learning_rate": 7.837870465382991e-06, "loss": 1.8642, "step": 2604 }, { "epoch": 0.16, "learning_rate": 7.843906561236193e-06, "loss": 1.4913, "step": 2606 }, { "epoch": 0.16, "learning_rate": 7.849942657089394e-06, "loss": 1.1016, "step": 2608 }, { "epoch": 0.16, "learning_rate": 7.855978752942597e-06, "loss": 1.4996, "step": 2610 }, { "epoch": 0.16, "learning_rate": 7.8620148487958e-06, "loss": 1.6813, "step": 2612 }, { "epoch": 0.16, "learning_rate": 7.868050944649002e-06, "loss": 1.6037, "step": 2614 }, { "epoch": 0.16, "learning_rate": 7.874087040502204e-06, "loss": 1.3399, "step": 2616 }, { "epoch": 0.16, "learning_rate": 7.880123136355405e-06, "loss": 2.112, "step": 2618 }, { "epoch": 0.16, "learning_rate": 7.886159232208608e-06, "loss": 1.596, "step": 2620 }, { "epoch": 0.16, "learning_rate": 7.89219532806181e-06, "loss": 1.5423, "step": 2622 }, { "epoch": 0.16, "learning_rate": 7.898231423915013e-06, "loss": 1.5366, "step": 2624 }, { "epoch": 0.16, "learning_rate": 7.904267519768215e-06, "loss": 1.3316, "step": 2626 }, { "epoch": 0.16, "learning_rate": 7.910303615621416e-06, "loss": 1.5215, "step": 2628 }, { "epoch": 0.16, "learning_rate": 7.916339711474618e-06, "loss": 1.3532, "step": 2630 }, { "epoch": 0.16, "learning_rate": 7.922375807327821e-06, "loss": 1.5145, "step": 2632 }, { "epoch": 0.16, "learning_rate": 7.928411903181022e-06, "loss": 2.0513, "step": 2634 }, { "epoch": 0.16, "learning_rate": 7.934447999034226e-06, "loss": 1.3716, "step": 2636 }, { "epoch": 0.16, "learning_rate": 7.940484094887427e-06, "loss": 1.6854, "step": 2638 }, { "epoch": 0.16, "learning_rate": 7.94652019074063e-06, "loss": 1.5598, "step": 2640 }, { "epoch": 0.16, "learning_rate": 7.952556286593832e-06, "loss": 1.3845, "step": 2642 }, { "epoch": 0.16, "learning_rate": 7.958592382447033e-06, "loss": 1.0641, "step": 2644 }, { "epoch": 0.16, "learning_rate": 7.964628478300235e-06, "loss": 1.3382, "step": 2646 }, { "epoch": 0.16, "learning_rate": 7.97066457415344e-06, "loss": 1.4262, "step": 2648 }, { "epoch": 0.16, "learning_rate": 7.97670067000664e-06, "loss": 1.4624, "step": 2650 }, { "epoch": 0.16, "learning_rate": 7.982736765859843e-06, "loss": 1.2917, "step": 2652 }, { "epoch": 0.16, "learning_rate": 7.988772861713044e-06, "loss": 1.7731, "step": 2654 }, { "epoch": 0.16, "learning_rate": 7.994808957566246e-06, "loss": 1.4027, "step": 2656 }, { "epoch": 0.16, "learning_rate": 8.000845053419449e-06, "loss": 1.2566, "step": 2658 }, { "epoch": 0.16, "learning_rate": 8.006881149272651e-06, "loss": 1.1567, "step": 2660 }, { "epoch": 0.16, "learning_rate": 8.012917245125854e-06, "loss": 1.6781, "step": 2662 }, { "epoch": 0.16, "learning_rate": 8.018953340979054e-06, "loss": 1.356, "step": 2664 }, { "epoch": 0.16, "learning_rate": 8.024989436832257e-06, "loss": 1.7874, "step": 2666 }, { "epoch": 0.16, "learning_rate": 8.03102553268546e-06, "loss": 1.3544, "step": 2668 }, { "epoch": 0.16, "learning_rate": 8.037061628538662e-06, "loss": 1.5545, "step": 2670 }, { "epoch": 0.16, "learning_rate": 8.043097724391864e-06, "loss": 1.5, "step": 2672 }, { "epoch": 0.16, "learning_rate": 8.049133820245065e-06, "loss": 1.9878, "step": 2674 }, { "epoch": 0.16, "learning_rate": 8.055169916098268e-06, "loss": 1.5201, "step": 2676 }, { "epoch": 0.16, "learning_rate": 8.06120601195147e-06, "loss": 1.392, "step": 2678 }, { "epoch": 0.16, "learning_rate": 8.067242107804673e-06, "loss": 1.576, "step": 2680 }, { "epoch": 0.16, "learning_rate": 8.073278203657875e-06, "loss": 1.4556, "step": 2682 }, { "epoch": 0.16, "learning_rate": 8.079314299511076e-06, "loss": 1.2824, "step": 2684 }, { "epoch": 0.16, "learning_rate": 8.085350395364279e-06, "loss": 1.3255, "step": 2686 }, { "epoch": 0.16, "learning_rate": 8.091386491217481e-06, "loss": 1.6864, "step": 2688 }, { "epoch": 0.16, "learning_rate": 8.097422587070682e-06, "loss": 1.7809, "step": 2690 }, { "epoch": 0.16, "learning_rate": 8.103458682923886e-06, "loss": 1.3945, "step": 2692 }, { "epoch": 0.16, "learning_rate": 8.109494778777087e-06, "loss": 1.5676, "step": 2694 }, { "epoch": 0.16, "learning_rate": 8.11553087463029e-06, "loss": 1.2591, "step": 2696 }, { "epoch": 0.16, "learning_rate": 8.121566970483492e-06, "loss": 1.3357, "step": 2698 }, { "epoch": 0.16, "learning_rate": 8.127603066336693e-06, "loss": 1.4997, "step": 2700 }, { "epoch": 0.16, "learning_rate": 8.133639162189895e-06, "loss": 1.425, "step": 2702 }, { "epoch": 0.16, "learning_rate": 8.1396752580431e-06, "loss": 1.3469, "step": 2704 }, { "epoch": 0.16, "learning_rate": 8.1457113538963e-06, "loss": 1.5654, "step": 2706 }, { "epoch": 0.16, "learning_rate": 8.151747449749503e-06, "loss": 1.9396, "step": 2708 }, { "epoch": 0.16, "learning_rate": 8.157783545602704e-06, "loss": 1.8839, "step": 2710 }, { "epoch": 0.16, "learning_rate": 8.163819641455906e-06, "loss": 1.692, "step": 2712 }, { "epoch": 0.16, "learning_rate": 8.169855737309109e-06, "loss": 1.5039, "step": 2714 }, { "epoch": 0.16, "learning_rate": 8.175891833162311e-06, "loss": 1.3876, "step": 2716 }, { "epoch": 0.16, "learning_rate": 8.181927929015514e-06, "loss": 1.7446, "step": 2718 }, { "epoch": 0.16, "learning_rate": 8.187964024868715e-06, "loss": 1.5083, "step": 2720 }, { "epoch": 0.16, "learning_rate": 8.194000120721917e-06, "loss": 1.4737, "step": 2722 }, { "epoch": 0.16, "learning_rate": 8.20003621657512e-06, "loss": 1.4574, "step": 2724 }, { "epoch": 0.16, "learning_rate": 8.206072312428322e-06, "loss": 1.4449, "step": 2726 }, { "epoch": 0.16, "learning_rate": 8.212108408281525e-06, "loss": 1.0657, "step": 2728 }, { "epoch": 0.16, "learning_rate": 8.218144504134725e-06, "loss": 1.3961, "step": 2730 }, { "epoch": 0.16, "learning_rate": 8.224180599987928e-06, "loss": 1.7627, "step": 2732 }, { "epoch": 0.17, "learning_rate": 8.23021669584113e-06, "loss": 1.43, "step": 2734 }, { "epoch": 0.17, "learning_rate": 8.236252791694333e-06, "loss": 1.2663, "step": 2736 }, { "epoch": 0.17, "learning_rate": 8.242288887547536e-06, "loss": 1.498, "step": 2738 }, { "epoch": 0.17, "learning_rate": 8.248324983400736e-06, "loss": 1.9162, "step": 2740 }, { "epoch": 0.17, "learning_rate": 8.254361079253939e-06, "loss": 1.6312, "step": 2742 }, { "epoch": 0.17, "learning_rate": 8.260397175107141e-06, "loss": 1.3778, "step": 2744 }, { "epoch": 0.17, "learning_rate": 8.266433270960342e-06, "loss": 1.4636, "step": 2746 }, { "epoch": 0.17, "learning_rate": 8.272469366813546e-06, "loss": 1.5903, "step": 2748 }, { "epoch": 0.17, "learning_rate": 8.278505462666747e-06, "loss": 1.9409, "step": 2750 }, { "epoch": 0.17, "learning_rate": 8.28454155851995e-06, "loss": 1.9197, "step": 2752 }, { "epoch": 0.17, "learning_rate": 8.290577654373152e-06, "loss": 1.5042, "step": 2754 }, { "epoch": 0.17, "learning_rate": 8.296613750226353e-06, "loss": 1.639, "step": 2756 }, { "epoch": 0.17, "learning_rate": 8.302649846079556e-06, "loss": 1.6864, "step": 2758 }, { "epoch": 0.17, "learning_rate": 8.30868594193276e-06, "loss": 1.4995, "step": 2760 }, { "epoch": 0.17, "learning_rate": 8.31472203778596e-06, "loss": 1.5885, "step": 2762 }, { "epoch": 0.17, "learning_rate": 8.320758133639163e-06, "loss": 1.6485, "step": 2764 }, { "epoch": 0.17, "learning_rate": 8.326794229492364e-06, "loss": 1.6079, "step": 2766 }, { "epoch": 0.17, "learning_rate": 8.332830325345566e-06, "loss": 1.5926, "step": 2768 }, { "epoch": 0.17, "learning_rate": 8.338866421198769e-06, "loss": 2.1585, "step": 2770 }, { "epoch": 0.17, "learning_rate": 8.344902517051971e-06, "loss": 1.5177, "step": 2772 }, { "epoch": 0.17, "learning_rate": 8.350938612905174e-06, "loss": 1.2483, "step": 2774 }, { "epoch": 0.17, "learning_rate": 8.356974708758375e-06, "loss": 1.079, "step": 2776 }, { "epoch": 0.17, "learning_rate": 8.363010804611577e-06, "loss": 1.6636, "step": 2778 }, { "epoch": 0.17, "learning_rate": 8.36904690046478e-06, "loss": 1.5975, "step": 2780 }, { "epoch": 0.17, "learning_rate": 8.375082996317982e-06, "loss": 1.2847, "step": 2782 }, { "epoch": 0.17, "learning_rate": 8.381119092171185e-06, "loss": 1.8898, "step": 2784 }, { "epoch": 0.17, "learning_rate": 8.387155188024386e-06, "loss": 1.7319, "step": 2786 }, { "epoch": 0.17, "learning_rate": 8.393191283877588e-06, "loss": 1.4391, "step": 2788 }, { "epoch": 0.17, "learning_rate": 8.39922737973079e-06, "loss": 1.4525, "step": 2790 }, { "epoch": 0.17, "learning_rate": 8.405263475583993e-06, "loss": 1.5994, "step": 2792 }, { "epoch": 0.17, "learning_rate": 8.411299571437196e-06, "loss": 1.8803, "step": 2794 }, { "epoch": 0.17, "learning_rate": 8.417335667290397e-06, "loss": 1.3505, "step": 2796 }, { "epoch": 0.17, "learning_rate": 8.423371763143599e-06, "loss": 1.3691, "step": 2798 }, { "epoch": 0.17, "learning_rate": 8.429407858996802e-06, "loss": 1.434, "step": 2800 }, { "epoch": 0.17, "learning_rate": 8.435443954850002e-06, "loss": 1.6098, "step": 2802 }, { "epoch": 0.17, "learning_rate": 8.441480050703207e-06, "loss": 1.5152, "step": 2804 }, { "epoch": 0.17, "learning_rate": 8.447516146556407e-06, "loss": 1.9154, "step": 2806 }, { "epoch": 0.17, "learning_rate": 8.45355224240961e-06, "loss": 1.2862, "step": 2808 }, { "epoch": 0.17, "learning_rate": 8.459588338262812e-06, "loss": 1.5905, "step": 2810 }, { "epoch": 0.17, "learning_rate": 8.465624434116013e-06, "loss": 2.0089, "step": 2812 }, { "epoch": 0.17, "learning_rate": 8.471660529969216e-06, "loss": 1.2824, "step": 2814 }, { "epoch": 0.17, "learning_rate": 8.47769662582242e-06, "loss": 1.7498, "step": 2816 }, { "epoch": 0.17, "learning_rate": 8.48373272167562e-06, "loss": 1.2659, "step": 2818 }, { "epoch": 0.17, "learning_rate": 8.489768817528823e-06, "loss": 1.2578, "step": 2820 }, { "epoch": 0.17, "learning_rate": 8.495804913382024e-06, "loss": 1.3455, "step": 2822 }, { "epoch": 0.17, "learning_rate": 8.501841009235227e-06, "loss": 1.4037, "step": 2824 }, { "epoch": 0.17, "learning_rate": 8.507877105088429e-06, "loss": 1.3753, "step": 2826 }, { "epoch": 0.17, "learning_rate": 8.513913200941632e-06, "loss": 1.5157, "step": 2828 }, { "epoch": 0.17, "learning_rate": 8.519949296794834e-06, "loss": 1.7347, "step": 2830 }, { "epoch": 0.17, "learning_rate": 8.525985392648035e-06, "loss": 1.3601, "step": 2832 }, { "epoch": 0.17, "learning_rate": 8.532021488501237e-06, "loss": 1.4061, "step": 2834 }, { "epoch": 0.17, "learning_rate": 8.53805758435444e-06, "loss": 2.0113, "step": 2836 }, { "epoch": 0.17, "learning_rate": 8.544093680207643e-06, "loss": 1.2745, "step": 2838 }, { "epoch": 0.17, "learning_rate": 8.550129776060845e-06, "loss": 1.4288, "step": 2840 }, { "epoch": 0.17, "learning_rate": 8.556165871914046e-06, "loss": 2.1322, "step": 2842 }, { "epoch": 0.17, "learning_rate": 8.562201967767248e-06, "loss": 1.7625, "step": 2844 }, { "epoch": 0.17, "learning_rate": 8.568238063620451e-06, "loss": 1.5627, "step": 2846 }, { "epoch": 0.17, "learning_rate": 8.574274159473652e-06, "loss": 1.5439, "step": 2848 }, { "epoch": 0.17, "learning_rate": 8.580310255326856e-06, "loss": 1.6115, "step": 2850 }, { "epoch": 0.17, "learning_rate": 8.586346351180057e-06, "loss": 1.9973, "step": 2852 }, { "epoch": 0.17, "learning_rate": 8.59238244703326e-06, "loss": 1.5014, "step": 2854 }, { "epoch": 0.17, "learning_rate": 8.598418542886462e-06, "loss": 1.5751, "step": 2856 }, { "epoch": 0.17, "learning_rate": 8.604454638739663e-06, "loss": 1.7708, "step": 2858 }, { "epoch": 0.17, "learning_rate": 8.610490734592867e-06, "loss": 1.3468, "step": 2860 }, { "epoch": 0.17, "learning_rate": 8.616526830446068e-06, "loss": 1.3193, "step": 2862 }, { "epoch": 0.17, "learning_rate": 8.62256292629927e-06, "loss": 1.5115, "step": 2864 }, { "epoch": 0.17, "learning_rate": 8.628599022152473e-06, "loss": 1.4495, "step": 2866 }, { "epoch": 0.17, "learning_rate": 8.634635118005673e-06, "loss": 1.6172, "step": 2868 }, { "epoch": 0.17, "learning_rate": 8.640671213858876e-06, "loss": 1.6967, "step": 2870 }, { "epoch": 0.17, "learning_rate": 8.64670730971208e-06, "loss": 1.1217, "step": 2872 }, { "epoch": 0.17, "learning_rate": 8.652743405565281e-06, "loss": 1.4497, "step": 2874 }, { "epoch": 0.17, "learning_rate": 8.658779501418483e-06, "loss": 1.471, "step": 2876 }, { "epoch": 0.17, "learning_rate": 8.664815597271684e-06, "loss": 1.2902, "step": 2878 }, { "epoch": 0.17, "learning_rate": 8.670851693124887e-06, "loss": 1.5898, "step": 2880 }, { "epoch": 0.17, "learning_rate": 8.67688778897809e-06, "loss": 1.3872, "step": 2882 }, { "epoch": 0.17, "learning_rate": 8.682923884831292e-06, "loss": 1.3058, "step": 2884 }, { "epoch": 0.17, "learning_rate": 8.688959980684494e-06, "loss": 1.4592, "step": 2886 }, { "epoch": 0.17, "learning_rate": 8.694996076537695e-06, "loss": 1.4556, "step": 2888 }, { "epoch": 0.17, "learning_rate": 8.701032172390898e-06, "loss": 1.6874, "step": 2890 }, { "epoch": 0.17, "learning_rate": 8.7070682682441e-06, "loss": 1.3896, "step": 2892 }, { "epoch": 0.17, "learning_rate": 8.713104364097303e-06, "loss": 1.571, "step": 2894 }, { "epoch": 0.17, "learning_rate": 8.719140459950505e-06, "loss": 1.2893, "step": 2896 }, { "epoch": 0.17, "learning_rate": 8.725176555803706e-06, "loss": 1.6869, "step": 2898 }, { "epoch": 0.18, "learning_rate": 8.731212651656909e-06, "loss": 1.6615, "step": 2900 }, { "epoch": 0.18, "learning_rate": 8.737248747510111e-06, "loss": 1.4881, "step": 2902 }, { "epoch": 0.18, "learning_rate": 8.743284843363312e-06, "loss": 1.5477, "step": 2904 }, { "epoch": 0.18, "learning_rate": 8.749320939216516e-06, "loss": 1.4627, "step": 2906 }, { "epoch": 0.18, "learning_rate": 8.755357035069717e-06, "loss": 1.3849, "step": 2908 }, { "epoch": 0.18, "learning_rate": 8.76139313092292e-06, "loss": 1.2478, "step": 2910 }, { "epoch": 0.18, "learning_rate": 8.767429226776122e-06, "loss": 1.2376, "step": 2912 }, { "epoch": 0.18, "learning_rate": 8.773465322629323e-06, "loss": 1.4917, "step": 2914 }, { "epoch": 0.18, "learning_rate": 8.779501418482525e-06, "loss": 1.6135, "step": 2916 }, { "epoch": 0.18, "learning_rate": 8.785537514335728e-06, "loss": 1.5346, "step": 2918 }, { "epoch": 0.18, "learning_rate": 8.79157361018893e-06, "loss": 1.621, "step": 2920 }, { "epoch": 0.18, "learning_rate": 8.797609706042133e-06, "loss": 1.779, "step": 2922 }, { "epoch": 0.18, "learning_rate": 8.803645801895334e-06, "loss": 1.3751, "step": 2924 }, { "epoch": 0.18, "learning_rate": 8.809681897748536e-06, "loss": 1.3356, "step": 2926 }, { "epoch": 0.18, "learning_rate": 8.81571799360174e-06, "loss": 1.1841, "step": 2928 }, { "epoch": 0.18, "learning_rate": 8.821754089454941e-06, "loss": 1.5451, "step": 2930 }, { "epoch": 0.18, "learning_rate": 8.827790185308144e-06, "loss": 1.5201, "step": 2932 }, { "epoch": 0.18, "learning_rate": 8.833826281161344e-06, "loss": 1.6265, "step": 2934 }, { "epoch": 0.18, "learning_rate": 8.839862377014547e-06, "loss": 1.4243, "step": 2936 }, { "epoch": 0.18, "learning_rate": 8.84589847286775e-06, "loss": 1.3662, "step": 2938 }, { "epoch": 0.18, "learning_rate": 8.851934568720952e-06, "loss": 1.4974, "step": 2940 }, { "epoch": 0.18, "learning_rate": 8.857970664574155e-06, "loss": 1.6542, "step": 2942 }, { "epoch": 0.18, "learning_rate": 8.864006760427355e-06, "loss": 1.6843, "step": 2944 }, { "epoch": 0.18, "learning_rate": 8.870042856280558e-06, "loss": 1.5851, "step": 2946 }, { "epoch": 0.18, "learning_rate": 8.87607895213376e-06, "loss": 1.8421, "step": 2948 }, { "epoch": 0.18, "learning_rate": 8.882115047986963e-06, "loss": 1.5557, "step": 2950 }, { "epoch": 0.18, "learning_rate": 8.888151143840165e-06, "loss": 1.6304, "step": 2952 }, { "epoch": 0.18, "learning_rate": 8.894187239693366e-06, "loss": 1.6033, "step": 2954 }, { "epoch": 0.18, "learning_rate": 8.900223335546569e-06, "loss": 1.4856, "step": 2956 }, { "epoch": 0.18, "learning_rate": 8.906259431399771e-06, "loss": 1.7806, "step": 2958 }, { "epoch": 0.18, "learning_rate": 8.912295527252972e-06, "loss": 1.8553, "step": 2960 }, { "epoch": 0.18, "learning_rate": 8.918331623106176e-06, "loss": 1.9486, "step": 2962 }, { "epoch": 0.18, "learning_rate": 8.924367718959377e-06, "loss": 1.5039, "step": 2964 }, { "epoch": 0.18, "learning_rate": 8.93040381481258e-06, "loss": 1.3797, "step": 2966 }, { "epoch": 0.18, "learning_rate": 8.936439910665782e-06, "loss": 1.6634, "step": 2968 }, { "epoch": 0.18, "learning_rate": 8.942476006518983e-06, "loss": 1.2401, "step": 2970 }, { "epoch": 0.18, "learning_rate": 8.948512102372185e-06, "loss": 1.7238, "step": 2972 }, { "epoch": 0.18, "learning_rate": 8.954548198225388e-06, "loss": 1.4484, "step": 2974 }, { "epoch": 0.18, "learning_rate": 8.96058429407859e-06, "loss": 1.2282, "step": 2976 }, { "epoch": 0.18, "learning_rate": 8.966620389931793e-06, "loss": 1.9563, "step": 2978 }, { "epoch": 0.18, "learning_rate": 8.972656485784994e-06, "loss": 1.4331, "step": 2980 }, { "epoch": 0.18, "learning_rate": 8.978692581638196e-06, "loss": 1.8961, "step": 2982 }, { "epoch": 0.18, "learning_rate": 8.984728677491399e-06, "loss": 1.3442, "step": 2984 }, { "epoch": 0.18, "learning_rate": 8.990764773344601e-06, "loss": 1.4189, "step": 2986 }, { "epoch": 0.18, "learning_rate": 8.996800869197804e-06, "loss": 1.5755, "step": 2988 }, { "epoch": 0.18, "learning_rate": 9.002836965051005e-06, "loss": 1.7578, "step": 2990 }, { "epoch": 0.18, "learning_rate": 9.008873060904207e-06, "loss": 1.6108, "step": 2992 }, { "epoch": 0.18, "learning_rate": 9.01490915675741e-06, "loss": 1.3323, "step": 2994 }, { "epoch": 0.18, "learning_rate": 9.020945252610612e-06, "loss": 1.5291, "step": 2996 }, { "epoch": 0.18, "learning_rate": 9.026981348463815e-06, "loss": 1.3573, "step": 2998 }, { "epoch": 0.18, "learning_rate": 9.033017444317016e-06, "loss": 1.3664, "step": 3000 }, { "epoch": 0.18, "learning_rate": 9.039053540170218e-06, "loss": 1.5041, "step": 3002 }, { "epoch": 0.18, "learning_rate": 9.04508963602342e-06, "loss": 1.9593, "step": 3004 }, { "epoch": 0.18, "learning_rate": 9.051125731876623e-06, "loss": 1.974, "step": 3006 }, { "epoch": 0.18, "learning_rate": 9.057161827729826e-06, "loss": 1.3523, "step": 3008 }, { "epoch": 0.18, "learning_rate": 9.063197923583026e-06, "loss": 1.4138, "step": 3010 }, { "epoch": 0.18, "learning_rate": 9.069234019436229e-06, "loss": 1.4651, "step": 3012 }, { "epoch": 0.18, "learning_rate": 9.075270115289431e-06, "loss": 1.6572, "step": 3014 }, { "epoch": 0.18, "learning_rate": 9.081306211142632e-06, "loss": 1.5739, "step": 3016 }, { "epoch": 0.18, "learning_rate": 9.087342306995836e-06, "loss": 1.8284, "step": 3018 }, { "epoch": 0.18, "learning_rate": 9.093378402849037e-06, "loss": 1.285, "step": 3020 }, { "epoch": 0.18, "learning_rate": 9.09941449870224e-06, "loss": 1.3916, "step": 3022 }, { "epoch": 0.18, "learning_rate": 9.105450594555442e-06, "loss": 1.7892, "step": 3024 }, { "epoch": 0.18, "learning_rate": 9.111486690408643e-06, "loss": 1.7985, "step": 3026 }, { "epoch": 0.18, "learning_rate": 9.117522786261846e-06, "loss": 1.637, "step": 3028 }, { "epoch": 0.18, "learning_rate": 9.123558882115048e-06, "loss": 1.5546, "step": 3030 }, { "epoch": 0.18, "learning_rate": 9.12959497796825e-06, "loss": 1.4578, "step": 3032 }, { "epoch": 0.18, "learning_rate": 9.135631073821453e-06, "loss": 1.5081, "step": 3034 }, { "epoch": 0.18, "learning_rate": 9.141667169674654e-06, "loss": 1.6426, "step": 3036 }, { "epoch": 0.18, "learning_rate": 9.147703265527857e-06, "loss": 1.5715, "step": 3038 }, { "epoch": 0.18, "learning_rate": 9.153739361381059e-06, "loss": 1.4095, "step": 3040 }, { "epoch": 0.18, "learning_rate": 9.159775457234262e-06, "loss": 2.2009, "step": 3042 }, { "epoch": 0.18, "learning_rate": 9.165811553087464e-06, "loss": 1.6373, "step": 3044 }, { "epoch": 0.18, "learning_rate": 9.171847648940665e-06, "loss": 1.6314, "step": 3046 }, { "epoch": 0.18, "learning_rate": 9.177883744793867e-06, "loss": 1.5896, "step": 3048 }, { "epoch": 0.18, "learning_rate": 9.18391984064707e-06, "loss": 1.3919, "step": 3050 }, { "epoch": 0.18, "learning_rate": 9.189955936500272e-06, "loss": 1.1854, "step": 3052 }, { "epoch": 0.18, "learning_rate": 9.195992032353475e-06, "loss": 1.5411, "step": 3054 }, { "epoch": 0.18, "learning_rate": 9.202028128206676e-06, "loss": 1.2986, "step": 3056 }, { "epoch": 0.18, "learning_rate": 9.208064224059878e-06, "loss": 1.4284, "step": 3058 }, { "epoch": 0.18, "learning_rate": 9.21410031991308e-06, "loss": 1.263, "step": 3060 }, { "epoch": 0.18, "learning_rate": 9.220136415766283e-06, "loss": 1.4706, "step": 3062 }, { "epoch": 0.18, "learning_rate": 9.226172511619486e-06, "loss": 1.6752, "step": 3064 }, { "epoch": 0.19, "learning_rate": 9.232208607472687e-06, "loss": 1.5054, "step": 3066 }, { "epoch": 0.19, "learning_rate": 9.238244703325889e-06, "loss": 1.9759, "step": 3068 }, { "epoch": 0.19, "learning_rate": 9.244280799179092e-06, "loss": 1.5197, "step": 3070 }, { "epoch": 0.19, "learning_rate": 9.250316895032292e-06, "loss": 1.2958, "step": 3072 }, { "epoch": 0.19, "learning_rate": 9.256352990885497e-06, "loss": 1.8925, "step": 3074 }, { "epoch": 0.19, "learning_rate": 9.262389086738697e-06, "loss": 1.545, "step": 3076 }, { "epoch": 0.19, "learning_rate": 9.2684251825919e-06, "loss": 1.7296, "step": 3078 }, { "epoch": 0.19, "learning_rate": 9.274461278445103e-06, "loss": 1.8176, "step": 3080 }, { "epoch": 0.19, "learning_rate": 9.280497374298303e-06, "loss": 1.6266, "step": 3082 }, { "epoch": 0.19, "learning_rate": 9.286533470151506e-06, "loss": 1.2769, "step": 3084 }, { "epoch": 0.19, "learning_rate": 9.292569566004708e-06, "loss": 1.5123, "step": 3086 }, { "epoch": 0.19, "learning_rate": 9.298605661857911e-06, "loss": 1.1211, "step": 3088 }, { "epoch": 0.19, "learning_rate": 9.304641757711113e-06, "loss": 1.4951, "step": 3090 }, { "epoch": 0.19, "learning_rate": 9.310677853564314e-06, "loss": 1.2678, "step": 3092 }, { "epoch": 0.19, "learning_rate": 9.316713949417517e-06, "loss": 1.5341, "step": 3094 }, { "epoch": 0.19, "learning_rate": 9.32275004527072e-06, "loss": 1.1585, "step": 3096 }, { "epoch": 0.19, "learning_rate": 9.328786141123922e-06, "loss": 1.7906, "step": 3098 }, { "epoch": 0.19, "learning_rate": 9.334822236977124e-06, "loss": 1.1679, "step": 3100 }, { "epoch": 0.19, "learning_rate": 9.340858332830325e-06, "loss": 1.6782, "step": 3102 }, { "epoch": 0.19, "learning_rate": 9.346894428683528e-06, "loss": 1.521, "step": 3104 }, { "epoch": 0.19, "learning_rate": 9.35293052453673e-06, "loss": 1.5568, "step": 3106 }, { "epoch": 0.19, "learning_rate": 9.358966620389933e-06, "loss": 1.7986, "step": 3108 }, { "epoch": 0.19, "learning_rate": 9.365002716243135e-06, "loss": 1.3524, "step": 3110 }, { "epoch": 0.19, "learning_rate": 9.371038812096336e-06, "loss": 1.5547, "step": 3112 }, { "epoch": 0.19, "learning_rate": 9.377074907949538e-06, "loss": 1.5514, "step": 3114 }, { "epoch": 0.19, "learning_rate": 9.383111003802741e-06, "loss": 1.1178, "step": 3116 }, { "epoch": 0.19, "learning_rate": 9.389147099655943e-06, "loss": 1.1397, "step": 3118 }, { "epoch": 0.19, "learning_rate": 9.395183195509146e-06, "loss": 1.2046, "step": 3120 }, { "epoch": 0.19, "learning_rate": 9.401219291362347e-06, "loss": 1.2033, "step": 3122 }, { "epoch": 0.19, "learning_rate": 9.40725538721555e-06, "loss": 1.6622, "step": 3124 }, { "epoch": 0.19, "learning_rate": 9.413291483068752e-06, "loss": 1.3484, "step": 3126 }, { "epoch": 0.19, "learning_rate": 9.419327578921953e-06, "loss": 1.6449, "step": 3128 }, { "epoch": 0.19, "learning_rate": 9.425363674775157e-06, "loss": 1.4201, "step": 3130 }, { "epoch": 0.19, "learning_rate": 9.431399770628358e-06, "loss": 1.436, "step": 3132 }, { "epoch": 0.19, "learning_rate": 9.43743586648156e-06, "loss": 1.5881, "step": 3134 }, { "epoch": 0.19, "learning_rate": 9.443471962334763e-06, "loss": 1.7289, "step": 3136 }, { "epoch": 0.19, "learning_rate": 9.449508058187964e-06, "loss": 1.3782, "step": 3138 }, { "epoch": 0.19, "learning_rate": 9.455544154041166e-06, "loss": 1.4471, "step": 3140 }, { "epoch": 0.19, "learning_rate": 9.461580249894369e-06, "loss": 1.3725, "step": 3142 }, { "epoch": 0.19, "learning_rate": 9.467616345747571e-06, "loss": 1.026, "step": 3144 }, { "epoch": 0.19, "learning_rate": 9.473652441600774e-06, "loss": 1.7023, "step": 3146 }, { "epoch": 0.19, "learning_rate": 9.479688537453974e-06, "loss": 1.3949, "step": 3148 }, { "epoch": 0.19, "learning_rate": 9.485724633307177e-06, "loss": 1.5402, "step": 3150 }, { "epoch": 0.19, "learning_rate": 9.49176072916038e-06, "loss": 1.1498, "step": 3152 }, { "epoch": 0.19, "learning_rate": 9.497796825013582e-06, "loss": 1.1913, "step": 3154 }, { "epoch": 0.19, "learning_rate": 9.503832920866784e-06, "loss": 1.8346, "step": 3156 }, { "epoch": 0.19, "learning_rate": 9.509869016719985e-06, "loss": 1.5995, "step": 3158 }, { "epoch": 0.19, "learning_rate": 9.515905112573188e-06, "loss": 1.2549, "step": 3160 }, { "epoch": 0.19, "learning_rate": 9.52194120842639e-06, "loss": 1.3578, "step": 3162 }, { "epoch": 0.19, "learning_rate": 9.527977304279593e-06, "loss": 1.3838, "step": 3164 }, { "epoch": 0.19, "learning_rate": 9.534013400132795e-06, "loss": 1.4354, "step": 3166 }, { "epoch": 0.19, "learning_rate": 9.540049495985996e-06, "loss": 1.2952, "step": 3168 }, { "epoch": 0.19, "learning_rate": 9.546085591839199e-06, "loss": 1.6494, "step": 3170 }, { "epoch": 0.19, "learning_rate": 9.552121687692401e-06, "loss": 1.6803, "step": 3172 }, { "epoch": 0.19, "learning_rate": 9.558157783545604e-06, "loss": 1.695, "step": 3174 }, { "epoch": 0.19, "learning_rate": 9.564193879398806e-06, "loss": 1.3998, "step": 3176 }, { "epoch": 0.19, "learning_rate": 9.570229975252007e-06, "loss": 1.2496, "step": 3178 }, { "epoch": 0.19, "learning_rate": 9.57626607110521e-06, "loss": 1.6163, "step": 3180 }, { "epoch": 0.19, "learning_rate": 9.582302166958412e-06, "loss": 1.8084, "step": 3182 }, { "epoch": 0.19, "learning_rate": 9.588338262811613e-06, "loss": 1.3763, "step": 3184 }, { "epoch": 0.19, "learning_rate": 9.594374358664817e-06, "loss": 1.9224, "step": 3186 }, { "epoch": 0.19, "learning_rate": 9.600410454518018e-06, "loss": 1.049, "step": 3188 }, { "epoch": 0.19, "learning_rate": 9.60644655037122e-06, "loss": 1.5791, "step": 3190 }, { "epoch": 0.19, "learning_rate": 9.612482646224423e-06, "loss": 1.3012, "step": 3192 }, { "epoch": 0.19, "learning_rate": 9.618518742077624e-06, "loss": 1.4554, "step": 3194 }, { "epoch": 0.19, "learning_rate": 9.624554837930826e-06, "loss": 1.7548, "step": 3196 }, { "epoch": 0.19, "learning_rate": 9.630590933784029e-06, "loss": 1.5833, "step": 3198 }, { "epoch": 0.19, "learning_rate": 9.636627029637231e-06, "loss": 2.1773, "step": 3200 }, { "epoch": 0.19, "learning_rate": 9.642663125490434e-06, "loss": 1.4597, "step": 3202 }, { "epoch": 0.19, "learning_rate": 9.648699221343635e-06, "loss": 1.5556, "step": 3204 }, { "epoch": 0.19, "learning_rate": 9.654735317196837e-06, "loss": 1.7984, "step": 3206 }, { "epoch": 0.19, "learning_rate": 9.66077141305004e-06, "loss": 1.5751, "step": 3208 }, { "epoch": 0.19, "learning_rate": 9.666807508903242e-06, "loss": 1.677, "step": 3210 }, { "epoch": 0.19, "learning_rate": 9.672843604756445e-06, "loss": 1.6015, "step": 3212 }, { "epoch": 0.19, "learning_rate": 9.678879700609645e-06, "loss": 1.4835, "step": 3214 }, { "epoch": 0.19, "learning_rate": 9.684915796462848e-06, "loss": 1.7808, "step": 3216 }, { "epoch": 0.19, "learning_rate": 9.69095189231605e-06, "loss": 1.4541, "step": 3218 }, { "epoch": 0.19, "learning_rate": 9.696987988169253e-06, "loss": 1.4249, "step": 3220 }, { "epoch": 0.19, "learning_rate": 9.703024084022455e-06, "loss": 1.4125, "step": 3222 }, { "epoch": 0.19, "learning_rate": 9.709060179875656e-06, "loss": 1.5911, "step": 3224 }, { "epoch": 0.19, "learning_rate": 9.715096275728859e-06, "loss": 1.6858, "step": 3226 }, { "epoch": 0.19, "learning_rate": 9.721132371582061e-06, "loss": 1.2661, "step": 3228 }, { "epoch": 0.19, "learning_rate": 9.727168467435264e-06, "loss": 1.6645, "step": 3230 }, { "epoch": 0.2, "learning_rate": 9.733204563288466e-06, "loss": 1.4296, "step": 3232 }, { "epoch": 0.2, "learning_rate": 9.739240659141667e-06, "loss": 1.3864, "step": 3234 }, { "epoch": 0.2, "learning_rate": 9.74527675499487e-06, "loss": 1.3798, "step": 3236 }, { "epoch": 0.2, "learning_rate": 9.751312850848072e-06, "loss": 1.5576, "step": 3238 }, { "epoch": 0.2, "learning_rate": 9.757348946701273e-06, "loss": 1.7857, "step": 3240 }, { "epoch": 0.2, "learning_rate": 9.763385042554477e-06, "loss": 1.5518, "step": 3242 }, { "epoch": 0.2, "learning_rate": 9.769421138407678e-06, "loss": 1.9504, "step": 3244 }, { "epoch": 0.2, "learning_rate": 9.77545723426088e-06, "loss": 1.44, "step": 3246 }, { "epoch": 0.2, "learning_rate": 9.781493330114083e-06, "loss": 1.5064, "step": 3248 }, { "epoch": 0.2, "learning_rate": 9.787529425967284e-06, "loss": 1.5812, "step": 3250 }, { "epoch": 0.2, "learning_rate": 9.793565521820486e-06, "loss": 1.3732, "step": 3252 }, { "epoch": 0.2, "learning_rate": 9.799601617673689e-06, "loss": 1.7358, "step": 3254 }, { "epoch": 0.2, "learning_rate": 9.805637713526891e-06, "loss": 1.7698, "step": 3256 }, { "epoch": 0.2, "learning_rate": 9.811673809380094e-06, "loss": 1.0198, "step": 3258 }, { "epoch": 0.2, "learning_rate": 9.817709905233295e-06, "loss": 1.6883, "step": 3260 }, { "epoch": 0.2, "learning_rate": 9.823746001086497e-06, "loss": 1.7338, "step": 3262 }, { "epoch": 0.2, "learning_rate": 9.8297820969397e-06, "loss": 1.3321, "step": 3264 }, { "epoch": 0.2, "learning_rate": 9.835818192792902e-06, "loss": 1.3664, "step": 3266 }, { "epoch": 0.2, "learning_rate": 9.841854288646105e-06, "loss": 1.2913, "step": 3268 }, { "epoch": 0.2, "learning_rate": 9.847890384499306e-06, "loss": 1.3024, "step": 3270 }, { "epoch": 0.2, "learning_rate": 9.853926480352508e-06, "loss": 1.3893, "step": 3272 }, { "epoch": 0.2, "learning_rate": 9.85996257620571e-06, "loss": 1.2955, "step": 3274 }, { "epoch": 0.2, "learning_rate": 9.865998672058913e-06, "loss": 1.6909, "step": 3276 }, { "epoch": 0.2, "learning_rate": 9.872034767912116e-06, "loss": 1.513, "step": 3278 }, { "epoch": 0.2, "learning_rate": 9.878070863765316e-06, "loss": 1.3313, "step": 3280 }, { "epoch": 0.2, "learning_rate": 9.884106959618519e-06, "loss": 1.596, "step": 3282 }, { "epoch": 0.2, "learning_rate": 9.890143055471722e-06, "loss": 1.1853, "step": 3284 }, { "epoch": 0.2, "learning_rate": 9.896179151324924e-06, "loss": 1.1267, "step": 3286 }, { "epoch": 0.2, "learning_rate": 9.902215247178127e-06, "loss": 1.1312, "step": 3288 }, { "epoch": 0.2, "learning_rate": 9.908251343031327e-06, "loss": 1.5925, "step": 3290 }, { "epoch": 0.2, "learning_rate": 9.91428743888453e-06, "loss": 1.6285, "step": 3292 }, { "epoch": 0.2, "learning_rate": 9.920323534737732e-06, "loss": 1.4374, "step": 3294 }, { "epoch": 0.2, "learning_rate": 9.926359630590933e-06, "loss": 1.5693, "step": 3296 }, { "epoch": 0.2, "learning_rate": 9.932395726444137e-06, "loss": 1.5605, "step": 3298 }, { "epoch": 0.2, "learning_rate": 9.938431822297338e-06, "loss": 1.5053, "step": 3300 }, { "epoch": 0.2, "learning_rate": 9.94446791815054e-06, "loss": 1.3699, "step": 3302 }, { "epoch": 0.2, "learning_rate": 9.950504014003743e-06, "loss": 1.5552, "step": 3304 }, { "epoch": 0.2, "learning_rate": 9.956540109856944e-06, "loss": 1.5472, "step": 3306 }, { "epoch": 0.2, "learning_rate": 9.962576205710147e-06, "loss": 1.6222, "step": 3308 }, { "epoch": 0.2, "learning_rate": 9.968612301563349e-06, "loss": 1.333, "step": 3310 }, { "epoch": 0.2, "learning_rate": 9.974648397416552e-06, "loss": 1.4137, "step": 3312 }, { "epoch": 0.2, "learning_rate": 9.980684493269754e-06, "loss": 1.4385, "step": 3314 }, { "epoch": 0.2, "learning_rate": 9.986720589122955e-06, "loss": 1.7265, "step": 3316 }, { "epoch": 0.2, "learning_rate": 9.992756684976157e-06, "loss": 1.3434, "step": 3318 }, { "epoch": 0.2, "learning_rate": 9.99879278082936e-06, "loss": 1.5653, "step": 3320 }, { "epoch": 0.2, "learning_rate": 1.0004828876682562e-05, "loss": 1.5378, "step": 3322 }, { "epoch": 0.2, "learning_rate": 1.0010864972535765e-05, "loss": 1.4273, "step": 3324 }, { "epoch": 0.2, "learning_rate": 1.0016901068388966e-05, "loss": 1.3665, "step": 3326 }, { "epoch": 0.2, "learning_rate": 1.0022937164242168e-05, "loss": 1.8401, "step": 3328 }, { "epoch": 0.2, "learning_rate": 1.002897326009537e-05, "loss": 1.4119, "step": 3330 }, { "epoch": 0.2, "learning_rate": 1.0035009355948573e-05, "loss": 1.3018, "step": 3332 }, { "epoch": 0.2, "learning_rate": 1.0041045451801776e-05, "loss": 1.5171, "step": 3334 }, { "epoch": 0.2, "learning_rate": 1.0047081547654977e-05, "loss": 1.4706, "step": 3336 }, { "epoch": 0.2, "learning_rate": 1.005311764350818e-05, "loss": 1.3719, "step": 3338 }, { "epoch": 0.2, "learning_rate": 1.0059153739361382e-05, "loss": 1.346, "step": 3340 }, { "epoch": 0.2, "learning_rate": 1.0065189835214584e-05, "loss": 1.5064, "step": 3342 }, { "epoch": 0.2, "learning_rate": 1.0071225931067787e-05, "loss": 1.268, "step": 3344 }, { "epoch": 0.2, "learning_rate": 1.0077262026920988e-05, "loss": 1.5651, "step": 3346 }, { "epoch": 0.2, "learning_rate": 1.008329812277419e-05, "loss": 1.4869, "step": 3348 }, { "epoch": 0.2, "learning_rate": 1.0089334218627393e-05, "loss": 1.3865, "step": 3350 }, { "epoch": 0.2, "learning_rate": 1.0095370314480593e-05, "loss": 1.4945, "step": 3352 }, { "epoch": 0.2, "learning_rate": 1.0101406410333798e-05, "loss": 1.7275, "step": 3354 }, { "epoch": 0.2, "learning_rate": 1.0107442506186998e-05, "loss": 1.5412, "step": 3356 }, { "epoch": 0.2, "learning_rate": 1.0113478602040201e-05, "loss": 1.2441, "step": 3358 }, { "epoch": 0.2, "learning_rate": 1.0119514697893403e-05, "loss": 1.1822, "step": 3360 }, { "epoch": 0.2, "learning_rate": 1.0125550793746604e-05, "loss": 1.2256, "step": 3362 }, { "epoch": 0.2, "learning_rate": 1.0131586889599807e-05, "loss": 1.1388, "step": 3364 }, { "epoch": 0.2, "learning_rate": 1.013762298545301e-05, "loss": 1.607, "step": 3366 }, { "epoch": 0.2, "learning_rate": 1.0143659081306212e-05, "loss": 1.6103, "step": 3368 }, { "epoch": 0.2, "learning_rate": 1.0149695177159414e-05, "loss": 1.4647, "step": 3370 }, { "epoch": 0.2, "learning_rate": 1.0155731273012615e-05, "loss": 1.2327, "step": 3372 }, { "epoch": 0.2, "learning_rate": 1.0161767368865818e-05, "loss": 1.7236, "step": 3374 }, { "epoch": 0.2, "learning_rate": 1.016780346471902e-05, "loss": 1.5809, "step": 3376 }, { "epoch": 0.2, "learning_rate": 1.0173839560572223e-05, "loss": 1.4647, "step": 3378 }, { "epoch": 0.2, "learning_rate": 1.0179875656425425e-05, "loss": 1.1351, "step": 3380 }, { "epoch": 0.2, "learning_rate": 1.0185911752278626e-05, "loss": 1.6043, "step": 3382 }, { "epoch": 0.2, "learning_rate": 1.0191947848131829e-05, "loss": 1.3203, "step": 3384 }, { "epoch": 0.2, "learning_rate": 1.0197983943985031e-05, "loss": 1.3516, "step": 3386 }, { "epoch": 0.2, "learning_rate": 1.0204020039838234e-05, "loss": 1.141, "step": 3388 }, { "epoch": 0.2, "learning_rate": 1.0210056135691436e-05, "loss": 1.3526, "step": 3390 }, { "epoch": 0.2, "learning_rate": 1.0216092231544637e-05, "loss": 1.4176, "step": 3392 }, { "epoch": 0.2, "learning_rate": 1.022212832739784e-05, "loss": 1.3581, "step": 3394 }, { "epoch": 0.2, "learning_rate": 1.0228164423251042e-05, "loss": 1.2249, "step": 3396 }, { "epoch": 0.21, "learning_rate": 1.0234200519104244e-05, "loss": 1.6679, "step": 3398 }, { "epoch": 0.21, "learning_rate": 1.0240236614957447e-05, "loss": 1.2752, "step": 3400 }, { "epoch": 0.21, "learning_rate": 1.0246272710810648e-05, "loss": 1.3419, "step": 3402 }, { "epoch": 0.21, "learning_rate": 1.025230880666385e-05, "loss": 1.6507, "step": 3404 }, { "epoch": 0.21, "learning_rate": 1.0258344902517053e-05, "loss": 1.3887, "step": 3406 }, { "epoch": 0.21, "learning_rate": 1.0264380998370254e-05, "loss": 1.3788, "step": 3408 }, { "epoch": 0.21, "learning_rate": 1.0270417094223458e-05, "loss": 1.3314, "step": 3410 }, { "epoch": 0.21, "learning_rate": 1.0276453190076659e-05, "loss": 1.6318, "step": 3412 }, { "epoch": 0.21, "learning_rate": 1.0282489285929861e-05, "loss": 1.5529, "step": 3414 }, { "epoch": 0.21, "learning_rate": 1.0288525381783064e-05, "loss": 1.2622, "step": 3416 }, { "epoch": 0.21, "learning_rate": 1.0294561477636264e-05, "loss": 1.3543, "step": 3418 }, { "epoch": 0.21, "learning_rate": 1.0300597573489467e-05, "loss": 1.5919, "step": 3420 }, { "epoch": 0.21, "learning_rate": 1.030663366934267e-05, "loss": 1.2115, "step": 3422 }, { "epoch": 0.21, "learning_rate": 1.0312669765195872e-05, "loss": 1.5361, "step": 3424 }, { "epoch": 0.21, "learning_rate": 1.0318705861049075e-05, "loss": 1.3774, "step": 3426 }, { "epoch": 0.21, "learning_rate": 1.0324741956902275e-05, "loss": 1.5925, "step": 3428 }, { "epoch": 0.21, "learning_rate": 1.0330778052755478e-05, "loss": 1.3254, "step": 3430 }, { "epoch": 0.21, "learning_rate": 1.033681414860868e-05, "loss": 1.2797, "step": 3432 }, { "epoch": 0.21, "learning_rate": 1.0342850244461883e-05, "loss": 1.2768, "step": 3434 }, { "epoch": 0.21, "learning_rate": 1.0348886340315085e-05, "loss": 1.3549, "step": 3436 }, { "epoch": 0.21, "learning_rate": 1.0354922436168286e-05, "loss": 1.4673, "step": 3438 }, { "epoch": 0.21, "learning_rate": 1.0360958532021489e-05, "loss": 1.396, "step": 3440 }, { "epoch": 0.21, "learning_rate": 1.0366994627874691e-05, "loss": 1.5254, "step": 3442 }, { "epoch": 0.21, "learning_rate": 1.0373030723727894e-05, "loss": 1.3642, "step": 3444 }, { "epoch": 0.21, "learning_rate": 1.0379066819581096e-05, "loss": 1.3987, "step": 3446 }, { "epoch": 0.21, "learning_rate": 1.0385102915434297e-05, "loss": 1.4337, "step": 3448 }, { "epoch": 0.21, "learning_rate": 1.03911390112875e-05, "loss": 1.2472, "step": 3450 }, { "epoch": 0.21, "learning_rate": 1.0397175107140702e-05, "loss": 1.3069, "step": 3452 }, { "epoch": 0.21, "learning_rate": 1.0403211202993905e-05, "loss": 1.2846, "step": 3454 }, { "epoch": 0.21, "learning_rate": 1.0409247298847107e-05, "loss": 1.5974, "step": 3456 }, { "epoch": 0.21, "learning_rate": 1.0415283394700308e-05, "loss": 1.5804, "step": 3458 }, { "epoch": 0.21, "learning_rate": 1.042131949055351e-05, "loss": 1.3579, "step": 3460 }, { "epoch": 0.21, "learning_rate": 1.0427355586406713e-05, "loss": 1.5209, "step": 3462 }, { "epoch": 0.21, "learning_rate": 1.0433391682259914e-05, "loss": 1.8259, "step": 3464 }, { "epoch": 0.21, "learning_rate": 1.0439427778113118e-05, "loss": 1.3939, "step": 3466 }, { "epoch": 0.21, "learning_rate": 1.0445463873966319e-05, "loss": 1.2708, "step": 3468 }, { "epoch": 0.21, "learning_rate": 1.0451499969819521e-05, "loss": 1.2402, "step": 3470 }, { "epoch": 0.21, "learning_rate": 1.0457536065672724e-05, "loss": 1.2018, "step": 3472 }, { "epoch": 0.21, "learning_rate": 1.0463572161525925e-05, "loss": 1.3292, "step": 3474 }, { "epoch": 0.21, "learning_rate": 1.0469608257379127e-05, "loss": 1.5722, "step": 3476 }, { "epoch": 0.21, "learning_rate": 1.047564435323233e-05, "loss": 1.3382, "step": 3478 }, { "epoch": 0.21, "learning_rate": 1.0481680449085532e-05, "loss": 1.3438, "step": 3480 }, { "epoch": 0.21, "learning_rate": 1.0487716544938735e-05, "loss": 1.5036, "step": 3482 }, { "epoch": 0.21, "learning_rate": 1.0493752640791936e-05, "loss": 1.8005, "step": 3484 }, { "epoch": 0.21, "learning_rate": 1.0499788736645138e-05, "loss": 1.5671, "step": 3486 }, { "epoch": 0.21, "learning_rate": 1.050582483249834e-05, "loss": 1.326, "step": 3488 }, { "epoch": 0.21, "learning_rate": 1.0511860928351543e-05, "loss": 1.2198, "step": 3490 }, { "epoch": 0.21, "learning_rate": 1.0517897024204746e-05, "loss": 1.4691, "step": 3492 }, { "epoch": 0.21, "learning_rate": 1.0523933120057946e-05, "loss": 1.7841, "step": 3494 }, { "epoch": 0.21, "learning_rate": 1.0529969215911149e-05, "loss": 1.7712, "step": 3496 }, { "epoch": 0.21, "learning_rate": 1.0536005311764351e-05, "loss": 1.6565, "step": 3498 }, { "epoch": 0.21, "learning_rate": 1.0542041407617554e-05, "loss": 1.6443, "step": 3500 }, { "epoch": 0.21, "learning_rate": 1.0548077503470756e-05, "loss": 1.6913, "step": 3502 }, { "epoch": 0.21, "learning_rate": 1.0554113599323957e-05, "loss": 1.3733, "step": 3504 }, { "epoch": 0.21, "learning_rate": 1.056014969517716e-05, "loss": 1.6177, "step": 3506 }, { "epoch": 0.21, "learning_rate": 1.0566185791030362e-05, "loss": 1.6161, "step": 3508 }, { "epoch": 0.21, "learning_rate": 1.0572221886883565e-05, "loss": 1.4739, "step": 3510 }, { "epoch": 0.21, "learning_rate": 1.0578257982736767e-05, "loss": 1.2181, "step": 3512 }, { "epoch": 0.21, "learning_rate": 1.0584294078589968e-05, "loss": 1.447, "step": 3514 }, { "epoch": 0.21, "learning_rate": 1.059033017444317e-05, "loss": 1.3488, "step": 3516 }, { "epoch": 0.21, "learning_rate": 1.0596366270296373e-05, "loss": 1.3168, "step": 3518 }, { "epoch": 0.21, "learning_rate": 1.0602402366149574e-05, "loss": 1.2423, "step": 3520 }, { "epoch": 0.21, "learning_rate": 1.0608438462002778e-05, "loss": 1.2222, "step": 3522 }, { "epoch": 0.21, "learning_rate": 1.0614474557855979e-05, "loss": 1.658, "step": 3524 }, { "epoch": 0.21, "learning_rate": 1.0620510653709182e-05, "loss": 1.2872, "step": 3526 }, { "epoch": 0.21, "learning_rate": 1.0626546749562384e-05, "loss": 1.7557, "step": 3528 }, { "epoch": 0.21, "learning_rate": 1.0632582845415585e-05, "loss": 1.2873, "step": 3530 }, { "epoch": 0.21, "learning_rate": 1.0638618941268787e-05, "loss": 1.6187, "step": 3532 }, { "epoch": 0.21, "learning_rate": 1.064465503712199e-05, "loss": 1.6299, "step": 3534 }, { "epoch": 0.21, "learning_rate": 1.0650691132975192e-05, "loss": 1.4494, "step": 3536 }, { "epoch": 0.21, "learning_rate": 1.0656727228828395e-05, "loss": 1.2529, "step": 3538 }, { "epoch": 0.21, "learning_rate": 1.0662763324681596e-05, "loss": 1.6176, "step": 3540 }, { "epoch": 0.21, "learning_rate": 1.0668799420534798e-05, "loss": 1.5702, "step": 3542 }, { "epoch": 0.21, "learning_rate": 1.0674835516388e-05, "loss": 1.2649, "step": 3544 }, { "epoch": 0.21, "learning_rate": 1.0680871612241203e-05, "loss": 1.6941, "step": 3546 }, { "epoch": 0.21, "learning_rate": 1.0686907708094406e-05, "loss": 1.3465, "step": 3548 }, { "epoch": 0.21, "learning_rate": 1.0692943803947607e-05, "loss": 1.3473, "step": 3550 }, { "epoch": 0.21, "learning_rate": 1.0698979899800809e-05, "loss": 1.4092, "step": 3552 }, { "epoch": 0.21, "learning_rate": 1.0705015995654012e-05, "loss": 1.4253, "step": 3554 }, { "epoch": 0.21, "learning_rate": 1.0711052091507214e-05, "loss": 1.0597, "step": 3556 }, { "epoch": 0.21, "learning_rate": 1.0717088187360417e-05, "loss": 1.7525, "step": 3558 }, { "epoch": 0.21, "learning_rate": 1.0723124283213617e-05, "loss": 1.4238, "step": 3560 }, { "epoch": 0.21, "learning_rate": 1.072916037906682e-05, "loss": 1.6262, "step": 3562 }, { "epoch": 0.22, "learning_rate": 1.0735196474920022e-05, "loss": 1.9463, "step": 3564 }, { "epoch": 0.22, "learning_rate": 1.0741232570773225e-05, "loss": 1.5687, "step": 3566 }, { "epoch": 0.22, "learning_rate": 1.0747268666626428e-05, "loss": 1.4492, "step": 3568 }, { "epoch": 0.22, "learning_rate": 1.0753304762479628e-05, "loss": 1.457, "step": 3570 }, { "epoch": 0.22, "learning_rate": 1.075934085833283e-05, "loss": 1.4863, "step": 3572 }, { "epoch": 0.22, "learning_rate": 1.0765376954186033e-05, "loss": 1.5374, "step": 3574 }, { "epoch": 0.22, "learning_rate": 1.0771413050039234e-05, "loss": 1.5384, "step": 3576 }, { "epoch": 0.22, "learning_rate": 1.0777449145892438e-05, "loss": 1.3573, "step": 3578 }, { "epoch": 0.22, "learning_rate": 1.078348524174564e-05, "loss": 1.1611, "step": 3580 }, { "epoch": 0.22, "learning_rate": 1.0789521337598842e-05, "loss": 1.3887, "step": 3582 }, { "epoch": 0.22, "learning_rate": 1.0795557433452044e-05, "loss": 1.6862, "step": 3584 }, { "epoch": 0.22, "learning_rate": 1.0801593529305245e-05, "loss": 1.4332, "step": 3586 }, { "epoch": 0.22, "learning_rate": 1.0807629625158448e-05, "loss": 1.4504, "step": 3588 }, { "epoch": 0.22, "learning_rate": 1.081366572101165e-05, "loss": 1.5727, "step": 3590 }, { "epoch": 0.22, "learning_rate": 1.0819701816864853e-05, "loss": 1.1774, "step": 3592 }, { "epoch": 0.22, "learning_rate": 1.0825737912718055e-05, "loss": 1.1552, "step": 3594 }, { "epoch": 0.22, "learning_rate": 1.0831774008571256e-05, "loss": 1.0652, "step": 3596 }, { "epoch": 0.22, "learning_rate": 1.0837810104424458e-05, "loss": 1.4079, "step": 3598 }, { "epoch": 0.22, "learning_rate": 1.0843846200277661e-05, "loss": 1.2733, "step": 3600 }, { "epoch": 0.22, "learning_rate": 1.0849882296130863e-05, "loss": 1.2903, "step": 3602 }, { "epoch": 0.22, "learning_rate": 1.0855918391984066e-05, "loss": 1.2197, "step": 3604 }, { "epoch": 0.22, "learning_rate": 1.0861954487837267e-05, "loss": 1.8555, "step": 3606 }, { "epoch": 0.22, "learning_rate": 1.086799058369047e-05, "loss": 1.296, "step": 3608 }, { "epoch": 0.22, "learning_rate": 1.0874026679543672e-05, "loss": 1.7469, "step": 3610 }, { "epoch": 0.22, "learning_rate": 1.0880062775396874e-05, "loss": 1.3279, "step": 3612 }, { "epoch": 0.22, "learning_rate": 1.0886098871250077e-05, "loss": 1.3543, "step": 3614 }, { "epoch": 0.22, "learning_rate": 1.0892134967103278e-05, "loss": 1.519, "step": 3616 }, { "epoch": 0.22, "learning_rate": 1.089817106295648e-05, "loss": 1.0618, "step": 3618 }, { "epoch": 0.22, "learning_rate": 1.0904207158809683e-05, "loss": 1.6561, "step": 3620 }, { "epoch": 0.22, "learning_rate": 1.0910243254662885e-05, "loss": 1.3497, "step": 3622 }, { "epoch": 0.22, "learning_rate": 1.0916279350516088e-05, "loss": 1.4718, "step": 3624 }, { "epoch": 0.22, "learning_rate": 1.0922315446369289e-05, "loss": 1.7229, "step": 3626 }, { "epoch": 0.22, "learning_rate": 1.0928351542222491e-05, "loss": 2.0753, "step": 3628 }, { "epoch": 0.22, "learning_rate": 1.0934387638075694e-05, "loss": 1.7088, "step": 3630 }, { "epoch": 0.22, "learning_rate": 1.0940423733928894e-05, "loss": 1.2229, "step": 3632 }, { "epoch": 0.22, "learning_rate": 1.0946459829782099e-05, "loss": 1.6336, "step": 3634 }, { "epoch": 0.22, "learning_rate": 1.09524959256353e-05, "loss": 1.1315, "step": 3636 }, { "epoch": 0.22, "learning_rate": 1.0958532021488502e-05, "loss": 1.4247, "step": 3638 }, { "epoch": 0.22, "learning_rate": 1.0964568117341704e-05, "loss": 1.8156, "step": 3640 }, { "epoch": 0.22, "learning_rate": 1.0970604213194905e-05, "loss": 1.3483, "step": 3642 }, { "epoch": 0.22, "learning_rate": 1.0976640309048108e-05, "loss": 1.3399, "step": 3644 }, { "epoch": 0.22, "learning_rate": 1.098267640490131e-05, "loss": 1.6202, "step": 3646 }, { "epoch": 0.22, "learning_rate": 1.0988712500754513e-05, "loss": 1.5257, "step": 3648 }, { "epoch": 0.22, "learning_rate": 1.0994748596607715e-05, "loss": 1.3087, "step": 3650 }, { "epoch": 0.22, "learning_rate": 1.1000784692460916e-05, "loss": 1.4393, "step": 3652 }, { "epoch": 0.22, "learning_rate": 1.1006820788314119e-05, "loss": 1.4949, "step": 3654 }, { "epoch": 0.22, "learning_rate": 1.1012856884167321e-05, "loss": 1.5612, "step": 3656 }, { "epoch": 0.22, "learning_rate": 1.1018892980020524e-05, "loss": 1.4255, "step": 3658 }, { "epoch": 0.22, "learning_rate": 1.1024929075873726e-05, "loss": 1.2572, "step": 3660 }, { "epoch": 0.22, "learning_rate": 1.1030965171726927e-05, "loss": 1.2974, "step": 3662 }, { "epoch": 0.22, "learning_rate": 1.103700126758013e-05, "loss": 1.5043, "step": 3664 }, { "epoch": 0.22, "learning_rate": 1.1043037363433332e-05, "loss": 1.5608, "step": 3666 }, { "epoch": 0.22, "learning_rate": 1.1049073459286534e-05, "loss": 1.5783, "step": 3668 }, { "epoch": 0.22, "learning_rate": 1.1055109555139737e-05, "loss": 1.5477, "step": 3670 }, { "epoch": 0.22, "learning_rate": 1.1061145650992938e-05, "loss": 1.3035, "step": 3672 }, { "epoch": 0.22, "learning_rate": 1.106718174684614e-05, "loss": 1.1164, "step": 3674 }, { "epoch": 0.22, "learning_rate": 1.1073217842699343e-05, "loss": 1.2615, "step": 3676 }, { "epoch": 0.22, "learning_rate": 1.1079253938552545e-05, "loss": 1.1794, "step": 3678 }, { "epoch": 0.22, "learning_rate": 1.1085290034405748e-05, "loss": 1.2905, "step": 3680 }, { "epoch": 0.22, "learning_rate": 1.1091326130258949e-05, "loss": 1.1455, "step": 3682 }, { "epoch": 0.22, "learning_rate": 1.1097362226112151e-05, "loss": 1.2939, "step": 3684 }, { "epoch": 0.22, "learning_rate": 1.1103398321965354e-05, "loss": 1.2886, "step": 3686 }, { "epoch": 0.22, "learning_rate": 1.1109434417818555e-05, "loss": 1.139, "step": 3688 }, { "epoch": 0.22, "learning_rate": 1.1115470513671759e-05, "loss": 1.5698, "step": 3690 }, { "epoch": 0.22, "learning_rate": 1.112150660952496e-05, "loss": 1.4063, "step": 3692 }, { "epoch": 0.22, "learning_rate": 1.1127542705378162e-05, "loss": 1.0861, "step": 3694 }, { "epoch": 0.22, "learning_rate": 1.1133578801231365e-05, "loss": 1.9002, "step": 3696 }, { "epoch": 0.22, "learning_rate": 1.1139614897084565e-05, "loss": 1.6905, "step": 3698 }, { "epoch": 0.22, "learning_rate": 1.1145650992937768e-05, "loss": 1.4821, "step": 3700 }, { "epoch": 0.22, "learning_rate": 1.115168708879097e-05, "loss": 1.3427, "step": 3702 }, { "epoch": 0.22, "learning_rate": 1.1157723184644173e-05, "loss": 1.4296, "step": 3704 }, { "epoch": 0.22, "learning_rate": 1.1163759280497375e-05, "loss": 1.3766, "step": 3706 }, { "epoch": 0.22, "learning_rate": 1.1169795376350576e-05, "loss": 1.2458, "step": 3708 }, { "epoch": 0.22, "learning_rate": 1.1175831472203779e-05, "loss": 1.3619, "step": 3710 }, { "epoch": 0.22, "learning_rate": 1.1181867568056981e-05, "loss": 1.2367, "step": 3712 }, { "epoch": 0.22, "learning_rate": 1.1187903663910184e-05, "loss": 1.4983, "step": 3714 }, { "epoch": 0.22, "learning_rate": 1.1193939759763386e-05, "loss": 1.1185, "step": 3716 }, { "epoch": 0.22, "learning_rate": 1.1199975855616587e-05, "loss": 1.6442, "step": 3718 }, { "epoch": 0.22, "learning_rate": 1.120601195146979e-05, "loss": 1.802, "step": 3720 }, { "epoch": 0.22, "learning_rate": 1.1212048047322992e-05, "loss": 1.4835, "step": 3722 }, { "epoch": 0.22, "learning_rate": 1.1218084143176195e-05, "loss": 1.3382, "step": 3724 }, { "epoch": 0.22, "learning_rate": 1.1224120239029397e-05, "loss": 1.5729, "step": 3726 }, { "epoch": 0.23, "learning_rate": 1.1230156334882598e-05, "loss": 1.3504, "step": 3728 }, { "epoch": 0.23, "learning_rate": 1.12361924307358e-05, "loss": 1.4869, "step": 3730 }, { "epoch": 0.23, "learning_rate": 1.1242228526589003e-05, "loss": 1.5771, "step": 3732 }, { "epoch": 0.23, "learning_rate": 1.1248264622442206e-05, "loss": 1.8543, "step": 3734 }, { "epoch": 0.23, "learning_rate": 1.1254300718295408e-05, "loss": 1.488, "step": 3736 }, { "epoch": 0.23, "learning_rate": 1.1260336814148609e-05, "loss": 1.4708, "step": 3738 }, { "epoch": 0.23, "learning_rate": 1.1266372910001811e-05, "loss": 1.2328, "step": 3740 }, { "epoch": 0.23, "learning_rate": 1.1272409005855014e-05, "loss": 1.6034, "step": 3742 }, { "epoch": 0.23, "learning_rate": 1.1278445101708215e-05, "loss": 1.1915, "step": 3744 }, { "epoch": 0.23, "learning_rate": 1.1284481197561419e-05, "loss": 1.303, "step": 3746 }, { "epoch": 0.23, "learning_rate": 1.129051729341462e-05, "loss": 1.3926, "step": 3748 }, { "epoch": 0.23, "learning_rate": 1.1296553389267822e-05, "loss": 1.2746, "step": 3750 }, { "epoch": 0.23, "learning_rate": 1.1302589485121025e-05, "loss": 1.2344, "step": 3752 }, { "epoch": 0.23, "learning_rate": 1.1308625580974226e-05, "loss": 1.7212, "step": 3754 }, { "epoch": 0.23, "learning_rate": 1.1314661676827428e-05, "loss": 1.3596, "step": 3756 }, { "epoch": 0.23, "learning_rate": 1.132069777268063e-05, "loss": 1.5007, "step": 3758 }, { "epoch": 0.23, "learning_rate": 1.1326733868533833e-05, "loss": 1.508, "step": 3760 }, { "epoch": 0.23, "learning_rate": 1.1332769964387036e-05, "loss": 1.7525, "step": 3762 }, { "epoch": 0.23, "learning_rate": 1.1338806060240236e-05, "loss": 1.6424, "step": 3764 }, { "epoch": 0.23, "learning_rate": 1.1344842156093439e-05, "loss": 1.1313, "step": 3766 }, { "epoch": 0.23, "learning_rate": 1.1350878251946641e-05, "loss": 1.357, "step": 3768 }, { "epoch": 0.23, "learning_rate": 1.1356914347799844e-05, "loss": 1.373, "step": 3770 }, { "epoch": 0.23, "learning_rate": 1.1362950443653047e-05, "loss": 1.5013, "step": 3772 }, { "epoch": 0.23, "learning_rate": 1.1368986539506247e-05, "loss": 1.7327, "step": 3774 }, { "epoch": 0.23, "learning_rate": 1.137502263535945e-05, "loss": 1.5814, "step": 3776 }, { "epoch": 0.23, "learning_rate": 1.1381058731212652e-05, "loss": 1.4796, "step": 3778 }, { "epoch": 0.23, "learning_rate": 1.1387094827065855e-05, "loss": 1.4702, "step": 3780 }, { "epoch": 0.23, "learning_rate": 1.1393130922919057e-05, "loss": 1.2729, "step": 3782 }, { "epoch": 0.23, "learning_rate": 1.1399167018772258e-05, "loss": 1.4254, "step": 3784 }, { "epoch": 0.23, "learning_rate": 1.140520311462546e-05, "loss": 1.5625, "step": 3786 }, { "epoch": 0.23, "learning_rate": 1.1411239210478663e-05, "loss": 1.2401, "step": 3788 }, { "epoch": 0.23, "learning_rate": 1.1417275306331866e-05, "loss": 1.5142, "step": 3790 }, { "epoch": 0.23, "learning_rate": 1.1423311402185068e-05, "loss": 1.3843, "step": 3792 }, { "epoch": 0.23, "learning_rate": 1.1429347498038269e-05, "loss": 1.4188, "step": 3794 }, { "epoch": 0.23, "learning_rate": 1.1435383593891472e-05, "loss": 1.4543, "step": 3796 }, { "epoch": 0.23, "learning_rate": 1.1441419689744674e-05, "loss": 1.5324, "step": 3798 }, { "epoch": 0.23, "learning_rate": 1.1447455785597875e-05, "loss": 1.3733, "step": 3800 }, { "epoch": 0.23, "learning_rate": 1.1453491881451079e-05, "loss": 1.3362, "step": 3802 }, { "epoch": 0.23, "learning_rate": 1.145952797730428e-05, "loss": 1.5677, "step": 3804 }, { "epoch": 0.23, "learning_rate": 1.1465564073157482e-05, "loss": 1.6416, "step": 3806 }, { "epoch": 0.23, "learning_rate": 1.1471600169010685e-05, "loss": 1.0665, "step": 3808 }, { "epoch": 0.23, "learning_rate": 1.1477636264863886e-05, "loss": 1.2546, "step": 3810 }, { "epoch": 0.23, "learning_rate": 1.1483672360717088e-05, "loss": 1.6499, "step": 3812 }, { "epoch": 0.23, "learning_rate": 1.148970845657029e-05, "loss": 1.5254, "step": 3814 }, { "epoch": 0.23, "learning_rate": 1.1495744552423493e-05, "loss": 1.8158, "step": 3816 }, { "epoch": 0.23, "learning_rate": 1.1501780648276696e-05, "loss": 1.205, "step": 3818 }, { "epoch": 0.23, "learning_rate": 1.1507816744129897e-05, "loss": 1.5039, "step": 3820 }, { "epoch": 0.23, "learning_rate": 1.15138528399831e-05, "loss": 1.4865, "step": 3822 }, { "epoch": 0.23, "learning_rate": 1.1519888935836302e-05, "loss": 1.8791, "step": 3824 }, { "epoch": 0.23, "learning_rate": 1.1525925031689504e-05, "loss": 1.3089, "step": 3826 }, { "epoch": 0.23, "learning_rate": 1.1531961127542707e-05, "loss": 1.759, "step": 3828 }, { "epoch": 0.23, "learning_rate": 1.1537997223395908e-05, "loss": 1.4826, "step": 3830 }, { "epoch": 0.23, "learning_rate": 1.154403331924911e-05, "loss": 0.8964, "step": 3832 }, { "epoch": 0.23, "learning_rate": 1.1550069415102313e-05, "loss": 1.6092, "step": 3834 }, { "epoch": 0.23, "learning_rate": 1.1556105510955515e-05, "loss": 1.5928, "step": 3836 }, { "epoch": 0.23, "learning_rate": 1.1562141606808718e-05, "loss": 1.4827, "step": 3838 }, { "epoch": 0.23, "learning_rate": 1.1568177702661918e-05, "loss": 1.461, "step": 3840 }, { "epoch": 0.23, "learning_rate": 1.1574213798515121e-05, "loss": 1.5405, "step": 3842 }, { "epoch": 0.23, "learning_rate": 1.1580249894368323e-05, "loss": 1.4442, "step": 3844 }, { "epoch": 0.23, "learning_rate": 1.1586285990221526e-05, "loss": 1.4439, "step": 3846 }, { "epoch": 0.23, "learning_rate": 1.1592322086074728e-05, "loss": 1.7273, "step": 3848 }, { "epoch": 0.23, "learning_rate": 1.159835818192793e-05, "loss": 1.3486, "step": 3850 }, { "epoch": 0.23, "learning_rate": 1.1604394277781132e-05, "loss": 1.6073, "step": 3852 }, { "epoch": 0.23, "learning_rate": 1.1610430373634334e-05, "loss": 1.507, "step": 3854 }, { "epoch": 0.23, "learning_rate": 1.1616466469487535e-05, "loss": 1.4445, "step": 3856 }, { "epoch": 0.23, "learning_rate": 1.162250256534074e-05, "loss": 1.4085, "step": 3858 }, { "epoch": 0.23, "learning_rate": 1.162853866119394e-05, "loss": 1.2621, "step": 3860 }, { "epoch": 0.23, "learning_rate": 1.1634574757047143e-05, "loss": 1.534, "step": 3862 }, { "epoch": 0.23, "learning_rate": 1.1640610852900345e-05, "loss": 1.5755, "step": 3864 }, { "epoch": 0.23, "learning_rate": 1.1646646948753546e-05, "loss": 1.3976, "step": 3866 }, { "epoch": 0.23, "learning_rate": 1.1652683044606748e-05, "loss": 1.7464, "step": 3868 }, { "epoch": 0.23, "learning_rate": 1.1658719140459951e-05, "loss": 1.2375, "step": 3870 }, { "epoch": 0.23, "learning_rate": 1.1664755236313154e-05, "loss": 0.9242, "step": 3872 }, { "epoch": 0.23, "learning_rate": 1.1670791332166356e-05, "loss": 1.4736, "step": 3874 }, { "epoch": 0.23, "learning_rate": 1.1676827428019557e-05, "loss": 1.3465, "step": 3876 }, { "epoch": 0.23, "learning_rate": 1.168286352387276e-05, "loss": 1.2631, "step": 3878 }, { "epoch": 0.23, "learning_rate": 1.1688899619725962e-05, "loss": 1.209, "step": 3880 }, { "epoch": 0.23, "learning_rate": 1.1694935715579164e-05, "loss": 1.4093, "step": 3882 }, { "epoch": 0.23, "learning_rate": 1.1700971811432367e-05, "loss": 1.4146, "step": 3884 }, { "epoch": 0.23, "learning_rate": 1.1707007907285568e-05, "loss": 1.2825, "step": 3886 }, { "epoch": 0.23, "learning_rate": 1.171304400313877e-05, "loss": 1.4302, "step": 3888 }, { "epoch": 0.23, "learning_rate": 1.1719080098991973e-05, "loss": 1.4046, "step": 3890 }, { "epoch": 0.23, "learning_rate": 1.1725116194845175e-05, "loss": 1.0678, "step": 3892 }, { "epoch": 0.24, "learning_rate": 1.1731152290698378e-05, "loss": 1.5198, "step": 3894 }, { "epoch": 0.24, "learning_rate": 1.1737188386551579e-05, "loss": 1.2917, "step": 3896 }, { "epoch": 0.24, "learning_rate": 1.1743224482404781e-05, "loss": 1.2514, "step": 3898 }, { "epoch": 0.24, "learning_rate": 1.1749260578257984e-05, "loss": 0.9033, "step": 3900 }, { "epoch": 0.24, "learning_rate": 1.1755296674111186e-05, "loss": 1.3858, "step": 3902 }, { "epoch": 0.24, "learning_rate": 1.1761332769964389e-05, "loss": 1.5308, "step": 3904 }, { "epoch": 0.24, "learning_rate": 1.176736886581759e-05, "loss": 1.3414, "step": 3906 }, { "epoch": 0.24, "learning_rate": 1.1773404961670792e-05, "loss": 1.1809, "step": 3908 }, { "epoch": 0.24, "learning_rate": 1.1779441057523994e-05, "loss": 1.4359, "step": 3910 }, { "epoch": 0.24, "learning_rate": 1.1785477153377195e-05, "loss": 0.9574, "step": 3912 }, { "epoch": 0.24, "learning_rate": 1.17915132492304e-05, "loss": 1.5627, "step": 3914 }, { "epoch": 0.24, "learning_rate": 1.17975493450836e-05, "loss": 1.5122, "step": 3916 }, { "epoch": 0.24, "learning_rate": 1.1803585440936803e-05, "loss": 1.1805, "step": 3918 }, { "epoch": 0.24, "learning_rate": 1.1809621536790005e-05, "loss": 1.5753, "step": 3920 }, { "epoch": 0.24, "learning_rate": 1.1815657632643206e-05, "loss": 1.3771, "step": 3922 }, { "epoch": 0.24, "learning_rate": 1.1821693728496409e-05, "loss": 1.4437, "step": 3924 }, { "epoch": 0.24, "learning_rate": 1.1827729824349611e-05, "loss": 1.5159, "step": 3926 }, { "epoch": 0.24, "learning_rate": 1.1833765920202814e-05, "loss": 1.0567, "step": 3928 }, { "epoch": 0.24, "learning_rate": 1.1839802016056016e-05, "loss": 1.2466, "step": 3930 }, { "epoch": 0.24, "learning_rate": 1.1845838111909217e-05, "loss": 1.3042, "step": 3932 }, { "epoch": 0.24, "learning_rate": 1.185187420776242e-05, "loss": 1.4184, "step": 3934 }, { "epoch": 0.24, "learning_rate": 1.1857910303615622e-05, "loss": 1.3742, "step": 3936 }, { "epoch": 0.24, "learning_rate": 1.1863946399468825e-05, "loss": 1.2517, "step": 3938 }, { "epoch": 0.24, "learning_rate": 1.1869982495322027e-05, "loss": 1.1906, "step": 3940 }, { "epoch": 0.24, "learning_rate": 1.1876018591175228e-05, "loss": 1.3005, "step": 3942 }, { "epoch": 0.24, "learning_rate": 1.188205468702843e-05, "loss": 1.5638, "step": 3944 }, { "epoch": 0.24, "learning_rate": 1.1888090782881631e-05, "loss": 1.1771, "step": 3946 }, { "epoch": 0.24, "learning_rate": 1.1894126878734835e-05, "loss": 1.1453, "step": 3948 }, { "epoch": 0.24, "learning_rate": 1.1900162974588038e-05, "loss": 1.6041, "step": 3950 }, { "epoch": 0.24, "learning_rate": 1.1906199070441239e-05, "loss": 1.7086, "step": 3952 }, { "epoch": 0.24, "learning_rate": 1.1912235166294441e-05, "loss": 1.3433, "step": 3954 }, { "epoch": 0.24, "learning_rate": 1.1918271262147644e-05, "loss": 1.5494, "step": 3956 }, { "epoch": 0.24, "learning_rate": 1.1924307358000846e-05, "loss": 1.4047, "step": 3958 }, { "epoch": 0.24, "learning_rate": 1.1930343453854049e-05, "loss": 1.245, "step": 3960 }, { "epoch": 0.24, "learning_rate": 1.193637954970725e-05, "loss": 1.1867, "step": 3962 }, { "epoch": 0.24, "learning_rate": 1.1942415645560452e-05, "loss": 1.3588, "step": 3964 }, { "epoch": 0.24, "learning_rate": 1.1948451741413655e-05, "loss": 1.5209, "step": 3966 }, { "epoch": 0.24, "learning_rate": 1.1954487837266855e-05, "loss": 1.0676, "step": 3968 }, { "epoch": 0.24, "learning_rate": 1.196052393312006e-05, "loss": 1.1465, "step": 3970 }, { "epoch": 0.24, "learning_rate": 1.196656002897326e-05, "loss": 1.4665, "step": 3972 }, { "epoch": 0.24, "learning_rate": 1.1972596124826463e-05, "loss": 1.2106, "step": 3974 }, { "epoch": 0.24, "learning_rate": 1.1978632220679666e-05, "loss": 1.2489, "step": 3976 }, { "epoch": 0.24, "learning_rate": 1.1984668316532866e-05, "loss": 1.7214, "step": 3978 }, { "epoch": 0.24, "learning_rate": 1.1990704412386069e-05, "loss": 1.6047, "step": 3980 }, { "epoch": 0.24, "learning_rate": 1.1996740508239271e-05, "loss": 1.2776, "step": 3982 }, { "epoch": 0.24, "learning_rate": 1.2002776604092474e-05, "loss": 1.5036, "step": 3984 }, { "epoch": 0.24, "learning_rate": 1.2008812699945676e-05, "loss": 1.2637, "step": 3986 }, { "epoch": 0.24, "learning_rate": 1.2014848795798877e-05, "loss": 1.4192, "step": 3988 }, { "epoch": 0.24, "learning_rate": 1.202088489165208e-05, "loss": 1.4035, "step": 3990 }, { "epoch": 0.24, "learning_rate": 1.2026920987505282e-05, "loss": 1.2281, "step": 3992 }, { "epoch": 0.24, "learning_rate": 1.2032957083358485e-05, "loss": 1.1843, "step": 3994 }, { "epoch": 0.24, "learning_rate": 1.2038993179211687e-05, "loss": 1.5836, "step": 3996 }, { "epoch": 0.24, "learning_rate": 1.2045029275064888e-05, "loss": 1.4646, "step": 3998 }, { "epoch": 0.24, "learning_rate": 1.205106537091809e-05, "loss": 1.1841, "step": 4000 }, { "epoch": 0.24, "learning_rate": 1.2057101466771291e-05, "loss": 0.9083, "step": 4002 }, { "epoch": 0.24, "learning_rate": 1.2063137562624496e-05, "loss": 1.2178, "step": 4004 }, { "epoch": 0.24, "learning_rate": 1.2069173658477698e-05, "loss": 1.2938, "step": 4006 }, { "epoch": 0.24, "learning_rate": 1.2075209754330899e-05, "loss": 1.5037, "step": 4008 }, { "epoch": 0.24, "learning_rate": 1.2081245850184101e-05, "loss": 1.2206, "step": 4010 }, { "epoch": 0.24, "learning_rate": 1.2087281946037302e-05, "loss": 1.3562, "step": 4012 }, { "epoch": 0.24, "learning_rate": 1.2093318041890505e-05, "loss": 1.6286, "step": 4014 }, { "epoch": 0.24, "learning_rate": 1.2099354137743709e-05, "loss": 1.2853, "step": 4016 }, { "epoch": 0.24, "learning_rate": 1.210539023359691e-05, "loss": 1.3344, "step": 4018 }, { "epoch": 0.24, "learning_rate": 1.2111426329450112e-05, "loss": 1.5533, "step": 4020 }, { "epoch": 0.24, "learning_rate": 1.2117462425303315e-05, "loss": 1.3074, "step": 4022 }, { "epoch": 0.24, "learning_rate": 1.2123498521156516e-05, "loss": 1.2716, "step": 4024 }, { "epoch": 0.24, "learning_rate": 1.212953461700972e-05, "loss": 1.237, "step": 4026 }, { "epoch": 0.24, "learning_rate": 1.213557071286292e-05, "loss": 1.7989, "step": 4028 }, { "epoch": 0.24, "learning_rate": 1.2141606808716123e-05, "loss": 1.4791, "step": 4030 }, { "epoch": 0.24, "learning_rate": 1.2147642904569326e-05, "loss": 1.3001, "step": 4032 }, { "epoch": 0.24, "learning_rate": 1.2153679000422527e-05, "loss": 1.0078, "step": 4034 }, { "epoch": 0.24, "learning_rate": 1.2159715096275729e-05, "loss": 1.1989, "step": 4036 }, { "epoch": 0.24, "learning_rate": 1.2165751192128932e-05, "loss": 1.3737, "step": 4038 }, { "epoch": 0.24, "learning_rate": 1.2171787287982134e-05, "loss": 1.4808, "step": 4040 }, { "epoch": 0.24, "learning_rate": 1.2177823383835337e-05, "loss": 1.2064, "step": 4042 }, { "epoch": 0.24, "learning_rate": 1.2183859479688537e-05, "loss": 1.3992, "step": 4044 }, { "epoch": 0.24, "learning_rate": 1.218989557554174e-05, "loss": 1.1841, "step": 4046 }, { "epoch": 0.24, "learning_rate": 1.2195931671394942e-05, "loss": 1.2566, "step": 4048 }, { "epoch": 0.24, "learning_rate": 1.2201967767248145e-05, "loss": 1.28, "step": 4050 }, { "epoch": 0.24, "learning_rate": 1.2208003863101347e-05, "loss": 0.9944, "step": 4052 }, { "epoch": 0.24, "learning_rate": 1.2214039958954548e-05, "loss": 1.5073, "step": 4054 }, { "epoch": 0.24, "learning_rate": 1.222007605480775e-05, "loss": 1.8236, "step": 4056 }, { "epoch": 0.24, "learning_rate": 1.2226112150660952e-05, "loss": 1.0627, "step": 4058 }, { "epoch": 0.25, "learning_rate": 1.2232148246514156e-05, "loss": 1.0713, "step": 4060 }, { "epoch": 0.25, "learning_rate": 1.2238184342367358e-05, "loss": 1.4522, "step": 4062 }, { "epoch": 0.25, "learning_rate": 1.224422043822056e-05, "loss": 1.4165, "step": 4064 }, { "epoch": 0.25, "learning_rate": 1.2250256534073762e-05, "loss": 1.3812, "step": 4066 }, { "epoch": 0.25, "learning_rate": 1.2256292629926962e-05, "loss": 1.2075, "step": 4068 }, { "epoch": 0.25, "learning_rate": 1.2262328725780165e-05, "loss": 1.2876, "step": 4070 }, { "epoch": 0.25, "learning_rate": 1.226836482163337e-05, "loss": 1.2071, "step": 4072 }, { "epoch": 0.25, "learning_rate": 1.227440091748657e-05, "loss": 1.7025, "step": 4074 }, { "epoch": 0.25, "learning_rate": 1.2280437013339773e-05, "loss": 1.5251, "step": 4076 }, { "epoch": 0.25, "learning_rate": 1.2286473109192975e-05, "loss": 1.4079, "step": 4078 }, { "epoch": 0.25, "learning_rate": 1.2292509205046176e-05, "loss": 1.5786, "step": 4080 }, { "epoch": 0.25, "learning_rate": 1.2298545300899378e-05, "loss": 1.2517, "step": 4082 }, { "epoch": 0.25, "learning_rate": 1.2304581396752581e-05, "loss": 1.5226, "step": 4084 }, { "epoch": 0.25, "learning_rate": 1.2310617492605783e-05, "loss": 1.476, "step": 4086 }, { "epoch": 0.25, "learning_rate": 1.2316653588458986e-05, "loss": 1.3581, "step": 4088 }, { "epoch": 0.25, "learning_rate": 1.2322689684312187e-05, "loss": 1.3525, "step": 4090 }, { "epoch": 0.25, "learning_rate": 1.232872578016539e-05, "loss": 1.1625, "step": 4092 }, { "epoch": 0.25, "learning_rate": 1.2334761876018592e-05, "loss": 1.3283, "step": 4094 }, { "epoch": 0.25, "learning_rate": 1.2340797971871794e-05, "loss": 1.4137, "step": 4096 }, { "epoch": 0.25, "learning_rate": 1.2346834067724997e-05, "loss": 1.8282, "step": 4098 }, { "epoch": 0.25, "learning_rate": 1.2352870163578198e-05, "loss": 1.0092, "step": 4100 }, { "epoch": 0.25, "learning_rate": 1.23589062594314e-05, "loss": 1.4459, "step": 4102 }, { "epoch": 0.25, "learning_rate": 1.2364942355284603e-05, "loss": 1.1501, "step": 4104 }, { "epoch": 0.25, "learning_rate": 1.2370978451137805e-05, "loss": 1.128, "step": 4106 }, { "epoch": 0.25, "learning_rate": 1.2377014546991008e-05, "loss": 1.3193, "step": 4108 }, { "epoch": 0.25, "learning_rate": 1.2383050642844208e-05, "loss": 1.2075, "step": 4110 }, { "epoch": 0.25, "learning_rate": 1.2389086738697411e-05, "loss": 1.4065, "step": 4112 }, { "epoch": 0.25, "learning_rate": 1.2395122834550612e-05, "loss": 1.2142, "step": 4114 }, { "epoch": 0.25, "learning_rate": 1.2401158930403816e-05, "loss": 1.0082, "step": 4116 }, { "epoch": 0.25, "learning_rate": 1.2407195026257019e-05, "loss": 1.3598, "step": 4118 }, { "epoch": 0.25, "learning_rate": 1.241323112211022e-05, "loss": 1.1555, "step": 4120 }, { "epoch": 0.25, "learning_rate": 1.2419267217963422e-05, "loss": 2.1894, "step": 4122 }, { "epoch": 0.25, "learning_rate": 1.2425303313816623e-05, "loss": 0.8688, "step": 4124 }, { "epoch": 0.25, "learning_rate": 1.2431339409669825e-05, "loss": 1.321, "step": 4126 }, { "epoch": 0.25, "learning_rate": 1.243737550552303e-05, "loss": 1.6525, "step": 4128 }, { "epoch": 0.25, "learning_rate": 1.244341160137623e-05, "loss": 1.1605, "step": 4130 }, { "epoch": 0.25, "learning_rate": 1.2449447697229433e-05, "loss": 1.2643, "step": 4132 }, { "epoch": 0.25, "learning_rate": 1.2455483793082635e-05, "loss": 1.2395, "step": 4134 }, { "epoch": 0.25, "learning_rate": 1.2461519888935836e-05, "loss": 1.6608, "step": 4136 }, { "epoch": 0.25, "learning_rate": 1.2467555984789039e-05, "loss": 1.4215, "step": 4138 }, { "epoch": 0.25, "learning_rate": 1.2473592080642241e-05, "loss": 1.2136, "step": 4140 }, { "epoch": 0.25, "learning_rate": 1.2479628176495444e-05, "loss": 1.4211, "step": 4142 }, { "epoch": 0.25, "learning_rate": 1.2485664272348646e-05, "loss": 1.2612, "step": 4144 }, { "epoch": 0.25, "learning_rate": 1.2491700368201847e-05, "loss": 1.1406, "step": 4146 }, { "epoch": 0.25, "learning_rate": 1.249773646405505e-05, "loss": 1.1594, "step": 4148 }, { "epoch": 0.25, "learning_rate": 1.2503772559908252e-05, "loss": 1.0646, "step": 4150 }, { "epoch": 0.25, "learning_rate": 1.2509808655761454e-05, "loss": 1.3466, "step": 4152 }, { "epoch": 0.25, "learning_rate": 1.2515844751614655e-05, "loss": 1.4155, "step": 4154 }, { "epoch": 0.25, "learning_rate": 1.252188084746786e-05, "loss": 1.4159, "step": 4156 }, { "epoch": 0.25, "learning_rate": 1.252791694332106e-05, "loss": 1.4197, "step": 4158 }, { "epoch": 0.25, "learning_rate": 1.2533953039174263e-05, "loss": 1.1713, "step": 4160 }, { "epoch": 0.25, "learning_rate": 1.2539989135027467e-05, "loss": 1.3132, "step": 4162 }, { "epoch": 0.25, "learning_rate": 1.2546025230880668e-05, "loss": 1.1746, "step": 4164 }, { "epoch": 0.25, "learning_rate": 1.2552061326733869e-05, "loss": 1.6145, "step": 4166 }, { "epoch": 0.25, "learning_rate": 1.255809742258707e-05, "loss": 1.6107, "step": 4168 }, { "epoch": 0.25, "learning_rate": 1.2564133518440274e-05, "loss": 1.2231, "step": 4170 }, { "epoch": 0.25, "learning_rate": 1.2570169614293476e-05, "loss": 1.5702, "step": 4172 }, { "epoch": 0.25, "learning_rate": 1.2576205710146677e-05, "loss": 1.6068, "step": 4174 }, { "epoch": 0.25, "learning_rate": 1.2582241805999881e-05, "loss": 1.4021, "step": 4176 }, { "epoch": 0.25, "learning_rate": 1.2588277901853082e-05, "loss": 2.0414, "step": 4178 }, { "epoch": 0.25, "learning_rate": 1.2594313997706283e-05, "loss": 1.5027, "step": 4180 }, { "epoch": 0.25, "learning_rate": 1.2600350093559487e-05, "loss": 1.4024, "step": 4182 }, { "epoch": 0.25, "learning_rate": 1.260638618941269e-05, "loss": 1.2333, "step": 4184 }, { "epoch": 0.25, "learning_rate": 1.261242228526589e-05, "loss": 1.6208, "step": 4186 }, { "epoch": 0.25, "learning_rate": 1.2618458381119091e-05, "loss": 1.1093, "step": 4188 }, { "epoch": 0.25, "learning_rate": 1.2624494476972295e-05, "loss": 1.3214, "step": 4190 }, { "epoch": 0.25, "learning_rate": 1.2630530572825496e-05, "loss": 1.0165, "step": 4192 }, { "epoch": 0.25, "learning_rate": 1.2636566668678699e-05, "loss": 1.5148, "step": 4194 }, { "epoch": 0.25, "learning_rate": 1.2642602764531903e-05, "loss": 1.2489, "step": 4196 }, { "epoch": 0.25, "learning_rate": 1.2648638860385104e-05, "loss": 1.3642, "step": 4198 }, { "epoch": 0.25, "learning_rate": 1.2654674956238305e-05, "loss": 1.5017, "step": 4200 }, { "epoch": 0.25, "learning_rate": 1.2660711052091509e-05, "loss": 1.1716, "step": 4202 }, { "epoch": 0.25, "learning_rate": 1.266674714794471e-05, "loss": 1.1478, "step": 4204 }, { "epoch": 0.25, "learning_rate": 1.2672783243797912e-05, "loss": 1.6393, "step": 4206 }, { "epoch": 0.25, "learning_rate": 1.2678819339651113e-05, "loss": 2.1988, "step": 4208 }, { "epoch": 0.25, "learning_rate": 1.2684855435504317e-05, "loss": 1.2693, "step": 4210 }, { "epoch": 0.25, "learning_rate": 1.2690891531357518e-05, "loss": 1.3559, "step": 4212 }, { "epoch": 0.25, "learning_rate": 1.2696927627210719e-05, "loss": 1.3149, "step": 4214 }, { "epoch": 0.25, "learning_rate": 1.2702963723063923e-05, "loss": 1.3628, "step": 4216 }, { "epoch": 0.25, "learning_rate": 1.2708999818917126e-05, "loss": 1.2818, "step": 4218 }, { "epoch": 0.25, "learning_rate": 1.2715035914770326e-05, "loss": 1.3947, "step": 4220 }, { "epoch": 0.25, "learning_rate": 1.272107201062353e-05, "loss": 1.1555, "step": 4222 }, { "epoch": 0.25, "learning_rate": 1.2727108106476731e-05, "loss": 1.4235, "step": 4224 }, { "epoch": 0.26, "learning_rate": 1.2733144202329932e-05, "loss": 1.4463, "step": 4226 }, { "epoch": 0.26, "learning_rate": 1.2739180298183136e-05, "loss": 1.2969, "step": 4228 }, { "epoch": 0.26, "learning_rate": 1.2745216394036339e-05, "loss": 1.5966, "step": 4230 }, { "epoch": 0.26, "learning_rate": 1.275125248988954e-05, "loss": 1.4009, "step": 4232 }, { "epoch": 0.26, "learning_rate": 1.275728858574274e-05, "loss": 1.3127, "step": 4234 }, { "epoch": 0.26, "learning_rate": 1.2763324681595945e-05, "loss": 1.7498, "step": 4236 }, { "epoch": 0.26, "learning_rate": 1.2769360777449146e-05, "loss": 1.6811, "step": 4238 }, { "epoch": 0.26, "learning_rate": 1.2775396873302348e-05, "loss": 0.927, "step": 4240 }, { "epoch": 0.26, "learning_rate": 1.2781432969155552e-05, "loss": 0.9093, "step": 4242 }, { "epoch": 0.26, "learning_rate": 1.2787469065008753e-05, "loss": 1.357, "step": 4244 }, { "epoch": 0.26, "learning_rate": 1.2793505160861954e-05, "loss": 1.5957, "step": 4246 }, { "epoch": 0.26, "learning_rate": 1.2799541256715158e-05, "loss": 1.3656, "step": 4248 }, { "epoch": 0.26, "learning_rate": 1.2805577352568359e-05, "loss": 1.4371, "step": 4250 }, { "epoch": 0.26, "learning_rate": 1.2811613448421561e-05, "loss": 1.1559, "step": 4252 }, { "epoch": 0.26, "learning_rate": 1.2817649544274762e-05, "loss": 1.1721, "step": 4254 }, { "epoch": 0.26, "learning_rate": 1.2823685640127966e-05, "loss": 1.808, "step": 4256 }, { "epoch": 0.26, "learning_rate": 1.2829721735981167e-05, "loss": 1.4839, "step": 4258 }, { "epoch": 0.26, "learning_rate": 1.283575783183437e-05, "loss": 1.6812, "step": 4260 }, { "epoch": 0.26, "learning_rate": 1.2841793927687572e-05, "loss": 1.799, "step": 4262 }, { "epoch": 0.26, "learning_rate": 1.2847830023540775e-05, "loss": 1.7404, "step": 4264 }, { "epoch": 0.26, "learning_rate": 1.2853866119393976e-05, "loss": 1.2393, "step": 4266 }, { "epoch": 0.26, "learning_rate": 1.285990221524718e-05, "loss": 1.4642, "step": 4268 }, { "epoch": 0.26, "learning_rate": 1.286593831110038e-05, "loss": 1.4192, "step": 4270 }, { "epoch": 0.26, "learning_rate": 1.2871974406953583e-05, "loss": 1.4167, "step": 4272 }, { "epoch": 0.26, "learning_rate": 1.2878010502806786e-05, "loss": 1.4123, "step": 4274 }, { "epoch": 0.26, "learning_rate": 1.2884046598659988e-05, "loss": 1.1318, "step": 4276 }, { "epoch": 0.26, "learning_rate": 1.2890082694513189e-05, "loss": 1.5448, "step": 4278 }, { "epoch": 0.26, "learning_rate": 1.289611879036639e-05, "loss": 1.3393, "step": 4280 }, { "epoch": 0.26, "learning_rate": 1.2902154886219594e-05, "loss": 1.1821, "step": 4282 }, { "epoch": 0.26, "learning_rate": 1.2908190982072797e-05, "loss": 1.5442, "step": 4284 }, { "epoch": 0.26, "learning_rate": 1.2914227077925997e-05, "loss": 1.5635, "step": 4286 }, { "epoch": 0.26, "learning_rate": 1.2920263173779202e-05, "loss": 1.7732, "step": 4288 }, { "epoch": 0.26, "learning_rate": 1.2926299269632402e-05, "loss": 1.2164, "step": 4290 }, { "epoch": 0.26, "learning_rate": 1.2932335365485603e-05, "loss": 1.7008, "step": 4292 }, { "epoch": 0.26, "learning_rate": 1.2938371461338807e-05, "loss": 1.4802, "step": 4294 }, { "epoch": 0.26, "learning_rate": 1.294440755719201e-05, "loss": 1.8867, "step": 4296 }, { "epoch": 0.26, "learning_rate": 1.295044365304521e-05, "loss": 1.3901, "step": 4298 }, { "epoch": 0.26, "learning_rate": 1.2956479748898412e-05, "loss": 0.9978, "step": 4300 }, { "epoch": 0.26, "learning_rate": 1.2962515844751616e-05, "loss": 1.1942, "step": 4302 }, { "epoch": 0.26, "learning_rate": 1.2968551940604817e-05, "loss": 1.6235, "step": 4304 }, { "epoch": 0.26, "learning_rate": 1.2974588036458019e-05, "loss": 1.2074, "step": 4306 }, { "epoch": 0.26, "learning_rate": 1.2980624132311223e-05, "loss": 1.3319, "step": 4308 }, { "epoch": 0.26, "learning_rate": 1.2986660228164424e-05, "loss": 1.3863, "step": 4310 }, { "epoch": 0.26, "learning_rate": 1.2992696324017625e-05, "loss": 1.5216, "step": 4312 }, { "epoch": 0.26, "learning_rate": 1.299873241987083e-05, "loss": 1.1153, "step": 4314 }, { "epoch": 0.26, "learning_rate": 1.300476851572403e-05, "loss": 1.016, "step": 4316 }, { "epoch": 0.26, "learning_rate": 1.3010804611577233e-05, "loss": 1.232, "step": 4318 }, { "epoch": 0.26, "learning_rate": 1.3016840707430433e-05, "loss": 1.2223, "step": 4320 }, { "epoch": 0.26, "learning_rate": 1.3022876803283638e-05, "loss": 1.4798, "step": 4322 }, { "epoch": 0.26, "learning_rate": 1.3028912899136838e-05, "loss": 1.6608, "step": 4324 }, { "epoch": 0.26, "learning_rate": 1.303494899499004e-05, "loss": 1.3691, "step": 4326 }, { "epoch": 0.26, "learning_rate": 1.3040985090843243e-05, "loss": 1.0101, "step": 4328 }, { "epoch": 0.26, "learning_rate": 1.3047021186696446e-05, "loss": 1.1885, "step": 4330 }, { "epoch": 0.26, "learning_rate": 1.3053057282549647e-05, "loss": 1.1623, "step": 4332 }, { "epoch": 0.26, "learning_rate": 1.3059093378402851e-05, "loss": 1.6737, "step": 4334 }, { "epoch": 0.26, "learning_rate": 1.3065129474256052e-05, "loss": 1.1323, "step": 4336 }, { "epoch": 0.26, "learning_rate": 1.3071165570109253e-05, "loss": 1.3638, "step": 4338 }, { "epoch": 0.26, "learning_rate": 1.3077201665962457e-05, "loss": 1.4452, "step": 4340 }, { "epoch": 0.26, "learning_rate": 1.308323776181566e-05, "loss": 1.6934, "step": 4342 }, { "epoch": 0.26, "learning_rate": 1.308927385766886e-05, "loss": 1.6824, "step": 4344 }, { "epoch": 0.26, "learning_rate": 1.3095309953522061e-05, "loss": 1.1391, "step": 4346 }, { "epoch": 0.26, "learning_rate": 1.3101346049375265e-05, "loss": 1.2809, "step": 4348 }, { "epoch": 0.26, "learning_rate": 1.3107382145228466e-05, "loss": 1.5735, "step": 4350 }, { "epoch": 0.26, "learning_rate": 1.3113418241081668e-05, "loss": 1.1094, "step": 4352 }, { "epoch": 0.26, "learning_rate": 1.3119454336934873e-05, "loss": 1.2264, "step": 4354 }, { "epoch": 0.26, "learning_rate": 1.3125490432788073e-05, "loss": 1.3434, "step": 4356 }, { "epoch": 0.26, "learning_rate": 1.3131526528641274e-05, "loss": 1.3017, "step": 4358 }, { "epoch": 0.26, "learning_rate": 1.3137562624494479e-05, "loss": 1.2239, "step": 4360 }, { "epoch": 0.26, "learning_rate": 1.314359872034768e-05, "loss": 1.3939, "step": 4362 }, { "epoch": 0.26, "learning_rate": 1.3149634816200882e-05, "loss": 1.3839, "step": 4364 }, { "epoch": 0.26, "learning_rate": 1.3155670912054083e-05, "loss": 1.2398, "step": 4366 }, { "epoch": 0.26, "learning_rate": 1.3161707007907287e-05, "loss": 1.8698, "step": 4368 }, { "epoch": 0.26, "learning_rate": 1.3167743103760488e-05, "loss": 1.1646, "step": 4370 }, { "epoch": 0.26, "learning_rate": 1.317377919961369e-05, "loss": 1.1607, "step": 4372 }, { "epoch": 0.26, "learning_rate": 1.3179815295466893e-05, "loss": 1.2448, "step": 4374 }, { "epoch": 0.26, "learning_rate": 1.3185851391320095e-05, "loss": 1.4082, "step": 4376 }, { "epoch": 0.26, "learning_rate": 1.3191887487173296e-05, "loss": 1.5273, "step": 4378 }, { "epoch": 0.26, "learning_rate": 1.31979235830265e-05, "loss": 1.6386, "step": 4380 }, { "epoch": 0.26, "learning_rate": 1.3203959678879701e-05, "loss": 1.3077, "step": 4382 }, { "epoch": 0.26, "learning_rate": 1.3209995774732904e-05, "loss": 1.6277, "step": 4384 }, { "epoch": 0.26, "learning_rate": 1.3216031870586106e-05, "loss": 1.1232, "step": 4386 }, { "epoch": 0.26, "learning_rate": 1.3222067966439309e-05, "loss": 1.3711, "step": 4388 }, { "epoch": 0.26, "learning_rate": 1.322810406229251e-05, "loss": 1.3355, "step": 4390 }, { "epoch": 0.27, "learning_rate": 1.323414015814571e-05, "loss": 1.709, "step": 4392 }, { "epoch": 0.27, "learning_rate": 1.3240176253998914e-05, "loss": 1.0052, "step": 4394 }, { "epoch": 0.27, "learning_rate": 1.3246212349852117e-05, "loss": 1.1165, "step": 4396 }, { "epoch": 0.27, "learning_rate": 1.3252248445705318e-05, "loss": 1.2667, "step": 4398 }, { "epoch": 0.27, "learning_rate": 1.3258284541558522e-05, "loss": 1.2788, "step": 4400 }, { "epoch": 0.27, "learning_rate": 1.3264320637411723e-05, "loss": 1.3535, "step": 4402 }, { "epoch": 0.27, "learning_rate": 1.3270356733264924e-05, "loss": 1.354, "step": 4404 }, { "epoch": 0.27, "learning_rate": 1.3276392829118128e-05, "loss": 1.3405, "step": 4406 }, { "epoch": 0.27, "learning_rate": 1.328242892497133e-05, "loss": 1.5974, "step": 4408 }, { "epoch": 0.27, "learning_rate": 1.3288465020824531e-05, "loss": 1.6967, "step": 4410 }, { "epoch": 0.27, "learning_rate": 1.3294501116677732e-05, "loss": 1.6365, "step": 4412 }, { "epoch": 0.27, "learning_rate": 1.3300537212530936e-05, "loss": 1.642, "step": 4414 }, { "epoch": 0.27, "learning_rate": 1.3306573308384137e-05, "loss": 1.1805, "step": 4416 }, { "epoch": 0.27, "learning_rate": 1.331260940423734e-05, "loss": 1.1318, "step": 4418 }, { "epoch": 0.27, "learning_rate": 1.3318645500090544e-05, "loss": 1.5448, "step": 4420 }, { "epoch": 0.27, "learning_rate": 1.3324681595943745e-05, "loss": 1.2766, "step": 4422 }, { "epoch": 0.27, "learning_rate": 1.3330717691796945e-05, "loss": 1.4095, "step": 4424 }, { "epoch": 0.27, "learning_rate": 1.333675378765015e-05, "loss": 1.6134, "step": 4426 }, { "epoch": 0.27, "learning_rate": 1.334278988350335e-05, "loss": 1.5534, "step": 4428 }, { "epoch": 0.27, "learning_rate": 1.3348825979356553e-05, "loss": 1.186, "step": 4430 }, { "epoch": 0.27, "learning_rate": 1.3354862075209754e-05, "loss": 1.4594, "step": 4432 }, { "epoch": 0.27, "learning_rate": 1.3360898171062958e-05, "loss": 1.0784, "step": 4434 }, { "epoch": 0.27, "learning_rate": 1.3366934266916159e-05, "loss": 1.4736, "step": 4436 }, { "epoch": 0.27, "learning_rate": 1.337297036276936e-05, "loss": 1.4281, "step": 4438 }, { "epoch": 0.27, "learning_rate": 1.3379006458622564e-05, "loss": 1.3016, "step": 4440 }, { "epoch": 0.27, "learning_rate": 1.3385042554475766e-05, "loss": 1.2306, "step": 4442 }, { "epoch": 0.27, "learning_rate": 1.3391078650328967e-05, "loss": 1.1832, "step": 4444 }, { "epoch": 0.27, "learning_rate": 1.3397114746182171e-05, "loss": 1.2065, "step": 4446 }, { "epoch": 0.27, "learning_rate": 1.3403150842035372e-05, "loss": 1.02, "step": 4448 }, { "epoch": 0.27, "learning_rate": 1.3409186937888573e-05, "loss": 1.7301, "step": 4450 }, { "epoch": 0.27, "learning_rate": 1.3415223033741777e-05, "loss": 1.7441, "step": 4452 }, { "epoch": 0.27, "learning_rate": 1.342125912959498e-05, "loss": 1.9777, "step": 4454 }, { "epoch": 0.27, "learning_rate": 1.342729522544818e-05, "loss": 1.2533, "step": 4456 }, { "epoch": 0.27, "learning_rate": 1.3433331321301381e-05, "loss": 1.398, "step": 4458 }, { "epoch": 0.27, "learning_rate": 1.3439367417154586e-05, "loss": 1.3104, "step": 4460 }, { "epoch": 0.27, "learning_rate": 1.3445403513007786e-05, "loss": 1.0643, "step": 4462 }, { "epoch": 0.27, "learning_rate": 1.3451439608860989e-05, "loss": 1.052, "step": 4464 }, { "epoch": 0.27, "learning_rate": 1.3457475704714193e-05, "loss": 1.5247, "step": 4466 }, { "epoch": 0.27, "learning_rate": 1.3463511800567394e-05, "loss": 1.2155, "step": 4468 }, { "epoch": 0.27, "learning_rate": 1.3469547896420595e-05, "loss": 1.7148, "step": 4470 }, { "epoch": 0.27, "learning_rate": 1.3475583992273799e-05, "loss": 1.0141, "step": 4472 }, { "epoch": 0.27, "learning_rate": 1.3481620088127e-05, "loss": 1.3403, "step": 4474 }, { "epoch": 0.27, "learning_rate": 1.3487656183980202e-05, "loss": 1.1906, "step": 4476 }, { "epoch": 0.27, "learning_rate": 1.3493692279833403e-05, "loss": 1.29, "step": 4478 }, { "epoch": 0.27, "learning_rate": 1.3499728375686607e-05, "loss": 1.2431, "step": 4480 }, { "epoch": 0.27, "learning_rate": 1.3505764471539808e-05, "loss": 1.2702, "step": 4482 }, { "epoch": 0.27, "learning_rate": 1.351180056739301e-05, "loss": 1.5594, "step": 4484 }, { "epoch": 0.27, "learning_rate": 1.3517836663246213e-05, "loss": 1.2269, "step": 4486 }, { "epoch": 0.27, "learning_rate": 1.3523872759099416e-05, "loss": 1.2838, "step": 4488 }, { "epoch": 0.27, "learning_rate": 1.3529908854952616e-05, "loss": 1.24, "step": 4490 }, { "epoch": 0.27, "learning_rate": 1.353594495080582e-05, "loss": 1.2326, "step": 4492 }, { "epoch": 0.27, "learning_rate": 1.3541981046659021e-05, "loss": 1.2974, "step": 4494 }, { "epoch": 0.27, "learning_rate": 1.3548017142512224e-05, "loss": 1.0885, "step": 4496 }, { "epoch": 0.27, "learning_rate": 1.3554053238365425e-05, "loss": 1.2361, "step": 4498 }, { "epoch": 0.27, "learning_rate": 1.3560089334218629e-05, "loss": 1.3434, "step": 4500 }, { "epoch": 0.27, "learning_rate": 1.356612543007183e-05, "loss": 1.6135, "step": 4502 }, { "epoch": 0.27, "learning_rate": 1.357216152592503e-05, "loss": 1.3235, "step": 4504 }, { "epoch": 0.27, "learning_rate": 1.3578197621778235e-05, "loss": 1.9081, "step": 4506 }, { "epoch": 0.27, "learning_rate": 1.3584233717631437e-05, "loss": 1.1885, "step": 4508 }, { "epoch": 0.27, "learning_rate": 1.3590269813484638e-05, "loss": 1.2983, "step": 4510 }, { "epoch": 0.27, "learning_rate": 1.3596305909337842e-05, "loss": 1.2273, "step": 4512 }, { "epoch": 0.27, "learning_rate": 1.3602342005191043e-05, "loss": 1.333, "step": 4514 }, { "epoch": 0.27, "learning_rate": 1.3608378101044244e-05, "loss": 1.3431, "step": 4516 }, { "epoch": 0.27, "learning_rate": 1.3614414196897448e-05, "loss": 1.5192, "step": 4518 }, { "epoch": 0.27, "learning_rate": 1.362045029275065e-05, "loss": 1.4394, "step": 4520 }, { "epoch": 0.27, "learning_rate": 1.3626486388603852e-05, "loss": 1.4347, "step": 4522 }, { "epoch": 0.27, "learning_rate": 1.3632522484457052e-05, "loss": 1.0103, "step": 4524 }, { "epoch": 0.27, "learning_rate": 1.3638558580310257e-05, "loss": 1.2309, "step": 4526 }, { "epoch": 0.27, "learning_rate": 1.3644594676163457e-05, "loss": 1.0657, "step": 4528 }, { "epoch": 0.27, "learning_rate": 1.365063077201666e-05, "loss": 1.095, "step": 4530 }, { "epoch": 0.27, "learning_rate": 1.3656666867869864e-05, "loss": 1.4268, "step": 4532 }, { "epoch": 0.27, "learning_rate": 1.3662702963723065e-05, "loss": 1.3773, "step": 4534 }, { "epoch": 0.27, "learning_rate": 1.3668739059576266e-05, "loss": 1.307, "step": 4536 }, { "epoch": 0.27, "learning_rate": 1.367477515542947e-05, "loss": 1.192, "step": 4538 }, { "epoch": 0.27, "learning_rate": 1.368081125128267e-05, "loss": 1.3424, "step": 4540 }, { "epoch": 0.27, "learning_rate": 1.3686847347135873e-05, "loss": 1.384, "step": 4542 }, { "epoch": 0.27, "learning_rate": 1.3692883442989074e-05, "loss": 1.4875, "step": 4544 }, { "epoch": 0.27, "learning_rate": 1.3698919538842278e-05, "loss": 1.3387, "step": 4546 }, { "epoch": 0.27, "learning_rate": 1.3704955634695479e-05, "loss": 1.1219, "step": 4548 }, { "epoch": 0.27, "learning_rate": 1.371099173054868e-05, "loss": 1.1444, "step": 4550 }, { "epoch": 0.27, "learning_rate": 1.3717027826401884e-05, "loss": 1.3611, "step": 4552 }, { "epoch": 0.27, "learning_rate": 1.3723063922255087e-05, "loss": 1.6943, "step": 4554 }, { "epoch": 0.27, "learning_rate": 1.3729100018108287e-05, "loss": 1.0395, "step": 4556 }, { "epoch": 0.28, "learning_rate": 1.3735136113961492e-05, "loss": 1.7364, "step": 4558 }, { "epoch": 0.28, "learning_rate": 1.3741172209814693e-05, "loss": 1.3436, "step": 4560 }, { "epoch": 0.28, "learning_rate": 1.3747208305667893e-05, "loss": 1.4281, "step": 4562 }, { "epoch": 0.28, "learning_rate": 1.3753244401521098e-05, "loss": 1.425, "step": 4564 }, { "epoch": 0.28, "learning_rate": 1.37592804973743e-05, "loss": 0.9799, "step": 4566 }, { "epoch": 0.28, "learning_rate": 1.3765316593227501e-05, "loss": 1.5558, "step": 4568 }, { "epoch": 0.28, "learning_rate": 1.3771352689080702e-05, "loss": 1.281, "step": 4570 }, { "epoch": 0.28, "learning_rate": 1.3777388784933906e-05, "loss": 1.3486, "step": 4572 }, { "epoch": 0.28, "learning_rate": 1.3783424880787107e-05, "loss": 1.3329, "step": 4574 }, { "epoch": 0.28, "learning_rate": 1.378946097664031e-05, "loss": 1.4434, "step": 4576 }, { "epoch": 0.28, "learning_rate": 1.3795497072493513e-05, "loss": 1.3588, "step": 4578 }, { "epoch": 0.28, "learning_rate": 1.3801533168346714e-05, "loss": 1.3757, "step": 4580 }, { "epoch": 0.28, "learning_rate": 1.3807569264199915e-05, "loss": 1.8282, "step": 4582 }, { "epoch": 0.28, "learning_rate": 1.381360536005312e-05, "loss": 1.3072, "step": 4584 }, { "epoch": 0.28, "learning_rate": 1.381964145590632e-05, "loss": 1.3567, "step": 4586 }, { "epoch": 0.28, "learning_rate": 1.3825677551759523e-05, "loss": 1.2812, "step": 4588 }, { "epoch": 0.28, "learning_rate": 1.3831713647612723e-05, "loss": 1.413, "step": 4590 }, { "epoch": 0.28, "learning_rate": 1.3837749743465928e-05, "loss": 1.3903, "step": 4592 }, { "epoch": 0.28, "learning_rate": 1.3843785839319128e-05, "loss": 1.4637, "step": 4594 }, { "epoch": 0.28, "learning_rate": 1.3849821935172331e-05, "loss": 1.5442, "step": 4596 }, { "epoch": 0.28, "learning_rate": 1.3855858031025533e-05, "loss": 1.1945, "step": 4598 }, { "epoch": 0.28, "learning_rate": 1.3861894126878736e-05, "loss": 1.6792, "step": 4600 }, { "epoch": 0.28, "learning_rate": 1.3867930222731937e-05, "loss": 1.3635, "step": 4602 }, { "epoch": 0.28, "learning_rate": 1.3873966318585141e-05, "loss": 1.533, "step": 4604 }, { "epoch": 0.28, "learning_rate": 1.3880002414438342e-05, "loss": 1.1696, "step": 4606 }, { "epoch": 0.28, "learning_rate": 1.3886038510291544e-05, "loss": 1.5126, "step": 4608 }, { "epoch": 0.28, "learning_rate": 1.3892074606144745e-05, "loss": 1.1418, "step": 4610 }, { "epoch": 0.28, "learning_rate": 1.389811070199795e-05, "loss": 1.4374, "step": 4612 }, { "epoch": 0.28, "learning_rate": 1.390414679785115e-05, "loss": 1.7416, "step": 4614 }, { "epoch": 0.28, "learning_rate": 1.3910182893704351e-05, "loss": 1.0636, "step": 4616 }, { "epoch": 0.28, "learning_rate": 1.3916218989557555e-05, "loss": 1.3575, "step": 4618 }, { "epoch": 0.28, "learning_rate": 1.3922255085410758e-05, "loss": 1.1638, "step": 4620 }, { "epoch": 0.28, "learning_rate": 1.3928291181263959e-05, "loss": 0.8478, "step": 4622 }, { "epoch": 0.28, "learning_rate": 1.3934327277117163e-05, "loss": 1.4888, "step": 4624 }, { "epoch": 0.28, "learning_rate": 1.3940363372970364e-05, "loss": 1.4928, "step": 4626 }, { "epoch": 0.28, "learning_rate": 1.3946399468823564e-05, "loss": 1.4372, "step": 4628 }, { "epoch": 0.28, "learning_rate": 1.3952435564676769e-05, "loss": 1.5066, "step": 4630 }, { "epoch": 0.28, "learning_rate": 1.3958471660529971e-05, "loss": 1.2805, "step": 4632 }, { "epoch": 0.28, "learning_rate": 1.3964507756383172e-05, "loss": 1.5799, "step": 4634 }, { "epoch": 0.28, "learning_rate": 1.3970543852236373e-05, "loss": 1.2833, "step": 4636 }, { "epoch": 0.28, "learning_rate": 1.3976579948089577e-05, "loss": 1.4017, "step": 4638 }, { "epoch": 0.28, "learning_rate": 1.3982616043942778e-05, "loss": 0.8664, "step": 4640 }, { "epoch": 0.28, "learning_rate": 1.398865213979598e-05, "loss": 1.1944, "step": 4642 }, { "epoch": 0.28, "learning_rate": 1.3994688235649184e-05, "loss": 1.2919, "step": 4644 }, { "epoch": 0.28, "learning_rate": 1.4000724331502385e-05, "loss": 1.3504, "step": 4646 }, { "epoch": 0.28, "learning_rate": 1.4006760427355586e-05, "loss": 1.0064, "step": 4648 }, { "epoch": 0.28, "learning_rate": 1.401279652320879e-05, "loss": 1.6124, "step": 4650 }, { "epoch": 0.28, "learning_rate": 1.4018832619061991e-05, "loss": 1.2726, "step": 4652 }, { "epoch": 0.28, "learning_rate": 1.4024868714915194e-05, "loss": 1.2699, "step": 4654 }, { "epoch": 0.28, "learning_rate": 1.4030904810768394e-05, "loss": 1.2302, "step": 4656 }, { "epoch": 0.28, "learning_rate": 1.4036940906621599e-05, "loss": 1.1148, "step": 4658 }, { "epoch": 0.28, "learning_rate": 1.40429770024748e-05, "loss": 1.5204, "step": 4660 }, { "epoch": 0.28, "learning_rate": 1.4049013098328e-05, "loss": 1.1608, "step": 4662 }, { "epoch": 0.28, "learning_rate": 1.4055049194181205e-05, "loss": 1.0092, "step": 4664 }, { "epoch": 0.28, "learning_rate": 1.4061085290034407e-05, "loss": 1.2441, "step": 4666 }, { "epoch": 0.28, "learning_rate": 1.4067121385887608e-05, "loss": 1.3566, "step": 4668 }, { "epoch": 0.28, "learning_rate": 1.4073157481740812e-05, "loss": 0.9298, "step": 4670 }, { "epoch": 0.28, "learning_rate": 1.4079193577594013e-05, "loss": 1.5657, "step": 4672 }, { "epoch": 0.28, "learning_rate": 1.4085229673447214e-05, "loss": 1.1006, "step": 4674 }, { "epoch": 0.28, "learning_rate": 1.4091265769300418e-05, "loss": 1.4315, "step": 4676 }, { "epoch": 0.28, "learning_rate": 1.409730186515362e-05, "loss": 1.173, "step": 4678 }, { "epoch": 0.28, "learning_rate": 1.4103337961006821e-05, "loss": 1.531, "step": 4680 }, { "epoch": 0.28, "learning_rate": 1.4109374056860022e-05, "loss": 1.0989, "step": 4682 }, { "epoch": 0.28, "learning_rate": 1.4115410152713226e-05, "loss": 1.2613, "step": 4684 }, { "epoch": 0.28, "learning_rate": 1.4121446248566427e-05, "loss": 1.2669, "step": 4686 }, { "epoch": 0.28, "learning_rate": 1.412748234441963e-05, "loss": 1.2974, "step": 4688 }, { "epoch": 0.28, "learning_rate": 1.4133518440272834e-05, "loss": 1.635, "step": 4690 }, { "epoch": 0.28, "learning_rate": 1.4139554536126035e-05, "loss": 0.9973, "step": 4692 }, { "epoch": 0.28, "learning_rate": 1.4145590631979235e-05, "loss": 1.1181, "step": 4694 }, { "epoch": 0.28, "learning_rate": 1.415162672783244e-05, "loss": 1.2472, "step": 4696 }, { "epoch": 0.28, "learning_rate": 1.415766282368564e-05, "loss": 1.2356, "step": 4698 }, { "epoch": 0.28, "learning_rate": 1.4163698919538843e-05, "loss": 1.2426, "step": 4700 }, { "epoch": 0.28, "learning_rate": 1.4169735015392044e-05, "loss": 1.3514, "step": 4702 }, { "epoch": 0.28, "learning_rate": 1.4175771111245248e-05, "loss": 0.8425, "step": 4704 }, { "epoch": 0.28, "learning_rate": 1.4181807207098449e-05, "loss": 1.3856, "step": 4706 }, { "epoch": 0.28, "learning_rate": 1.4187843302951651e-05, "loss": 1.0709, "step": 4708 }, { "epoch": 0.28, "learning_rate": 1.4193879398804854e-05, "loss": 1.2958, "step": 4710 }, { "epoch": 0.28, "learning_rate": 1.4199915494658056e-05, "loss": 1.4713, "step": 4712 }, { "epoch": 0.28, "learning_rate": 1.4205951590511257e-05, "loss": 1.5658, "step": 4714 }, { "epoch": 0.28, "learning_rate": 1.4211987686364461e-05, "loss": 1.2307, "step": 4716 }, { "epoch": 0.28, "learning_rate": 1.4218023782217662e-05, "loss": 1.4299, "step": 4718 }, { "epoch": 0.28, "learning_rate": 1.4224059878070865e-05, "loss": 1.5112, "step": 4720 }, { "epoch": 0.29, "learning_rate": 1.4230095973924066e-05, "loss": 1.1667, "step": 4722 }, { "epoch": 0.29, "learning_rate": 1.423613206977727e-05, "loss": 1.1183, "step": 4724 }, { "epoch": 0.29, "learning_rate": 1.424216816563047e-05, "loss": 1.2885, "step": 4726 }, { "epoch": 0.29, "learning_rate": 1.4248204261483671e-05, "loss": 1.2211, "step": 4728 }, { "epoch": 0.29, "learning_rate": 1.4254240357336876e-05, "loss": 1.5667, "step": 4730 }, { "epoch": 0.29, "learning_rate": 1.4260276453190078e-05, "loss": 1.4572, "step": 4732 }, { "epoch": 0.29, "learning_rate": 1.4266312549043279e-05, "loss": 1.0524, "step": 4734 }, { "epoch": 0.29, "learning_rate": 1.4272348644896483e-05, "loss": 1.3833, "step": 4736 }, { "epoch": 0.29, "learning_rate": 1.4278384740749684e-05, "loss": 1.5913, "step": 4738 }, { "epoch": 0.29, "learning_rate": 1.4284420836602885e-05, "loss": 1.3788, "step": 4740 }, { "epoch": 0.29, "learning_rate": 1.4290456932456089e-05, "loss": 1.1308, "step": 4742 }, { "epoch": 0.29, "learning_rate": 1.4296493028309291e-05, "loss": 1.2179, "step": 4744 }, { "epoch": 0.29, "learning_rate": 1.4302529124162492e-05, "loss": 1.24, "step": 4746 }, { "epoch": 0.29, "learning_rate": 1.4308565220015693e-05, "loss": 1.4355, "step": 4748 }, { "epoch": 0.29, "learning_rate": 1.4314601315868897e-05, "loss": 1.227, "step": 4750 }, { "epoch": 0.29, "learning_rate": 1.4320637411722098e-05, "loss": 1.2212, "step": 4752 }, { "epoch": 0.29, "learning_rate": 1.43266735075753e-05, "loss": 1.34, "step": 4754 }, { "epoch": 0.29, "learning_rate": 1.4332709603428505e-05, "loss": 1.0712, "step": 4756 }, { "epoch": 0.29, "learning_rate": 1.4338745699281706e-05, "loss": 1.4881, "step": 4758 }, { "epoch": 0.29, "learning_rate": 1.4344781795134907e-05, "loss": 1.3744, "step": 4760 }, { "epoch": 0.29, "learning_rate": 1.435081789098811e-05, "loss": 1.2481, "step": 4762 }, { "epoch": 0.29, "learning_rate": 1.4356853986841312e-05, "loss": 1.1498, "step": 4764 }, { "epoch": 0.29, "learning_rate": 1.4362890082694514e-05, "loss": 1.5518, "step": 4766 }, { "epoch": 0.29, "learning_rate": 1.4368926178547715e-05, "loss": 1.2804, "step": 4768 }, { "epoch": 0.29, "learning_rate": 1.4374962274400919e-05, "loss": 1.1757, "step": 4770 }, { "epoch": 0.29, "learning_rate": 1.438099837025412e-05, "loss": 1.3833, "step": 4772 }, { "epoch": 0.29, "learning_rate": 1.438703446610732e-05, "loss": 1.1082, "step": 4774 }, { "epoch": 0.29, "learning_rate": 1.4393070561960525e-05, "loss": 1.3934, "step": 4776 }, { "epoch": 0.29, "learning_rate": 1.4399106657813727e-05, "loss": 1.32, "step": 4778 }, { "epoch": 0.29, "learning_rate": 1.4405142753666928e-05, "loss": 1.5184, "step": 4780 }, { "epoch": 0.29, "learning_rate": 1.4411178849520132e-05, "loss": 1.0944, "step": 4782 }, { "epoch": 0.29, "learning_rate": 1.4417214945373333e-05, "loss": 1.5676, "step": 4784 }, { "epoch": 0.29, "learning_rate": 1.4423251041226534e-05, "loss": 1.4347, "step": 4786 }, { "epoch": 0.29, "learning_rate": 1.4429287137079738e-05, "loss": 1.4061, "step": 4788 }, { "epoch": 0.29, "learning_rate": 1.443532323293294e-05, "loss": 1.2853, "step": 4790 }, { "epoch": 0.29, "learning_rate": 1.4441359328786142e-05, "loss": 0.9585, "step": 4792 }, { "epoch": 0.29, "learning_rate": 1.4447395424639342e-05, "loss": 1.7423, "step": 4794 }, { "epoch": 0.29, "learning_rate": 1.4453431520492547e-05, "loss": 1.3783, "step": 4796 }, { "epoch": 0.29, "learning_rate": 1.4459467616345747e-05, "loss": 1.2009, "step": 4798 }, { "epoch": 0.29, "learning_rate": 1.446550371219895e-05, "loss": 0.9288, "step": 4800 }, { "epoch": 0.29, "learning_rate": 1.4471539808052154e-05, "loss": 0.9516, "step": 4802 }, { "epoch": 0.29, "learning_rate": 1.4477575903905355e-05, "loss": 0.9965, "step": 4804 }, { "epoch": 0.29, "learning_rate": 1.4483611999758556e-05, "loss": 1.2276, "step": 4806 }, { "epoch": 0.29, "learning_rate": 1.448964809561176e-05, "loss": 1.6888, "step": 4808 }, { "epoch": 0.29, "learning_rate": 1.4495684191464961e-05, "loss": 1.2213, "step": 4810 }, { "epoch": 0.29, "learning_rate": 1.4501720287318163e-05, "loss": 1.3836, "step": 4812 }, { "epoch": 0.29, "learning_rate": 1.4507756383171364e-05, "loss": 1.0345, "step": 4814 }, { "epoch": 0.29, "learning_rate": 1.4513792479024568e-05, "loss": 1.1967, "step": 4816 }, { "epoch": 0.29, "learning_rate": 1.451982857487777e-05, "loss": 1.1886, "step": 4818 }, { "epoch": 0.29, "learning_rate": 1.4525864670730972e-05, "loss": 1.6932, "step": 4820 }, { "epoch": 0.29, "learning_rate": 1.4531900766584174e-05, "loss": 1.7379, "step": 4822 }, { "epoch": 0.29, "learning_rate": 1.4537936862437377e-05, "loss": 1.253, "step": 4824 }, { "epoch": 0.29, "learning_rate": 1.4543972958290578e-05, "loss": 1.3369, "step": 4826 }, { "epoch": 0.29, "learning_rate": 1.4550009054143782e-05, "loss": 1.0765, "step": 4828 }, { "epoch": 0.29, "learning_rate": 1.4556045149996983e-05, "loss": 1.3021, "step": 4830 }, { "epoch": 0.29, "learning_rate": 1.4562081245850185e-05, "loss": 1.3413, "step": 4832 }, { "epoch": 0.29, "learning_rate": 1.4568117341703386e-05, "loss": 1.2645, "step": 4834 }, { "epoch": 0.29, "learning_rate": 1.457415343755659e-05, "loss": 1.4206, "step": 4836 }, { "epoch": 0.29, "learning_rate": 1.4580189533409791e-05, "loss": 1.4247, "step": 4838 }, { "epoch": 0.29, "learning_rate": 1.4586225629262992e-05, "loss": 1.4997, "step": 4840 }, { "epoch": 0.29, "learning_rate": 1.4592261725116196e-05, "loss": 1.5148, "step": 4842 }, { "epoch": 0.29, "learning_rate": 1.4598297820969398e-05, "loss": 1.6282, "step": 4844 }, { "epoch": 0.29, "learning_rate": 1.46043339168226e-05, "loss": 1.078, "step": 4846 }, { "epoch": 0.29, "learning_rate": 1.4610370012675804e-05, "loss": 1.2036, "step": 4848 }, { "epoch": 0.29, "learning_rate": 1.4616406108529004e-05, "loss": 1.2553, "step": 4850 }, { "epoch": 0.29, "learning_rate": 1.4622442204382205e-05, "loss": 1.2238, "step": 4852 }, { "epoch": 0.29, "learning_rate": 1.462847830023541e-05, "loss": 1.1808, "step": 4854 }, { "epoch": 0.29, "learning_rate": 1.4634514396088612e-05, "loss": 1.5883, "step": 4856 }, { "epoch": 0.29, "learning_rate": 1.4640550491941813e-05, "loss": 1.1732, "step": 4858 }, { "epoch": 0.29, "learning_rate": 1.4646586587795014e-05, "loss": 1.1253, "step": 4860 }, { "epoch": 0.29, "learning_rate": 1.4652622683648218e-05, "loss": 1.5576, "step": 4862 }, { "epoch": 0.29, "learning_rate": 1.4658658779501419e-05, "loss": 1.3806, "step": 4864 }, { "epoch": 0.29, "learning_rate": 1.4664694875354621e-05, "loss": 1.1599, "step": 4866 }, { "epoch": 0.29, "learning_rate": 1.4670730971207825e-05, "loss": 0.9394, "step": 4868 }, { "epoch": 0.29, "learning_rate": 1.4676767067061026e-05, "loss": 1.2976, "step": 4870 }, { "epoch": 0.29, "learning_rate": 1.4682803162914227e-05, "loss": 1.0447, "step": 4872 }, { "epoch": 0.29, "learning_rate": 1.4688839258767431e-05, "loss": 1.8121, "step": 4874 }, { "epoch": 0.29, "learning_rate": 1.4694875354620632e-05, "loss": 1.377, "step": 4876 }, { "epoch": 0.29, "learning_rate": 1.4700911450473834e-05, "loss": 1.605, "step": 4878 }, { "epoch": 0.29, "learning_rate": 1.4706947546327035e-05, "loss": 1.198, "step": 4880 }, { "epoch": 0.29, "learning_rate": 1.471298364218024e-05, "loss": 1.1938, "step": 4882 }, { "epoch": 0.29, "learning_rate": 1.471901973803344e-05, "loss": 1.2462, "step": 4884 }, { "epoch": 0.29, "learning_rate": 1.4725055833886641e-05, "loss": 1.6968, "step": 4886 }, { "epoch": 0.3, "learning_rate": 1.4731091929739845e-05, "loss": 1.3078, "step": 4888 }, { "epoch": 0.3, "learning_rate": 1.4737128025593048e-05, "loss": 1.3704, "step": 4890 }, { "epoch": 0.3, "learning_rate": 1.4743164121446249e-05, "loss": 1.1797, "step": 4892 }, { "epoch": 0.3, "learning_rate": 1.4749200217299453e-05, "loss": 1.478, "step": 4894 }, { "epoch": 0.3, "learning_rate": 1.4755236313152654e-05, "loss": 1.5767, "step": 4896 }, { "epoch": 0.3, "learning_rate": 1.4761272409005854e-05, "loss": 1.564, "step": 4898 }, { "epoch": 0.3, "learning_rate": 1.4767308504859059e-05, "loss": 1.2935, "step": 4900 }, { "epoch": 0.3, "learning_rate": 1.4773344600712261e-05, "loss": 1.4437, "step": 4902 }, { "epoch": 0.3, "learning_rate": 1.4779380696565462e-05, "loss": 1.4226, "step": 4904 }, { "epoch": 0.3, "learning_rate": 1.4785416792418663e-05, "loss": 1.4423, "step": 4906 }, { "epoch": 0.3, "learning_rate": 1.4791452888271867e-05, "loss": 1.2185, "step": 4908 }, { "epoch": 0.3, "learning_rate": 1.4797488984125068e-05, "loss": 1.1877, "step": 4910 }, { "epoch": 0.3, "learning_rate": 1.480352507997827e-05, "loss": 1.2424, "step": 4912 }, { "epoch": 0.3, "learning_rate": 1.4809561175831475e-05, "loss": 1.4043, "step": 4914 }, { "epoch": 0.3, "learning_rate": 1.4815597271684675e-05, "loss": 1.2852, "step": 4916 }, { "epoch": 0.3, "learning_rate": 1.4821633367537876e-05, "loss": 1.2747, "step": 4918 }, { "epoch": 0.3, "learning_rate": 1.482766946339108e-05, "loss": 1.224, "step": 4920 }, { "epoch": 0.3, "learning_rate": 1.4833705559244281e-05, "loss": 1.3024, "step": 4922 }, { "epoch": 0.3, "learning_rate": 1.4839741655097484e-05, "loss": 1.1062, "step": 4924 }, { "epoch": 0.3, "learning_rate": 1.4845777750950685e-05, "loss": 1.0297, "step": 4926 }, { "epoch": 0.3, "learning_rate": 1.4851813846803889e-05, "loss": 1.4315, "step": 4928 }, { "epoch": 0.3, "learning_rate": 1.485784994265709e-05, "loss": 1.518, "step": 4930 }, { "epoch": 0.3, "learning_rate": 1.4863886038510292e-05, "loss": 1.4706, "step": 4932 }, { "epoch": 0.3, "learning_rate": 1.4869922134363495e-05, "loss": 1.4892, "step": 4934 }, { "epoch": 0.3, "learning_rate": 1.4875958230216697e-05, "loss": 1.2296, "step": 4936 }, { "epoch": 0.3, "learning_rate": 1.4881994326069898e-05, "loss": 1.1583, "step": 4938 }, { "epoch": 0.3, "learning_rate": 1.4888030421923102e-05, "loss": 1.0109, "step": 4940 }, { "epoch": 0.3, "learning_rate": 1.4894066517776303e-05, "loss": 1.3204, "step": 4942 }, { "epoch": 0.3, "learning_rate": 1.4900102613629505e-05, "loss": 1.3926, "step": 4944 }, { "epoch": 0.3, "learning_rate": 1.4906138709482706e-05, "loss": 0.9674, "step": 4946 }, { "epoch": 0.3, "learning_rate": 1.491217480533591e-05, "loss": 1.5004, "step": 4948 }, { "epoch": 0.3, "learning_rate": 1.4918210901189111e-05, "loss": 1.1782, "step": 4950 }, { "epoch": 0.3, "learning_rate": 1.4924246997042312e-05, "loss": 1.1822, "step": 4952 }, { "epoch": 0.3, "learning_rate": 1.4930283092895516e-05, "loss": 1.1623, "step": 4954 }, { "epoch": 0.3, "learning_rate": 1.4936319188748719e-05, "loss": 1.6452, "step": 4956 }, { "epoch": 0.3, "learning_rate": 1.494235528460192e-05, "loss": 1.1153, "step": 4958 }, { "epoch": 0.3, "learning_rate": 1.4948391380455124e-05, "loss": 1.3192, "step": 4960 }, { "epoch": 0.3, "learning_rate": 1.4954427476308325e-05, "loss": 1.0272, "step": 4962 }, { "epoch": 0.3, "learning_rate": 1.4960463572161526e-05, "loss": 1.309, "step": 4964 }, { "epoch": 0.3, "learning_rate": 1.496649966801473e-05, "loss": 1.4377, "step": 4966 }, { "epoch": 0.3, "learning_rate": 1.4972535763867932e-05, "loss": 1.1605, "step": 4968 }, { "epoch": 0.3, "learning_rate": 1.4978571859721133e-05, "loss": 1.2221, "step": 4970 }, { "epoch": 0.3, "learning_rate": 1.4984607955574334e-05, "loss": 1.1343, "step": 4972 }, { "epoch": 0.3, "learning_rate": 1.4990644051427538e-05, "loss": 1.2244, "step": 4974 }, { "epoch": 0.3, "learning_rate": 1.4996680147280739e-05, "loss": 1.2848, "step": 4976 }, { "epoch": 0.3, "learning_rate": 1.5002716243133941e-05, "loss": 1.2609, "step": 4978 }, { "epoch": 0.3, "learning_rate": 1.5008752338987146e-05, "loss": 1.3182, "step": 4980 }, { "epoch": 0.3, "learning_rate": 1.5014788434840346e-05, "loss": 1.1834, "step": 4982 }, { "epoch": 0.3, "learning_rate": 1.5020824530693547e-05, "loss": 1.4097, "step": 4984 }, { "epoch": 0.3, "learning_rate": 1.5026860626546751e-05, "loss": 1.2401, "step": 4986 }, { "epoch": 0.3, "learning_rate": 1.5032896722399952e-05, "loss": 1.39, "step": 4988 }, { "epoch": 0.3, "learning_rate": 1.5038932818253155e-05, "loss": 1.2079, "step": 4990 }, { "epoch": 0.3, "learning_rate": 1.5044968914106356e-05, "loss": 1.4649, "step": 4992 }, { "epoch": 0.3, "learning_rate": 1.505100500995956e-05, "loss": 1.4085, "step": 4994 }, { "epoch": 0.3, "learning_rate": 1.505704110581276e-05, "loss": 1.0034, "step": 4996 }, { "epoch": 0.3, "learning_rate": 1.5063077201665961e-05, "loss": 1.1527, "step": 4998 }, { "epoch": 0.3, "learning_rate": 1.5069113297519166e-05, "loss": 1.1363, "step": 5000 }, { "epoch": 0.3, "learning_rate": 1.5075149393372368e-05, "loss": 1.3143, "step": 5002 }, { "epoch": 0.3, "learning_rate": 1.5081185489225569e-05, "loss": 1.8935, "step": 5004 }, { "epoch": 0.3, "learning_rate": 1.5087221585078773e-05, "loss": 1.4853, "step": 5006 }, { "epoch": 0.3, "learning_rate": 1.5093257680931974e-05, "loss": 1.287, "step": 5008 }, { "epoch": 0.3, "learning_rate": 1.5099293776785175e-05, "loss": 1.2889, "step": 5010 }, { "epoch": 0.3, "learning_rate": 1.5105329872638377e-05, "loss": 1.6199, "step": 5012 }, { "epoch": 0.3, "learning_rate": 1.5111365968491582e-05, "loss": 1.5996, "step": 5014 }, { "epoch": 0.3, "learning_rate": 1.5117402064344782e-05, "loss": 1.1764, "step": 5016 }, { "epoch": 0.3, "learning_rate": 1.5123438160197983e-05, "loss": 1.2007, "step": 5018 }, { "epoch": 0.3, "learning_rate": 1.5129474256051187e-05, "loss": 1.5688, "step": 5020 }, { "epoch": 0.3, "learning_rate": 1.5135510351904388e-05, "loss": 1.3877, "step": 5022 }, { "epoch": 0.3, "learning_rate": 1.514154644775759e-05, "loss": 1.3982, "step": 5024 }, { "epoch": 0.3, "learning_rate": 1.5147582543610795e-05, "loss": 1.2615, "step": 5026 }, { "epoch": 0.3, "learning_rate": 1.5153618639463996e-05, "loss": 1.4956, "step": 5028 }, { "epoch": 0.3, "learning_rate": 1.5159654735317197e-05, "loss": 1.1903, "step": 5030 }, { "epoch": 0.3, "learning_rate": 1.51656908311704e-05, "loss": 1.8769, "step": 5032 }, { "epoch": 0.3, "learning_rate": 1.5171726927023602e-05, "loss": 1.1687, "step": 5034 }, { "epoch": 0.3, "learning_rate": 1.5177763022876804e-05, "loss": 1.3404, "step": 5036 }, { "epoch": 0.3, "learning_rate": 1.5183799118730005e-05, "loss": 1.1657, "step": 5038 }, { "epoch": 0.3, "learning_rate": 1.518983521458321e-05, "loss": 1.2684, "step": 5040 }, { "epoch": 0.3, "learning_rate": 1.519587131043641e-05, "loss": 1.2049, "step": 5042 }, { "epoch": 0.3, "learning_rate": 1.520190740628961e-05, "loss": 1.2935, "step": 5044 }, { "epoch": 0.3, "learning_rate": 1.5207943502142815e-05, "loss": 1.0716, "step": 5046 }, { "epoch": 0.3, "learning_rate": 1.5213979597996018e-05, "loss": 1.2851, "step": 5048 }, { "epoch": 0.3, "learning_rate": 1.5220015693849218e-05, "loss": 1.2453, "step": 5050 }, { "epoch": 0.3, "learning_rate": 1.5226051789702423e-05, "loss": 1.5928, "step": 5052 }, { "epoch": 0.31, "learning_rate": 1.5232087885555623e-05, "loss": 1.3351, "step": 5054 }, { "epoch": 0.31, "learning_rate": 1.5238123981408826e-05, "loss": 1.4993, "step": 5056 }, { "epoch": 0.31, "learning_rate": 1.5244160077262027e-05, "loss": 1.2115, "step": 5058 }, { "epoch": 0.31, "learning_rate": 1.5250196173115231e-05, "loss": 1.7222, "step": 5060 }, { "epoch": 0.31, "learning_rate": 1.5256232268968432e-05, "loss": 1.5563, "step": 5062 }, { "epoch": 0.31, "learning_rate": 1.5262268364821634e-05, "loss": 1.6354, "step": 5064 }, { "epoch": 0.31, "learning_rate": 1.526830446067484e-05, "loss": 1.7598, "step": 5066 }, { "epoch": 0.31, "learning_rate": 1.527434055652804e-05, "loss": 1.1466, "step": 5068 }, { "epoch": 0.31, "learning_rate": 1.528037665238124e-05, "loss": 1.1781, "step": 5070 }, { "epoch": 0.31, "learning_rate": 1.5286412748234444e-05, "loss": 1.6184, "step": 5072 }, { "epoch": 0.31, "learning_rate": 1.5292448844087645e-05, "loss": 1.0779, "step": 5074 }, { "epoch": 0.31, "learning_rate": 1.5298484939940846e-05, "loss": 1.6894, "step": 5076 }, { "epoch": 0.31, "learning_rate": 1.530452103579405e-05, "loss": 1.2537, "step": 5078 }, { "epoch": 0.31, "learning_rate": 1.531055713164725e-05, "loss": 1.2205, "step": 5080 }, { "epoch": 0.31, "learning_rate": 1.5316593227500452e-05, "loss": 1.2111, "step": 5082 }, { "epoch": 0.31, "learning_rate": 1.5322629323353656e-05, "loss": 1.2465, "step": 5084 }, { "epoch": 0.31, "learning_rate": 1.532866541920686e-05, "loss": 1.2633, "step": 5086 }, { "epoch": 0.31, "learning_rate": 1.533470151506006e-05, "loss": 1.3012, "step": 5088 }, { "epoch": 0.31, "learning_rate": 1.5340737610913262e-05, "loss": 1.7363, "step": 5090 }, { "epoch": 0.31, "learning_rate": 1.5346773706766466e-05, "loss": 1.1078, "step": 5092 }, { "epoch": 0.31, "learning_rate": 1.5352809802619667e-05, "loss": 1.4449, "step": 5094 }, { "epoch": 0.31, "learning_rate": 1.5358845898472868e-05, "loss": 1.3174, "step": 5096 }, { "epoch": 0.31, "learning_rate": 1.5364881994326072e-05, "loss": 1.2293, "step": 5098 }, { "epoch": 0.31, "learning_rate": 1.5370918090179273e-05, "loss": 1.4359, "step": 5100 }, { "epoch": 0.31, "learning_rate": 1.5376954186032473e-05, "loss": 1.3507, "step": 5102 }, { "epoch": 0.31, "learning_rate": 1.5382990281885674e-05, "loss": 1.1911, "step": 5104 }, { "epoch": 0.31, "learning_rate": 1.538902637773888e-05, "loss": 1.4087, "step": 5106 }, { "epoch": 0.31, "learning_rate": 1.5395062473592083e-05, "loss": 1.2025, "step": 5108 }, { "epoch": 0.31, "learning_rate": 1.5401098569445284e-05, "loss": 1.2311, "step": 5110 }, { "epoch": 0.31, "learning_rate": 1.5407134665298488e-05, "loss": 1.3657, "step": 5112 }, { "epoch": 0.31, "learning_rate": 1.541317076115169e-05, "loss": 1.125, "step": 5114 }, { "epoch": 0.31, "learning_rate": 1.541920685700489e-05, "loss": 1.1239, "step": 5116 }, { "epoch": 0.31, "learning_rate": 1.5425242952858094e-05, "loss": 1.3764, "step": 5118 }, { "epoch": 0.31, "learning_rate": 1.5431279048711294e-05, "loss": 0.8009, "step": 5120 }, { "epoch": 0.31, "learning_rate": 1.5437315144564495e-05, "loss": 2.2573, "step": 5122 }, { "epoch": 0.31, "learning_rate": 1.5443351240417696e-05, "loss": 1.327, "step": 5124 }, { "epoch": 0.31, "learning_rate": 1.54493873362709e-05, "loss": 1.485, "step": 5126 }, { "epoch": 0.31, "learning_rate": 1.54554234321241e-05, "loss": 1.1424, "step": 5128 }, { "epoch": 0.31, "learning_rate": 1.5461459527977305e-05, "loss": 1.8363, "step": 5130 }, { "epoch": 0.31, "learning_rate": 1.546749562383051e-05, "loss": 0.8483, "step": 5132 }, { "epoch": 0.31, "learning_rate": 1.547353171968371e-05, "loss": 1.0445, "step": 5134 }, { "epoch": 0.31, "learning_rate": 1.547956781553691e-05, "loss": 1.161, "step": 5136 }, { "epoch": 0.31, "learning_rate": 1.5485603911390115e-05, "loss": 1.5613, "step": 5138 }, { "epoch": 0.31, "learning_rate": 1.5491640007243316e-05, "loss": 1.226, "step": 5140 }, { "epoch": 0.31, "learning_rate": 1.5497676103096517e-05, "loss": 1.2032, "step": 5142 }, { "epoch": 0.31, "learning_rate": 1.550371219894972e-05, "loss": 1.365, "step": 5144 }, { "epoch": 0.31, "learning_rate": 1.5509748294802922e-05, "loss": 1.5767, "step": 5146 }, { "epoch": 0.31, "learning_rate": 1.5515784390656123e-05, "loss": 1.0227, "step": 5148 }, { "epoch": 0.31, "learning_rate": 1.5521820486509324e-05, "loss": 1.461, "step": 5150 }, { "epoch": 0.31, "learning_rate": 1.5527856582362528e-05, "loss": 1.286, "step": 5152 }, { "epoch": 0.31, "learning_rate": 1.5533892678215732e-05, "loss": 1.1339, "step": 5154 }, { "epoch": 0.31, "learning_rate": 1.5539928774068933e-05, "loss": 1.5894, "step": 5156 }, { "epoch": 0.31, "learning_rate": 1.5545964869922137e-05, "loss": 1.3423, "step": 5158 }, { "epoch": 0.31, "learning_rate": 1.5552000965775338e-05, "loss": 1.3439, "step": 5160 }, { "epoch": 0.31, "learning_rate": 1.555803706162854e-05, "loss": 1.8245, "step": 5162 }, { "epoch": 0.31, "learning_rate": 1.5564073157481743e-05, "loss": 1.1265, "step": 5164 }, { "epoch": 0.31, "learning_rate": 1.5570109253334944e-05, "loss": 1.0569, "step": 5166 }, { "epoch": 0.31, "learning_rate": 1.5576145349188145e-05, "loss": 1.2538, "step": 5168 }, { "epoch": 0.31, "learning_rate": 1.5582181445041345e-05, "loss": 1.4422, "step": 5170 }, { "epoch": 0.31, "learning_rate": 1.558821754089455e-05, "loss": 1.1067, "step": 5172 }, { "epoch": 0.31, "learning_rate": 1.559425363674775e-05, "loss": 1.255, "step": 5174 }, { "epoch": 0.31, "learning_rate": 1.5600289732600955e-05, "loss": 1.1313, "step": 5176 }, { "epoch": 0.31, "learning_rate": 1.560632582845416e-05, "loss": 1.507, "step": 5178 }, { "epoch": 0.31, "learning_rate": 1.561236192430736e-05, "loss": 1.6402, "step": 5180 }, { "epoch": 0.31, "learning_rate": 1.561839802016056e-05, "loss": 1.4152, "step": 5182 }, { "epoch": 0.31, "learning_rate": 1.5624434116013765e-05, "loss": 1.0532, "step": 5184 }, { "epoch": 0.31, "learning_rate": 1.5630470211866965e-05, "loss": 1.4956, "step": 5186 }, { "epoch": 0.31, "learning_rate": 1.5636506307720166e-05, "loss": 1.2717, "step": 5188 }, { "epoch": 0.31, "learning_rate": 1.564254240357337e-05, "loss": 1.3794, "step": 5190 }, { "epoch": 0.31, "learning_rate": 1.564857849942657e-05, "loss": 1.5881, "step": 5192 }, { "epoch": 0.31, "learning_rate": 1.5654614595279772e-05, "loss": 1.4751, "step": 5194 }, { "epoch": 0.31, "learning_rate": 1.5660650691132976e-05, "loss": 1.5195, "step": 5196 }, { "epoch": 0.31, "learning_rate": 1.566668678698618e-05, "loss": 1.4792, "step": 5198 }, { "epoch": 0.31, "learning_rate": 1.567272288283938e-05, "loss": 1.3013, "step": 5200 }, { "epoch": 0.31, "learning_rate": 1.5678758978692582e-05, "loss": 1.4163, "step": 5202 }, { "epoch": 0.31, "learning_rate": 1.5684795074545786e-05, "loss": 1.3954, "step": 5204 }, { "epoch": 0.31, "learning_rate": 1.5690831170398987e-05, "loss": 1.1196, "step": 5206 }, { "epoch": 0.31, "learning_rate": 1.5696867266252188e-05, "loss": 1.2158, "step": 5208 }, { "epoch": 0.31, "learning_rate": 1.5702903362105392e-05, "loss": 1.4229, "step": 5210 }, { "epoch": 0.31, "learning_rate": 1.5708939457958593e-05, "loss": 1.3357, "step": 5212 }, { "epoch": 0.31, "learning_rate": 1.5714975553811794e-05, "loss": 1.2755, "step": 5214 }, { "epoch": 0.31, "learning_rate": 1.5721011649664995e-05, "loss": 1.685, "step": 5216 }, { "epoch": 0.31, "learning_rate": 1.57270477455182e-05, "loss": 1.3459, "step": 5218 }, { "epoch": 0.32, "learning_rate": 1.5733083841371403e-05, "loss": 1.4051, "step": 5220 }, { "epoch": 0.32, "learning_rate": 1.5739119937224604e-05, "loss": 1.5137, "step": 5222 }, { "epoch": 0.32, "learning_rate": 1.5745156033077808e-05, "loss": 1.5213, "step": 5224 }, { "epoch": 0.32, "learning_rate": 1.575119212893101e-05, "loss": 1.1779, "step": 5226 }, { "epoch": 0.32, "learning_rate": 1.575722822478421e-05, "loss": 1.2285, "step": 5228 }, { "epoch": 0.32, "learning_rate": 1.5763264320637414e-05, "loss": 1.5594, "step": 5230 }, { "epoch": 0.32, "learning_rate": 1.5769300416490615e-05, "loss": 1.1386, "step": 5232 }, { "epoch": 0.32, "learning_rate": 1.5775336512343816e-05, "loss": 1.1233, "step": 5234 }, { "epoch": 0.32, "learning_rate": 1.5781372608197016e-05, "loss": 1.6705, "step": 5236 }, { "epoch": 0.32, "learning_rate": 1.578740870405022e-05, "loss": 1.273, "step": 5238 }, { "epoch": 0.32, "learning_rate": 1.579344479990342e-05, "loss": 1.1709, "step": 5240 }, { "epoch": 0.32, "learning_rate": 1.5799480895756626e-05, "loss": 1.359, "step": 5242 }, { "epoch": 0.32, "learning_rate": 1.580551699160983e-05, "loss": 1.1935, "step": 5244 }, { "epoch": 0.32, "learning_rate": 1.581155308746303e-05, "loss": 1.127, "step": 5246 }, { "epoch": 0.32, "learning_rate": 1.581758918331623e-05, "loss": 1.3111, "step": 5248 }, { "epoch": 0.32, "learning_rate": 1.5823625279169436e-05, "loss": 1.5482, "step": 5250 }, { "epoch": 0.32, "learning_rate": 1.5829661375022637e-05, "loss": 1.2753, "step": 5252 }, { "epoch": 0.32, "learning_rate": 1.5835697470875837e-05, "loss": 1.4209, "step": 5254 }, { "epoch": 0.32, "learning_rate": 1.584173356672904e-05, "loss": 1.126, "step": 5256 }, { "epoch": 0.32, "learning_rate": 1.5847769662582242e-05, "loss": 1.4035, "step": 5258 }, { "epoch": 0.32, "learning_rate": 1.5853805758435443e-05, "loss": 1.2472, "step": 5260 }, { "epoch": 0.32, "learning_rate": 1.5859841854288644e-05, "loss": 1.6055, "step": 5262 }, { "epoch": 0.32, "learning_rate": 1.5865877950141848e-05, "loss": 1.4155, "step": 5264 }, { "epoch": 0.32, "learning_rate": 1.5871914045995052e-05, "loss": 1.008, "step": 5266 }, { "epoch": 0.32, "learning_rate": 1.5877950141848253e-05, "loss": 1.4811, "step": 5268 }, { "epoch": 0.32, "learning_rate": 1.5883986237701457e-05, "loss": 0.879, "step": 5270 }, { "epoch": 0.32, "learning_rate": 1.5890022333554658e-05, "loss": 1.5044, "step": 5272 }, { "epoch": 0.32, "learning_rate": 1.589605842940786e-05, "loss": 1.3258, "step": 5274 }, { "epoch": 0.32, "learning_rate": 1.5902094525261063e-05, "loss": 1.2728, "step": 5276 }, { "epoch": 0.32, "learning_rate": 1.5908130621114264e-05, "loss": 1.6241, "step": 5278 }, { "epoch": 0.32, "learning_rate": 1.5914166716967465e-05, "loss": 1.1592, "step": 5280 }, { "epoch": 0.32, "learning_rate": 1.5920202812820666e-05, "loss": 1.6104, "step": 5282 }, { "epoch": 0.32, "learning_rate": 1.592623890867387e-05, "loss": 1.0683, "step": 5284 }, { "epoch": 0.32, "learning_rate": 1.593227500452707e-05, "loss": 1.8479, "step": 5286 }, { "epoch": 0.32, "learning_rate": 1.5938311100380275e-05, "loss": 1.2161, "step": 5288 }, { "epoch": 0.32, "learning_rate": 1.594434719623348e-05, "loss": 1.2097, "step": 5290 }, { "epoch": 0.32, "learning_rate": 1.595038329208668e-05, "loss": 1.1045, "step": 5292 }, { "epoch": 0.32, "learning_rate": 1.595641938793988e-05, "loss": 1.2838, "step": 5294 }, { "epoch": 0.32, "learning_rate": 1.5962455483793085e-05, "loss": 1.1923, "step": 5296 }, { "epoch": 0.32, "learning_rate": 1.5968491579646286e-05, "loss": 1.0508, "step": 5298 }, { "epoch": 0.32, "learning_rate": 1.5974527675499487e-05, "loss": 1.1152, "step": 5300 }, { "epoch": 0.32, "learning_rate": 1.598056377135269e-05, "loss": 1.4094, "step": 5302 }, { "epoch": 0.32, "learning_rate": 1.5986599867205892e-05, "loss": 1.1715, "step": 5304 }, { "epoch": 0.32, "learning_rate": 1.5992635963059093e-05, "loss": 1.2775, "step": 5306 }, { "epoch": 0.32, "learning_rate": 1.5998672058912297e-05, "loss": 1.3976, "step": 5308 }, { "epoch": 0.32, "learning_rate": 1.6004708154765498e-05, "loss": 1.4497, "step": 5310 }, { "epoch": 0.32, "learning_rate": 1.6010744250618702e-05, "loss": 1.2426, "step": 5312 }, { "epoch": 0.32, "learning_rate": 1.6016780346471903e-05, "loss": 1.0923, "step": 5314 }, { "epoch": 0.32, "learning_rate": 1.6022816442325107e-05, "loss": 1.2349, "step": 5316 }, { "epoch": 0.32, "learning_rate": 1.6028852538178308e-05, "loss": 1.6057, "step": 5318 }, { "epoch": 0.32, "learning_rate": 1.603488863403151e-05, "loss": 1.3405, "step": 5320 }, { "epoch": 0.32, "learning_rate": 1.6040924729884713e-05, "loss": 1.3521, "step": 5322 }, { "epoch": 0.32, "learning_rate": 1.6046960825737913e-05, "loss": 1.4098, "step": 5324 }, { "epoch": 0.32, "learning_rate": 1.6052996921591114e-05, "loss": 1.3702, "step": 5326 }, { "epoch": 0.32, "learning_rate": 1.6059033017444315e-05, "loss": 1.3464, "step": 5328 }, { "epoch": 0.32, "learning_rate": 1.606506911329752e-05, "loss": 1.3721, "step": 5330 }, { "epoch": 0.32, "learning_rate": 1.6071105209150723e-05, "loss": 0.9593, "step": 5332 }, { "epoch": 0.32, "learning_rate": 1.6077141305003924e-05, "loss": 0.8434, "step": 5334 }, { "epoch": 0.32, "learning_rate": 1.608317740085713e-05, "loss": 1.525, "step": 5336 }, { "epoch": 0.32, "learning_rate": 1.608921349671033e-05, "loss": 1.3845, "step": 5338 }, { "epoch": 0.32, "learning_rate": 1.609524959256353e-05, "loss": 1.3689, "step": 5340 }, { "epoch": 0.32, "learning_rate": 1.6101285688416734e-05, "loss": 1.3557, "step": 5342 }, { "epoch": 0.32, "learning_rate": 1.6107321784269935e-05, "loss": 1.2504, "step": 5344 }, { "epoch": 0.32, "learning_rate": 1.6113357880123136e-05, "loss": 1.4686, "step": 5346 }, { "epoch": 0.32, "learning_rate": 1.6119393975976337e-05, "loss": 1.1849, "step": 5348 }, { "epoch": 0.32, "learning_rate": 1.612543007182954e-05, "loss": 1.5837, "step": 5350 }, { "epoch": 0.32, "learning_rate": 1.6131466167682742e-05, "loss": 1.4081, "step": 5352 }, { "epoch": 0.32, "learning_rate": 1.6137502263535946e-05, "loss": 0.9146, "step": 5354 }, { "epoch": 0.32, "learning_rate": 1.614353835938915e-05, "loss": 1.1661, "step": 5356 }, { "epoch": 0.32, "learning_rate": 1.614957445524235e-05, "loss": 1.1623, "step": 5358 }, { "epoch": 0.32, "learning_rate": 1.6155610551095552e-05, "loss": 0.8722, "step": 5360 }, { "epoch": 0.32, "learning_rate": 1.6161646646948756e-05, "loss": 1.2835, "step": 5362 }, { "epoch": 0.32, "learning_rate": 1.6167682742801957e-05, "loss": 1.0045, "step": 5364 }, { "epoch": 0.32, "learning_rate": 1.6173718838655158e-05, "loss": 1.1307, "step": 5366 }, { "epoch": 0.32, "learning_rate": 1.6179754934508362e-05, "loss": 1.3174, "step": 5368 }, { "epoch": 0.32, "learning_rate": 1.6185791030361563e-05, "loss": 1.1932, "step": 5370 }, { "epoch": 0.32, "learning_rate": 1.6191827126214764e-05, "loss": 1.3972, "step": 5372 }, { "epoch": 0.32, "learning_rate": 1.6197863222067964e-05, "loss": 1.0453, "step": 5374 }, { "epoch": 0.32, "learning_rate": 1.620389931792117e-05, "loss": 1.1394, "step": 5376 }, { "epoch": 0.32, "learning_rate": 1.6209935413774373e-05, "loss": 1.3485, "step": 5378 }, { "epoch": 0.32, "learning_rate": 1.6215971509627574e-05, "loss": 1.3851, "step": 5380 }, { "epoch": 0.32, "learning_rate": 1.6222007605480778e-05, "loss": 1.0019, "step": 5382 }, { "epoch": 0.32, "learning_rate": 1.622804370133398e-05, "loss": 1.0814, "step": 5384 }, { "epoch": 0.33, "learning_rate": 1.623407979718718e-05, "loss": 1.0043, "step": 5386 }, { "epoch": 0.33, "learning_rate": 1.6240115893040384e-05, "loss": 0.9262, "step": 5388 }, { "epoch": 0.33, "learning_rate": 1.6246151988893584e-05, "loss": 1.4811, "step": 5390 }, { "epoch": 0.33, "learning_rate": 1.6252188084746785e-05, "loss": 1.2064, "step": 5392 }, { "epoch": 0.33, "learning_rate": 1.6258224180599986e-05, "loss": 1.4178, "step": 5394 }, { "epoch": 0.33, "learning_rate": 1.626426027645319e-05, "loss": 1.3192, "step": 5396 }, { "epoch": 0.33, "learning_rate": 1.627029637230639e-05, "loss": 1.7623, "step": 5398 }, { "epoch": 0.33, "learning_rate": 1.6276332468159595e-05, "loss": 1.5291, "step": 5400 }, { "epoch": 0.33, "learning_rate": 1.62823685640128e-05, "loss": 1.1505, "step": 5402 }, { "epoch": 0.33, "learning_rate": 1.6288404659866e-05, "loss": 1.1713, "step": 5404 }, { "epoch": 0.33, "learning_rate": 1.62944407557192e-05, "loss": 1.4428, "step": 5406 }, { "epoch": 0.33, "learning_rate": 1.6300476851572405e-05, "loss": 1.1305, "step": 5408 }, { "epoch": 0.33, "learning_rate": 1.6306512947425606e-05, "loss": 1.2112, "step": 5410 }, { "epoch": 0.33, "learning_rate": 1.6312549043278807e-05, "loss": 0.9607, "step": 5412 }, { "epoch": 0.33, "learning_rate": 1.6318585139132008e-05, "loss": 1.3175, "step": 5414 }, { "epoch": 0.33, "learning_rate": 1.6324621234985212e-05, "loss": 1.1952, "step": 5416 }, { "epoch": 0.33, "learning_rate": 1.6330657330838413e-05, "loss": 1.0719, "step": 5418 }, { "epoch": 0.33, "learning_rate": 1.6336693426691617e-05, "loss": 1.7934, "step": 5420 }, { "epoch": 0.33, "learning_rate": 1.6342729522544818e-05, "loss": 1.6839, "step": 5422 }, { "epoch": 0.33, "learning_rate": 1.6348765618398022e-05, "loss": 1.1577, "step": 5424 }, { "epoch": 0.33, "learning_rate": 1.6354801714251223e-05, "loss": 1.5373, "step": 5426 }, { "epoch": 0.33, "learning_rate": 1.6360837810104427e-05, "loss": 1.0799, "step": 5428 }, { "epoch": 0.33, "learning_rate": 1.6366873905957628e-05, "loss": 1.2575, "step": 5430 }, { "epoch": 0.33, "learning_rate": 1.637291000181083e-05, "loss": 1.3967, "step": 5432 }, { "epoch": 0.33, "learning_rate": 1.6378946097664033e-05, "loss": 1.2902, "step": 5434 }, { "epoch": 0.33, "learning_rate": 1.6384982193517234e-05, "loss": 1.103, "step": 5436 }, { "epoch": 0.33, "learning_rate": 1.6391018289370435e-05, "loss": 1.3468, "step": 5438 }, { "epoch": 0.33, "learning_rate": 1.6397054385223635e-05, "loss": 1.3651, "step": 5440 }, { "epoch": 0.33, "learning_rate": 1.640309048107684e-05, "loss": 1.5195, "step": 5442 }, { "epoch": 0.33, "learning_rate": 1.6409126576930044e-05, "loss": 1.4016, "step": 5444 }, { "epoch": 0.33, "learning_rate": 1.6415162672783245e-05, "loss": 1.2156, "step": 5446 }, { "epoch": 0.33, "learning_rate": 1.642119876863645e-05, "loss": 1.232, "step": 5448 }, { "epoch": 0.33, "learning_rate": 1.642723486448965e-05, "loss": 1.2177, "step": 5450 }, { "epoch": 0.33, "learning_rate": 1.643327096034285e-05, "loss": 1.2931, "step": 5452 }, { "epoch": 0.33, "learning_rate": 1.6439307056196055e-05, "loss": 1.7073, "step": 5454 }, { "epoch": 0.33, "learning_rate": 1.6445343152049256e-05, "loss": 1.1217, "step": 5456 }, { "epoch": 0.33, "learning_rate": 1.6451379247902456e-05, "loss": 1.1912, "step": 5458 }, { "epoch": 0.33, "learning_rate": 1.6457415343755657e-05, "loss": 1.1081, "step": 5460 }, { "epoch": 0.33, "learning_rate": 1.646345143960886e-05, "loss": 1.0105, "step": 5462 }, { "epoch": 0.33, "learning_rate": 1.6469487535462062e-05, "loss": 1.1694, "step": 5464 }, { "epoch": 0.33, "learning_rate": 1.6475523631315266e-05, "loss": 1.7295, "step": 5466 }, { "epoch": 0.33, "learning_rate": 1.648155972716847e-05, "loss": 1.5453, "step": 5468 }, { "epoch": 0.33, "learning_rate": 1.648759582302167e-05, "loss": 1.1487, "step": 5470 }, { "epoch": 0.33, "learning_rate": 1.6493631918874872e-05, "loss": 1.3077, "step": 5472 }, { "epoch": 0.33, "learning_rate": 1.6499668014728076e-05, "loss": 1.3285, "step": 5474 }, { "epoch": 0.33, "learning_rate": 1.6505704110581277e-05, "loss": 1.1681, "step": 5476 }, { "epoch": 0.33, "learning_rate": 1.6511740206434478e-05, "loss": 1.3207, "step": 5478 }, { "epoch": 0.33, "learning_rate": 1.6517776302287682e-05, "loss": 1.2764, "step": 5480 }, { "epoch": 0.33, "learning_rate": 1.6523812398140883e-05, "loss": 1.295, "step": 5482 }, { "epoch": 0.33, "learning_rate": 1.6529848493994084e-05, "loss": 1.666, "step": 5484 }, { "epoch": 0.33, "learning_rate": 1.6535884589847285e-05, "loss": 1.4463, "step": 5486 }, { "epoch": 0.33, "learning_rate": 1.654192068570049e-05, "loss": 1.2977, "step": 5488 }, { "epoch": 0.33, "learning_rate": 1.6547956781553693e-05, "loss": 1.3003, "step": 5490 }, { "epoch": 0.33, "learning_rate": 1.6553992877406894e-05, "loss": 1.5247, "step": 5492 }, { "epoch": 0.33, "learning_rate": 1.6560028973260098e-05, "loss": 1.5557, "step": 5494 }, { "epoch": 0.33, "learning_rate": 1.65660650691133e-05, "loss": 1.4108, "step": 5496 }, { "epoch": 0.33, "learning_rate": 1.65721011649665e-05, "loss": 1.3133, "step": 5498 }, { "epoch": 0.33, "learning_rate": 1.6578137260819704e-05, "loss": 1.4186, "step": 5500 }, { "epoch": 0.33, "learning_rate": 1.6584173356672905e-05, "loss": 0.9728, "step": 5502 }, { "epoch": 0.33, "learning_rate": 1.6590209452526106e-05, "loss": 1.2995, "step": 5504 }, { "epoch": 0.33, "learning_rate": 1.6596245548379307e-05, "loss": 1.1129, "step": 5506 }, { "epoch": 0.33, "learning_rate": 1.660228164423251e-05, "loss": 1.2985, "step": 5508 }, { "epoch": 0.33, "learning_rate": 1.660831774008571e-05, "loss": 1.0754, "step": 5510 }, { "epoch": 0.33, "learning_rate": 1.6614353835938916e-05, "loss": 1.327, "step": 5512 }, { "epoch": 0.33, "learning_rate": 1.662038993179212e-05, "loss": 0.9863, "step": 5514 }, { "epoch": 0.33, "learning_rate": 1.662642602764532e-05, "loss": 0.9812, "step": 5516 }, { "epoch": 0.33, "learning_rate": 1.663246212349852e-05, "loss": 1.3321, "step": 5518 }, { "epoch": 0.33, "learning_rate": 1.6638498219351726e-05, "loss": 1.2081, "step": 5520 }, { "epoch": 0.33, "learning_rate": 1.6644534315204927e-05, "loss": 1.4576, "step": 5522 }, { "epoch": 0.33, "learning_rate": 1.6650570411058127e-05, "loss": 1.3968, "step": 5524 }, { "epoch": 0.33, "learning_rate": 1.6656606506911328e-05, "loss": 1.3147, "step": 5526 }, { "epoch": 0.33, "learning_rate": 1.6662642602764532e-05, "loss": 1.105, "step": 5528 }, { "epoch": 0.33, "learning_rate": 1.6668678698617733e-05, "loss": 1.3317, "step": 5530 }, { "epoch": 0.33, "learning_rate": 1.6674714794470937e-05, "loss": 1.4394, "step": 5532 }, { "epoch": 0.33, "learning_rate": 1.6680750890324138e-05, "loss": 1.3, "step": 5534 }, { "epoch": 0.33, "learning_rate": 1.6686786986177343e-05, "loss": 1.4955, "step": 5536 }, { "epoch": 0.33, "learning_rate": 1.6692823082030543e-05, "loss": 1.1945, "step": 5538 }, { "epoch": 0.33, "learning_rate": 1.6698859177883748e-05, "loss": 1.236, "step": 5540 }, { "epoch": 0.33, "learning_rate": 1.670489527373695e-05, "loss": 1.301, "step": 5542 }, { "epoch": 0.33, "learning_rate": 1.671093136959015e-05, "loss": 1.5443, "step": 5544 }, { "epoch": 0.33, "learning_rate": 1.6716967465443353e-05, "loss": 0.8461, "step": 5546 }, { "epoch": 0.33, "learning_rate": 1.6723003561296554e-05, "loss": 1.1231, "step": 5548 }, { "epoch": 0.33, "learning_rate": 1.6729039657149755e-05, "loss": 1.844, "step": 5550 }, { "epoch": 0.34, "learning_rate": 1.6735075753002956e-05, "loss": 1.2524, "step": 5552 }, { "epoch": 0.34, "learning_rate": 1.674111184885616e-05, "loss": 1.607, "step": 5554 }, { "epoch": 0.34, "learning_rate": 1.6747147944709364e-05, "loss": 1.2398, "step": 5556 }, { "epoch": 0.34, "learning_rate": 1.6753184040562565e-05, "loss": 1.29, "step": 5558 }, { "epoch": 0.34, "learning_rate": 1.675620208848917e-05, "loss": 1.4103, "step": 5560 }, { "epoch": 0.34, "learning_rate": 1.676223818434237e-05, "loss": 1.2092, "step": 5562 }, { "epoch": 0.34, "learning_rate": 1.676827428019557e-05, "loss": 1.4661, "step": 5564 }, { "epoch": 0.34, "learning_rate": 1.677431037604877e-05, "loss": 1.3534, "step": 5566 }, { "epoch": 0.34, "learning_rate": 1.6780346471901976e-05, "loss": 1.5273, "step": 5568 }, { "epoch": 0.34, "learning_rate": 1.6786382567755176e-05, "loss": 1.4992, "step": 5570 }, { "epoch": 0.34, "learning_rate": 1.6792418663608377e-05, "loss": 1.091, "step": 5572 }, { "epoch": 0.34, "learning_rate": 1.679845475946158e-05, "loss": 1.5212, "step": 5574 }, { "epoch": 0.34, "learning_rate": 1.6804490855314782e-05, "loss": 1.5882, "step": 5576 }, { "epoch": 0.34, "learning_rate": 1.6810526951167986e-05, "loss": 1.1934, "step": 5578 }, { "epoch": 0.34, "learning_rate": 1.6816563047021187e-05, "loss": 1.8945, "step": 5580 }, { "epoch": 0.34, "learning_rate": 1.682259914287439e-05, "loss": 1.3645, "step": 5582 }, { "epoch": 0.34, "learning_rate": 1.6828635238727592e-05, "loss": 0.8923, "step": 5584 }, { "epoch": 0.34, "learning_rate": 1.6834671334580793e-05, "loss": 1.5158, "step": 5586 }, { "epoch": 0.34, "learning_rate": 1.6840707430433997e-05, "loss": 1.3952, "step": 5588 }, { "epoch": 0.34, "learning_rate": 1.6846743526287198e-05, "loss": 1.1713, "step": 5590 }, { "epoch": 0.34, "learning_rate": 1.68527796221404e-05, "loss": 1.2781, "step": 5592 }, { "epoch": 0.34, "learning_rate": 1.6858815717993603e-05, "loss": 1.3434, "step": 5594 }, { "epoch": 0.34, "learning_rate": 1.6864851813846804e-05, "loss": 1.3502, "step": 5596 }, { "epoch": 0.34, "learning_rate": 1.6870887909700005e-05, "loss": 1.2923, "step": 5598 }, { "epoch": 0.34, "learning_rate": 1.687692400555321e-05, "loss": 1.1087, "step": 5600 }, { "epoch": 0.34, "learning_rate": 1.6882960101406413e-05, "loss": 1.2988, "step": 5602 }, { "epoch": 0.34, "learning_rate": 1.6888996197259614e-05, "loss": 1.005, "step": 5604 }, { "epoch": 0.34, "learning_rate": 1.6895032293112815e-05, "loss": 1.3555, "step": 5606 }, { "epoch": 0.34, "learning_rate": 1.690106838896602e-05, "loss": 1.2781, "step": 5608 }, { "epoch": 0.34, "learning_rate": 1.690710448481922e-05, "loss": 1.377, "step": 5610 }, { "epoch": 0.34, "learning_rate": 1.691314058067242e-05, "loss": 1.4091, "step": 5612 }, { "epoch": 0.34, "learning_rate": 1.6919176676525625e-05, "loss": 0.8528, "step": 5614 }, { "epoch": 0.34, "learning_rate": 1.6925212772378826e-05, "loss": 0.9737, "step": 5616 }, { "epoch": 0.34, "learning_rate": 1.6931248868232026e-05, "loss": 1.0167, "step": 5618 }, { "epoch": 0.34, "learning_rate": 1.693728496408523e-05, "loss": 1.0599, "step": 5620 }, { "epoch": 0.34, "learning_rate": 1.694332105993843e-05, "loss": 1.1998, "step": 5622 }, { "epoch": 0.34, "learning_rate": 1.6949357155791636e-05, "loss": 1.2664, "step": 5624 }, { "epoch": 0.34, "learning_rate": 1.695539325164484e-05, "loss": 1.1514, "step": 5626 }, { "epoch": 0.34, "learning_rate": 1.696142934749804e-05, "loss": 1.3268, "step": 5628 }, { "epoch": 0.34, "learning_rate": 1.696746544335124e-05, "loss": 1.375, "step": 5630 }, { "epoch": 0.34, "learning_rate": 1.6973501539204442e-05, "loss": 1.1995, "step": 5632 }, { "epoch": 0.34, "learning_rate": 1.6979537635057647e-05, "loss": 1.307, "step": 5634 }, { "epoch": 0.34, "learning_rate": 1.6985573730910847e-05, "loss": 1.2958, "step": 5636 }, { "epoch": 0.34, "learning_rate": 1.6991609826764048e-05, "loss": 1.4866, "step": 5638 }, { "epoch": 0.34, "learning_rate": 1.6997645922617252e-05, "loss": 1.4892, "step": 5640 }, { "epoch": 0.34, "learning_rate": 1.7003682018470453e-05, "loss": 1.5202, "step": 5642 }, { "epoch": 0.34, "learning_rate": 1.7009718114323654e-05, "loss": 1.6423, "step": 5644 }, { "epoch": 0.34, "learning_rate": 1.7015754210176858e-05, "loss": 1.3462, "step": 5646 }, { "epoch": 0.34, "learning_rate": 1.7021790306030062e-05, "loss": 1.5522, "step": 5648 }, { "epoch": 0.34, "learning_rate": 1.7027826401883263e-05, "loss": 1.3486, "step": 5650 }, { "epoch": 0.34, "learning_rate": 1.7033862497736464e-05, "loss": 0.9145, "step": 5652 }, { "epoch": 0.34, "learning_rate": 1.703989859358967e-05, "loss": 1.2328, "step": 5654 }, { "epoch": 0.34, "learning_rate": 1.704593468944287e-05, "loss": 0.9377, "step": 5656 }, { "epoch": 0.34, "learning_rate": 1.705197078529607e-05, "loss": 1.3311, "step": 5658 }, { "epoch": 0.34, "learning_rate": 1.7058006881149274e-05, "loss": 1.3845, "step": 5660 }, { "epoch": 0.34, "learning_rate": 1.7064042977002475e-05, "loss": 1.3824, "step": 5662 }, { "epoch": 0.34, "learning_rate": 1.7070079072855676e-05, "loss": 1.1334, "step": 5664 }, { "epoch": 0.34, "learning_rate": 1.707611516870888e-05, "loss": 1.1248, "step": 5666 }, { "epoch": 0.34, "learning_rate": 1.708215126456208e-05, "loss": 1.2854, "step": 5668 }, { "epoch": 0.34, "learning_rate": 1.7088187360415285e-05, "loss": 1.0326, "step": 5670 }, { "epoch": 0.34, "learning_rate": 1.7094223456268486e-05, "loss": 1.5422, "step": 5672 }, { "epoch": 0.34, "learning_rate": 1.710025955212169e-05, "loss": 1.2094, "step": 5674 }, { "epoch": 0.34, "learning_rate": 1.710629564797489e-05, "loss": 1.1104, "step": 5676 }, { "epoch": 0.34, "learning_rate": 1.7112331743828092e-05, "loss": 1.3602, "step": 5678 }, { "epoch": 0.34, "learning_rate": 1.7118367839681296e-05, "loss": 1.4434, "step": 5680 }, { "epoch": 0.34, "learning_rate": 1.7124403935534497e-05, "loss": 1.4392, "step": 5682 }, { "epoch": 0.34, "learning_rate": 1.7130440031387698e-05, "loss": 1.0684, "step": 5684 }, { "epoch": 0.34, "learning_rate": 1.7136476127240902e-05, "loss": 1.2613, "step": 5686 }, { "epoch": 0.34, "learning_rate": 1.7142512223094103e-05, "loss": 1.3442, "step": 5688 }, { "epoch": 0.34, "learning_rate": 1.7148548318947303e-05, "loss": 0.8654, "step": 5690 }, { "epoch": 0.34, "learning_rate": 1.7154584414800508e-05, "loss": 1.3328, "step": 5692 }, { "epoch": 0.34, "learning_rate": 1.7160620510653712e-05, "loss": 1.3691, "step": 5694 }, { "epoch": 0.34, "learning_rate": 1.7166656606506913e-05, "loss": 1.4423, "step": 5696 }, { "epoch": 0.34, "learning_rate": 1.7172692702360113e-05, "loss": 1.3687, "step": 5698 }, { "epoch": 0.34, "learning_rate": 1.7178728798213318e-05, "loss": 1.6203, "step": 5700 }, { "epoch": 0.34, "learning_rate": 1.718476489406652e-05, "loss": 1.2744, "step": 5702 }, { "epoch": 0.34, "learning_rate": 1.719080098991972e-05, "loss": 1.4462, "step": 5704 }, { "epoch": 0.34, "learning_rate": 1.7196837085772923e-05, "loss": 0.9455, "step": 5706 }, { "epoch": 0.34, "learning_rate": 1.7202873181626124e-05, "loss": 1.1755, "step": 5708 }, { "epoch": 0.34, "learning_rate": 1.7208909277479325e-05, "loss": 1.0881, "step": 5710 }, { "epoch": 0.34, "learning_rate": 1.721494537333253e-05, "loss": 1.4747, "step": 5712 }, { "epoch": 0.34, "learning_rate": 1.7220981469185734e-05, "loss": 1.2607, "step": 5714 }, { "epoch": 0.35, "learning_rate": 1.7227017565038934e-05, "loss": 1.3368, "step": 5716 }, { "epoch": 0.35, "learning_rate": 1.7233053660892135e-05, "loss": 1.6115, "step": 5718 }, { "epoch": 0.35, "learning_rate": 1.723908975674534e-05, "loss": 1.0474, "step": 5720 }, { "epoch": 0.35, "learning_rate": 1.724512585259854e-05, "loss": 0.8811, "step": 5722 }, { "epoch": 0.35, "learning_rate": 1.725116194845174e-05, "loss": 1.1545, "step": 5724 }, { "epoch": 0.35, "learning_rate": 1.7257198044304945e-05, "loss": 1.4725, "step": 5726 }, { "epoch": 0.35, "learning_rate": 1.7263234140158146e-05, "loss": 1.0298, "step": 5728 }, { "epoch": 0.35, "learning_rate": 1.7269270236011347e-05, "loss": 0.9556, "step": 5730 }, { "epoch": 0.35, "learning_rate": 1.727530633186455e-05, "loss": 1.3701, "step": 5732 }, { "epoch": 0.35, "learning_rate": 1.7281342427717752e-05, "loss": 1.1854, "step": 5734 }, { "epoch": 0.35, "learning_rate": 1.7287378523570956e-05, "loss": 0.9369, "step": 5736 }, { "epoch": 0.35, "learning_rate": 1.729341461942416e-05, "loss": 1.093, "step": 5738 }, { "epoch": 0.35, "learning_rate": 1.729945071527736e-05, "loss": 1.4489, "step": 5740 }, { "epoch": 0.35, "learning_rate": 1.7305486811130562e-05, "loss": 1.2903, "step": 5742 }, { "epoch": 0.35, "learning_rate": 1.7311522906983763e-05, "loss": 1.3041, "step": 5744 }, { "epoch": 0.35, "learning_rate": 1.7317559002836967e-05, "loss": 1.2721, "step": 5746 }, { "epoch": 0.35, "learning_rate": 1.7323595098690168e-05, "loss": 1.1956, "step": 5748 }, { "epoch": 0.35, "learning_rate": 1.732963119454337e-05, "loss": 1.6073, "step": 5750 }, { "epoch": 0.35, "learning_rate": 1.7335667290396573e-05, "loss": 1.2624, "step": 5752 }, { "epoch": 0.35, "learning_rate": 1.7341703386249774e-05, "loss": 1.3437, "step": 5754 }, { "epoch": 0.35, "learning_rate": 1.7347739482102974e-05, "loss": 1.023, "step": 5756 }, { "epoch": 0.35, "learning_rate": 1.735377557795618e-05, "loss": 1.4042, "step": 5758 }, { "epoch": 0.35, "learning_rate": 1.7359811673809383e-05, "loss": 1.2123, "step": 5760 }, { "epoch": 0.35, "learning_rate": 1.7365847769662584e-05, "loss": 1.3636, "step": 5762 }, { "epoch": 0.35, "learning_rate": 1.7371883865515784e-05, "loss": 1.3169, "step": 5764 }, { "epoch": 0.35, "learning_rate": 1.737791996136899e-05, "loss": 0.9941, "step": 5766 }, { "epoch": 0.35, "learning_rate": 1.738395605722219e-05, "loss": 1.3236, "step": 5768 }, { "epoch": 0.35, "learning_rate": 1.738999215307539e-05, "loss": 1.1552, "step": 5770 }, { "epoch": 0.35, "learning_rate": 1.7396028248928595e-05, "loss": 1.1217, "step": 5772 }, { "epoch": 0.35, "learning_rate": 1.7402064344781795e-05, "loss": 1.1767, "step": 5774 }, { "epoch": 0.35, "learning_rate": 1.7408100440634996e-05, "loss": 1.0708, "step": 5776 }, { "epoch": 0.35, "learning_rate": 1.74141365364882e-05, "loss": 1.3368, "step": 5778 }, { "epoch": 0.35, "learning_rate": 1.74201726323414e-05, "loss": 1.1108, "step": 5780 }, { "epoch": 0.35, "learning_rate": 1.7426208728194605e-05, "loss": 1.7188, "step": 5782 }, { "epoch": 0.35, "learning_rate": 1.7432244824047806e-05, "loss": 1.2984, "step": 5784 }, { "epoch": 0.35, "learning_rate": 1.743828091990101e-05, "loss": 1.6159, "step": 5786 }, { "epoch": 0.35, "learning_rate": 1.744431701575421e-05, "loss": 1.355, "step": 5788 }, { "epoch": 0.35, "learning_rate": 1.7450353111607412e-05, "loss": 1.1815, "step": 5790 }, { "epoch": 0.35, "learning_rate": 1.7456389207460616e-05, "loss": 1.3242, "step": 5792 }, { "epoch": 0.35, "learning_rate": 1.7462425303313817e-05, "loss": 1.3533, "step": 5794 }, { "epoch": 0.35, "learning_rate": 1.7468461399167018e-05, "loss": 0.947, "step": 5796 }, { "epoch": 0.35, "learning_rate": 1.7474497495020222e-05, "loss": 1.3644, "step": 5798 }, { "epoch": 0.35, "learning_rate": 1.7480533590873423e-05, "loss": 1.3161, "step": 5800 }, { "epoch": 0.35, "learning_rate": 1.7486569686726624e-05, "loss": 1.5901, "step": 5802 }, { "epoch": 0.35, "learning_rate": 1.7492605782579828e-05, "loss": 1.7117, "step": 5804 }, { "epoch": 0.35, "learning_rate": 1.7498641878433032e-05, "loss": 1.5335, "step": 5806 }, { "epoch": 0.35, "learning_rate": 1.7504677974286233e-05, "loss": 1.054, "step": 5808 }, { "epoch": 0.35, "learning_rate": 1.7510714070139434e-05, "loss": 0.9561, "step": 5810 }, { "epoch": 0.35, "learning_rate": 1.7516750165992638e-05, "loss": 0.9989, "step": 5812 }, { "epoch": 0.35, "learning_rate": 1.752278626184584e-05, "loss": 1.2633, "step": 5814 }, { "epoch": 0.35, "learning_rate": 1.752882235769904e-05, "loss": 1.3971, "step": 5816 }, { "epoch": 0.35, "learning_rate": 1.7534858453552244e-05, "loss": 1.3924, "step": 5818 }, { "epoch": 0.35, "learning_rate": 1.7540894549405445e-05, "loss": 1.3449, "step": 5820 }, { "epoch": 0.35, "learning_rate": 1.7546930645258646e-05, "loss": 1.2574, "step": 5822 }, { "epoch": 0.35, "learning_rate": 1.755296674111185e-05, "loss": 1.2601, "step": 5824 }, { "epoch": 0.35, "learning_rate": 1.755900283696505e-05, "loss": 1.0026, "step": 5826 }, { "epoch": 0.35, "learning_rate": 1.7565038932818255e-05, "loss": 1.3027, "step": 5828 }, { "epoch": 0.35, "learning_rate": 1.7571075028671456e-05, "loss": 1.092, "step": 5830 }, { "epoch": 0.35, "learning_rate": 1.757711112452466e-05, "loss": 1.5281, "step": 5832 }, { "epoch": 0.35, "learning_rate": 1.758314722037786e-05, "loss": 1.5283, "step": 5834 }, { "epoch": 0.35, "learning_rate": 1.758918331623106e-05, "loss": 0.8723, "step": 5836 }, { "epoch": 0.35, "learning_rate": 1.7595219412084266e-05, "loss": 1.0591, "step": 5838 }, { "epoch": 0.35, "learning_rate": 1.7601255507937466e-05, "loss": 1.3547, "step": 5840 }, { "epoch": 0.35, "learning_rate": 1.7607291603790667e-05, "loss": 1.507, "step": 5842 }, { "epoch": 0.35, "learning_rate": 1.761332769964387e-05, "loss": 1.204, "step": 5844 }, { "epoch": 0.35, "learning_rate": 1.7619363795497072e-05, "loss": 1.0862, "step": 5846 }, { "epoch": 0.35, "learning_rate": 1.7625399891350276e-05, "loss": 1.239, "step": 5848 }, { "epoch": 0.35, "learning_rate": 1.763143598720348e-05, "loss": 1.1104, "step": 5850 }, { "epoch": 0.35, "learning_rate": 1.763747208305668e-05, "loss": 1.4304, "step": 5852 }, { "epoch": 0.35, "learning_rate": 1.7643508178909882e-05, "loss": 1.2561, "step": 5854 }, { "epoch": 0.35, "learning_rate": 1.7649544274763083e-05, "loss": 1.6747, "step": 5856 }, { "epoch": 0.35, "learning_rate": 1.7655580370616287e-05, "loss": 1.3777, "step": 5858 }, { "epoch": 0.35, "learning_rate": 1.7661616466469488e-05, "loss": 1.0224, "step": 5860 }, { "epoch": 0.35, "learning_rate": 1.766765256232269e-05, "loss": 1.1508, "step": 5862 }, { "epoch": 0.35, "learning_rate": 1.7673688658175893e-05, "loss": 1.3523, "step": 5864 }, { "epoch": 0.35, "learning_rate": 1.7679724754029094e-05, "loss": 1.1238, "step": 5866 }, { "epoch": 0.35, "learning_rate": 1.7685760849882295e-05, "loss": 1.1224, "step": 5868 }, { "epoch": 0.35, "learning_rate": 1.76917969457355e-05, "loss": 0.937, "step": 5870 }, { "epoch": 0.35, "learning_rate": 1.7697833041588703e-05, "loss": 1.5329, "step": 5872 }, { "epoch": 0.35, "learning_rate": 1.7703869137441904e-05, "loss": 0.9586, "step": 5874 }, { "epoch": 0.35, "learning_rate": 1.7709905233295105e-05, "loss": 1.1113, "step": 5876 }, { "epoch": 0.35, "learning_rate": 1.771594132914831e-05, "loss": 1.4579, "step": 5878 }, { "epoch": 0.35, "learning_rate": 1.772197742500151e-05, "loss": 1.2724, "step": 5880 }, { "epoch": 0.36, "learning_rate": 1.772801352085471e-05, "loss": 1.4227, "step": 5882 }, { "epoch": 0.36, "learning_rate": 1.7734049616707915e-05, "loss": 1.149, "step": 5884 }, { "epoch": 0.36, "learning_rate": 1.7740085712561116e-05, "loss": 1.1883, "step": 5886 }, { "epoch": 0.36, "learning_rate": 1.7746121808414317e-05, "loss": 1.1975, "step": 5888 }, { "epoch": 0.36, "learning_rate": 1.775215790426752e-05, "loss": 1.2993, "step": 5890 }, { "epoch": 0.36, "learning_rate": 1.775819400012072e-05, "loss": 1.1776, "step": 5892 }, { "epoch": 0.36, "learning_rate": 1.7764230095973926e-05, "loss": 1.1669, "step": 5894 }, { "epoch": 0.36, "learning_rate": 1.7770266191827127e-05, "loss": 1.2026, "step": 5896 }, { "epoch": 0.36, "learning_rate": 1.777630228768033e-05, "loss": 1.3934, "step": 5898 }, { "epoch": 0.36, "learning_rate": 1.778233838353353e-05, "loss": 1.1082, "step": 5900 }, { "epoch": 0.36, "learning_rate": 1.7788374479386732e-05, "loss": 1.3689, "step": 5902 }, { "epoch": 0.36, "learning_rate": 1.7794410575239937e-05, "loss": 1.1249, "step": 5904 }, { "epoch": 0.36, "learning_rate": 1.7800446671093137e-05, "loss": 1.3932, "step": 5906 }, { "epoch": 0.36, "learning_rate": 1.7806482766946338e-05, "loss": 1.1268, "step": 5908 }, { "epoch": 0.36, "learning_rate": 1.7812518862799543e-05, "loss": 1.1323, "step": 5910 }, { "epoch": 0.36, "learning_rate": 1.7818554958652743e-05, "loss": 1.1055, "step": 5912 }, { "epoch": 0.36, "learning_rate": 1.7824591054505944e-05, "loss": 1.1831, "step": 5914 }, { "epoch": 0.36, "learning_rate": 1.783062715035915e-05, "loss": 1.081, "step": 5916 }, { "epoch": 0.36, "learning_rate": 1.7836663246212353e-05, "loss": 1.0899, "step": 5918 }, { "epoch": 0.36, "learning_rate": 1.7842699342065553e-05, "loss": 1.1456, "step": 5920 }, { "epoch": 0.36, "learning_rate": 1.7848735437918754e-05, "loss": 1.1736, "step": 5922 }, { "epoch": 0.36, "learning_rate": 1.785477153377196e-05, "loss": 1.1975, "step": 5924 }, { "epoch": 0.36, "learning_rate": 1.786080762962516e-05, "loss": 1.3231, "step": 5926 }, { "epoch": 0.36, "learning_rate": 1.786684372547836e-05, "loss": 1.3181, "step": 5928 }, { "epoch": 0.36, "learning_rate": 1.7872879821331564e-05, "loss": 1.182, "step": 5930 }, { "epoch": 0.36, "learning_rate": 1.7878915917184765e-05, "loss": 1.253, "step": 5932 }, { "epoch": 0.36, "learning_rate": 1.7884952013037966e-05, "loss": 1.4369, "step": 5934 }, { "epoch": 0.36, "learning_rate": 1.789098810889117e-05, "loss": 1.5232, "step": 5936 }, { "epoch": 0.36, "learning_rate": 1.789702420474437e-05, "loss": 1.2018, "step": 5938 }, { "epoch": 0.36, "learning_rate": 1.7903060300597575e-05, "loss": 1.4627, "step": 5940 }, { "epoch": 0.36, "learning_rate": 1.7909096396450776e-05, "loss": 1.073, "step": 5942 }, { "epoch": 0.36, "learning_rate": 1.791513249230398e-05, "loss": 1.1211, "step": 5944 }, { "epoch": 0.36, "learning_rate": 1.792116858815718e-05, "loss": 1.2921, "step": 5946 }, { "epoch": 0.36, "learning_rate": 1.7927204684010382e-05, "loss": 1.4408, "step": 5948 }, { "epoch": 0.36, "learning_rate": 1.7933240779863586e-05, "loss": 1.4148, "step": 5950 }, { "epoch": 0.36, "learning_rate": 1.7939276875716787e-05, "loss": 1.2687, "step": 5952 }, { "epoch": 0.36, "learning_rate": 1.7945312971569988e-05, "loss": 1.3225, "step": 5954 }, { "epoch": 0.36, "learning_rate": 1.7951349067423192e-05, "loss": 1.2171, "step": 5956 }, { "epoch": 0.36, "learning_rate": 1.7957385163276393e-05, "loss": 1.4274, "step": 5958 }, { "epoch": 0.36, "learning_rate": 1.7963421259129597e-05, "loss": 1.1345, "step": 5960 }, { "epoch": 0.36, "learning_rate": 1.7969457354982798e-05, "loss": 1.6291, "step": 5962 }, { "epoch": 0.36, "learning_rate": 1.7975493450836002e-05, "loss": 1.4433, "step": 5964 }, { "epoch": 0.36, "learning_rate": 1.7981529546689203e-05, "loss": 1.291, "step": 5966 }, { "epoch": 0.36, "learning_rate": 1.7987565642542404e-05, "loss": 1.1198, "step": 5968 }, { "epoch": 0.36, "learning_rate": 1.7993601738395608e-05, "loss": 1.6072, "step": 5970 }, { "epoch": 0.36, "learning_rate": 1.799963783424881e-05, "loss": 1.2551, "step": 5972 }, { "epoch": 0.36, "learning_rate": 1.800567393010201e-05, "loss": 1.2114, "step": 5974 }, { "epoch": 0.36, "learning_rate": 1.8011710025955214e-05, "loss": 1.1637, "step": 5976 }, { "epoch": 0.36, "learning_rate": 1.8017746121808414e-05, "loss": 1.0767, "step": 5978 }, { "epoch": 0.36, "learning_rate": 1.8023782217661615e-05, "loss": 1.5267, "step": 5980 }, { "epoch": 0.36, "learning_rate": 1.802981831351482e-05, "loss": 1.3919, "step": 5982 }, { "epoch": 0.36, "learning_rate": 1.8035854409368024e-05, "loss": 1.4688, "step": 5984 }, { "epoch": 0.36, "learning_rate": 1.8041890505221224e-05, "loss": 1.1908, "step": 5986 }, { "epoch": 0.36, "learning_rate": 1.8047926601074425e-05, "loss": 1.2609, "step": 5988 }, { "epoch": 0.36, "learning_rate": 1.805396269692763e-05, "loss": 1.2145, "step": 5990 }, { "epoch": 0.36, "learning_rate": 1.805999879278083e-05, "loss": 1.6691, "step": 5992 }, { "epoch": 0.36, "learning_rate": 1.806603488863403e-05, "loss": 1.275, "step": 5994 }, { "epoch": 0.36, "learning_rate": 1.8072070984487235e-05, "loss": 1.104, "step": 5996 }, { "epoch": 0.36, "learning_rate": 1.8078107080340436e-05, "loss": 0.9576, "step": 5998 }, { "epoch": 0.36, "learning_rate": 1.8084143176193637e-05, "loss": 1.0053, "step": 6000 }, { "epoch": 0.36, "learning_rate": 1.809017927204684e-05, "loss": 1.2862, "step": 6002 }, { "epoch": 0.36, "learning_rate": 1.8096215367900042e-05, "loss": 1.2955, "step": 6004 }, { "epoch": 0.36, "learning_rate": 1.8102251463753246e-05, "loss": 1.2987, "step": 6006 }, { "epoch": 0.36, "learning_rate": 1.8108287559606447e-05, "loss": 1.0853, "step": 6008 }, { "epoch": 0.36, "learning_rate": 1.811432365545965e-05, "loss": 1.0203, "step": 6010 }, { "epoch": 0.36, "learning_rate": 1.8120359751312852e-05, "loss": 1.3351, "step": 6012 }, { "epoch": 0.36, "learning_rate": 1.8126395847166053e-05, "loss": 1.3061, "step": 6014 }, { "epoch": 0.36, "learning_rate": 1.8132431943019257e-05, "loss": 1.2073, "step": 6016 }, { "epoch": 0.36, "learning_rate": 1.8138468038872458e-05, "loss": 1.0344, "step": 6018 }, { "epoch": 0.36, "learning_rate": 1.814450413472566e-05, "loss": 1.197, "step": 6020 }, { "epoch": 0.36, "learning_rate": 1.8150540230578863e-05, "loss": 1.3186, "step": 6022 }, { "epoch": 0.36, "learning_rate": 1.8156576326432064e-05, "loss": 1.1133, "step": 6024 }, { "epoch": 0.36, "learning_rate": 1.8162612422285265e-05, "loss": 1.3416, "step": 6026 }, { "epoch": 0.36, "learning_rate": 1.816864851813847e-05, "loss": 1.3091, "step": 6028 }, { "epoch": 0.36, "learning_rate": 1.8174684613991673e-05, "loss": 1.1903, "step": 6030 }, { "epoch": 0.36, "learning_rate": 1.8180720709844874e-05, "loss": 1.3021, "step": 6032 }, { "epoch": 0.36, "learning_rate": 1.8186756805698075e-05, "loss": 1.1868, "step": 6034 }, { "epoch": 0.36, "learning_rate": 1.819279290155128e-05, "loss": 1.266, "step": 6036 }, { "epoch": 0.36, "learning_rate": 1.819882899740448e-05, "loss": 1.3837, "step": 6038 }, { "epoch": 0.36, "learning_rate": 1.820486509325768e-05, "loss": 1.589, "step": 6040 }, { "epoch": 0.36, "learning_rate": 1.8210901189110885e-05, "loss": 1.3578, "step": 6042 }, { "epoch": 0.36, "learning_rate": 1.8216937284964085e-05, "loss": 1.3749, "step": 6044 }, { "epoch": 0.36, "learning_rate": 1.8222973380817286e-05, "loss": 1.5917, "step": 6046 }, { "epoch": 0.37, "learning_rate": 1.822900947667049e-05, "loss": 1.2083, "step": 6048 }, { "epoch": 0.37, "learning_rate": 1.823504557252369e-05, "loss": 1.0929, "step": 6050 }, { "epoch": 0.37, "learning_rate": 1.8241081668376895e-05, "loss": 1.3411, "step": 6052 }, { "epoch": 0.37, "learning_rate": 1.8247117764230096e-05, "loss": 1.1078, "step": 6054 }, { "epoch": 0.37, "learning_rate": 1.82531538600833e-05, "loss": 1.5336, "step": 6056 }, { "epoch": 0.37, "learning_rate": 1.82591899559365e-05, "loss": 1.0306, "step": 6058 }, { "epoch": 0.37, "learning_rate": 1.8265226051789702e-05, "loss": 1.4208, "step": 6060 }, { "epoch": 0.37, "learning_rate": 1.8271262147642906e-05, "loss": 1.3413, "step": 6062 }, { "epoch": 0.37, "learning_rate": 1.8277298243496107e-05, "loss": 1.2506, "step": 6064 }, { "epoch": 0.37, "learning_rate": 1.8283334339349308e-05, "loss": 1.2602, "step": 6066 }, { "epoch": 0.37, "learning_rate": 1.8289370435202512e-05, "loss": 1.0362, "step": 6068 }, { "epoch": 0.37, "learning_rate": 1.8295406531055713e-05, "loss": 1.4991, "step": 6070 }, { "epoch": 0.37, "learning_rate": 1.8301442626908917e-05, "loss": 1.3889, "step": 6072 }, { "epoch": 0.37, "learning_rate": 1.8307478722762118e-05, "loss": 1.2258, "step": 6074 }, { "epoch": 0.37, "learning_rate": 1.8313514818615322e-05, "loss": 1.4142, "step": 6076 }, { "epoch": 0.37, "learning_rate": 1.8319550914468523e-05, "loss": 1.3891, "step": 6078 }, { "epoch": 0.37, "learning_rate": 1.8325587010321724e-05, "loss": 1.3108, "step": 6080 }, { "epoch": 0.37, "learning_rate": 1.8331623106174928e-05, "loss": 1.2508, "step": 6082 }, { "epoch": 0.37, "learning_rate": 1.833765920202813e-05, "loss": 1.2643, "step": 6084 }, { "epoch": 0.37, "learning_rate": 1.834369529788133e-05, "loss": 1.3829, "step": 6086 }, { "epoch": 0.37, "learning_rate": 1.8349731393734534e-05, "loss": 0.9522, "step": 6088 }, { "epoch": 0.37, "learning_rate": 1.8355767489587735e-05, "loss": 1.3329, "step": 6090 }, { "epoch": 0.37, "learning_rate": 1.8361803585440936e-05, "loss": 1.3375, "step": 6092 }, { "epoch": 0.37, "learning_rate": 1.836783968129414e-05, "loss": 1.3026, "step": 6094 }, { "epoch": 0.37, "learning_rate": 1.8373875777147344e-05, "loss": 1.2102, "step": 6096 }, { "epoch": 0.37, "learning_rate": 1.8379911873000545e-05, "loss": 1.3171, "step": 6098 }, { "epoch": 0.37, "learning_rate": 1.8385947968853746e-05, "loss": 1.1753, "step": 6100 }, { "epoch": 0.37, "learning_rate": 1.839198406470695e-05, "loss": 0.9463, "step": 6102 }, { "epoch": 0.37, "learning_rate": 1.839802016056015e-05, "loss": 1.3464, "step": 6104 }, { "epoch": 0.37, "learning_rate": 1.840405625641335e-05, "loss": 1.2036, "step": 6106 }, { "epoch": 0.37, "learning_rate": 1.8410092352266556e-05, "loss": 1.1027, "step": 6108 }, { "epoch": 0.37, "learning_rate": 1.8416128448119757e-05, "loss": 1.4538, "step": 6110 }, { "epoch": 0.37, "learning_rate": 1.8422164543972957e-05, "loss": 1.4828, "step": 6112 }, { "epoch": 0.37, "learning_rate": 1.842820063982616e-05, "loss": 1.0387, "step": 6114 }, { "epoch": 0.37, "learning_rate": 1.8434236735679362e-05, "loss": 1.6855, "step": 6116 }, { "epoch": 0.37, "learning_rate": 1.8440272831532567e-05, "loss": 0.8604, "step": 6118 }, { "epoch": 0.37, "learning_rate": 1.8446308927385767e-05, "loss": 1.1197, "step": 6120 }, { "epoch": 0.37, "learning_rate": 1.845234502323897e-05, "loss": 1.0887, "step": 6122 }, { "epoch": 0.37, "learning_rate": 1.8458381119092172e-05, "loss": 1.2653, "step": 6124 }, { "epoch": 0.37, "learning_rate": 1.8464417214945373e-05, "loss": 1.3188, "step": 6126 }, { "epoch": 0.37, "learning_rate": 1.8470453310798577e-05, "loss": 1.5347, "step": 6128 }, { "epoch": 0.37, "learning_rate": 1.8476489406651778e-05, "loss": 1.0578, "step": 6130 }, { "epoch": 0.37, "learning_rate": 1.848252550250498e-05, "loss": 1.1075, "step": 6132 }, { "epoch": 0.37, "learning_rate": 1.8488561598358183e-05, "loss": 1.2416, "step": 6134 }, { "epoch": 0.37, "learning_rate": 1.8494597694211384e-05, "loss": 1.1942, "step": 6136 }, { "epoch": 0.37, "learning_rate": 1.8500633790064585e-05, "loss": 1.1291, "step": 6138 }, { "epoch": 0.37, "learning_rate": 1.850666988591779e-05, "loss": 1.0746, "step": 6140 }, { "epoch": 0.37, "learning_rate": 1.8512705981770993e-05, "loss": 1.7916, "step": 6142 }, { "epoch": 0.37, "learning_rate": 1.8518742077624194e-05, "loss": 1.326, "step": 6144 }, { "epoch": 0.37, "learning_rate": 1.8524778173477395e-05, "loss": 1.2767, "step": 6146 }, { "epoch": 0.37, "learning_rate": 1.85308142693306e-05, "loss": 0.9949, "step": 6148 }, { "epoch": 0.37, "learning_rate": 1.85368503651838e-05, "loss": 1.2849, "step": 6150 }, { "epoch": 0.37, "learning_rate": 1.8542886461037e-05, "loss": 1.2716, "step": 6152 }, { "epoch": 0.37, "learning_rate": 1.8548922556890205e-05, "loss": 0.8559, "step": 6154 }, { "epoch": 0.37, "learning_rate": 1.8554958652743406e-05, "loss": 1.2526, "step": 6156 }, { "epoch": 0.37, "learning_rate": 1.8560994748596607e-05, "loss": 1.4878, "step": 6158 }, { "epoch": 0.37, "learning_rate": 1.856703084444981e-05, "loss": 1.2646, "step": 6160 }, { "epoch": 0.37, "learning_rate": 1.857306694030301e-05, "loss": 1.2186, "step": 6162 }, { "epoch": 0.37, "learning_rate": 1.8579103036156216e-05, "loss": 1.3182, "step": 6164 }, { "epoch": 0.37, "learning_rate": 1.8585139132009417e-05, "loss": 1.2848, "step": 6166 }, { "epoch": 0.37, "learning_rate": 1.859117522786262e-05, "loss": 1.3239, "step": 6168 }, { "epoch": 0.37, "learning_rate": 1.8597211323715822e-05, "loss": 1.0971, "step": 6170 }, { "epoch": 0.37, "learning_rate": 1.8603247419569023e-05, "loss": 1.0199, "step": 6172 }, { "epoch": 0.37, "learning_rate": 1.8609283515422227e-05, "loss": 0.8207, "step": 6174 }, { "epoch": 0.37, "learning_rate": 1.8615319611275428e-05, "loss": 1.0233, "step": 6176 }, { "epoch": 0.37, "learning_rate": 1.862135570712863e-05, "loss": 1.4272, "step": 6178 }, { "epoch": 0.37, "learning_rate": 1.8627391802981833e-05, "loss": 1.0346, "step": 6180 }, { "epoch": 0.37, "learning_rate": 1.8633427898835033e-05, "loss": 1.2869, "step": 6182 }, { "epoch": 0.37, "learning_rate": 1.8639463994688238e-05, "loss": 1.1086, "step": 6184 }, { "epoch": 0.37, "learning_rate": 1.864550009054144e-05, "loss": 1.0859, "step": 6186 }, { "epoch": 0.37, "learning_rate": 1.8651536186394643e-05, "loss": 1.3301, "step": 6188 }, { "epoch": 0.37, "learning_rate": 1.8657572282247843e-05, "loss": 1.37, "step": 6190 }, { "epoch": 0.37, "learning_rate": 1.8663608378101044e-05, "loss": 1.4405, "step": 6192 }, { "epoch": 0.37, "learning_rate": 1.866964447395425e-05, "loss": 1.2227, "step": 6194 }, { "epoch": 0.37, "learning_rate": 1.867568056980745e-05, "loss": 1.1866, "step": 6196 }, { "epoch": 0.37, "learning_rate": 1.868171666566065e-05, "loss": 1.0771, "step": 6198 }, { "epoch": 0.37, "learning_rate": 1.8687752761513854e-05, "loss": 1.157, "step": 6200 }, { "epoch": 0.37, "learning_rate": 1.8693788857367055e-05, "loss": 1.1475, "step": 6202 }, { "epoch": 0.37, "learning_rate": 1.8699824953220256e-05, "loss": 1.3549, "step": 6204 }, { "epoch": 0.37, "learning_rate": 1.870586104907346e-05, "loss": 1.3445, "step": 6206 }, { "epoch": 0.37, "learning_rate": 1.8711897144926664e-05, "loss": 1.505, "step": 6208 }, { "epoch": 0.37, "learning_rate": 1.8717933240779865e-05, "loss": 1.1703, "step": 6210 }, { "epoch": 0.37, "learning_rate": 1.8723969336633066e-05, "loss": 1.3861, "step": 6212 }, { "epoch": 0.38, "learning_rate": 1.873000543248627e-05, "loss": 1.2169, "step": 6214 }, { "epoch": 0.38, "learning_rate": 1.873604152833947e-05, "loss": 1.2378, "step": 6216 }, { "epoch": 0.38, "learning_rate": 1.8742077624192672e-05, "loss": 1.4195, "step": 6218 }, { "epoch": 0.38, "learning_rate": 1.8748113720045876e-05, "loss": 1.1385, "step": 6220 }, { "epoch": 0.38, "learning_rate": 1.8754149815899077e-05, "loss": 1.2207, "step": 6222 }, { "epoch": 0.38, "learning_rate": 1.8760185911752278e-05, "loss": 1.3467, "step": 6224 }, { "epoch": 0.38, "learning_rate": 1.8766222007605482e-05, "loss": 1.4208, "step": 6226 }, { "epoch": 0.38, "learning_rate": 1.8772258103458683e-05, "loss": 1.3209, "step": 6228 }, { "epoch": 0.38, "learning_rate": 1.8778294199311887e-05, "loss": 1.6146, "step": 6230 }, { "epoch": 0.38, "learning_rate": 1.8784330295165088e-05, "loss": 1.388, "step": 6232 }, { "epoch": 0.38, "learning_rate": 1.8790366391018292e-05, "loss": 1.4286, "step": 6234 }, { "epoch": 0.38, "learning_rate": 1.8796402486871493e-05, "loss": 0.7651, "step": 6236 }, { "epoch": 0.38, "learning_rate": 1.8802438582724694e-05, "loss": 1.2008, "step": 6238 }, { "epoch": 0.38, "learning_rate": 1.8808474678577898e-05, "loss": 1.3747, "step": 6240 }, { "epoch": 0.38, "learning_rate": 1.88145107744311e-05, "loss": 1.2507, "step": 6242 }, { "epoch": 0.38, "learning_rate": 1.88205468702843e-05, "loss": 1.4464, "step": 6244 }, { "epoch": 0.38, "learning_rate": 1.8826582966137504e-05, "loss": 1.4892, "step": 6246 }, { "epoch": 0.38, "learning_rate": 1.8832619061990704e-05, "loss": 1.0823, "step": 6248 }, { "epoch": 0.38, "learning_rate": 1.8838655157843905e-05, "loss": 1.2689, "step": 6250 }, { "epoch": 0.38, "learning_rate": 1.884469125369711e-05, "loss": 1.2152, "step": 6252 }, { "epoch": 0.38, "learning_rate": 1.8850727349550314e-05, "loss": 1.0155, "step": 6254 }, { "epoch": 0.38, "learning_rate": 1.8856763445403515e-05, "loss": 1.508, "step": 6256 }, { "epoch": 0.38, "learning_rate": 1.8862799541256715e-05, "loss": 1.0668, "step": 6258 }, { "epoch": 0.38, "learning_rate": 1.886883563710992e-05, "loss": 1.3251, "step": 6260 }, { "epoch": 0.38, "learning_rate": 1.887487173296312e-05, "loss": 1.2268, "step": 6262 }, { "epoch": 0.38, "learning_rate": 1.888090782881632e-05, "loss": 1.6359, "step": 6264 }, { "epoch": 0.38, "learning_rate": 1.8886943924669525e-05, "loss": 1.2018, "step": 6266 }, { "epoch": 0.38, "learning_rate": 1.8892980020522726e-05, "loss": 1.2222, "step": 6268 }, { "epoch": 0.38, "learning_rate": 1.8899016116375927e-05, "loss": 1.2188, "step": 6270 }, { "epoch": 0.38, "learning_rate": 1.890505221222913e-05, "loss": 1.4907, "step": 6272 }, { "epoch": 0.38, "learning_rate": 1.8911088308082332e-05, "loss": 1.7234, "step": 6274 }, { "epoch": 0.38, "learning_rate": 1.8917124403935536e-05, "loss": 1.4168, "step": 6276 }, { "epoch": 0.38, "learning_rate": 1.8923160499788737e-05, "loss": 1.2774, "step": 6278 }, { "epoch": 0.38, "learning_rate": 1.892919659564194e-05, "loss": 1.5652, "step": 6280 }, { "epoch": 0.38, "learning_rate": 1.8935232691495142e-05, "loss": 1.0162, "step": 6282 }, { "epoch": 0.38, "learning_rate": 1.8941268787348343e-05, "loss": 1.3223, "step": 6284 }, { "epoch": 0.38, "learning_rate": 1.8947304883201547e-05, "loss": 1.3331, "step": 6286 }, { "epoch": 0.38, "learning_rate": 1.8953340979054748e-05, "loss": 1.444, "step": 6288 }, { "epoch": 0.38, "learning_rate": 1.895937707490795e-05, "loss": 0.9313, "step": 6290 }, { "epoch": 0.38, "learning_rate": 1.8965413170761153e-05, "loss": 1.391, "step": 6292 }, { "epoch": 0.38, "learning_rate": 1.8971449266614354e-05, "loss": 1.3018, "step": 6294 }, { "epoch": 0.38, "learning_rate": 1.8977485362467558e-05, "loss": 1.5301, "step": 6296 }, { "epoch": 0.38, "learning_rate": 1.898352145832076e-05, "loss": 1.2729, "step": 6298 }, { "epoch": 0.38, "learning_rate": 1.8989557554173963e-05, "loss": 1.3661, "step": 6300 }, { "epoch": 0.38, "learning_rate": 1.8995593650027164e-05, "loss": 1.2204, "step": 6302 }, { "epoch": 0.38, "learning_rate": 1.9001629745880365e-05, "loss": 1.2583, "step": 6304 }, { "epoch": 0.38, "learning_rate": 1.900766584173357e-05, "loss": 1.0101, "step": 6306 }, { "epoch": 0.38, "learning_rate": 1.901370193758677e-05, "loss": 1.739, "step": 6308 }, { "epoch": 0.38, "learning_rate": 1.901973803343997e-05, "loss": 1.148, "step": 6310 }, { "epoch": 0.38, "learning_rate": 1.9025774129293175e-05, "loss": 1.0769, "step": 6312 }, { "epoch": 0.38, "learning_rate": 1.9031810225146376e-05, "loss": 1.2745, "step": 6314 }, { "epoch": 0.38, "learning_rate": 1.9037846320999576e-05, "loss": 1.1847, "step": 6316 }, { "epoch": 0.38, "learning_rate": 1.904388241685278e-05, "loss": 1.2122, "step": 6318 }, { "epoch": 0.38, "learning_rate": 1.9049918512705985e-05, "loss": 1.4856, "step": 6320 }, { "epoch": 0.38, "learning_rate": 1.9055954608559186e-05, "loss": 1.05, "step": 6322 }, { "epoch": 0.38, "learning_rate": 1.9061990704412386e-05, "loss": 1.4845, "step": 6324 }, { "epoch": 0.38, "learning_rate": 1.906802680026559e-05, "loss": 1.0641, "step": 6326 }, { "epoch": 0.38, "learning_rate": 1.907406289611879e-05, "loss": 0.891, "step": 6328 }, { "epoch": 0.38, "learning_rate": 1.9080098991971992e-05, "loss": 1.2506, "step": 6330 }, { "epoch": 0.38, "learning_rate": 1.9086135087825196e-05, "loss": 1.614, "step": 6332 }, { "epoch": 0.38, "learning_rate": 1.9092171183678397e-05, "loss": 1.5827, "step": 6334 }, { "epoch": 0.38, "learning_rate": 1.9098207279531598e-05, "loss": 1.128, "step": 6336 }, { "epoch": 0.38, "learning_rate": 1.9104243375384802e-05, "loss": 1.2174, "step": 6338 }, { "epoch": 0.38, "learning_rate": 1.9110279471238003e-05, "loss": 1.3866, "step": 6340 }, { "epoch": 0.38, "learning_rate": 1.9116315567091207e-05, "loss": 1.2103, "step": 6342 }, { "epoch": 0.38, "learning_rate": 1.9122351662944408e-05, "loss": 1.3225, "step": 6344 }, { "epoch": 0.38, "learning_rate": 1.9128387758797612e-05, "loss": 1.5581, "step": 6346 }, { "epoch": 0.38, "learning_rate": 1.9134423854650813e-05, "loss": 1.0967, "step": 6348 }, { "epoch": 0.38, "learning_rate": 1.9140459950504014e-05, "loss": 1.1914, "step": 6350 }, { "epoch": 0.38, "learning_rate": 1.9146496046357218e-05, "loss": 1.0631, "step": 6352 }, { "epoch": 0.38, "learning_rate": 1.915253214221042e-05, "loss": 1.3215, "step": 6354 }, { "epoch": 0.38, "learning_rate": 1.915856823806362e-05, "loss": 1.3225, "step": 6356 }, { "epoch": 0.38, "learning_rate": 1.9164604333916824e-05, "loss": 1.0233, "step": 6358 }, { "epoch": 0.38, "learning_rate": 1.9170640429770025e-05, "loss": 1.4662, "step": 6360 }, { "epoch": 0.38, "learning_rate": 1.9176676525623226e-05, "loss": 1.0648, "step": 6362 }, { "epoch": 0.38, "learning_rate": 1.918271262147643e-05, "loss": 1.2287, "step": 6364 }, { "epoch": 0.38, "learning_rate": 1.9188748717329634e-05, "loss": 1.1578, "step": 6366 }, { "epoch": 0.38, "learning_rate": 1.9194784813182835e-05, "loss": 1.0673, "step": 6368 }, { "epoch": 0.38, "learning_rate": 1.9200820909036036e-05, "loss": 1.6886, "step": 6370 }, { "epoch": 0.38, "learning_rate": 1.920685700488924e-05, "loss": 1.3598, "step": 6372 }, { "epoch": 0.38, "learning_rate": 1.921289310074244e-05, "loss": 1.1305, "step": 6374 }, { "epoch": 0.38, "learning_rate": 1.921892919659564e-05, "loss": 1.045, "step": 6376 }, { "epoch": 0.38, "learning_rate": 1.9224965292448846e-05, "loss": 1.1648, "step": 6378 }, { "epoch": 0.39, "learning_rate": 1.9231001388302047e-05, "loss": 1.4611, "step": 6380 }, { "epoch": 0.39, "learning_rate": 1.9237037484155247e-05, "loss": 1.1727, "step": 6382 }, { "epoch": 0.39, "learning_rate": 1.924307358000845e-05, "loss": 1.6303, "step": 6384 }, { "epoch": 0.39, "learning_rate": 1.9249109675861652e-05, "loss": 1.4222, "step": 6386 }, { "epoch": 0.39, "learning_rate": 1.9255145771714857e-05, "loss": 1.0476, "step": 6388 }, { "epoch": 0.39, "learning_rate": 1.9261181867568057e-05, "loss": 1.4266, "step": 6390 }, { "epoch": 0.39, "learning_rate": 1.926721796342126e-05, "loss": 1.2058, "step": 6392 }, { "epoch": 0.39, "learning_rate": 1.9273254059274462e-05, "loss": 1.2393, "step": 6394 }, { "epoch": 0.39, "learning_rate": 1.9279290155127663e-05, "loss": 1.0496, "step": 6396 }, { "epoch": 0.39, "learning_rate": 1.9285326250980868e-05, "loss": 1.1065, "step": 6398 }, { "epoch": 0.39, "learning_rate": 1.929136234683407e-05, "loss": 1.1052, "step": 6400 }, { "epoch": 0.39, "learning_rate": 1.929739844268727e-05, "loss": 0.9464, "step": 6402 }, { "epoch": 0.39, "learning_rate": 1.9303434538540473e-05, "loss": 1.421, "step": 6404 }, { "epoch": 0.39, "learning_rate": 1.9309470634393674e-05, "loss": 1.053, "step": 6406 }, { "epoch": 0.39, "learning_rate": 1.931550673024688e-05, "loss": 0.8931, "step": 6408 }, { "epoch": 0.39, "learning_rate": 1.932154282610008e-05, "loss": 1.8184, "step": 6410 }, { "epoch": 0.39, "learning_rate": 1.9327578921953283e-05, "loss": 1.2666, "step": 6412 }, { "epoch": 0.39, "learning_rate": 1.9333615017806484e-05, "loss": 1.2434, "step": 6414 }, { "epoch": 0.39, "learning_rate": 1.9339651113659685e-05, "loss": 1.5206, "step": 6416 }, { "epoch": 0.39, "learning_rate": 1.934568720951289e-05, "loss": 1.5359, "step": 6418 }, { "epoch": 0.39, "learning_rate": 1.935172330536609e-05, "loss": 1.2015, "step": 6420 }, { "epoch": 0.39, "learning_rate": 1.935775940121929e-05, "loss": 1.5882, "step": 6422 }, { "epoch": 0.39, "learning_rate": 1.9363795497072495e-05, "loss": 1.5987, "step": 6424 }, { "epoch": 0.39, "learning_rate": 1.9369831592925696e-05, "loss": 1.2727, "step": 6426 }, { "epoch": 0.39, "learning_rate": 1.9375867688778897e-05, "loss": 1.6167, "step": 6428 }, { "epoch": 0.39, "learning_rate": 1.93819037846321e-05, "loss": 1.0924, "step": 6430 }, { "epoch": 0.39, "learning_rate": 1.9387939880485305e-05, "loss": 1.0554, "step": 6432 }, { "epoch": 0.39, "learning_rate": 1.9393975976338506e-05, "loss": 1.4291, "step": 6434 }, { "epoch": 0.39, "learning_rate": 1.9400012072191707e-05, "loss": 1.3559, "step": 6436 }, { "epoch": 0.39, "learning_rate": 1.940604816804491e-05, "loss": 1.169, "step": 6438 }, { "epoch": 0.39, "learning_rate": 1.9412084263898112e-05, "loss": 1.5845, "step": 6440 }, { "epoch": 0.39, "learning_rate": 1.9418120359751313e-05, "loss": 1.3207, "step": 6442 }, { "epoch": 0.39, "learning_rate": 1.9424156455604517e-05, "loss": 1.339, "step": 6444 }, { "epoch": 0.39, "learning_rate": 1.9430192551457718e-05, "loss": 1.1374, "step": 6446 }, { "epoch": 0.39, "learning_rate": 1.943622864731092e-05, "loss": 1.5865, "step": 6448 }, { "epoch": 0.39, "learning_rate": 1.9442264743164123e-05, "loss": 0.9296, "step": 6450 }, { "epoch": 0.39, "learning_rate": 1.9448300839017323e-05, "loss": 1.4097, "step": 6452 }, { "epoch": 0.39, "learning_rate": 1.9454336934870528e-05, "loss": 1.2729, "step": 6454 }, { "epoch": 0.39, "learning_rate": 1.946037303072373e-05, "loss": 1.7283, "step": 6456 }, { "epoch": 0.39, "learning_rate": 1.9466409126576933e-05, "loss": 0.9549, "step": 6458 }, { "epoch": 0.39, "learning_rate": 1.9472445222430134e-05, "loss": 1.0615, "step": 6460 }, { "epoch": 0.39, "learning_rate": 1.9478481318283334e-05, "loss": 1.1485, "step": 6462 }, { "epoch": 0.39, "learning_rate": 1.948451741413654e-05, "loss": 1.23, "step": 6464 }, { "epoch": 0.39, "learning_rate": 1.949055350998974e-05, "loss": 1.5381, "step": 6466 }, { "epoch": 0.39, "learning_rate": 1.949658960584294e-05, "loss": 1.3493, "step": 6468 }, { "epoch": 0.39, "learning_rate": 1.9502625701696144e-05, "loss": 1.2158, "step": 6470 }, { "epoch": 0.39, "learning_rate": 1.9508661797549345e-05, "loss": 1.3801, "step": 6472 }, { "epoch": 0.39, "learning_rate": 1.9514697893402546e-05, "loss": 0.9857, "step": 6474 }, { "epoch": 0.39, "learning_rate": 1.952073398925575e-05, "loss": 1.2721, "step": 6476 }, { "epoch": 0.39, "learning_rate": 1.9526770085108954e-05, "loss": 1.2985, "step": 6478 }, { "epoch": 0.39, "learning_rate": 1.9532806180962155e-05, "loss": 1.2917, "step": 6480 }, { "epoch": 0.39, "learning_rate": 1.9538842276815356e-05, "loss": 1.0878, "step": 6482 }, { "epoch": 0.39, "learning_rate": 1.954487837266856e-05, "loss": 0.9316, "step": 6484 }, { "epoch": 0.39, "learning_rate": 1.955091446852176e-05, "loss": 1.562, "step": 6486 }, { "epoch": 0.39, "learning_rate": 1.9556950564374962e-05, "loss": 1.2688, "step": 6488 }, { "epoch": 0.39, "learning_rate": 1.9562986660228166e-05, "loss": 1.2984, "step": 6490 }, { "epoch": 0.39, "learning_rate": 1.9569022756081367e-05, "loss": 1.3822, "step": 6492 }, { "epoch": 0.39, "learning_rate": 1.9575058851934568e-05, "loss": 1.2623, "step": 6494 }, { "epoch": 0.39, "learning_rate": 1.9581094947787772e-05, "loss": 1.1557, "step": 6496 }, { "epoch": 0.39, "learning_rate": 1.9587131043640973e-05, "loss": 1.1125, "step": 6498 }, { "epoch": 0.39, "learning_rate": 1.9593167139494177e-05, "loss": 1.0925, "step": 6500 }, { "epoch": 0.39, "learning_rate": 1.9599203235347378e-05, "loss": 1.2223, "step": 6502 }, { "epoch": 0.39, "learning_rate": 1.9605239331200582e-05, "loss": 1.2561, "step": 6504 }, { "epoch": 0.39, "learning_rate": 1.9611275427053783e-05, "loss": 0.961, "step": 6506 }, { "epoch": 0.39, "learning_rate": 1.9617311522906984e-05, "loss": 1.332, "step": 6508 }, { "epoch": 0.39, "learning_rate": 1.9623347618760188e-05, "loss": 1.2056, "step": 6510 }, { "epoch": 0.39, "learning_rate": 1.962938371461339e-05, "loss": 1.0892, "step": 6512 }, { "epoch": 0.39, "learning_rate": 1.963541981046659e-05, "loss": 1.4333, "step": 6514 }, { "epoch": 0.39, "learning_rate": 1.9641455906319794e-05, "loss": 1.1951, "step": 6516 }, { "epoch": 0.39, "learning_rate": 1.9647492002172995e-05, "loss": 1.343, "step": 6518 }, { "epoch": 0.39, "learning_rate": 1.96535280980262e-05, "loss": 1.1867, "step": 6520 }, { "epoch": 0.39, "learning_rate": 1.96595641938794e-05, "loss": 1.1378, "step": 6522 }, { "epoch": 0.39, "learning_rate": 1.9665600289732604e-05, "loss": 1.3842, "step": 6524 }, { "epoch": 0.39, "learning_rate": 1.9671636385585805e-05, "loss": 1.0901, "step": 6526 }, { "epoch": 0.39, "learning_rate": 1.9677672481439005e-05, "loss": 1.0661, "step": 6528 }, { "epoch": 0.39, "learning_rate": 1.968370857729221e-05, "loss": 1.4551, "step": 6530 }, { "epoch": 0.39, "learning_rate": 1.968974467314541e-05, "loss": 1.4772, "step": 6532 }, { "epoch": 0.39, "learning_rate": 1.969578076899861e-05, "loss": 1.1517, "step": 6534 }, { "epoch": 0.39, "learning_rate": 1.9701816864851815e-05, "loss": 1.217, "step": 6536 }, { "epoch": 0.39, "learning_rate": 1.9707852960705016e-05, "loss": 1.0187, "step": 6538 }, { "epoch": 0.39, "learning_rate": 1.9713889056558217e-05, "loss": 1.0704, "step": 6540 }, { "epoch": 0.39, "learning_rate": 1.971992515241142e-05, "loss": 1.0535, "step": 6542 }, { "epoch": 0.39, "learning_rate": 1.9725961248264626e-05, "loss": 1.0608, "step": 6544 }, { "epoch": 0.4, "learning_rate": 1.9731997344117826e-05, "loss": 1.2197, "step": 6546 }, { "epoch": 0.4, "learning_rate": 1.9738033439971027e-05, "loss": 1.4942, "step": 6548 }, { "epoch": 0.4, "learning_rate": 1.974406953582423e-05, "loss": 1.6174, "step": 6550 }, { "epoch": 0.4, "learning_rate": 1.9750105631677432e-05, "loss": 1.2655, "step": 6552 }, { "epoch": 0.4, "learning_rate": 1.9756141727530633e-05, "loss": 0.8719, "step": 6554 }, { "epoch": 0.4, "learning_rate": 1.9762177823383837e-05, "loss": 1.4757, "step": 6556 }, { "epoch": 0.4, "learning_rate": 1.9768213919237038e-05, "loss": 1.295, "step": 6558 }, { "epoch": 0.4, "learning_rate": 1.977425001509024e-05, "loss": 0.8928, "step": 6560 }, { "epoch": 0.4, "learning_rate": 1.9780286110943443e-05, "loss": 1.3492, "step": 6562 }, { "epoch": 0.4, "learning_rate": 1.9786322206796644e-05, "loss": 1.2283, "step": 6564 }, { "epoch": 0.4, "learning_rate": 1.9792358302649848e-05, "loss": 1.0171, "step": 6566 }, { "epoch": 0.4, "learning_rate": 1.979839439850305e-05, "loss": 1.2026, "step": 6568 }, { "epoch": 0.4, "learning_rate": 1.9804430494356253e-05, "loss": 1.3499, "step": 6570 }, { "epoch": 0.4, "learning_rate": 1.9810466590209454e-05, "loss": 1.5191, "step": 6572 }, { "epoch": 0.4, "learning_rate": 1.9816502686062655e-05, "loss": 1.1804, "step": 6574 }, { "epoch": 0.4, "learning_rate": 1.982253878191586e-05, "loss": 1.3739, "step": 6576 }, { "epoch": 0.4, "learning_rate": 1.982857487776906e-05, "loss": 1.2302, "step": 6578 }, { "epoch": 0.4, "learning_rate": 1.983461097362226e-05, "loss": 1.3159, "step": 6580 }, { "epoch": 0.4, "learning_rate": 1.9840647069475465e-05, "loss": 1.5292, "step": 6582 }, { "epoch": 0.4, "learning_rate": 1.9846683165328666e-05, "loss": 1.3085, "step": 6584 }, { "epoch": 0.4, "learning_rate": 1.9852719261181866e-05, "loss": 1.094, "step": 6586 }, { "epoch": 0.4, "learning_rate": 1.985875535703507e-05, "loss": 1.3682, "step": 6588 }, { "epoch": 0.4, "learning_rate": 1.9864791452888275e-05, "loss": 1.0502, "step": 6590 }, { "epoch": 0.4, "learning_rate": 1.9870827548741476e-05, "loss": 1.4761, "step": 6592 }, { "epoch": 0.4, "learning_rate": 1.9876863644594676e-05, "loss": 1.1997, "step": 6594 }, { "epoch": 0.4, "learning_rate": 1.988289974044788e-05, "loss": 1.2164, "step": 6596 }, { "epoch": 0.4, "learning_rate": 1.988893583630108e-05, "loss": 1.236, "step": 6598 }, { "epoch": 0.4, "learning_rate": 1.9894971932154282e-05, "loss": 1.5476, "step": 6600 }, { "epoch": 0.4, "learning_rate": 1.9901008028007487e-05, "loss": 1.0351, "step": 6602 }, { "epoch": 0.4, "learning_rate": 1.9907044123860687e-05, "loss": 1.2792, "step": 6604 }, { "epoch": 0.4, "learning_rate": 1.9913080219713888e-05, "loss": 1.3107, "step": 6606 }, { "epoch": 0.4, "learning_rate": 1.9919116315567092e-05, "loss": 1.1416, "step": 6608 }, { "epoch": 0.4, "learning_rate": 1.9925152411420293e-05, "loss": 1.5425, "step": 6610 }, { "epoch": 0.4, "learning_rate": 1.9931188507273497e-05, "loss": 1.155, "step": 6612 }, { "epoch": 0.4, "learning_rate": 1.9937224603126698e-05, "loss": 1.1135, "step": 6614 }, { "epoch": 0.4, "learning_rate": 1.9943260698979902e-05, "loss": 1.0552, "step": 6616 }, { "epoch": 0.4, "learning_rate": 1.9949296794833103e-05, "loss": 1.284, "step": 6618 }, { "epoch": 0.4, "learning_rate": 1.9955332890686304e-05, "loss": 1.0419, "step": 6620 }, { "epoch": 0.4, "learning_rate": 1.9961368986539508e-05, "loss": 1.121, "step": 6622 }, { "epoch": 0.4, "learning_rate": 1.996740508239271e-05, "loss": 1.2268, "step": 6624 }, { "epoch": 0.4, "learning_rate": 1.997344117824591e-05, "loss": 1.1699, "step": 6626 }, { "epoch": 0.4, "learning_rate": 1.9979477274099114e-05, "loss": 1.6158, "step": 6628 }, { "epoch": 0.4, "learning_rate": 1.9985513369952315e-05, "loss": 1.1396, "step": 6630 }, { "epoch": 0.4, "learning_rate": 1.999154946580552e-05, "loss": 1.7365, "step": 6632 }, { "epoch": 0.4, "learning_rate": 1.999758556165872e-05, "loss": 1.048, "step": 6634 }, { "epoch": 0.4, "learning_rate": 2.0003621657511924e-05, "loss": 1.2331, "step": 6636 }, { "epoch": 0.4, "learning_rate": 2.0009657753365125e-05, "loss": 1.4095, "step": 6638 }, { "epoch": 0.4, "learning_rate": 2.0015693849218326e-05, "loss": 1.0954, "step": 6640 }, { "epoch": 0.4, "learning_rate": 2.002172994507153e-05, "loss": 1.2311, "step": 6642 }, { "epoch": 0.4, "learning_rate": 2.002776604092473e-05, "loss": 1.2294, "step": 6644 }, { "epoch": 0.4, "learning_rate": 2.003380213677793e-05, "loss": 1.2132, "step": 6646 }, { "epoch": 0.4, "learning_rate": 2.0039838232631136e-05, "loss": 1.0673, "step": 6648 }, { "epoch": 0.4, "learning_rate": 2.0045874328484337e-05, "loss": 0.9353, "step": 6650 }, { "epoch": 0.4, "learning_rate": 2.0051910424337537e-05, "loss": 1.5369, "step": 6652 }, { "epoch": 0.4, "learning_rate": 2.005794652019074e-05, "loss": 1.6725, "step": 6654 }, { "epoch": 0.4, "learning_rate": 2.0063982616043946e-05, "loss": 1.3886, "step": 6656 }, { "epoch": 0.4, "learning_rate": 2.0070018711897147e-05, "loss": 1.1944, "step": 6658 }, { "epoch": 0.4, "learning_rate": 2.0076054807750348e-05, "loss": 1.2207, "step": 6660 }, { "epoch": 0.4, "learning_rate": 2.0082090903603552e-05, "loss": 1.179, "step": 6662 }, { "epoch": 0.4, "learning_rate": 2.0088126999456753e-05, "loss": 1.2024, "step": 6664 }, { "epoch": 0.4, "learning_rate": 2.0094163095309953e-05, "loss": 1.177, "step": 6666 }, { "epoch": 0.4, "learning_rate": 2.0100199191163158e-05, "loss": 1.1517, "step": 6668 }, { "epoch": 0.4, "learning_rate": 2.010623528701636e-05, "loss": 1.1943, "step": 6670 }, { "epoch": 0.4, "learning_rate": 2.011227138286956e-05, "loss": 1.5286, "step": 6672 }, { "epoch": 0.4, "learning_rate": 2.0118307478722763e-05, "loss": 1.4603, "step": 6674 }, { "epoch": 0.4, "learning_rate": 2.0124343574575964e-05, "loss": 1.4967, "step": 6676 }, { "epoch": 0.4, "learning_rate": 2.013037967042917e-05, "loss": 1.3822, "step": 6678 }, { "epoch": 0.4, "learning_rate": 2.013641576628237e-05, "loss": 1.2121, "step": 6680 }, { "epoch": 0.4, "learning_rate": 2.0142451862135573e-05, "loss": 1.0986, "step": 6682 }, { "epoch": 0.4, "learning_rate": 2.0148487957988774e-05, "loss": 1.0205, "step": 6684 }, { "epoch": 0.4, "learning_rate": 2.0154524053841975e-05, "loss": 1.0676, "step": 6686 }, { "epoch": 0.4, "learning_rate": 2.016056014969518e-05, "loss": 1.2425, "step": 6688 }, { "epoch": 0.4, "learning_rate": 2.016659624554838e-05, "loss": 1.4322, "step": 6690 }, { "epoch": 0.4, "learning_rate": 2.017263234140158e-05, "loss": 1.3233, "step": 6692 }, { "epoch": 0.4, "learning_rate": 2.0178668437254785e-05, "loss": 1.2023, "step": 6694 }, { "epoch": 0.4, "learning_rate": 2.0184704533107986e-05, "loss": 1.0965, "step": 6696 }, { "epoch": 0.4, "learning_rate": 2.0190740628961187e-05, "loss": 1.2482, "step": 6698 }, { "epoch": 0.4, "learning_rate": 2.019677672481439e-05, "loss": 1.3513, "step": 6700 }, { "epoch": 0.4, "learning_rate": 2.0202812820667595e-05, "loss": 1.2845, "step": 6702 }, { "epoch": 0.4, "learning_rate": 2.0208848916520796e-05, "loss": 1.2952, "step": 6704 }, { "epoch": 0.4, "learning_rate": 2.0214885012373997e-05, "loss": 1.0907, "step": 6706 }, { "epoch": 0.4, "learning_rate": 2.02209211082272e-05, "loss": 1.2984, "step": 6708 }, { "epoch": 0.4, "learning_rate": 2.0226957204080402e-05, "loss": 1.2721, "step": 6710 }, { "epoch": 0.41, "learning_rate": 2.0232993299933603e-05, "loss": 1.1057, "step": 6712 }, { "epoch": 0.41, "learning_rate": 2.0239029395786807e-05, "loss": 1.6216, "step": 6714 }, { "epoch": 0.41, "learning_rate": 2.0245065491640008e-05, "loss": 1.3809, "step": 6716 }, { "epoch": 0.41, "learning_rate": 2.025110158749321e-05, "loss": 1.304, "step": 6718 }, { "epoch": 0.41, "learning_rate": 2.0257137683346413e-05, "loss": 1.6635, "step": 6720 }, { "epoch": 0.41, "learning_rate": 2.0263173779199614e-05, "loss": 1.0496, "step": 6722 }, { "epoch": 0.41, "learning_rate": 2.0269209875052818e-05, "loss": 0.9955, "step": 6724 }, { "epoch": 0.41, "learning_rate": 2.027524597090602e-05, "loss": 1.3569, "step": 6726 }, { "epoch": 0.41, "learning_rate": 2.0281282066759223e-05, "loss": 0.9552, "step": 6728 }, { "epoch": 0.41, "learning_rate": 2.0287318162612424e-05, "loss": 2.1179, "step": 6730 }, { "epoch": 0.41, "learning_rate": 2.0293354258465624e-05, "loss": 1.3799, "step": 6732 }, { "epoch": 0.41, "learning_rate": 2.029939035431883e-05, "loss": 1.1799, "step": 6734 }, { "epoch": 0.41, "learning_rate": 2.030542645017203e-05, "loss": 1.3566, "step": 6736 }, { "epoch": 0.41, "learning_rate": 2.031146254602523e-05, "loss": 1.3008, "step": 6738 }, { "epoch": 0.41, "learning_rate": 2.0317498641878434e-05, "loss": 1.1092, "step": 6740 }, { "epoch": 0.41, "learning_rate": 2.0323534737731635e-05, "loss": 1.4922, "step": 6742 }, { "epoch": 0.41, "learning_rate": 2.032957083358484e-05, "loss": 1.6949, "step": 6744 }, { "epoch": 0.41, "learning_rate": 2.033560692943804e-05, "loss": 1.6347, "step": 6746 }, { "epoch": 0.41, "learning_rate": 2.0341643025291245e-05, "loss": 1.1508, "step": 6748 }, { "epoch": 0.41, "learning_rate": 2.0347679121144445e-05, "loss": 1.2564, "step": 6750 }, { "epoch": 0.41, "learning_rate": 2.0353715216997646e-05, "loss": 1.2399, "step": 6752 }, { "epoch": 0.41, "learning_rate": 2.035975131285085e-05, "loss": 1.6229, "step": 6754 }, { "epoch": 0.41, "learning_rate": 2.036578740870405e-05, "loss": 0.9953, "step": 6756 }, { "epoch": 0.41, "learning_rate": 2.0371823504557252e-05, "loss": 1.4605, "step": 6758 }, { "epoch": 0.41, "learning_rate": 2.0377859600410456e-05, "loss": 1.2497, "step": 6760 }, { "epoch": 0.41, "learning_rate": 2.0383895696263657e-05, "loss": 1.6368, "step": 6762 }, { "epoch": 0.41, "learning_rate": 2.0389931792116858e-05, "loss": 1.1935, "step": 6764 }, { "epoch": 0.41, "learning_rate": 2.0395967887970062e-05, "loss": 1.0954, "step": 6766 }, { "epoch": 0.41, "learning_rate": 2.0402003983823266e-05, "loss": 1.2045, "step": 6768 }, { "epoch": 0.41, "learning_rate": 2.0408040079676467e-05, "loss": 1.4533, "step": 6770 }, { "epoch": 0.41, "learning_rate": 2.0414076175529668e-05, "loss": 1.3614, "step": 6772 }, { "epoch": 0.41, "learning_rate": 2.0420112271382872e-05, "loss": 1.3807, "step": 6774 }, { "epoch": 0.41, "learning_rate": 2.0426148367236073e-05, "loss": 1.102, "step": 6776 }, { "epoch": 0.41, "learning_rate": 2.0432184463089274e-05, "loss": 1.2505, "step": 6778 }, { "epoch": 0.41, "learning_rate": 2.0438220558942478e-05, "loss": 0.9107, "step": 6780 }, { "epoch": 0.41, "learning_rate": 2.044425665479568e-05, "loss": 2.0283, "step": 6782 }, { "epoch": 0.41, "learning_rate": 2.045029275064888e-05, "loss": 0.9599, "step": 6784 }, { "epoch": 0.41, "learning_rate": 2.0456328846502084e-05, "loss": 0.9963, "step": 6786 }, { "epoch": 0.41, "learning_rate": 2.0462364942355285e-05, "loss": 1.0882, "step": 6788 }, { "epoch": 0.41, "learning_rate": 2.046840103820849e-05, "loss": 1.0468, "step": 6790 }, { "epoch": 0.41, "learning_rate": 2.047443713406169e-05, "loss": 1.2262, "step": 6792 }, { "epoch": 0.41, "learning_rate": 2.0480473229914894e-05, "loss": 1.3808, "step": 6794 }, { "epoch": 0.41, "learning_rate": 2.0486509325768095e-05, "loss": 1.1023, "step": 6796 }, { "epoch": 0.41, "learning_rate": 2.0492545421621295e-05, "loss": 0.8823, "step": 6798 }, { "epoch": 0.41, "learning_rate": 2.04985815174745e-05, "loss": 1.3089, "step": 6800 }, { "epoch": 0.41, "learning_rate": 2.05046176133277e-05, "loss": 1.4129, "step": 6802 }, { "epoch": 0.41, "learning_rate": 2.05106537091809e-05, "loss": 1.4887, "step": 6804 }, { "epoch": 0.41, "learning_rate": 2.0516689805034106e-05, "loss": 1.3807, "step": 6806 }, { "epoch": 0.41, "learning_rate": 2.0522725900887306e-05, "loss": 1.2895, "step": 6808 }, { "epoch": 0.41, "learning_rate": 2.0528761996740507e-05, "loss": 1.0904, "step": 6810 }, { "epoch": 0.41, "learning_rate": 2.053479809259371e-05, "loss": 1.0504, "step": 6812 }, { "epoch": 0.41, "learning_rate": 2.0540834188446916e-05, "loss": 1.0895, "step": 6814 }, { "epoch": 0.41, "learning_rate": 2.0546870284300116e-05, "loss": 1.1249, "step": 6816 }, { "epoch": 0.41, "learning_rate": 2.0552906380153317e-05, "loss": 1.1188, "step": 6818 }, { "epoch": 0.41, "learning_rate": 2.055894247600652e-05, "loss": 1.5192, "step": 6820 }, { "epoch": 0.41, "learning_rate": 2.0564978571859722e-05, "loss": 1.2301, "step": 6822 }, { "epoch": 0.41, "learning_rate": 2.0571014667712923e-05, "loss": 1.2395, "step": 6824 }, { "epoch": 0.41, "learning_rate": 2.0577050763566127e-05, "loss": 1.2764, "step": 6826 }, { "epoch": 0.41, "learning_rate": 2.0583086859419328e-05, "loss": 1.0696, "step": 6828 }, { "epoch": 0.41, "learning_rate": 2.058912295527253e-05, "loss": 1.2785, "step": 6830 }, { "epoch": 0.41, "learning_rate": 2.0595159051125733e-05, "loss": 1.173, "step": 6832 }, { "epoch": 0.41, "learning_rate": 2.0601195146978934e-05, "loss": 1.264, "step": 6834 }, { "epoch": 0.41, "learning_rate": 2.0607231242832138e-05, "loss": 1.1657, "step": 6836 }, { "epoch": 0.41, "learning_rate": 2.061326733868534e-05, "loss": 1.2457, "step": 6838 }, { "epoch": 0.41, "learning_rate": 2.0619303434538543e-05, "loss": 1.216, "step": 6840 }, { "epoch": 0.41, "learning_rate": 2.0625339530391744e-05, "loss": 1.24, "step": 6842 }, { "epoch": 0.41, "learning_rate": 2.0631375626244945e-05, "loss": 1.0774, "step": 6844 }, { "epoch": 0.41, "learning_rate": 2.063741172209815e-05, "loss": 1.427, "step": 6846 }, { "epoch": 0.41, "learning_rate": 2.064344781795135e-05, "loss": 0.976, "step": 6848 }, { "epoch": 0.41, "learning_rate": 2.064948391380455e-05, "loss": 1.1939, "step": 6850 }, { "epoch": 0.41, "learning_rate": 2.0655520009657755e-05, "loss": 1.373, "step": 6852 }, { "epoch": 0.41, "learning_rate": 2.0661556105510956e-05, "loss": 1.0791, "step": 6854 }, { "epoch": 0.41, "learning_rate": 2.0667592201364157e-05, "loss": 1.3345, "step": 6856 }, { "epoch": 0.41, "learning_rate": 2.067362829721736e-05, "loss": 1.0116, "step": 6858 }, { "epoch": 0.41, "learning_rate": 2.0679664393070565e-05, "loss": 1.1556, "step": 6860 }, { "epoch": 0.41, "learning_rate": 2.0685700488923766e-05, "loss": 1.5645, "step": 6862 }, { "epoch": 0.41, "learning_rate": 2.0691736584776967e-05, "loss": 1.1894, "step": 6864 }, { "epoch": 0.41, "learning_rate": 2.069777268063017e-05, "loss": 1.0744, "step": 6866 }, { "epoch": 0.41, "learning_rate": 2.070380877648337e-05, "loss": 1.6243, "step": 6868 }, { "epoch": 0.41, "learning_rate": 2.0709844872336572e-05, "loss": 1.2962, "step": 6870 }, { "epoch": 0.41, "learning_rate": 2.0715880968189777e-05, "loss": 1.4599, "step": 6872 }, { "epoch": 0.41, "learning_rate": 2.0721917064042977e-05, "loss": 1.1861, "step": 6874 }, { "epoch": 0.42, "learning_rate": 2.0727953159896178e-05, "loss": 1.3368, "step": 6876 }, { "epoch": 0.42, "learning_rate": 2.0733989255749382e-05, "loss": 1.9533, "step": 6878 }, { "epoch": 0.42, "learning_rate": 2.0740025351602587e-05, "loss": 1.3398, "step": 6880 }, { "epoch": 0.42, "learning_rate": 2.0746061447455787e-05, "loss": 1.5674, "step": 6882 }, { "epoch": 0.42, "learning_rate": 2.0752097543308988e-05, "loss": 1.218, "step": 6884 }, { "epoch": 0.42, "learning_rate": 2.0758133639162193e-05, "loss": 1.354, "step": 6886 }, { "epoch": 0.42, "learning_rate": 2.0764169735015393e-05, "loss": 1.7464, "step": 6888 }, { "epoch": 0.42, "learning_rate": 2.0770205830868594e-05, "loss": 1.7029, "step": 6890 }, { "epoch": 0.42, "learning_rate": 2.07762419267218e-05, "loss": 1.328, "step": 6892 }, { "epoch": 0.42, "learning_rate": 2.0782278022575e-05, "loss": 1.6973, "step": 6894 }, { "epoch": 0.42, "learning_rate": 2.07883141184282e-05, "loss": 1.071, "step": 6896 }, { "epoch": 0.42, "learning_rate": 2.0794350214281404e-05, "loss": 1.5642, "step": 6898 }, { "epoch": 0.42, "learning_rate": 2.0800386310134605e-05, "loss": 1.2357, "step": 6900 }, { "epoch": 0.42, "learning_rate": 2.080642240598781e-05, "loss": 1.1492, "step": 6902 }, { "epoch": 0.42, "learning_rate": 2.081245850184101e-05, "loss": 2.0775, "step": 6904 }, { "epoch": 0.42, "learning_rate": 2.0818494597694214e-05, "loss": 1.1932, "step": 6906 }, { "epoch": 0.42, "learning_rate": 2.0824530693547415e-05, "loss": 1.1709, "step": 6908 }, { "epoch": 0.42, "learning_rate": 2.0830566789400616e-05, "loss": 1.2716, "step": 6910 }, { "epoch": 0.42, "learning_rate": 2.083660288525382e-05, "loss": 1.3626, "step": 6912 }, { "epoch": 0.42, "learning_rate": 2.084263898110702e-05, "loss": 1.5924, "step": 6914 }, { "epoch": 0.42, "learning_rate": 2.0848675076960222e-05, "loss": 1.1444, "step": 6916 }, { "epoch": 0.42, "learning_rate": 2.0854711172813426e-05, "loss": 1.1442, "step": 6918 }, { "epoch": 0.42, "learning_rate": 2.0860747268666627e-05, "loss": 1.2084, "step": 6920 }, { "epoch": 0.42, "learning_rate": 2.0866783364519828e-05, "loss": 1.4875, "step": 6922 }, { "epoch": 0.42, "learning_rate": 2.0872819460373032e-05, "loss": 1.7229, "step": 6924 }, { "epoch": 0.42, "learning_rate": 2.0878855556226236e-05, "loss": 1.1349, "step": 6926 }, { "epoch": 0.42, "learning_rate": 2.0884891652079437e-05, "loss": 1.5772, "step": 6928 }, { "epoch": 0.42, "learning_rate": 2.0890927747932638e-05, "loss": 1.578, "step": 6930 }, { "epoch": 0.42, "learning_rate": 2.0896963843785842e-05, "loss": 1.2324, "step": 6932 }, { "epoch": 0.42, "learning_rate": 2.0902999939639043e-05, "loss": 1.2696, "step": 6934 }, { "epoch": 0.42, "learning_rate": 2.0909036035492243e-05, "loss": 1.2579, "step": 6936 }, { "epoch": 0.42, "learning_rate": 2.0915072131345448e-05, "loss": 1.3158, "step": 6938 }, { "epoch": 0.42, "learning_rate": 2.092110822719865e-05, "loss": 1.0851, "step": 6940 }, { "epoch": 0.42, "learning_rate": 2.092714432305185e-05, "loss": 1.4331, "step": 6942 }, { "epoch": 0.42, "learning_rate": 2.0933180418905054e-05, "loss": 1.5676, "step": 6944 }, { "epoch": 0.42, "learning_rate": 2.0939216514758254e-05, "loss": 1.2829, "step": 6946 }, { "epoch": 0.42, "learning_rate": 2.094525261061146e-05, "loss": 1.1772, "step": 6948 }, { "epoch": 0.42, "learning_rate": 2.095128870646466e-05, "loss": 1.325, "step": 6950 }, { "epoch": 0.42, "learning_rate": 2.0957324802317864e-05, "loss": 1.3413, "step": 6952 }, { "epoch": 0.42, "learning_rate": 2.0963360898171064e-05, "loss": 1.4205, "step": 6954 }, { "epoch": 0.42, "learning_rate": 2.0969396994024265e-05, "loss": 1.277, "step": 6956 }, { "epoch": 0.42, "learning_rate": 2.097543308987747e-05, "loss": 1.3345, "step": 6958 }, { "epoch": 0.42, "learning_rate": 2.098146918573067e-05, "loss": 1.5656, "step": 6960 }, { "epoch": 0.42, "learning_rate": 2.098750528158387e-05, "loss": 1.2236, "step": 6962 }, { "epoch": 0.42, "learning_rate": 2.0993541377437075e-05, "loss": 1.1081, "step": 6964 }, { "epoch": 0.42, "learning_rate": 2.0999577473290276e-05, "loss": 1.2607, "step": 6966 }, { "epoch": 0.42, "learning_rate": 2.1005613569143477e-05, "loss": 1.4263, "step": 6968 }, { "epoch": 0.42, "learning_rate": 2.101164966499668e-05, "loss": 1.6015, "step": 6970 }, { "epoch": 0.42, "learning_rate": 2.1017685760849885e-05, "loss": 1.4657, "step": 6972 }, { "epoch": 0.42, "learning_rate": 2.1023721856703086e-05, "loss": 1.3489, "step": 6974 }, { "epoch": 0.42, "learning_rate": 2.1029757952556287e-05, "loss": 1.605, "step": 6976 }, { "epoch": 0.42, "learning_rate": 2.103579404840949e-05, "loss": 1.3067, "step": 6978 }, { "epoch": 0.42, "learning_rate": 2.1041830144262692e-05, "loss": 1.5179, "step": 6980 }, { "epoch": 0.42, "learning_rate": 2.1047866240115893e-05, "loss": 1.3656, "step": 6982 }, { "epoch": 0.42, "learning_rate": 2.1053902335969097e-05, "loss": 1.0224, "step": 6984 }, { "epoch": 0.42, "learning_rate": 2.1059938431822298e-05, "loss": 1.5963, "step": 6986 }, { "epoch": 0.42, "learning_rate": 2.10659745276755e-05, "loss": 1.4694, "step": 6988 }, { "epoch": 0.42, "learning_rate": 2.1072010623528703e-05, "loss": 0.9756, "step": 6990 }, { "epoch": 0.42, "learning_rate": 2.1078046719381904e-05, "loss": 1.3264, "step": 6992 }, { "epoch": 0.42, "learning_rate": 2.1084082815235108e-05, "loss": 0.9978, "step": 6994 }, { "epoch": 0.42, "learning_rate": 2.109011891108831e-05, "loss": 1.4128, "step": 6996 }, { "epoch": 0.42, "learning_rate": 2.1096155006941513e-05, "loss": 1.1991, "step": 6998 }, { "epoch": 0.42, "learning_rate": 2.1102191102794714e-05, "loss": 1.0495, "step": 7000 }, { "epoch": 0.42, "learning_rate": 2.1108227198647915e-05, "loss": 1.3867, "step": 7002 }, { "epoch": 0.42, "learning_rate": 2.111426329450112e-05, "loss": 0.9597, "step": 7004 }, { "epoch": 0.42, "learning_rate": 2.112029939035432e-05, "loss": 1.3847, "step": 7006 }, { "epoch": 0.42, "learning_rate": 2.112633548620752e-05, "loss": 1.2576, "step": 7008 }, { "epoch": 0.42, "learning_rate": 2.1132371582060725e-05, "loss": 1.2122, "step": 7010 }, { "epoch": 0.42, "learning_rate": 2.1138407677913925e-05, "loss": 1.2995, "step": 7012 }, { "epoch": 0.42, "learning_rate": 2.114444377376713e-05, "loss": 1.6268, "step": 7014 }, { "epoch": 0.42, "learning_rate": 2.115047986962033e-05, "loss": 1.4599, "step": 7016 }, { "epoch": 0.42, "learning_rate": 2.1156515965473535e-05, "loss": 1.2443, "step": 7018 }, { "epoch": 0.42, "learning_rate": 2.1162552061326735e-05, "loss": 0.9616, "step": 7020 }, { "epoch": 0.42, "learning_rate": 2.1168588157179936e-05, "loss": 1.2133, "step": 7022 }, { "epoch": 0.42, "learning_rate": 2.117462425303314e-05, "loss": 1.1519, "step": 7024 }, { "epoch": 0.42, "learning_rate": 2.118066034888634e-05, "loss": 1.3297, "step": 7026 }, { "epoch": 0.42, "learning_rate": 2.1186696444739542e-05, "loss": 1.0729, "step": 7028 }, { "epoch": 0.42, "learning_rate": 2.1192732540592746e-05, "loss": 1.3142, "step": 7030 }, { "epoch": 0.42, "learning_rate": 2.1198768636445947e-05, "loss": 1.3317, "step": 7032 }, { "epoch": 0.42, "learning_rate": 2.1204804732299148e-05, "loss": 1.3936, "step": 7034 }, { "epoch": 0.42, "learning_rate": 2.1210840828152352e-05, "loss": 1.2426, "step": 7036 }, { "epoch": 0.42, "learning_rate": 2.1216876924005556e-05, "loss": 1.0664, "step": 7038 }, { "epoch": 0.42, "learning_rate": 2.1222913019858757e-05, "loss": 1.4611, "step": 7040 }, { "epoch": 0.43, "learning_rate": 2.1228949115711958e-05, "loss": 1.1823, "step": 7042 }, { "epoch": 0.43, "learning_rate": 2.1234985211565162e-05, "loss": 1.1141, "step": 7044 }, { "epoch": 0.43, "learning_rate": 2.1241021307418363e-05, "loss": 1.3474, "step": 7046 }, { "epoch": 0.43, "learning_rate": 2.1247057403271564e-05, "loss": 1.3329, "step": 7048 }, { "epoch": 0.43, "learning_rate": 2.1253093499124768e-05, "loss": 0.9536, "step": 7050 }, { "epoch": 0.43, "learning_rate": 2.125912959497797e-05, "loss": 1.4419, "step": 7052 }, { "epoch": 0.43, "learning_rate": 2.126516569083117e-05, "loss": 1.0984, "step": 7054 }, { "epoch": 0.43, "learning_rate": 2.1271201786684374e-05, "loss": 1.2905, "step": 7056 }, { "epoch": 0.43, "learning_rate": 2.1277237882537575e-05, "loss": 1.6069, "step": 7058 }, { "epoch": 0.43, "learning_rate": 2.128327397839078e-05, "loss": 1.7079, "step": 7060 }, { "epoch": 0.43, "learning_rate": 2.128931007424398e-05, "loss": 1.3582, "step": 7062 }, { "epoch": 0.43, "learning_rate": 2.1295346170097184e-05, "loss": 1.0624, "step": 7064 }, { "epoch": 0.43, "learning_rate": 2.1301382265950385e-05, "loss": 1.4815, "step": 7066 }, { "epoch": 0.43, "learning_rate": 2.1307418361803586e-05, "loss": 1.7056, "step": 7068 }, { "epoch": 0.43, "learning_rate": 2.131345445765679e-05, "loss": 1.9886, "step": 7070 }, { "epoch": 0.43, "learning_rate": 2.131949055350999e-05, "loss": 1.5096, "step": 7072 }, { "epoch": 0.43, "learning_rate": 2.132552664936319e-05, "loss": 1.7683, "step": 7074 }, { "epoch": 0.43, "learning_rate": 2.1331562745216396e-05, "loss": 1.1203, "step": 7076 }, { "epoch": 0.43, "learning_rate": 2.1337598841069596e-05, "loss": 1.32, "step": 7078 }, { "epoch": 0.43, "learning_rate": 2.1343634936922797e-05, "loss": 1.3485, "step": 7080 }, { "epoch": 0.43, "learning_rate": 2.1349671032776e-05, "loss": 0.9212, "step": 7082 }, { "epoch": 0.43, "learning_rate": 2.1355707128629206e-05, "loss": 1.1613, "step": 7084 }, { "epoch": 0.43, "learning_rate": 2.1361743224482406e-05, "loss": 1.2544, "step": 7086 }, { "epoch": 0.43, "learning_rate": 2.1367779320335607e-05, "loss": 1.2306, "step": 7088 }, { "epoch": 0.43, "learning_rate": 2.137381541618881e-05, "loss": 1.3054, "step": 7090 }, { "epoch": 0.43, "learning_rate": 2.1379851512042012e-05, "loss": 1.1107, "step": 7092 }, { "epoch": 0.43, "learning_rate": 2.1385887607895213e-05, "loss": 1.2844, "step": 7094 }, { "epoch": 0.43, "learning_rate": 2.1391923703748417e-05, "loss": 1.1628, "step": 7096 }, { "epoch": 0.43, "learning_rate": 2.1397959799601618e-05, "loss": 1.3944, "step": 7098 }, { "epoch": 0.43, "learning_rate": 2.140399589545482e-05, "loss": 1.4739, "step": 7100 }, { "epoch": 0.43, "learning_rate": 2.1410031991308023e-05, "loss": 1.3143, "step": 7102 }, { "epoch": 0.43, "learning_rate": 2.1416068087161224e-05, "loss": 1.3582, "step": 7104 }, { "epoch": 0.43, "learning_rate": 2.1422104183014428e-05, "loss": 1.0098, "step": 7106 }, { "epoch": 0.43, "learning_rate": 2.142814027886763e-05, "loss": 1.6914, "step": 7108 }, { "epoch": 0.43, "learning_rate": 2.1434176374720833e-05, "loss": 1.1738, "step": 7110 }, { "epoch": 0.43, "learning_rate": 2.1440212470574034e-05, "loss": 1.1573, "step": 7112 }, { "epoch": 0.43, "learning_rate": 2.1446248566427235e-05, "loss": 1.0107, "step": 7114 }, { "epoch": 0.43, "learning_rate": 2.145228466228044e-05, "loss": 1.1673, "step": 7116 }, { "epoch": 0.43, "learning_rate": 2.145832075813364e-05, "loss": 0.9745, "step": 7118 }, { "epoch": 0.43, "learning_rate": 2.146435685398684e-05, "loss": 0.9876, "step": 7120 }, { "epoch": 0.43, "learning_rate": 2.1470392949840045e-05, "loss": 1.3073, "step": 7122 }, { "epoch": 0.43, "learning_rate": 2.1476429045693246e-05, "loss": 1.1286, "step": 7124 }, { "epoch": 0.43, "learning_rate": 2.148246514154645e-05, "loss": 1.1513, "step": 7126 }, { "epoch": 0.43, "learning_rate": 2.148850123739965e-05, "loss": 1.1207, "step": 7128 }, { "epoch": 0.43, "learning_rate": 2.1494537333252855e-05, "loss": 1.6232, "step": 7130 }, { "epoch": 0.43, "learning_rate": 2.1500573429106056e-05, "loss": 1.5315, "step": 7132 }, { "epoch": 0.43, "learning_rate": 2.1506609524959257e-05, "loss": 1.3841, "step": 7134 }, { "epoch": 0.43, "learning_rate": 2.151264562081246e-05, "loss": 1.0713, "step": 7136 }, { "epoch": 0.43, "learning_rate": 2.151868171666566e-05, "loss": 1.2498, "step": 7138 }, { "epoch": 0.43, "learning_rate": 2.1524717812518862e-05, "loss": 1.0771, "step": 7140 }, { "epoch": 0.43, "learning_rate": 2.1530753908372067e-05, "loss": 1.2462, "step": 7142 }, { "epoch": 0.43, "learning_rate": 2.1536790004225268e-05, "loss": 1.4241, "step": 7144 }, { "epoch": 0.43, "learning_rate": 2.154282610007847e-05, "loss": 1.3793, "step": 7146 }, { "epoch": 0.43, "learning_rate": 2.1548862195931673e-05, "loss": 1.3666, "step": 7148 }, { "epoch": 0.43, "learning_rate": 2.1554898291784877e-05, "loss": 1.0233, "step": 7150 }, { "epoch": 0.43, "learning_rate": 2.1560934387638078e-05, "loss": 1.2336, "step": 7152 }, { "epoch": 0.43, "learning_rate": 2.156697048349128e-05, "loss": 1.7705, "step": 7154 }, { "epoch": 0.43, "learning_rate": 2.1573006579344483e-05, "loss": 1.2488, "step": 7156 }, { "epoch": 0.43, "learning_rate": 2.1579042675197683e-05, "loss": 0.8836, "step": 7158 }, { "epoch": 0.43, "learning_rate": 2.1585078771050884e-05, "loss": 1.4072, "step": 7160 }, { "epoch": 0.43, "learning_rate": 2.159111486690409e-05, "loss": 1.1527, "step": 7162 }, { "epoch": 0.43, "learning_rate": 2.159715096275729e-05, "loss": 1.2095, "step": 7164 }, { "epoch": 0.43, "learning_rate": 2.160318705861049e-05, "loss": 1.104, "step": 7166 }, { "epoch": 0.43, "learning_rate": 2.1609223154463694e-05, "loss": 1.1277, "step": 7168 }, { "epoch": 0.43, "learning_rate": 2.1615259250316895e-05, "loss": 1.3234, "step": 7170 }, { "epoch": 0.43, "learning_rate": 2.16212953461701e-05, "loss": 1.2856, "step": 7172 }, { "epoch": 0.43, "learning_rate": 2.16273314420233e-05, "loss": 1.1766, "step": 7174 }, { "epoch": 0.43, "learning_rate": 2.1633367537876504e-05, "loss": 1.1303, "step": 7176 }, { "epoch": 0.43, "learning_rate": 2.1639403633729705e-05, "loss": 1.2315, "step": 7178 }, { "epoch": 0.43, "learning_rate": 2.1645439729582906e-05, "loss": 1.2946, "step": 7180 }, { "epoch": 0.43, "learning_rate": 2.165147582543611e-05, "loss": 1.2764, "step": 7182 }, { "epoch": 0.43, "learning_rate": 2.165751192128931e-05, "loss": 1.0217, "step": 7184 }, { "epoch": 0.43, "learning_rate": 2.1663548017142512e-05, "loss": 1.5151, "step": 7186 }, { "epoch": 0.43, "learning_rate": 2.1669584112995716e-05, "loss": 1.4445, "step": 7188 }, { "epoch": 0.43, "learning_rate": 2.1675620208848917e-05, "loss": 1.0207, "step": 7190 }, { "epoch": 0.43, "learning_rate": 2.1681656304702118e-05, "loss": 0.9216, "step": 7192 }, { "epoch": 0.43, "learning_rate": 2.1687692400555322e-05, "loss": 1.1763, "step": 7194 }, { "epoch": 0.43, "learning_rate": 2.1693728496408526e-05, "loss": 1.2243, "step": 7196 }, { "epoch": 0.43, "learning_rate": 2.1699764592261727e-05, "loss": 1.4099, "step": 7198 }, { "epoch": 0.43, "learning_rate": 2.1705800688114928e-05, "loss": 1.0413, "step": 7200 }, { "epoch": 0.43, "learning_rate": 2.1711836783968132e-05, "loss": 1.4017, "step": 7202 }, { "epoch": 0.43, "learning_rate": 2.1717872879821333e-05, "loss": 1.0302, "step": 7204 }, { "epoch": 0.43, "learning_rate": 2.1723908975674534e-05, "loss": 1.609, "step": 7206 }, { "epoch": 0.44, "learning_rate": 2.1729945071527738e-05, "loss": 1.219, "step": 7208 }, { "epoch": 0.44, "learning_rate": 2.173598116738094e-05, "loss": 1.3161, "step": 7210 }, { "epoch": 0.44, "learning_rate": 2.174201726323414e-05, "loss": 1.1956, "step": 7212 }, { "epoch": 0.44, "learning_rate": 2.1748053359087344e-05, "loss": 1.094, "step": 7214 }, { "epoch": 0.44, "learning_rate": 2.1754089454940544e-05, "loss": 0.9879, "step": 7216 }, { "epoch": 0.44, "learning_rate": 2.176012555079375e-05, "loss": 1.189, "step": 7218 }, { "epoch": 0.44, "learning_rate": 2.176616164664695e-05, "loss": 1.1035, "step": 7220 }, { "epoch": 0.44, "learning_rate": 2.1772197742500154e-05, "loss": 1.4081, "step": 7222 }, { "epoch": 0.44, "learning_rate": 2.1778233838353354e-05, "loss": 0.9309, "step": 7224 }, { "epoch": 0.44, "learning_rate": 2.1784269934206555e-05, "loss": 1.2513, "step": 7226 }, { "epoch": 0.44, "learning_rate": 2.179030603005976e-05, "loss": 1.5642, "step": 7228 }, { "epoch": 0.44, "learning_rate": 2.179634212591296e-05, "loss": 1.378, "step": 7230 }, { "epoch": 0.44, "learning_rate": 2.180237822176616e-05, "loss": 1.1646, "step": 7232 }, { "epoch": 0.44, "learning_rate": 2.1808414317619365e-05, "loss": 1.1204, "step": 7234 }, { "epoch": 0.44, "learning_rate": 2.1814450413472566e-05, "loss": 1.2225, "step": 7236 }, { "epoch": 0.44, "learning_rate": 2.182048650932577e-05, "loss": 1.1521, "step": 7238 }, { "epoch": 0.44, "learning_rate": 2.182652260517897e-05, "loss": 1.2691, "step": 7240 }, { "epoch": 0.44, "learning_rate": 2.1832558701032175e-05, "loss": 1.1801, "step": 7242 }, { "epoch": 0.44, "learning_rate": 2.1838594796885376e-05, "loss": 1.2915, "step": 7244 }, { "epoch": 0.44, "learning_rate": 2.1844630892738577e-05, "loss": 1.0733, "step": 7246 }, { "epoch": 0.44, "learning_rate": 2.185066698859178e-05, "loss": 1.3176, "step": 7248 }, { "epoch": 0.44, "learning_rate": 2.1856703084444982e-05, "loss": 1.1778, "step": 7250 }, { "epoch": 0.44, "learning_rate": 2.1862739180298183e-05, "loss": 1.3625, "step": 7252 }, { "epoch": 0.44, "learning_rate": 2.1868775276151387e-05, "loss": 1.2797, "step": 7254 }, { "epoch": 0.44, "learning_rate": 2.1874811372004588e-05, "loss": 1.3457, "step": 7256 }, { "epoch": 0.44, "learning_rate": 2.188084746785779e-05, "loss": 1.6095, "step": 7258 }, { "epoch": 0.44, "learning_rate": 2.1886883563710993e-05, "loss": 1.359, "step": 7260 }, { "epoch": 0.44, "learning_rate": 2.1892919659564197e-05, "loss": 1.4914, "step": 7262 }, { "epoch": 0.44, "learning_rate": 2.1898955755417398e-05, "loss": 1.028, "step": 7264 }, { "epoch": 0.44, "learning_rate": 2.19049918512706e-05, "loss": 1.0177, "step": 7266 }, { "epoch": 0.44, "learning_rate": 2.1911027947123803e-05, "loss": 1.3641, "step": 7268 }, { "epoch": 0.44, "learning_rate": 2.1917064042977004e-05, "loss": 1.3123, "step": 7270 }, { "epoch": 0.44, "learning_rate": 2.1923100138830205e-05, "loss": 0.9542, "step": 7272 }, { "epoch": 0.44, "learning_rate": 2.192913623468341e-05, "loss": 1.16, "step": 7274 }, { "epoch": 0.44, "learning_rate": 2.193517233053661e-05, "loss": 1.3816, "step": 7276 }, { "epoch": 0.44, "learning_rate": 2.194120842638981e-05, "loss": 1.3488, "step": 7278 }, { "epoch": 0.44, "learning_rate": 2.1947244522243015e-05, "loss": 1.056, "step": 7280 }, { "epoch": 0.44, "learning_rate": 2.1953280618096215e-05, "loss": 1.2458, "step": 7282 }, { "epoch": 0.44, "learning_rate": 2.195931671394942e-05, "loss": 1.0576, "step": 7284 }, { "epoch": 0.44, "learning_rate": 2.196535280980262e-05, "loss": 1.3885, "step": 7286 }, { "epoch": 0.44, "learning_rate": 2.1971388905655825e-05, "loss": 1.2395, "step": 7288 }, { "epoch": 0.44, "learning_rate": 2.1977425001509026e-05, "loss": 1.1355, "step": 7290 }, { "epoch": 0.44, "learning_rate": 2.1983461097362226e-05, "loss": 1.0762, "step": 7292 }, { "epoch": 0.44, "learning_rate": 2.198949719321543e-05, "loss": 1.2666, "step": 7294 }, { "epoch": 0.44, "learning_rate": 2.199553328906863e-05, "loss": 0.9229, "step": 7296 }, { "epoch": 0.44, "learning_rate": 2.2001569384921832e-05, "loss": 1.2614, "step": 7298 }, { "epoch": 0.44, "learning_rate": 2.2007605480775036e-05, "loss": 1.0149, "step": 7300 }, { "epoch": 0.44, "learning_rate": 2.2013641576628237e-05, "loss": 1.767, "step": 7302 }, { "epoch": 0.44, "learning_rate": 2.2019677672481438e-05, "loss": 1.1625, "step": 7304 }, { "epoch": 0.44, "learning_rate": 2.2025713768334642e-05, "loss": 1.4209, "step": 7306 }, { "epoch": 0.44, "learning_rate": 2.2031749864187846e-05, "loss": 1.2581, "step": 7308 }, { "epoch": 0.44, "learning_rate": 2.2037785960041047e-05, "loss": 1.3005, "step": 7310 }, { "epoch": 0.44, "learning_rate": 2.2043822055894248e-05, "loss": 1.3601, "step": 7312 }, { "epoch": 0.44, "learning_rate": 2.2049858151747452e-05, "loss": 1.7435, "step": 7314 }, { "epoch": 0.44, "learning_rate": 2.2055894247600653e-05, "loss": 1.1143, "step": 7316 }, { "epoch": 0.44, "learning_rate": 2.2061930343453854e-05, "loss": 1.6547, "step": 7318 }, { "epoch": 0.44, "learning_rate": 2.2067966439307058e-05, "loss": 1.3211, "step": 7320 }, { "epoch": 0.44, "learning_rate": 2.207400253516026e-05, "loss": 1.2322, "step": 7322 }, { "epoch": 0.44, "learning_rate": 2.208003863101346e-05, "loss": 1.4171, "step": 7324 }, { "epoch": 0.44, "learning_rate": 2.2086074726866664e-05, "loss": 1.2755, "step": 7326 }, { "epoch": 0.44, "learning_rate": 2.2092110822719865e-05, "loss": 1.4301, "step": 7328 }, { "epoch": 0.44, "learning_rate": 2.209814691857307e-05, "loss": 1.3017, "step": 7330 }, { "epoch": 0.44, "learning_rate": 2.210418301442627e-05, "loss": 1.3997, "step": 7332 }, { "epoch": 0.44, "learning_rate": 2.2110219110279474e-05, "loss": 1.1296, "step": 7334 }, { "epoch": 0.44, "learning_rate": 2.2116255206132675e-05, "loss": 1.9907, "step": 7336 }, { "epoch": 0.44, "learning_rate": 2.2122291301985876e-05, "loss": 1.2798, "step": 7338 }, { "epoch": 0.44, "learning_rate": 2.212832739783908e-05, "loss": 1.1271, "step": 7340 }, { "epoch": 0.44, "learning_rate": 2.213436349369228e-05, "loss": 1.1292, "step": 7342 }, { "epoch": 0.44, "learning_rate": 2.214039958954548e-05, "loss": 1.5037, "step": 7344 }, { "epoch": 0.44, "learning_rate": 2.2146435685398686e-05, "loss": 1.2229, "step": 7346 }, { "epoch": 0.44, "learning_rate": 2.2152471781251887e-05, "loss": 0.9661, "step": 7348 }, { "epoch": 0.44, "learning_rate": 2.215850787710509e-05, "loss": 1.2631, "step": 7350 }, { "epoch": 0.44, "learning_rate": 2.216454397295829e-05, "loss": 1.1906, "step": 7352 }, { "epoch": 0.44, "learning_rate": 2.2170580068811496e-05, "loss": 1.3496, "step": 7354 }, { "epoch": 0.44, "learning_rate": 2.2176616164664697e-05, "loss": 1.1156, "step": 7356 }, { "epoch": 0.44, "learning_rate": 2.2182652260517897e-05, "loss": 1.3896, "step": 7358 }, { "epoch": 0.44, "learning_rate": 2.21886883563711e-05, "loss": 1.4025, "step": 7360 }, { "epoch": 0.44, "learning_rate": 2.2194724452224302e-05, "loss": 1.2695, "step": 7362 }, { "epoch": 0.44, "learning_rate": 2.2200760548077503e-05, "loss": 0.9757, "step": 7364 }, { "epoch": 0.44, "learning_rate": 2.2206796643930707e-05, "loss": 1.3657, "step": 7366 }, { "epoch": 0.44, "learning_rate": 2.2212832739783908e-05, "loss": 1.3028, "step": 7368 }, { "epoch": 0.44, "learning_rate": 2.221886883563711e-05, "loss": 1.4046, "step": 7370 }, { "epoch": 0.44, "learning_rate": 2.2224904931490313e-05, "loss": 1.1719, "step": 7372 }, { "epoch": 0.45, "learning_rate": 2.2230941027343517e-05, "loss": 1.132, "step": 7374 }, { "epoch": 0.45, "learning_rate": 2.223697712319672e-05, "loss": 0.9231, "step": 7376 }, { "epoch": 0.45, "learning_rate": 2.224301321904992e-05, "loss": 1.2355, "step": 7378 }, { "epoch": 0.45, "learning_rate": 2.2249049314903123e-05, "loss": 1.4846, "step": 7380 }, { "epoch": 0.45, "learning_rate": 2.2255085410756324e-05, "loss": 0.9158, "step": 7382 }, { "epoch": 0.45, "learning_rate": 2.2261121506609525e-05, "loss": 1.2585, "step": 7384 }, { "epoch": 0.45, "learning_rate": 2.226715760246273e-05, "loss": 1.1907, "step": 7386 }, { "epoch": 0.45, "learning_rate": 2.227319369831593e-05, "loss": 1.3436, "step": 7388 }, { "epoch": 0.45, "learning_rate": 2.227922979416913e-05, "loss": 1.3789, "step": 7390 }, { "epoch": 0.45, "learning_rate": 2.2285265890022335e-05, "loss": 1.4508, "step": 7392 }, { "epoch": 0.45, "learning_rate": 2.2291301985875536e-05, "loss": 1.3329, "step": 7394 }, { "epoch": 0.45, "learning_rate": 2.229733808172874e-05, "loss": 1.1938, "step": 7396 }, { "epoch": 0.45, "learning_rate": 2.230337417758194e-05, "loss": 1.1749, "step": 7398 }, { "epoch": 0.45, "learning_rate": 2.2309410273435145e-05, "loss": 1.3854, "step": 7400 }, { "epoch": 0.45, "learning_rate": 2.2315446369288346e-05, "loss": 1.548, "step": 7402 }, { "epoch": 0.45, "learning_rate": 2.2321482465141547e-05, "loss": 1.3593, "step": 7404 }, { "epoch": 0.45, "learning_rate": 2.232751856099475e-05, "loss": 1.1806, "step": 7406 }, { "epoch": 0.45, "learning_rate": 2.2333554656847952e-05, "loss": 1.0424, "step": 7408 }, { "epoch": 0.45, "learning_rate": 2.2339590752701153e-05, "loss": 1.0917, "step": 7410 }, { "epoch": 0.45, "learning_rate": 2.2345626848554357e-05, "loss": 1.5071, "step": 7412 }, { "epoch": 0.45, "learning_rate": 2.2351662944407558e-05, "loss": 1.0153, "step": 7414 }, { "epoch": 0.45, "learning_rate": 2.235769904026076e-05, "loss": 1.2119, "step": 7416 }, { "epoch": 0.45, "learning_rate": 2.2363735136113963e-05, "loss": 1.5109, "step": 7418 }, { "epoch": 0.45, "learning_rate": 2.2369771231967167e-05, "loss": 1.2523, "step": 7420 }, { "epoch": 0.45, "learning_rate": 2.2375807327820368e-05, "loss": 1.1473, "step": 7422 }, { "epoch": 0.45, "learning_rate": 2.238184342367357e-05, "loss": 0.9248, "step": 7424 }, { "epoch": 0.45, "learning_rate": 2.2387879519526773e-05, "loss": 1.849, "step": 7426 }, { "epoch": 0.45, "learning_rate": 2.2393915615379973e-05, "loss": 1.2927, "step": 7428 }, { "epoch": 0.45, "learning_rate": 2.2399951711233174e-05, "loss": 1.4862, "step": 7430 }, { "epoch": 0.45, "learning_rate": 2.240598780708638e-05, "loss": 1.3635, "step": 7432 }, { "epoch": 0.45, "learning_rate": 2.241202390293958e-05, "loss": 1.535, "step": 7434 }, { "epoch": 0.45, "learning_rate": 2.241805999879278e-05, "loss": 1.2584, "step": 7436 }, { "epoch": 0.45, "learning_rate": 2.2424096094645984e-05, "loss": 1.0754, "step": 7438 }, { "epoch": 0.45, "learning_rate": 2.2430132190499185e-05, "loss": 1.1944, "step": 7440 }, { "epoch": 0.45, "learning_rate": 2.243616828635239e-05, "loss": 1.0399, "step": 7442 }, { "epoch": 0.45, "learning_rate": 2.244220438220559e-05, "loss": 1.4775, "step": 7444 }, { "epoch": 0.45, "learning_rate": 2.2448240478058794e-05, "loss": 0.9922, "step": 7446 }, { "epoch": 0.45, "learning_rate": 2.2454276573911995e-05, "loss": 1.2468, "step": 7448 }, { "epoch": 0.45, "learning_rate": 2.2460312669765196e-05, "loss": 0.9527, "step": 7450 }, { "epoch": 0.45, "learning_rate": 2.24663487656184e-05, "loss": 1.1738, "step": 7452 }, { "epoch": 0.45, "learning_rate": 2.24723848614716e-05, "loss": 1.0794, "step": 7454 }, { "epoch": 0.45, "learning_rate": 2.2478420957324802e-05, "loss": 0.9804, "step": 7456 }, { "epoch": 0.45, "learning_rate": 2.2484457053178006e-05, "loss": 1.1486, "step": 7458 }, { "epoch": 0.45, "learning_rate": 2.2490493149031207e-05, "loss": 1.4107, "step": 7460 }, { "epoch": 0.45, "learning_rate": 2.249652924488441e-05, "loss": 1.2672, "step": 7462 }, { "epoch": 0.45, "learning_rate": 2.2502565340737612e-05, "loss": 1.1381, "step": 7464 }, { "epoch": 0.45, "learning_rate": 2.2508601436590816e-05, "loss": 1.3972, "step": 7466 }, { "epoch": 0.45, "learning_rate": 2.2514637532444017e-05, "loss": 1.0689, "step": 7468 }, { "epoch": 0.45, "learning_rate": 2.2520673628297218e-05, "loss": 1.4942, "step": 7470 }, { "epoch": 0.45, "learning_rate": 2.2526709724150422e-05, "loss": 1.18, "step": 7472 }, { "epoch": 0.45, "learning_rate": 2.2532745820003623e-05, "loss": 1.1493, "step": 7474 }, { "epoch": 0.45, "learning_rate": 2.2538781915856824e-05, "loss": 1.0895, "step": 7476 }, { "epoch": 0.45, "learning_rate": 2.2544818011710028e-05, "loss": 1.4657, "step": 7478 }, { "epoch": 0.45, "learning_rate": 2.255085410756323e-05, "loss": 1.4356, "step": 7480 }, { "epoch": 0.45, "learning_rate": 2.255689020341643e-05, "loss": 1.1632, "step": 7482 }, { "epoch": 0.45, "learning_rate": 2.2562926299269634e-05, "loss": 1.3207, "step": 7484 }, { "epoch": 0.45, "learning_rate": 2.2568962395122838e-05, "loss": 1.3832, "step": 7486 }, { "epoch": 0.45, "learning_rate": 2.257499849097604e-05, "loss": 1.4398, "step": 7488 }, { "epoch": 0.45, "learning_rate": 2.258103458682924e-05, "loss": 1.1397, "step": 7490 }, { "epoch": 0.45, "learning_rate": 2.2587070682682444e-05, "loss": 1.066, "step": 7492 }, { "epoch": 0.45, "learning_rate": 2.2593106778535645e-05, "loss": 1.5564, "step": 7494 }, { "epoch": 0.45, "learning_rate": 2.2599142874388845e-05, "loss": 1.1316, "step": 7496 }, { "epoch": 0.45, "learning_rate": 2.260517897024205e-05, "loss": 1.494, "step": 7498 }, { "epoch": 0.45, "learning_rate": 2.261121506609525e-05, "loss": 1.3519, "step": 7500 }, { "epoch": 0.45, "learning_rate": 2.261725116194845e-05, "loss": 1.2087, "step": 7502 }, { "epoch": 0.45, "learning_rate": 2.2623287257801652e-05, "loss": 1.3707, "step": 7504 }, { "epoch": 0.45, "learning_rate": 2.2629323353654856e-05, "loss": 1.4997, "step": 7506 }, { "epoch": 0.45, "learning_rate": 2.263535944950806e-05, "loss": 0.961, "step": 7508 }, { "epoch": 0.45, "learning_rate": 2.264139554536126e-05, "loss": 1.3023, "step": 7510 }, { "epoch": 0.45, "learning_rate": 2.2647431641214465e-05, "loss": 1.1644, "step": 7512 }, { "epoch": 0.45, "learning_rate": 2.2653467737067666e-05, "loss": 1.3698, "step": 7514 }, { "epoch": 0.45, "learning_rate": 2.2659503832920867e-05, "loss": 1.0121, "step": 7516 }, { "epoch": 0.45, "learning_rate": 2.266553992877407e-05, "loss": 1.5366, "step": 7518 }, { "epoch": 0.45, "learning_rate": 2.2671576024627272e-05, "loss": 1.481, "step": 7520 }, { "epoch": 0.45, "learning_rate": 2.2677612120480473e-05, "loss": 1.1461, "step": 7522 }, { "epoch": 0.45, "learning_rate": 2.2683648216333677e-05, "loss": 1.4833, "step": 7524 }, { "epoch": 0.45, "learning_rate": 2.2689684312186878e-05, "loss": 1.2962, "step": 7526 }, { "epoch": 0.45, "learning_rate": 2.269572040804008e-05, "loss": 1.3661, "step": 7528 }, { "epoch": 0.45, "learning_rate": 2.2701756503893283e-05, "loss": 1.4922, "step": 7530 }, { "epoch": 0.45, "learning_rate": 2.2707792599746487e-05, "loss": 1.5316, "step": 7532 }, { "epoch": 0.45, "learning_rate": 2.2713828695599688e-05, "loss": 1.1278, "step": 7534 }, { "epoch": 0.45, "learning_rate": 2.271986479145289e-05, "loss": 1.0753, "step": 7536 }, { "epoch": 0.45, "learning_rate": 2.2725900887306093e-05, "loss": 1.4399, "step": 7538 }, { "epoch": 0.46, "learning_rate": 2.2731936983159294e-05, "loss": 1.0019, "step": 7540 }, { "epoch": 0.46, "learning_rate": 2.2737973079012495e-05, "loss": 1.1708, "step": 7542 }, { "epoch": 0.46, "learning_rate": 2.27440091748657e-05, "loss": 1.5918, "step": 7544 }, { "epoch": 0.46, "learning_rate": 2.27500452707189e-05, "loss": 1.1144, "step": 7546 }, { "epoch": 0.46, "learning_rate": 2.27560813665721e-05, "loss": 1.456, "step": 7548 }, { "epoch": 0.46, "learning_rate": 2.2762117462425305e-05, "loss": 1.4615, "step": 7550 }, { "epoch": 0.46, "learning_rate": 2.2768153558278506e-05, "loss": 1.546, "step": 7552 }, { "epoch": 0.46, "learning_rate": 2.277418965413171e-05, "loss": 1.025, "step": 7554 }, { "epoch": 0.46, "learning_rate": 2.278022574998491e-05, "loss": 1.4867, "step": 7556 }, { "epoch": 0.46, "learning_rate": 2.2786261845838115e-05, "loss": 1.2152, "step": 7558 }, { "epoch": 0.46, "learning_rate": 2.2792297941691316e-05, "loss": 1.3351, "step": 7560 }, { "epoch": 0.46, "learning_rate": 2.2798334037544516e-05, "loss": 1.5978, "step": 7562 }, { "epoch": 0.46, "learning_rate": 2.280437013339772e-05, "loss": 1.0352, "step": 7564 }, { "epoch": 0.46, "learning_rate": 2.281040622925092e-05, "loss": 1.1664, "step": 7566 }, { "epoch": 0.46, "learning_rate": 2.2816442325104122e-05, "loss": 1.2958, "step": 7568 }, { "epoch": 0.46, "learning_rate": 2.2822478420957326e-05, "loss": 1.0718, "step": 7570 }, { "epoch": 0.46, "learning_rate": 2.2828514516810527e-05, "loss": 1.5704, "step": 7572 }, { "epoch": 0.46, "learning_rate": 2.283455061266373e-05, "loss": 1.3208, "step": 7574 }, { "epoch": 0.46, "learning_rate": 2.2840586708516932e-05, "loss": 1.1484, "step": 7576 }, { "epoch": 0.46, "learning_rate": 2.2846622804370137e-05, "loss": 1.3601, "step": 7578 }, { "epoch": 0.46, "learning_rate": 2.2852658900223337e-05, "loss": 1.2284, "step": 7580 }, { "epoch": 0.46, "learning_rate": 2.2858694996076538e-05, "loss": 1.0348, "step": 7582 }, { "epoch": 0.46, "learning_rate": 2.2864731091929742e-05, "loss": 1.5303, "step": 7584 }, { "epoch": 0.46, "learning_rate": 2.2870767187782943e-05, "loss": 1.2648, "step": 7586 }, { "epoch": 0.46, "learning_rate": 2.2876803283636144e-05, "loss": 1.2268, "step": 7588 }, { "epoch": 0.46, "learning_rate": 2.2882839379489348e-05, "loss": 1.0852, "step": 7590 }, { "epoch": 0.46, "learning_rate": 2.288887547534255e-05, "loss": 1.3058, "step": 7592 }, { "epoch": 0.46, "learning_rate": 2.289491157119575e-05, "loss": 1.4148, "step": 7594 }, { "epoch": 0.46, "learning_rate": 2.2900947667048954e-05, "loss": 0.9596, "step": 7596 }, { "epoch": 0.46, "learning_rate": 2.2906983762902158e-05, "loss": 1.3183, "step": 7598 }, { "epoch": 0.46, "learning_rate": 2.291301985875536e-05, "loss": 1.3488, "step": 7600 }, { "epoch": 0.46, "learning_rate": 2.291905595460856e-05, "loss": 1.2659, "step": 7602 }, { "epoch": 0.46, "learning_rate": 2.2925092050461764e-05, "loss": 1.2853, "step": 7604 }, { "epoch": 0.46, "learning_rate": 2.2931128146314965e-05, "loss": 1.3392, "step": 7606 }, { "epoch": 0.46, "learning_rate": 2.2937164242168166e-05, "loss": 1.1865, "step": 7608 }, { "epoch": 0.46, "learning_rate": 2.294320033802137e-05, "loss": 1.2166, "step": 7610 }, { "epoch": 0.46, "learning_rate": 2.294923643387457e-05, "loss": 1.4719, "step": 7612 }, { "epoch": 0.46, "learning_rate": 2.295527252972777e-05, "loss": 1.2885, "step": 7614 }, { "epoch": 0.46, "learning_rate": 2.2961308625580972e-05, "loss": 1.3437, "step": 7616 }, { "epoch": 0.46, "learning_rate": 2.2967344721434177e-05, "loss": 1.4894, "step": 7618 }, { "epoch": 0.46, "learning_rate": 2.297338081728738e-05, "loss": 1.2165, "step": 7620 }, { "epoch": 0.46, "learning_rate": 2.297941691314058e-05, "loss": 1.4231, "step": 7622 }, { "epoch": 0.46, "learning_rate": 2.2985453008993786e-05, "loss": 1.4801, "step": 7624 }, { "epoch": 0.46, "learning_rate": 2.2991489104846987e-05, "loss": 1.1308, "step": 7626 }, { "epoch": 0.46, "learning_rate": 2.2997525200700187e-05, "loss": 1.5099, "step": 7628 }, { "epoch": 0.46, "learning_rate": 2.300356129655339e-05, "loss": 1.7068, "step": 7630 }, { "epoch": 0.46, "learning_rate": 2.3009597392406593e-05, "loss": 1.2977, "step": 7632 }, { "epoch": 0.46, "learning_rate": 2.3015633488259793e-05, "loss": 1.2573, "step": 7634 }, { "epoch": 0.46, "learning_rate": 2.3021669584112998e-05, "loss": 1.3687, "step": 7636 }, { "epoch": 0.46, "learning_rate": 2.30277056799662e-05, "loss": 1.8953, "step": 7638 }, { "epoch": 0.46, "learning_rate": 2.30337417758194e-05, "loss": 1.3253, "step": 7640 }, { "epoch": 0.46, "learning_rate": 2.3039777871672603e-05, "loss": 1.1576, "step": 7642 }, { "epoch": 0.46, "learning_rate": 2.3045813967525808e-05, "loss": 1.0794, "step": 7644 }, { "epoch": 0.46, "learning_rate": 2.305185006337901e-05, "loss": 1.29, "step": 7646 }, { "epoch": 0.46, "learning_rate": 2.305788615923221e-05, "loss": 1.3378, "step": 7648 }, { "epoch": 0.46, "learning_rate": 2.3063922255085413e-05, "loss": 1.1165, "step": 7650 }, { "epoch": 0.46, "learning_rate": 2.3069958350938614e-05, "loss": 1.427, "step": 7652 }, { "epoch": 0.46, "learning_rate": 2.3075994446791815e-05, "loss": 0.8342, "step": 7654 }, { "epoch": 0.46, "learning_rate": 2.308203054264502e-05, "loss": 1.2783, "step": 7656 }, { "epoch": 0.46, "learning_rate": 2.308806663849822e-05, "loss": 0.9238, "step": 7658 }, { "epoch": 0.46, "learning_rate": 2.309410273435142e-05, "loss": 1.1381, "step": 7660 }, { "epoch": 0.46, "learning_rate": 2.3100138830204625e-05, "loss": 1.2843, "step": 7662 }, { "epoch": 0.46, "learning_rate": 2.3106174926057826e-05, "loss": 1.5156, "step": 7664 }, { "epoch": 0.46, "learning_rate": 2.311221102191103e-05, "loss": 1.5345, "step": 7666 }, { "epoch": 0.46, "learning_rate": 2.311824711776423e-05, "loss": 1.377, "step": 7668 }, { "epoch": 0.46, "learning_rate": 2.3124283213617435e-05, "loss": 1.3092, "step": 7670 }, { "epoch": 0.46, "learning_rate": 2.3130319309470636e-05, "loss": 1.2412, "step": 7672 }, { "epoch": 0.46, "learning_rate": 2.3136355405323837e-05, "loss": 1.1227, "step": 7674 }, { "epoch": 0.46, "learning_rate": 2.314239150117704e-05, "loss": 1.5051, "step": 7676 }, { "epoch": 0.46, "learning_rate": 2.3148427597030242e-05, "loss": 1.1314, "step": 7678 }, { "epoch": 0.46, "learning_rate": 2.3154463692883443e-05, "loss": 1.1832, "step": 7680 }, { "epoch": 0.46, "learning_rate": 2.3160499788736647e-05, "loss": 1.3473, "step": 7682 }, { "epoch": 0.46, "learning_rate": 2.3166535884589848e-05, "loss": 1.5209, "step": 7684 }, { "epoch": 0.46, "learning_rate": 2.3172571980443052e-05, "loss": 0.7704, "step": 7686 }, { "epoch": 0.46, "learning_rate": 2.3178608076296253e-05, "loss": 1.0043, "step": 7688 }, { "epoch": 0.46, "learning_rate": 2.3184644172149457e-05, "loss": 1.7125, "step": 7690 }, { "epoch": 0.46, "learning_rate": 2.3190680268002658e-05, "loss": 1.1493, "step": 7692 }, { "epoch": 0.46, "learning_rate": 2.319671636385586e-05, "loss": 1.0935, "step": 7694 }, { "epoch": 0.46, "learning_rate": 2.3202752459709063e-05, "loss": 1.1606, "step": 7696 }, { "epoch": 0.46, "learning_rate": 2.3208788555562264e-05, "loss": 1.11, "step": 7698 }, { "epoch": 0.46, "learning_rate": 2.3214824651415464e-05, "loss": 0.9881, "step": 7700 }, { "epoch": 0.46, "learning_rate": 2.322086074726867e-05, "loss": 0.9313, "step": 7702 }, { "epoch": 0.46, "learning_rate": 2.322689684312187e-05, "loss": 1.0652, "step": 7704 }, { "epoch": 0.47, "learning_rate": 2.323293293897507e-05, "loss": 1.4187, "step": 7706 }, { "epoch": 0.47, "learning_rate": 2.3238969034828274e-05, "loss": 1.1188, "step": 7708 }, { "epoch": 0.47, "learning_rate": 2.324500513068148e-05, "loss": 1.4251, "step": 7710 }, { "epoch": 0.47, "learning_rate": 2.325104122653468e-05, "loss": 1.0911, "step": 7712 }, { "epoch": 0.47, "learning_rate": 2.325707732238788e-05, "loss": 0.9498, "step": 7714 }, { "epoch": 0.47, "learning_rate": 2.3263113418241084e-05, "loss": 1.4373, "step": 7716 }, { "epoch": 0.47, "learning_rate": 2.3269149514094285e-05, "loss": 1.1149, "step": 7718 }, { "epoch": 0.47, "learning_rate": 2.3275185609947486e-05, "loss": 1.0698, "step": 7720 }, { "epoch": 0.47, "learning_rate": 2.328122170580069e-05, "loss": 1.179, "step": 7722 }, { "epoch": 0.47, "learning_rate": 2.328725780165389e-05, "loss": 1.1386, "step": 7724 }, { "epoch": 0.47, "learning_rate": 2.3293293897507092e-05, "loss": 1.6898, "step": 7726 }, { "epoch": 0.47, "learning_rate": 2.3299329993360293e-05, "loss": 1.1083, "step": 7728 }, { "epoch": 0.47, "learning_rate": 2.3305366089213497e-05, "loss": 0.9786, "step": 7730 }, { "epoch": 0.47, "learning_rate": 2.33114021850667e-05, "loss": 1.1949, "step": 7732 }, { "epoch": 0.47, "learning_rate": 2.3317438280919902e-05, "loss": 1.2037, "step": 7734 }, { "epoch": 0.47, "learning_rate": 2.3323474376773106e-05, "loss": 1.047, "step": 7736 }, { "epoch": 0.47, "learning_rate": 2.3329510472626307e-05, "loss": 1.4091, "step": 7738 }, { "epoch": 0.47, "learning_rate": 2.3335546568479508e-05, "loss": 1.1854, "step": 7740 }, { "epoch": 0.47, "learning_rate": 2.3341582664332712e-05, "loss": 1.0384, "step": 7742 }, { "epoch": 0.47, "learning_rate": 2.3347618760185913e-05, "loss": 1.2249, "step": 7744 }, { "epoch": 0.47, "learning_rate": 2.3353654856039114e-05, "loss": 0.8852, "step": 7746 }, { "epoch": 0.47, "learning_rate": 2.3359690951892318e-05, "loss": 1.1901, "step": 7748 }, { "epoch": 0.47, "learning_rate": 2.336572704774552e-05, "loss": 1.0229, "step": 7750 }, { "epoch": 0.47, "learning_rate": 2.337176314359872e-05, "loss": 1.0646, "step": 7752 }, { "epoch": 0.47, "learning_rate": 2.3377799239451924e-05, "loss": 1.2075, "step": 7754 }, { "epoch": 0.47, "learning_rate": 2.3383835335305128e-05, "loss": 1.1141, "step": 7756 }, { "epoch": 0.47, "learning_rate": 2.338987143115833e-05, "loss": 1.0939, "step": 7758 }, { "epoch": 0.47, "learning_rate": 2.339590752701153e-05, "loss": 1.2186, "step": 7760 }, { "epoch": 0.47, "learning_rate": 2.3401943622864734e-05, "loss": 1.2136, "step": 7762 }, { "epoch": 0.47, "learning_rate": 2.3407979718717935e-05, "loss": 1.2726, "step": 7764 }, { "epoch": 0.47, "learning_rate": 2.3414015814571135e-05, "loss": 1.3245, "step": 7766 }, { "epoch": 0.47, "learning_rate": 2.342005191042434e-05, "loss": 0.8634, "step": 7768 }, { "epoch": 0.47, "learning_rate": 2.342608800627754e-05, "loss": 1.2316, "step": 7770 }, { "epoch": 0.47, "learning_rate": 2.343212410213074e-05, "loss": 1.0207, "step": 7772 }, { "epoch": 0.47, "learning_rate": 2.3438160197983945e-05, "loss": 1.2531, "step": 7774 }, { "epoch": 0.47, "learning_rate": 2.3444196293837146e-05, "loss": 1.2218, "step": 7776 }, { "epoch": 0.47, "learning_rate": 2.345023238969035e-05, "loss": 1.2519, "step": 7778 }, { "epoch": 0.47, "learning_rate": 2.345626848554355e-05, "loss": 1.3082, "step": 7780 }, { "epoch": 0.47, "learning_rate": 2.3462304581396756e-05, "loss": 1.2112, "step": 7782 }, { "epoch": 0.47, "learning_rate": 2.3468340677249956e-05, "loss": 1.4998, "step": 7784 }, { "epoch": 0.47, "learning_rate": 2.3474376773103157e-05, "loss": 1.347, "step": 7786 }, { "epoch": 0.47, "learning_rate": 2.348041286895636e-05, "loss": 1.4089, "step": 7788 }, { "epoch": 0.47, "learning_rate": 2.3486448964809562e-05, "loss": 1.4496, "step": 7790 }, { "epoch": 0.47, "learning_rate": 2.3492485060662763e-05, "loss": 1.3493, "step": 7792 }, { "epoch": 0.47, "learning_rate": 2.3498521156515967e-05, "loss": 1.4989, "step": 7794 }, { "epoch": 0.47, "learning_rate": 2.3504557252369168e-05, "loss": 1.1825, "step": 7796 }, { "epoch": 0.47, "learning_rate": 2.3510593348222372e-05, "loss": 1.4092, "step": 7798 }, { "epoch": 0.47, "learning_rate": 2.3516629444075573e-05, "loss": 1.045, "step": 7800 }, { "epoch": 0.47, "learning_rate": 2.3522665539928777e-05, "loss": 1.297, "step": 7802 }, { "epoch": 0.47, "learning_rate": 2.3528701635781978e-05, "loss": 1.0694, "step": 7804 }, { "epoch": 0.47, "learning_rate": 2.353473773163518e-05, "loss": 1.1923, "step": 7806 }, { "epoch": 0.47, "learning_rate": 2.3540773827488383e-05, "loss": 1.0704, "step": 7808 }, { "epoch": 0.47, "learning_rate": 2.3546809923341584e-05, "loss": 1.3009, "step": 7810 }, { "epoch": 0.47, "learning_rate": 2.3552846019194785e-05, "loss": 1.3196, "step": 7812 }, { "epoch": 0.47, "learning_rate": 2.355888211504799e-05, "loss": 1.1357, "step": 7814 }, { "epoch": 0.47, "learning_rate": 2.356491821090119e-05, "loss": 1.2637, "step": 7816 }, { "epoch": 0.47, "learning_rate": 2.357095430675439e-05, "loss": 1.582, "step": 7818 }, { "epoch": 0.47, "learning_rate": 2.3576990402607595e-05, "loss": 1.2996, "step": 7820 }, { "epoch": 0.47, "learning_rate": 2.35830264984608e-05, "loss": 1.0378, "step": 7822 }, { "epoch": 0.47, "learning_rate": 2.3589062594314e-05, "loss": 1.1607, "step": 7824 }, { "epoch": 0.47, "learning_rate": 2.35950986901672e-05, "loss": 0.8585, "step": 7826 }, { "epoch": 0.47, "learning_rate": 2.3601134786020405e-05, "loss": 1.2217, "step": 7828 }, { "epoch": 0.47, "learning_rate": 2.3607170881873606e-05, "loss": 1.3363, "step": 7830 }, { "epoch": 0.47, "learning_rate": 2.3613206977726807e-05, "loss": 1.1185, "step": 7832 }, { "epoch": 0.47, "learning_rate": 2.361924307358001e-05, "loss": 1.0628, "step": 7834 }, { "epoch": 0.47, "learning_rate": 2.362527916943321e-05, "loss": 1.1525, "step": 7836 }, { "epoch": 0.47, "learning_rate": 2.3631315265286412e-05, "loss": 1.2565, "step": 7838 }, { "epoch": 0.47, "learning_rate": 2.3637351361139613e-05, "loss": 1.7836, "step": 7840 }, { "epoch": 0.47, "learning_rate": 2.3643387456992817e-05, "loss": 1.8883, "step": 7842 }, { "epoch": 0.47, "learning_rate": 2.364942355284602e-05, "loss": 1.2655, "step": 7844 }, { "epoch": 0.47, "learning_rate": 2.3655459648699222e-05, "loss": 1.0941, "step": 7846 }, { "epoch": 0.47, "learning_rate": 2.3661495744552427e-05, "loss": 1.1486, "step": 7848 }, { "epoch": 0.47, "learning_rate": 2.3667531840405627e-05, "loss": 1.4934, "step": 7850 }, { "epoch": 0.47, "learning_rate": 2.3673567936258828e-05, "loss": 1.0556, "step": 7852 }, { "epoch": 0.47, "learning_rate": 2.3679604032112032e-05, "loss": 1.7329, "step": 7854 }, { "epoch": 0.47, "learning_rate": 2.3685640127965233e-05, "loss": 1.3049, "step": 7856 }, { "epoch": 0.47, "learning_rate": 2.3691676223818434e-05, "loss": 1.3106, "step": 7858 }, { "epoch": 0.47, "learning_rate": 2.3697712319671638e-05, "loss": 1.0594, "step": 7860 }, { "epoch": 0.47, "learning_rate": 2.370374841552484e-05, "loss": 1.1478, "step": 7862 }, { "epoch": 0.47, "learning_rate": 2.370978451137804e-05, "loss": 1.31, "step": 7864 }, { "epoch": 0.47, "learning_rate": 2.3715820607231244e-05, "loss": 1.2496, "step": 7866 }, { "epoch": 0.47, "learning_rate": 2.372185670308445e-05, "loss": 1.1462, "step": 7868 }, { "epoch": 0.48, "learning_rate": 2.372789279893765e-05, "loss": 1.2146, "step": 7870 }, { "epoch": 0.48, "learning_rate": 2.373392889479085e-05, "loss": 1.407, "step": 7872 }, { "epoch": 0.48, "learning_rate": 2.3739964990644054e-05, "loss": 1.2038, "step": 7874 }, { "epoch": 0.48, "learning_rate": 2.3746001086497255e-05, "loss": 1.3716, "step": 7876 }, { "epoch": 0.48, "learning_rate": 2.3752037182350456e-05, "loss": 1.1737, "step": 7878 }, { "epoch": 0.48, "learning_rate": 2.375807327820366e-05, "loss": 1.1182, "step": 7880 }, { "epoch": 0.48, "learning_rate": 2.376410937405686e-05, "loss": 1.3778, "step": 7882 }, { "epoch": 0.48, "learning_rate": 2.377014546991006e-05, "loss": 1.1038, "step": 7884 }, { "epoch": 0.48, "learning_rate": 2.3776181565763262e-05, "loss": 1.4492, "step": 7886 }, { "epoch": 0.48, "learning_rate": 2.3782217661616467e-05, "loss": 0.9988, "step": 7888 }, { "epoch": 0.48, "learning_rate": 2.378825375746967e-05, "loss": 1.4341, "step": 7890 }, { "epoch": 0.48, "learning_rate": 2.3794289853322872e-05, "loss": 1.2908, "step": 7892 }, { "epoch": 0.48, "learning_rate": 2.3800325949176076e-05, "loss": 1.3353, "step": 7894 }, { "epoch": 0.48, "learning_rate": 2.3806362045029277e-05, "loss": 1.1913, "step": 7896 }, { "epoch": 0.48, "learning_rate": 2.3812398140882478e-05, "loss": 1.254, "step": 7898 }, { "epoch": 0.48, "learning_rate": 2.3818434236735682e-05, "loss": 1.0482, "step": 7900 }, { "epoch": 0.48, "learning_rate": 2.3824470332588883e-05, "loss": 1.2311, "step": 7902 }, { "epoch": 0.48, "learning_rate": 2.3830506428442083e-05, "loss": 1.0023, "step": 7904 }, { "epoch": 0.48, "learning_rate": 2.3836542524295288e-05, "loss": 1.1563, "step": 7906 }, { "epoch": 0.48, "learning_rate": 2.384257862014849e-05, "loss": 1.4703, "step": 7908 }, { "epoch": 0.48, "learning_rate": 2.3848614716001693e-05, "loss": 1.1473, "step": 7910 }, { "epoch": 0.48, "learning_rate": 2.3854650811854893e-05, "loss": 1.5312, "step": 7912 }, { "epoch": 0.48, "learning_rate": 2.3860686907708098e-05, "loss": 1.1669, "step": 7914 }, { "epoch": 0.48, "learning_rate": 2.38667230035613e-05, "loss": 1.2865, "step": 7916 }, { "epoch": 0.48, "learning_rate": 2.38727590994145e-05, "loss": 1.0932, "step": 7918 }, { "epoch": 0.48, "learning_rate": 2.3878795195267704e-05, "loss": 1.2432, "step": 7920 }, { "epoch": 0.48, "learning_rate": 2.3884831291120904e-05, "loss": 1.1472, "step": 7922 }, { "epoch": 0.48, "learning_rate": 2.3890867386974105e-05, "loss": 1.4828, "step": 7924 }, { "epoch": 0.48, "learning_rate": 2.389690348282731e-05, "loss": 1.2938, "step": 7926 }, { "epoch": 0.48, "learning_rate": 2.390293957868051e-05, "loss": 1.2874, "step": 7928 }, { "epoch": 0.48, "learning_rate": 2.390897567453371e-05, "loss": 1.2863, "step": 7930 }, { "epoch": 0.48, "learning_rate": 2.3915011770386915e-05, "loss": 0.8692, "step": 7932 }, { "epoch": 0.48, "learning_rate": 2.392104786624012e-05, "loss": 1.3554, "step": 7934 }, { "epoch": 0.48, "learning_rate": 2.392708396209332e-05, "loss": 1.18, "step": 7936 }, { "epoch": 0.48, "learning_rate": 2.393312005794652e-05, "loss": 1.268, "step": 7938 }, { "epoch": 0.48, "learning_rate": 2.3939156153799725e-05, "loss": 0.9415, "step": 7940 }, { "epoch": 0.48, "learning_rate": 2.3945192249652926e-05, "loss": 1.1147, "step": 7942 }, { "epoch": 0.48, "learning_rate": 2.3951228345506127e-05, "loss": 1.2341, "step": 7944 }, { "epoch": 0.48, "learning_rate": 2.395726444135933e-05, "loss": 1.3994, "step": 7946 }, { "epoch": 0.48, "learning_rate": 2.3963300537212532e-05, "loss": 1.3919, "step": 7948 }, { "epoch": 0.48, "learning_rate": 2.3969336633065733e-05, "loss": 1.3597, "step": 7950 }, { "epoch": 0.48, "learning_rate": 2.3975372728918934e-05, "loss": 1.3451, "step": 7952 }, { "epoch": 0.48, "learning_rate": 2.3981408824772138e-05, "loss": 1.047, "step": 7954 }, { "epoch": 0.48, "learning_rate": 2.3987444920625342e-05, "loss": 1.2596, "step": 7956 }, { "epoch": 0.48, "learning_rate": 2.3993481016478543e-05, "loss": 1.01, "step": 7958 }, { "epoch": 0.48, "learning_rate": 2.3999517112331747e-05, "loss": 1.11, "step": 7960 }, { "epoch": 0.48, "learning_rate": 2.4005553208184948e-05, "loss": 1.4703, "step": 7962 }, { "epoch": 0.48, "learning_rate": 2.401158930403815e-05, "loss": 1.3955, "step": 7964 }, { "epoch": 0.48, "learning_rate": 2.4017625399891353e-05, "loss": 0.9957, "step": 7966 }, { "epoch": 0.48, "learning_rate": 2.4023661495744554e-05, "loss": 1.3286, "step": 7968 }, { "epoch": 0.48, "learning_rate": 2.4029697591597754e-05, "loss": 1.1318, "step": 7970 }, { "epoch": 0.48, "learning_rate": 2.403573368745096e-05, "loss": 1.2401, "step": 7972 }, { "epoch": 0.48, "learning_rate": 2.404176978330416e-05, "loss": 1.2278, "step": 7974 }, { "epoch": 0.48, "learning_rate": 2.404780587915736e-05, "loss": 1.6405, "step": 7976 }, { "epoch": 0.48, "learning_rate": 2.4053841975010565e-05, "loss": 1.3614, "step": 7978 }, { "epoch": 0.48, "learning_rate": 2.405987807086377e-05, "loss": 1.214, "step": 7980 }, { "epoch": 0.48, "learning_rate": 2.406591416671697e-05, "loss": 1.2007, "step": 7982 }, { "epoch": 0.48, "learning_rate": 2.407195026257017e-05, "loss": 1.131, "step": 7984 }, { "epoch": 0.48, "learning_rate": 2.4077986358423375e-05, "loss": 1.5335, "step": 7986 }, { "epoch": 0.48, "learning_rate": 2.4084022454276575e-05, "loss": 1.6938, "step": 7988 }, { "epoch": 0.48, "learning_rate": 2.4090058550129776e-05, "loss": 1.1213, "step": 7990 }, { "epoch": 0.48, "learning_rate": 2.409609464598298e-05, "loss": 0.934, "step": 7992 }, { "epoch": 0.48, "learning_rate": 2.410213074183618e-05, "loss": 1.377, "step": 7994 }, { "epoch": 0.48, "learning_rate": 2.4108166837689382e-05, "loss": 1.2162, "step": 7996 }, { "epoch": 0.48, "learning_rate": 2.4114202933542583e-05, "loss": 0.9823, "step": 7998 }, { "epoch": 0.48, "learning_rate": 2.4120239029395787e-05, "loss": 1.0492, "step": 8000 }, { "epoch": 0.48, "learning_rate": 2.412627512524899e-05, "loss": 1.2287, "step": 8002 }, { "epoch": 0.48, "learning_rate": 2.4132311221102192e-05, "loss": 1.4295, "step": 8004 }, { "epoch": 0.48, "learning_rate": 2.4138347316955396e-05, "loss": 0.9215, "step": 8006 }, { "epoch": 0.48, "learning_rate": 2.4144383412808597e-05, "loss": 1.4005, "step": 8008 }, { "epoch": 0.48, "learning_rate": 2.4150419508661798e-05, "loss": 1.1571, "step": 8010 }, { "epoch": 0.48, "learning_rate": 2.4156455604515002e-05, "loss": 1.2925, "step": 8012 }, { "epoch": 0.48, "learning_rate": 2.4162491700368203e-05, "loss": 1.0693, "step": 8014 }, { "epoch": 0.48, "learning_rate": 2.4168527796221404e-05, "loss": 1.5826, "step": 8016 }, { "epoch": 0.48, "learning_rate": 2.4174563892074605e-05, "loss": 1.2175, "step": 8018 }, { "epoch": 0.48, "learning_rate": 2.418059998792781e-05, "loss": 1.3609, "step": 8020 }, { "epoch": 0.48, "learning_rate": 2.418663608378101e-05, "loss": 1.4975, "step": 8022 }, { "epoch": 0.48, "learning_rate": 2.4192672179634214e-05, "loss": 1.1528, "step": 8024 }, { "epoch": 0.48, "learning_rate": 2.4198708275487418e-05, "loss": 1.245, "step": 8026 }, { "epoch": 0.48, "learning_rate": 2.420474437134062e-05, "loss": 1.3135, "step": 8028 }, { "epoch": 0.48, "learning_rate": 2.421078046719382e-05, "loss": 1.3052, "step": 8030 }, { "epoch": 0.48, "learning_rate": 2.4216816563047024e-05, "loss": 1.6073, "step": 8032 }, { "epoch": 0.48, "learning_rate": 2.4222852658900225e-05, "loss": 1.2094, "step": 8034 }, { "epoch": 0.49, "learning_rate": 2.4228888754753426e-05, "loss": 1.2038, "step": 8036 }, { "epoch": 0.49, "learning_rate": 2.423492485060663e-05, "loss": 1.426, "step": 8038 }, { "epoch": 0.49, "learning_rate": 2.424096094645983e-05, "loss": 1.4219, "step": 8040 }, { "epoch": 0.49, "learning_rate": 2.424699704231303e-05, "loss": 1.1233, "step": 8042 }, { "epoch": 0.49, "learning_rate": 2.4253033138166236e-05, "loss": 0.9688, "step": 8044 }, { "epoch": 0.49, "learning_rate": 2.425906923401944e-05, "loss": 1.3473, "step": 8046 }, { "epoch": 0.49, "learning_rate": 2.426510532987264e-05, "loss": 1.5428, "step": 8048 }, { "epoch": 0.49, "learning_rate": 2.427114142572584e-05, "loss": 1.0206, "step": 8050 }, { "epoch": 0.49, "learning_rate": 2.4277177521579046e-05, "loss": 1.1927, "step": 8052 }, { "epoch": 0.49, "learning_rate": 2.4283213617432246e-05, "loss": 1.2005, "step": 8054 }, { "epoch": 0.49, "learning_rate": 2.4289249713285447e-05, "loss": 1.0041, "step": 8056 }, { "epoch": 0.49, "learning_rate": 2.429528580913865e-05, "loss": 1.131, "step": 8058 }, { "epoch": 0.49, "learning_rate": 2.4301321904991852e-05, "loss": 1.0117, "step": 8060 }, { "epoch": 0.49, "learning_rate": 2.4307358000845053e-05, "loss": 1.3552, "step": 8062 }, { "epoch": 0.49, "learning_rate": 2.4313394096698254e-05, "loss": 1.1978, "step": 8064 }, { "epoch": 0.49, "learning_rate": 2.4319430192551458e-05, "loss": 1.2475, "step": 8066 }, { "epoch": 0.49, "learning_rate": 2.4325466288404662e-05, "loss": 1.1923, "step": 8068 }, { "epoch": 0.49, "learning_rate": 2.4331502384257863e-05, "loss": 1.2845, "step": 8070 }, { "epoch": 0.49, "learning_rate": 2.4337538480111067e-05, "loss": 1.0189, "step": 8072 }, { "epoch": 0.49, "learning_rate": 2.4343574575964268e-05, "loss": 1.3304, "step": 8074 }, { "epoch": 0.49, "learning_rate": 2.434961067181747e-05, "loss": 1.5541, "step": 8076 }, { "epoch": 0.49, "learning_rate": 2.4355646767670673e-05, "loss": 1.2705, "step": 8078 }, { "epoch": 0.49, "learning_rate": 2.4361682863523874e-05, "loss": 1.1541, "step": 8080 }, { "epoch": 0.49, "learning_rate": 2.4367718959377075e-05, "loss": 1.5112, "step": 8082 }, { "epoch": 0.49, "learning_rate": 2.437375505523028e-05, "loss": 1.1359, "step": 8084 }, { "epoch": 0.49, "learning_rate": 2.437979115108348e-05, "loss": 1.2354, "step": 8086 }, { "epoch": 0.49, "learning_rate": 2.438582724693668e-05, "loss": 1.7119, "step": 8088 }, { "epoch": 0.49, "learning_rate": 2.4391863342789885e-05, "loss": 1.1614, "step": 8090 }, { "epoch": 0.49, "learning_rate": 2.439789943864309e-05, "loss": 1.3462, "step": 8092 }, { "epoch": 0.49, "learning_rate": 2.440393553449629e-05, "loss": 1.1741, "step": 8094 }, { "epoch": 0.49, "learning_rate": 2.440997163034949e-05, "loss": 0.9818, "step": 8096 }, { "epoch": 0.49, "learning_rate": 2.4416007726202695e-05, "loss": 0.989, "step": 8098 }, { "epoch": 0.49, "learning_rate": 2.4422043822055896e-05, "loss": 1.5631, "step": 8100 }, { "epoch": 0.49, "learning_rate": 2.4428079917909097e-05, "loss": 1.4022, "step": 8102 }, { "epoch": 0.49, "learning_rate": 2.44341160137623e-05, "loss": 1.0406, "step": 8104 }, { "epoch": 0.49, "learning_rate": 2.44401521096155e-05, "loss": 0.9567, "step": 8106 }, { "epoch": 0.49, "learning_rate": 2.4446188205468702e-05, "loss": 1.2954, "step": 8108 }, { "epoch": 0.49, "learning_rate": 2.4452224301321903e-05, "loss": 1.0838, "step": 8110 }, { "epoch": 0.49, "learning_rate": 2.4458260397175107e-05, "loss": 1.614, "step": 8112 }, { "epoch": 0.49, "learning_rate": 2.446429649302831e-05, "loss": 1.0856, "step": 8114 }, { "epoch": 0.49, "learning_rate": 2.4470332588881512e-05, "loss": 0.8301, "step": 8116 }, { "epoch": 0.49, "learning_rate": 2.4476368684734717e-05, "loss": 1.1622, "step": 8118 }, { "epoch": 0.49, "learning_rate": 2.4482404780587918e-05, "loss": 1.2716, "step": 8120 }, { "epoch": 0.49, "learning_rate": 2.448844087644112e-05, "loss": 1.3558, "step": 8122 }, { "epoch": 0.49, "learning_rate": 2.4494476972294323e-05, "loss": 1.4982, "step": 8124 }, { "epoch": 0.49, "learning_rate": 2.4500513068147523e-05, "loss": 0.9478, "step": 8126 }, { "epoch": 0.49, "learning_rate": 2.4506549164000724e-05, "loss": 0.9525, "step": 8128 }, { "epoch": 0.49, "learning_rate": 2.4512585259853925e-05, "loss": 1.111, "step": 8130 }, { "epoch": 0.49, "learning_rate": 2.451862135570713e-05, "loss": 1.1919, "step": 8132 }, { "epoch": 0.49, "learning_rate": 2.452465745156033e-05, "loss": 1.451, "step": 8134 }, { "epoch": 0.49, "learning_rate": 2.4530693547413534e-05, "loss": 1.0074, "step": 8136 }, { "epoch": 0.49, "learning_rate": 2.453672964326674e-05, "loss": 1.3697, "step": 8138 }, { "epoch": 0.49, "learning_rate": 2.454276573911994e-05, "loss": 1.0531, "step": 8140 }, { "epoch": 0.49, "learning_rate": 2.454880183497314e-05, "loss": 1.3429, "step": 8142 }, { "epoch": 0.49, "learning_rate": 2.4554837930826344e-05, "loss": 1.0151, "step": 8144 }, { "epoch": 0.49, "learning_rate": 2.4560874026679545e-05, "loss": 1.5932, "step": 8146 }, { "epoch": 0.49, "learning_rate": 2.4566910122532746e-05, "loss": 1.2146, "step": 8148 }, { "epoch": 0.49, "learning_rate": 2.457294621838595e-05, "loss": 1.5849, "step": 8150 }, { "epoch": 0.49, "learning_rate": 2.457898231423915e-05, "loss": 1.7305, "step": 8152 }, { "epoch": 0.49, "learning_rate": 2.4585018410092352e-05, "loss": 1.5472, "step": 8154 }, { "epoch": 0.49, "learning_rate": 2.4591054505945556e-05, "loss": 1.2133, "step": 8156 }, { "epoch": 0.49, "learning_rate": 2.4597090601798757e-05, "loss": 1.2363, "step": 8158 }, { "epoch": 0.49, "learning_rate": 2.460312669765196e-05, "loss": 1.3851, "step": 8160 }, { "epoch": 0.49, "learning_rate": 2.4609162793505162e-05, "loss": 1.2349, "step": 8162 }, { "epoch": 0.49, "learning_rate": 2.4615198889358366e-05, "loss": 1.444, "step": 8164 }, { "epoch": 0.49, "learning_rate": 2.4621234985211567e-05, "loss": 1.5373, "step": 8166 }, { "epoch": 0.49, "learning_rate": 2.4627271081064768e-05, "loss": 1.3871, "step": 8168 }, { "epoch": 0.49, "learning_rate": 2.4633307176917972e-05, "loss": 1.4707, "step": 8170 }, { "epoch": 0.49, "learning_rate": 2.4639343272771173e-05, "loss": 0.9979, "step": 8172 }, { "epoch": 0.49, "learning_rate": 2.4645379368624373e-05, "loss": 1.5179, "step": 8174 }, { "epoch": 0.49, "learning_rate": 2.4651415464477574e-05, "loss": 1.4264, "step": 8176 }, { "epoch": 0.49, "learning_rate": 2.465745156033078e-05, "loss": 1.467, "step": 8178 }, { "epoch": 0.49, "learning_rate": 2.4663487656183983e-05, "loss": 1.238, "step": 8180 }, { "epoch": 0.49, "learning_rate": 2.4669523752037184e-05, "loss": 1.3607, "step": 8182 }, { "epoch": 0.49, "learning_rate": 2.4675559847890388e-05, "loss": 1.2781, "step": 8184 }, { "epoch": 0.49, "learning_rate": 2.468159594374359e-05, "loss": 1.3228, "step": 8186 }, { "epoch": 0.49, "learning_rate": 2.468763203959679e-05, "loss": 0.9978, "step": 8188 }, { "epoch": 0.49, "learning_rate": 2.4693668135449994e-05, "loss": 1.1831, "step": 8190 }, { "epoch": 0.49, "learning_rate": 2.4699704231303194e-05, "loss": 1.4436, "step": 8192 }, { "epoch": 0.49, "learning_rate": 2.4705740327156395e-05, "loss": 1.1629, "step": 8194 }, { "epoch": 0.49, "learning_rate": 2.47117764230096e-05, "loss": 1.4904, "step": 8196 }, { "epoch": 0.49, "learning_rate": 2.47178125188628e-05, "loss": 1.4835, "step": 8198 }, { "epoch": 0.49, "learning_rate": 2.4723848614716e-05, "loss": 1.3904, "step": 8200 }, { "epoch": 0.5, "learning_rate": 2.4729884710569205e-05, "loss": 1.1589, "step": 8202 }, { "epoch": 0.5, "learning_rate": 2.473592080642241e-05, "loss": 1.3883, "step": 8204 }, { "epoch": 0.5, "learning_rate": 2.474195690227561e-05, "loss": 0.7981, "step": 8206 }, { "epoch": 0.5, "learning_rate": 2.474799299812881e-05, "loss": 1.3663, "step": 8208 }, { "epoch": 0.5, "learning_rate": 2.4754029093982015e-05, "loss": 1.1499, "step": 8210 }, { "epoch": 0.5, "learning_rate": 2.4760065189835216e-05, "loss": 1.1142, "step": 8212 }, { "epoch": 0.5, "learning_rate": 2.4766101285688417e-05, "loss": 1.3853, "step": 8214 }, { "epoch": 0.5, "learning_rate": 2.477213738154162e-05, "loss": 1.5789, "step": 8216 }, { "epoch": 0.5, "learning_rate": 2.4778173477394822e-05, "loss": 1.2145, "step": 8218 }, { "epoch": 0.5, "learning_rate": 2.4784209573248023e-05, "loss": 1.035, "step": 8220 }, { "epoch": 0.5, "learning_rate": 2.4790245669101224e-05, "loss": 1.6133, "step": 8222 }, { "epoch": 0.5, "learning_rate": 2.4796281764954428e-05, "loss": 1.5703, "step": 8224 }, { "epoch": 0.5, "learning_rate": 2.4802317860807632e-05, "loss": 0.9054, "step": 8226 }, { "epoch": 0.5, "learning_rate": 2.4808353956660833e-05, "loss": 1.5533, "step": 8228 }, { "epoch": 0.5, "learning_rate": 2.4814390052514037e-05, "loss": 1.3515, "step": 8230 }, { "epoch": 0.5, "learning_rate": 2.4820426148367238e-05, "loss": 1.0341, "step": 8232 }, { "epoch": 0.5, "learning_rate": 2.482646224422044e-05, "loss": 1.0885, "step": 8234 }, { "epoch": 0.5, "learning_rate": 2.4832498340073643e-05, "loss": 1.3888, "step": 8236 }, { "epoch": 0.5, "learning_rate": 2.4838534435926844e-05, "loss": 1.3319, "step": 8238 }, { "epoch": 0.5, "learning_rate": 2.4844570531780045e-05, "loss": 1.2565, "step": 8240 }, { "epoch": 0.5, "learning_rate": 2.4850606627633245e-05, "loss": 1.5048, "step": 8242 }, { "epoch": 0.5, "learning_rate": 2.485664272348645e-05, "loss": 1.3156, "step": 8244 }, { "epoch": 0.5, "learning_rate": 2.486267881933965e-05, "loss": 1.1864, "step": 8246 }, { "epoch": 0.5, "learning_rate": 2.4868714915192855e-05, "loss": 1.2027, "step": 8248 }, { "epoch": 0.5, "learning_rate": 2.487475101104606e-05, "loss": 1.0728, "step": 8250 }, { "epoch": 0.5, "learning_rate": 2.488078710689926e-05, "loss": 1.3531, "step": 8252 }, { "epoch": 0.5, "learning_rate": 2.488682320275246e-05, "loss": 1.2667, "step": 8254 }, { "epoch": 0.5, "learning_rate": 2.4892859298605665e-05, "loss": 1.088, "step": 8256 }, { "epoch": 0.5, "learning_rate": 2.4898895394458865e-05, "loss": 1.1623, "step": 8258 }, { "epoch": 0.5, "learning_rate": 2.4904931490312066e-05, "loss": 1.2012, "step": 8260 }, { "epoch": 0.5, "learning_rate": 2.491096758616527e-05, "loss": 1.2922, "step": 8262 }, { "epoch": 0.5, "learning_rate": 2.491700368201847e-05, "loss": 1.375, "step": 8264 }, { "epoch": 0.5, "learning_rate": 2.4923039777871672e-05, "loss": 1.1001, "step": 8266 }, { "epoch": 0.5, "learning_rate": 2.4929075873724876e-05, "loss": 1.4759, "step": 8268 }, { "epoch": 0.5, "learning_rate": 2.4935111969578077e-05, "loss": 1.1124, "step": 8270 }, { "epoch": 0.5, "learning_rate": 2.494114806543128e-05, "loss": 1.1629, "step": 8272 }, { "epoch": 0.5, "learning_rate": 2.4947184161284482e-05, "loss": 1.1225, "step": 8274 }, { "epoch": 0.5, "learning_rate": 2.4953220257137686e-05, "loss": 1.04, "step": 8276 }, { "epoch": 0.5, "learning_rate": 2.4959256352990887e-05, "loss": 1.4306, "step": 8278 }, { "epoch": 0.5, "learning_rate": 2.4965292448844088e-05, "loss": 1.134, "step": 8280 }, { "epoch": 0.5, "learning_rate": 2.4971328544697292e-05, "loss": 1.2404, "step": 8282 }, { "epoch": 0.5, "learning_rate": 2.4977364640550493e-05, "loss": 1.3123, "step": 8284 }, { "epoch": 0.5, "learning_rate": 2.4983400736403694e-05, "loss": 0.7623, "step": 8286 }, { "epoch": 0.5, "learning_rate": 2.4989436832256895e-05, "loss": 1.397, "step": 8288 }, { "epoch": 0.5, "learning_rate": 2.49954729281101e-05, "loss": 1.3015, "step": 8290 }, { "epoch": 0.5, "learning_rate": 2.5001509023963303e-05, "loss": 1.3054, "step": 8292 }, { "epoch": 0.5, "learning_rate": 2.5007545119816504e-05, "loss": 1.1595, "step": 8294 }, { "epoch": 0.5, "learning_rate": 2.5013581215669708e-05, "loss": 1.1508, "step": 8296 }, { "epoch": 0.5, "learning_rate": 2.501961731152291e-05, "loss": 1.0219, "step": 8298 }, { "epoch": 0.5, "learning_rate": 2.502565340737611e-05, "loss": 1.3752, "step": 8300 }, { "epoch": 0.5, "learning_rate": 2.503168950322931e-05, "loss": 0.9807, "step": 8302 }, { "epoch": 0.5, "learning_rate": 2.503772559908251e-05, "loss": 1.109, "step": 8304 }, { "epoch": 0.5, "learning_rate": 2.504376169493572e-05, "loss": 1.2516, "step": 8306 }, { "epoch": 0.5, "learning_rate": 2.504979779078892e-05, "loss": 0.9, "step": 8308 }, { "epoch": 0.5, "learning_rate": 2.505583388664212e-05, "loss": 1.5247, "step": 8310 }, { "epoch": 0.5, "learning_rate": 2.506186998249532e-05, "loss": 1.5929, "step": 8312 }, { "epoch": 0.5, "learning_rate": 2.5067906078348526e-05, "loss": 1.153, "step": 8314 }, { "epoch": 0.5, "learning_rate": 2.5073942174201726e-05, "loss": 1.1323, "step": 8316 }, { "epoch": 0.5, "learning_rate": 2.5079978270054934e-05, "loss": 1.232, "step": 8318 }, { "epoch": 0.5, "learning_rate": 2.5086014365908135e-05, "loss": 1.3044, "step": 8320 }, { "epoch": 0.5, "learning_rate": 2.5092050461761336e-05, "loss": 1.3192, "step": 8322 }, { "epoch": 0.5, "learning_rate": 2.5098086557614537e-05, "loss": 1.3468, "step": 8324 }, { "epoch": 0.5, "learning_rate": 2.5104122653467737e-05, "loss": 1.315, "step": 8326 }, { "epoch": 0.5, "learning_rate": 2.5110158749320938e-05, "loss": 0.7684, "step": 8328 }, { "epoch": 0.5, "learning_rate": 2.511619484517414e-05, "loss": 1.2417, "step": 8330 }, { "epoch": 0.5, "learning_rate": 2.5122230941027347e-05, "loss": 1.0214, "step": 8332 }, { "epoch": 0.5, "learning_rate": 2.5128267036880547e-05, "loss": 1.1619, "step": 8334 }, { "epoch": 0.5, "learning_rate": 2.5134303132733748e-05, "loss": 1.0866, "step": 8336 }, { "epoch": 0.5, "learning_rate": 2.5140339228586952e-05, "loss": 1.2247, "step": 8338 }, { "epoch": 0.5, "learning_rate": 2.5146375324440153e-05, "loss": 1.2577, "step": 8340 }, { "epoch": 0.5, "learning_rate": 2.5152411420293354e-05, "loss": 1.4086, "step": 8342 }, { "epoch": 0.5, "learning_rate": 2.5158447516146555e-05, "loss": 1.3927, "step": 8344 }, { "epoch": 0.5, "learning_rate": 2.5164483611999762e-05, "loss": 1.1095, "step": 8346 }, { "epoch": 0.5, "learning_rate": 2.5170519707852963e-05, "loss": 1.641, "step": 8348 }, { "epoch": 0.5, "learning_rate": 2.5176555803706164e-05, "loss": 1.2543, "step": 8350 }, { "epoch": 0.5, "learning_rate": 2.5182591899559365e-05, "loss": 1.2353, "step": 8352 }, { "epoch": 0.5, "learning_rate": 2.5188627995412566e-05, "loss": 1.1918, "step": 8354 }, { "epoch": 0.5, "learning_rate": 2.519466409126577e-05, "loss": 1.1285, "step": 8356 }, { "epoch": 0.5, "learning_rate": 2.5200700187118974e-05, "loss": 1.3958, "step": 8358 }, { "epoch": 0.5, "learning_rate": 2.5206736282972175e-05, "loss": 1.038, "step": 8360 }, { "epoch": 0.5, "learning_rate": 2.521277237882538e-05, "loss": 1.4192, "step": 8362 }, { "epoch": 0.5, "learning_rate": 2.521880847467858e-05, "loss": 1.4572, "step": 8364 }, { "epoch": 0.5, "learning_rate": 2.522484457053178e-05, "loss": 1.2869, "step": 8366 }, { "epoch": 0.51, "learning_rate": 2.523088066638498e-05, "loss": 1.8324, "step": 8368 }, { "epoch": 0.51, "learning_rate": 2.5236916762238182e-05, "loss": 1.4663, "step": 8370 }, { "epoch": 0.51, "learning_rate": 2.524295285809139e-05, "loss": 1.4226, "step": 8372 }, { "epoch": 0.51, "learning_rate": 2.524898895394459e-05, "loss": 1.332, "step": 8374 }, { "epoch": 0.51, "learning_rate": 2.525502504979779e-05, "loss": 1.1697, "step": 8376 }, { "epoch": 0.51, "learning_rate": 2.5261061145650993e-05, "loss": 1.3968, "step": 8378 }, { "epoch": 0.51, "learning_rate": 2.5267097241504197e-05, "loss": 1.3351, "step": 8380 }, { "epoch": 0.51, "learning_rate": 2.5273133337357398e-05, "loss": 1.2235, "step": 8382 }, { "epoch": 0.51, "learning_rate": 2.5279169433210602e-05, "loss": 1.0164, "step": 8384 }, { "epoch": 0.51, "learning_rate": 2.5285205529063806e-05, "loss": 1.3175, "step": 8386 }, { "epoch": 0.51, "learning_rate": 2.5291241624917007e-05, "loss": 1.169, "step": 8388 }, { "epoch": 0.51, "learning_rate": 2.5297277720770208e-05, "loss": 1.0399, "step": 8390 }, { "epoch": 0.51, "learning_rate": 2.530331381662341e-05, "loss": 1.0355, "step": 8392 }, { "epoch": 0.51, "learning_rate": 2.530934991247661e-05, "loss": 1.0098, "step": 8394 }, { "epoch": 0.51, "learning_rate": 2.531538600832981e-05, "loss": 1.6367, "step": 8396 }, { "epoch": 0.51, "learning_rate": 2.5321422104183018e-05, "loss": 1.0981, "step": 8398 }, { "epoch": 0.51, "learning_rate": 2.532745820003622e-05, "loss": 1.5598, "step": 8400 }, { "epoch": 0.51, "learning_rate": 2.533349429588942e-05, "loss": 0.7722, "step": 8402 }, { "epoch": 0.51, "learning_rate": 2.5339530391742623e-05, "loss": 1.4129, "step": 8404 }, { "epoch": 0.51, "learning_rate": 2.5345566487595824e-05, "loss": 1.2984, "step": 8406 }, { "epoch": 0.51, "learning_rate": 2.5351602583449025e-05, "loss": 1.3598, "step": 8408 }, { "epoch": 0.51, "learning_rate": 2.5357638679302226e-05, "loss": 1.3899, "step": 8410 }, { "epoch": 0.51, "learning_rate": 2.5363674775155434e-05, "loss": 1.2319, "step": 8412 }, { "epoch": 0.51, "learning_rate": 2.5369710871008634e-05, "loss": 1.5493, "step": 8414 }, { "epoch": 0.51, "learning_rate": 2.5375746966861835e-05, "loss": 1.5896, "step": 8416 }, { "epoch": 0.51, "learning_rate": 2.5381783062715036e-05, "loss": 1.0484, "step": 8418 }, { "epoch": 0.51, "learning_rate": 2.5387819158568237e-05, "loss": 1.0843, "step": 8420 }, { "epoch": 0.51, "learning_rate": 2.5393855254421438e-05, "loss": 1.079, "step": 8422 }, { "epoch": 0.51, "learning_rate": 2.5399891350274645e-05, "loss": 1.7479, "step": 8424 }, { "epoch": 0.51, "learning_rate": 2.5405927446127846e-05, "loss": 0.8881, "step": 8426 }, { "epoch": 0.51, "learning_rate": 2.541196354198105e-05, "loss": 1.2213, "step": 8428 }, { "epoch": 0.51, "learning_rate": 2.541799963783425e-05, "loss": 1.0246, "step": 8430 }, { "epoch": 0.51, "learning_rate": 2.5424035733687452e-05, "loss": 1.8298, "step": 8432 }, { "epoch": 0.51, "learning_rate": 2.5430071829540653e-05, "loss": 1.3444, "step": 8434 }, { "epoch": 0.51, "learning_rate": 2.5436107925393854e-05, "loss": 0.9531, "step": 8436 }, { "epoch": 0.51, "learning_rate": 2.544214402124706e-05, "loss": 1.035, "step": 8438 }, { "epoch": 0.51, "learning_rate": 2.5448180117100262e-05, "loss": 1.4635, "step": 8440 }, { "epoch": 0.51, "learning_rate": 2.5454216212953463e-05, "loss": 1.1336, "step": 8442 }, { "epoch": 0.51, "learning_rate": 2.5460252308806664e-05, "loss": 1.3023, "step": 8444 }, { "epoch": 0.51, "learning_rate": 2.5466288404659864e-05, "loss": 1.1973, "step": 8446 }, { "epoch": 0.51, "learning_rate": 2.547232450051307e-05, "loss": 1.0708, "step": 8448 }, { "epoch": 0.51, "learning_rate": 2.5478360596366273e-05, "loss": 1.0778, "step": 8450 }, { "epoch": 0.51, "learning_rate": 2.5484396692219477e-05, "loss": 1.2633, "step": 8452 }, { "epoch": 0.51, "learning_rate": 2.5490432788072678e-05, "loss": 1.3192, "step": 8454 }, { "epoch": 0.51, "learning_rate": 2.549646888392588e-05, "loss": 0.9598, "step": 8456 }, { "epoch": 0.51, "learning_rate": 2.550250497977908e-05, "loss": 1.1255, "step": 8458 }, { "epoch": 0.51, "learning_rate": 2.550854107563228e-05, "loss": 1.221, "step": 8460 }, { "epoch": 0.51, "learning_rate": 2.551457717148548e-05, "loss": 1.5649, "step": 8462 }, { "epoch": 0.51, "learning_rate": 2.552061326733869e-05, "loss": 1.1861, "step": 8464 }, { "epoch": 0.51, "learning_rate": 2.552664936319189e-05, "loss": 1.4341, "step": 8466 }, { "epoch": 0.51, "learning_rate": 2.553268545904509e-05, "loss": 1.3062, "step": 8468 }, { "epoch": 0.51, "learning_rate": 2.553872155489829e-05, "loss": 1.1588, "step": 8470 }, { "epoch": 0.51, "learning_rate": 2.5544757650751495e-05, "loss": 1.1979, "step": 8472 }, { "epoch": 0.51, "learning_rate": 2.5550793746604696e-05, "loss": 1.2956, "step": 8474 }, { "epoch": 0.51, "learning_rate": 2.5556829842457897e-05, "loss": 1.0413, "step": 8476 }, { "epoch": 0.51, "learning_rate": 2.5562865938311105e-05, "loss": 1.006, "step": 8478 }, { "epoch": 0.51, "learning_rate": 2.5568902034164305e-05, "loss": 1.3387, "step": 8480 }, { "epoch": 0.51, "learning_rate": 2.5574938130017506e-05, "loss": 1.0609, "step": 8482 }, { "epoch": 0.51, "learning_rate": 2.5580974225870707e-05, "loss": 1.1438, "step": 8484 }, { "epoch": 0.51, "learning_rate": 2.5587010321723908e-05, "loss": 1.6005, "step": 8486 }, { "epoch": 0.51, "learning_rate": 2.559304641757711e-05, "loss": 1.3127, "step": 8488 }, { "epoch": 0.51, "learning_rate": 2.5599082513430316e-05, "loss": 1.232, "step": 8490 }, { "epoch": 0.51, "learning_rate": 2.5605118609283517e-05, "loss": 1.2017, "step": 8492 }, { "epoch": 0.51, "learning_rate": 2.5611154705136718e-05, "loss": 0.9465, "step": 8494 }, { "epoch": 0.51, "learning_rate": 2.5617190800989922e-05, "loss": 1.0485, "step": 8496 }, { "epoch": 0.51, "learning_rate": 2.5623226896843123e-05, "loss": 1.1585, "step": 8498 }, { "epoch": 0.51, "learning_rate": 2.5629262992696324e-05, "loss": 1.4197, "step": 8500 }, { "epoch": 0.51, "learning_rate": 2.5635299088549525e-05, "loss": 1.3985, "step": 8502 }, { "epoch": 0.51, "learning_rate": 2.5641335184402732e-05, "loss": 1.3522, "step": 8504 }, { "epoch": 0.51, "learning_rate": 2.5647371280255933e-05, "loss": 1.2921, "step": 8506 }, { "epoch": 0.51, "learning_rate": 2.5653407376109134e-05, "loss": 1.401, "step": 8508 }, { "epoch": 0.51, "learning_rate": 2.5659443471962335e-05, "loss": 0.8908, "step": 8510 }, { "epoch": 0.51, "learning_rate": 2.5665479567815535e-05, "loss": 1.355, "step": 8512 }, { "epoch": 0.51, "learning_rate": 2.567151566366874e-05, "loss": 1.0864, "step": 8514 }, { "epoch": 0.51, "learning_rate": 2.5677551759521944e-05, "loss": 1.3265, "step": 8516 }, { "epoch": 0.51, "learning_rate": 2.5683587855375145e-05, "loss": 1.2289, "step": 8518 }, { "epoch": 0.51, "learning_rate": 2.568962395122835e-05, "loss": 1.1609, "step": 8520 }, { "epoch": 0.51, "learning_rate": 2.569566004708155e-05, "loss": 1.5138, "step": 8522 }, { "epoch": 0.51, "learning_rate": 2.570169614293475e-05, "loss": 1.4912, "step": 8524 }, { "epoch": 0.51, "learning_rate": 2.570773223878795e-05, "loss": 1.0714, "step": 8526 }, { "epoch": 0.51, "learning_rate": 2.5713768334641152e-05, "loss": 1.7318, "step": 8528 }, { "epoch": 0.51, "learning_rate": 2.571980443049436e-05, "loss": 1.3372, "step": 8530 }, { "epoch": 0.51, "learning_rate": 2.572584052634756e-05, "loss": 1.3375, "step": 8532 }, { "epoch": 0.52, "learning_rate": 2.573187662220076e-05, "loss": 1.1825, "step": 8534 }, { "epoch": 0.52, "learning_rate": 2.5737912718053962e-05, "loss": 1.4345, "step": 8536 }, { "epoch": 0.52, "learning_rate": 2.5743948813907166e-05, "loss": 0.8748, "step": 8538 }, { "epoch": 0.52, "learning_rate": 2.5749984909760367e-05, "loss": 1.0943, "step": 8540 }, { "epoch": 0.52, "learning_rate": 2.575602100561357e-05, "loss": 0.9699, "step": 8542 }, { "epoch": 0.52, "learning_rate": 2.5762057101466776e-05, "loss": 1.2349, "step": 8544 }, { "epoch": 0.52, "learning_rate": 2.5768093197319976e-05, "loss": 0.8947, "step": 8546 }, { "epoch": 0.52, "learning_rate": 2.5774129293173177e-05, "loss": 1.3185, "step": 8548 }, { "epoch": 0.52, "learning_rate": 2.5780165389026378e-05, "loss": 1.1593, "step": 8550 }, { "epoch": 0.52, "learning_rate": 2.578620148487958e-05, "loss": 1.3929, "step": 8552 }, { "epoch": 0.52, "learning_rate": 2.579223758073278e-05, "loss": 1.1933, "step": 8554 }, { "epoch": 0.52, "learning_rate": 2.5798273676585987e-05, "loss": 1.3483, "step": 8556 }, { "epoch": 0.52, "learning_rate": 2.5804309772439188e-05, "loss": 1.1305, "step": 8558 }, { "epoch": 0.52, "learning_rate": 2.581034586829239e-05, "loss": 1.3287, "step": 8560 }, { "epoch": 0.52, "learning_rate": 2.5816381964145593e-05, "loss": 1.2629, "step": 8562 }, { "epoch": 0.52, "learning_rate": 2.5822418059998794e-05, "loss": 1.4044, "step": 8564 }, { "epoch": 0.52, "learning_rate": 2.5828454155851995e-05, "loss": 1.3835, "step": 8566 }, { "epoch": 0.52, "learning_rate": 2.5834490251705196e-05, "loss": 0.8832, "step": 8568 }, { "epoch": 0.52, "learning_rate": 2.5840526347558403e-05, "loss": 1.1992, "step": 8570 }, { "epoch": 0.52, "learning_rate": 2.5846562443411604e-05, "loss": 1.5066, "step": 8572 }, { "epoch": 0.52, "learning_rate": 2.5852598539264805e-05, "loss": 1.3273, "step": 8574 }, { "epoch": 0.52, "learning_rate": 2.5858634635118006e-05, "loss": 1.0443, "step": 8576 }, { "epoch": 0.52, "learning_rate": 2.5864670730971207e-05, "loss": 1.2134, "step": 8578 }, { "epoch": 0.52, "learning_rate": 2.587070682682441e-05, "loss": 1.4535, "step": 8580 }, { "epoch": 0.52, "learning_rate": 2.5876742922677615e-05, "loss": 1.0187, "step": 8582 }, { "epoch": 0.52, "learning_rate": 2.5882779018530816e-05, "loss": 1.0119, "step": 8584 }, { "epoch": 0.52, "learning_rate": 2.588881511438402e-05, "loss": 1.0741, "step": 8586 }, { "epoch": 0.52, "learning_rate": 2.589485121023722e-05, "loss": 1.5303, "step": 8588 }, { "epoch": 0.52, "learning_rate": 2.590088730609042e-05, "loss": 1.1584, "step": 8590 }, { "epoch": 0.52, "learning_rate": 2.5906923401943622e-05, "loss": 1.2186, "step": 8592 }, { "epoch": 0.52, "learning_rate": 2.5912959497796823e-05, "loss": 1.7965, "step": 8594 }, { "epoch": 0.52, "learning_rate": 2.591899559365003e-05, "loss": 1.1506, "step": 8596 }, { "epoch": 0.52, "learning_rate": 2.592503168950323e-05, "loss": 1.6207, "step": 8598 }, { "epoch": 0.52, "learning_rate": 2.5931067785356432e-05, "loss": 1.4934, "step": 8600 }, { "epoch": 0.52, "learning_rate": 2.5937103881209633e-05, "loss": 1.2242, "step": 8602 }, { "epoch": 0.52, "learning_rate": 2.5943139977062837e-05, "loss": 1.1769, "step": 8604 }, { "epoch": 0.52, "learning_rate": 2.5949176072916038e-05, "loss": 1.3391, "step": 8606 }, { "epoch": 0.52, "learning_rate": 2.5955212168769242e-05, "loss": 0.9964, "step": 8608 }, { "epoch": 0.52, "learning_rate": 2.5961248264622447e-05, "loss": 1.5496, "step": 8610 }, { "epoch": 0.52, "learning_rate": 2.5967284360475648e-05, "loss": 1.3653, "step": 8612 }, { "epoch": 0.52, "learning_rate": 2.597332045632885e-05, "loss": 1.0139, "step": 8614 }, { "epoch": 0.52, "learning_rate": 2.597935655218205e-05, "loss": 0.9998, "step": 8616 }, { "epoch": 0.52, "learning_rate": 2.598539264803525e-05, "loss": 1.2568, "step": 8618 }, { "epoch": 0.52, "learning_rate": 2.599142874388845e-05, "loss": 1.2381, "step": 8620 }, { "epoch": 0.52, "learning_rate": 2.599746483974166e-05, "loss": 1.3861, "step": 8622 }, { "epoch": 0.52, "learning_rate": 2.600350093559486e-05, "loss": 1.1623, "step": 8624 }, { "epoch": 0.52, "learning_rate": 2.600953703144806e-05, "loss": 1.3056, "step": 8626 }, { "epoch": 0.52, "learning_rate": 2.6015573127301264e-05, "loss": 1.1005, "step": 8628 }, { "epoch": 0.52, "learning_rate": 2.6021609223154465e-05, "loss": 1.006, "step": 8630 }, { "epoch": 0.52, "learning_rate": 2.6027645319007666e-05, "loss": 1.0919, "step": 8632 }, { "epoch": 0.52, "learning_rate": 2.6033681414860867e-05, "loss": 1.416, "step": 8634 }, { "epoch": 0.52, "learning_rate": 2.6039717510714074e-05, "loss": 1.1653, "step": 8636 }, { "epoch": 0.52, "learning_rate": 2.6045753606567275e-05, "loss": 1.0967, "step": 8638 }, { "epoch": 0.52, "learning_rate": 2.6051789702420476e-05, "loss": 1.0584, "step": 8640 }, { "epoch": 0.52, "learning_rate": 2.6057825798273677e-05, "loss": 1.6, "step": 8642 }, { "epoch": 0.52, "learning_rate": 2.6063861894126878e-05, "loss": 1.4511, "step": 8644 }, { "epoch": 0.52, "learning_rate": 2.606989798998008e-05, "loss": 1.1146, "step": 8646 }, { "epoch": 0.52, "learning_rate": 2.6075934085833286e-05, "loss": 1.7178, "step": 8648 }, { "epoch": 0.52, "learning_rate": 2.6081970181686487e-05, "loss": 1.2858, "step": 8650 }, { "epoch": 0.52, "learning_rate": 2.608800627753969e-05, "loss": 1.0609, "step": 8652 }, { "epoch": 0.52, "learning_rate": 2.6094042373392892e-05, "loss": 1.227, "step": 8654 }, { "epoch": 0.52, "learning_rate": 2.6100078469246093e-05, "loss": 1.2504, "step": 8656 }, { "epoch": 0.52, "learning_rate": 2.6106114565099293e-05, "loss": 1.296, "step": 8658 }, { "epoch": 0.52, "learning_rate": 2.6112150660952494e-05, "loss": 1.7869, "step": 8660 }, { "epoch": 0.52, "learning_rate": 2.6118186756805702e-05, "loss": 1.2613, "step": 8662 }, { "epoch": 0.52, "learning_rate": 2.6124222852658903e-05, "loss": 1.401, "step": 8664 }, { "epoch": 0.52, "learning_rate": 2.6130258948512104e-05, "loss": 1.2967, "step": 8666 }, { "epoch": 0.52, "learning_rate": 2.6136295044365304e-05, "loss": 1.1725, "step": 8668 }, { "epoch": 0.52, "learning_rate": 2.6142331140218505e-05, "loss": 1.2612, "step": 8670 }, { "epoch": 0.52, "learning_rate": 2.614836723607171e-05, "loss": 1.5737, "step": 8672 }, { "epoch": 0.52, "learning_rate": 2.6154403331924914e-05, "loss": 1.1867, "step": 8674 }, { "epoch": 0.52, "learning_rate": 2.6160439427778118e-05, "loss": 1.3457, "step": 8676 }, { "epoch": 0.52, "learning_rate": 2.616647552363132e-05, "loss": 1.3109, "step": 8678 }, { "epoch": 0.52, "learning_rate": 2.617251161948452e-05, "loss": 1.3032, "step": 8680 }, { "epoch": 0.52, "learning_rate": 2.617854771533772e-05, "loss": 1.3888, "step": 8682 }, { "epoch": 0.52, "learning_rate": 2.618458381119092e-05, "loss": 1.6345, "step": 8684 }, { "epoch": 0.52, "learning_rate": 2.6190619907044122e-05, "loss": 1.3176, "step": 8686 }, { "epoch": 0.52, "learning_rate": 2.619665600289733e-05, "loss": 1.3791, "step": 8688 }, { "epoch": 0.52, "learning_rate": 2.620269209875053e-05, "loss": 1.0416, "step": 8690 }, { "epoch": 0.52, "learning_rate": 2.620872819460373e-05, "loss": 1.3804, "step": 8692 }, { "epoch": 0.52, "learning_rate": 2.6214764290456932e-05, "loss": 1.2349, "step": 8694 }, { "epoch": 0.52, "learning_rate": 2.6220800386310136e-05, "loss": 1.1829, "step": 8696 }, { "epoch": 0.52, "learning_rate": 2.6226836482163337e-05, "loss": 1.271, "step": 8698 }, { "epoch": 0.53, "learning_rate": 2.6232872578016538e-05, "loss": 1.7041, "step": 8700 }, { "epoch": 0.53, "learning_rate": 2.6238908673869745e-05, "loss": 1.1982, "step": 8702 }, { "epoch": 0.53, "learning_rate": 2.6244944769722946e-05, "loss": 1.2023, "step": 8704 }, { "epoch": 0.53, "learning_rate": 2.6250980865576147e-05, "loss": 1.0958, "step": 8706 }, { "epoch": 0.53, "learning_rate": 2.6257016961429348e-05, "loss": 1.5344, "step": 8708 }, { "epoch": 0.53, "learning_rate": 2.626305305728255e-05, "loss": 1.5019, "step": 8710 }, { "epoch": 0.53, "learning_rate": 2.626908915313575e-05, "loss": 1.3603, "step": 8712 }, { "epoch": 0.53, "learning_rate": 2.6275125248988957e-05, "loss": 1.1679, "step": 8714 }, { "epoch": 0.53, "learning_rate": 2.6281161344842158e-05, "loss": 1.096, "step": 8716 }, { "epoch": 0.53, "learning_rate": 2.628719744069536e-05, "loss": 1.1101, "step": 8718 }, { "epoch": 0.53, "learning_rate": 2.6293233536548563e-05, "loss": 1.2186, "step": 8720 }, { "epoch": 0.53, "learning_rate": 2.6299269632401764e-05, "loss": 1.323, "step": 8722 }, { "epoch": 0.53, "learning_rate": 2.6305305728254965e-05, "loss": 1.2707, "step": 8724 }, { "epoch": 0.53, "learning_rate": 2.6311341824108165e-05, "loss": 1.1902, "step": 8726 }, { "epoch": 0.53, "learning_rate": 2.6317377919961373e-05, "loss": 1.2822, "step": 8728 }, { "epoch": 0.53, "learning_rate": 2.6323414015814574e-05, "loss": 1.3757, "step": 8730 }, { "epoch": 0.53, "learning_rate": 2.6329450111667775e-05, "loss": 0.9413, "step": 8732 }, { "epoch": 0.53, "learning_rate": 2.6335486207520975e-05, "loss": 1.1264, "step": 8734 }, { "epoch": 0.53, "learning_rate": 2.6341522303374176e-05, "loss": 0.9889, "step": 8736 }, { "epoch": 0.53, "learning_rate": 2.634755839922738e-05, "loss": 1.2525, "step": 8738 }, { "epoch": 0.53, "learning_rate": 2.6353594495080585e-05, "loss": 0.9457, "step": 8740 }, { "epoch": 0.53, "learning_rate": 2.6359630590933785e-05, "loss": 1.0974, "step": 8742 }, { "epoch": 0.53, "learning_rate": 2.636566668678699e-05, "loss": 1.4915, "step": 8744 }, { "epoch": 0.53, "learning_rate": 2.636868473471359e-05, "loss": 1.749, "step": 8746 }, { "epoch": 0.53, "learning_rate": 2.637472083056679e-05, "loss": 1.0251, "step": 8748 }, { "epoch": 0.53, "learning_rate": 2.638075692641999e-05, "loss": 1.1753, "step": 8750 }, { "epoch": 0.53, "learning_rate": 2.6386793022273193e-05, "loss": 1.478, "step": 8752 }, { "epoch": 0.53, "learning_rate": 2.63928291181264e-05, "loss": 1.1923, "step": 8754 }, { "epoch": 0.53, "learning_rate": 2.63988652139796e-05, "loss": 1.1524, "step": 8756 }, { "epoch": 0.53, "learning_rate": 2.6404901309832802e-05, "loss": 1.2102, "step": 8758 }, { "epoch": 0.53, "learning_rate": 2.6410937405686003e-05, "loss": 1.4084, "step": 8760 }, { "epoch": 0.53, "learning_rate": 2.6416973501539203e-05, "loss": 1.0131, "step": 8762 }, { "epoch": 0.53, "learning_rate": 2.6423009597392408e-05, "loss": 0.976, "step": 8764 }, { "epoch": 0.53, "learning_rate": 2.642904569324561e-05, "loss": 1.5589, "step": 8766 }, { "epoch": 0.53, "learning_rate": 2.6435081789098816e-05, "loss": 1.2872, "step": 8768 }, { "epoch": 0.53, "learning_rate": 2.6441117884952017e-05, "loss": 1.4412, "step": 8770 }, { "epoch": 0.53, "learning_rate": 2.6447153980805218e-05, "loss": 1.1389, "step": 8772 }, { "epoch": 0.53, "learning_rate": 2.645319007665842e-05, "loss": 1.2114, "step": 8774 }, { "epoch": 0.53, "learning_rate": 2.645922617251162e-05, "loss": 1.6595, "step": 8776 }, { "epoch": 0.53, "learning_rate": 2.646526226836482e-05, "loss": 1.4783, "step": 8778 }, { "epoch": 0.53, "learning_rate": 2.647129836421802e-05, "loss": 1.309, "step": 8780 }, { "epoch": 0.53, "learning_rate": 2.647733446007123e-05, "loss": 1.2077, "step": 8782 }, { "epoch": 0.53, "learning_rate": 2.648337055592443e-05, "loss": 0.9658, "step": 8784 }, { "epoch": 0.53, "learning_rate": 2.648940665177763e-05, "loss": 1.8621, "step": 8786 }, { "epoch": 0.53, "learning_rate": 2.6495442747630834e-05, "loss": 1.2226, "step": 8788 }, { "epoch": 0.53, "learning_rate": 2.6501478843484035e-05, "loss": 1.5, "step": 8790 }, { "epoch": 0.53, "learning_rate": 2.6507514939337236e-05, "loss": 1.1354, "step": 8792 }, { "epoch": 0.53, "learning_rate": 2.6513551035190444e-05, "loss": 1.6683, "step": 8794 }, { "epoch": 0.53, "learning_rate": 2.6519587131043644e-05, "loss": 1.4728, "step": 8796 }, { "epoch": 0.53, "learning_rate": 2.6525623226896845e-05, "loss": 1.3628, "step": 8798 }, { "epoch": 0.53, "learning_rate": 2.6531659322750046e-05, "loss": 1.1918, "step": 8800 }, { "epoch": 0.53, "learning_rate": 2.6537695418603247e-05, "loss": 1.3535, "step": 8802 }, { "epoch": 0.53, "learning_rate": 2.6543731514456448e-05, "loss": 1.2088, "step": 8804 }, { "epoch": 0.53, "learning_rate": 2.6549767610309652e-05, "loss": 1.1978, "step": 8806 }, { "epoch": 0.53, "learning_rate": 2.6555803706162856e-05, "loss": 1.0595, "step": 8808 }, { "epoch": 0.53, "learning_rate": 2.6561839802016057e-05, "loss": 1.6281, "step": 8810 }, { "epoch": 0.53, "learning_rate": 2.656787589786926e-05, "loss": 1.4943, "step": 8812 }, { "epoch": 0.53, "learning_rate": 2.6573911993722462e-05, "loss": 1.4156, "step": 8814 }, { "epoch": 0.53, "learning_rate": 2.6579948089575663e-05, "loss": 1.2745, "step": 8816 }, { "epoch": 0.53, "learning_rate": 2.6585984185428864e-05, "loss": 0.786, "step": 8818 }, { "epoch": 0.53, "learning_rate": 2.659202028128207e-05, "loss": 1.317, "step": 8820 }, { "epoch": 0.53, "learning_rate": 2.6598056377135272e-05, "loss": 1.3512, "step": 8822 }, { "epoch": 0.53, "learning_rate": 2.6604092472988473e-05, "loss": 0.9038, "step": 8824 }, { "epoch": 0.53, "learning_rate": 2.6610128568841674e-05, "loss": 1.2308, "step": 8826 }, { "epoch": 0.53, "learning_rate": 2.6616164664694874e-05, "loss": 1.4264, "step": 8828 }, { "epoch": 0.53, "learning_rate": 2.662220076054808e-05, "loss": 1.2087, "step": 8830 }, { "epoch": 0.53, "learning_rate": 2.662823685640128e-05, "loss": 1.2791, "step": 8832 }, { "epoch": 0.53, "learning_rate": 2.6634272952254487e-05, "loss": 1.2842, "step": 8834 }, { "epoch": 0.53, "learning_rate": 2.6640309048107688e-05, "loss": 1.1349, "step": 8836 }, { "epoch": 0.53, "learning_rate": 2.664634514396089e-05, "loss": 1.1128, "step": 8838 }, { "epoch": 0.53, "learning_rate": 2.665238123981409e-05, "loss": 1.1146, "step": 8840 }, { "epoch": 0.53, "learning_rate": 2.665841733566729e-05, "loss": 1.0546, "step": 8842 }, { "epoch": 0.53, "learning_rate": 2.666445343152049e-05, "loss": 1.3383, "step": 8844 }, { "epoch": 0.53, "learning_rate": 2.6670489527373692e-05, "loss": 1.2305, "step": 8846 }, { "epoch": 0.53, "learning_rate": 2.66765256232269e-05, "loss": 1.3503, "step": 8848 }, { "epoch": 0.53, "learning_rate": 2.66825617190801e-05, "loss": 1.1436, "step": 8850 }, { "epoch": 0.53, "learning_rate": 2.66885978149333e-05, "loss": 1.1782, "step": 8852 }, { "epoch": 0.53, "learning_rate": 2.6694633910786505e-05, "loss": 0.9298, "step": 8854 }, { "epoch": 0.53, "learning_rate": 2.6700670006639706e-05, "loss": 1.7384, "step": 8856 }, { "epoch": 0.53, "learning_rate": 2.6706706102492907e-05, "loss": 1.3351, "step": 8858 }, { "epoch": 0.53, "learning_rate": 2.6712742198346115e-05, "loss": 1.246, "step": 8860 }, { "epoch": 0.53, "learning_rate": 2.6718778294199315e-05, "loss": 1.1446, "step": 8862 }, { "epoch": 0.54, "learning_rate": 2.6724814390052516e-05, "loss": 1.3124, "step": 8864 }, { "epoch": 0.54, "learning_rate": 2.6730850485905717e-05, "loss": 1.4694, "step": 8866 }, { "epoch": 0.54, "learning_rate": 2.6736886581758918e-05, "loss": 1.1613, "step": 8868 }, { "epoch": 0.54, "learning_rate": 2.674292267761212e-05, "loss": 1.2629, "step": 8870 }, { "epoch": 0.54, "learning_rate": 2.6748958773465323e-05, "loss": 1.507, "step": 8872 }, { "epoch": 0.54, "learning_rate": 2.6754994869318527e-05, "loss": 1.0537, "step": 8874 }, { "epoch": 0.54, "learning_rate": 2.6761030965171728e-05, "loss": 1.0763, "step": 8876 }, { "epoch": 0.54, "learning_rate": 2.6767067061024932e-05, "loss": 1.6143, "step": 8878 }, { "epoch": 0.54, "learning_rate": 2.6773103156878133e-05, "loss": 1.2922, "step": 8880 }, { "epoch": 0.54, "learning_rate": 2.6779139252731334e-05, "loss": 1.4091, "step": 8882 }, { "epoch": 0.54, "learning_rate": 2.6785175348584535e-05, "loss": 1.1006, "step": 8884 }, { "epoch": 0.54, "learning_rate": 2.6791211444437742e-05, "loss": 1.4061, "step": 8886 }, { "epoch": 0.54, "learning_rate": 2.6797247540290943e-05, "loss": 1.3515, "step": 8888 }, { "epoch": 0.54, "learning_rate": 2.6803283636144144e-05, "loss": 1.1773, "step": 8890 }, { "epoch": 0.54, "learning_rate": 2.6809319731997345e-05, "loss": 1.2813, "step": 8892 }, { "epoch": 0.54, "learning_rate": 2.6815355827850545e-05, "loss": 1.3358, "step": 8894 }, { "epoch": 0.54, "learning_rate": 2.682139192370375e-05, "loss": 1.3735, "step": 8896 }, { "epoch": 0.54, "learning_rate": 2.682742801955695e-05, "loss": 1.314, "step": 8898 }, { "epoch": 0.54, "learning_rate": 2.6833464115410155e-05, "loss": 1.1362, "step": 8900 }, { "epoch": 0.54, "learning_rate": 2.683950021126336e-05, "loss": 1.4431, "step": 8902 }, { "epoch": 0.54, "learning_rate": 2.684553630711656e-05, "loss": 1.3918, "step": 8904 }, { "epoch": 0.54, "learning_rate": 2.685157240296976e-05, "loss": 1.1813, "step": 8906 }, { "epoch": 0.54, "learning_rate": 2.685760849882296e-05, "loss": 1.1595, "step": 8908 }, { "epoch": 0.54, "learning_rate": 2.6863644594676162e-05, "loss": 0.9604, "step": 8910 }, { "epoch": 0.54, "learning_rate": 2.686968069052937e-05, "loss": 1.3084, "step": 8912 }, { "epoch": 0.54, "learning_rate": 2.687571678638257e-05, "loss": 1.1524, "step": 8914 }, { "epoch": 0.54, "learning_rate": 2.688175288223577e-05, "loss": 1.3945, "step": 8916 }, { "epoch": 0.54, "learning_rate": 2.6887788978088972e-05, "loss": 1.6816, "step": 8918 }, { "epoch": 0.54, "learning_rate": 2.6893825073942176e-05, "loss": 0.8698, "step": 8920 }, { "epoch": 0.54, "learning_rate": 2.6899861169795377e-05, "loss": 1.2862, "step": 8922 }, { "epoch": 0.54, "learning_rate": 2.6905897265648578e-05, "loss": 1.4316, "step": 8924 }, { "epoch": 0.54, "learning_rate": 2.6911933361501786e-05, "loss": 1.0502, "step": 8926 }, { "epoch": 0.54, "learning_rate": 2.6917969457354987e-05, "loss": 1.062, "step": 8928 }, { "epoch": 0.54, "learning_rate": 2.6924005553208187e-05, "loss": 1.6408, "step": 8930 }, { "epoch": 0.54, "learning_rate": 2.6930041649061388e-05, "loss": 1.3802, "step": 8932 }, { "epoch": 0.54, "learning_rate": 2.693607774491459e-05, "loss": 1.3654, "step": 8934 }, { "epoch": 0.54, "learning_rate": 2.694211384076779e-05, "loss": 1.1905, "step": 8936 }, { "epoch": 0.54, "learning_rate": 2.694814993662099e-05, "loss": 1.4178, "step": 8938 }, { "epoch": 0.54, "learning_rate": 2.6954186032474198e-05, "loss": 1.2105, "step": 8940 }, { "epoch": 0.54, "learning_rate": 2.69602221283274e-05, "loss": 1.2165, "step": 8942 }, { "epoch": 0.54, "learning_rate": 2.6966258224180603e-05, "loss": 1.3568, "step": 8944 }, { "epoch": 0.54, "learning_rate": 2.6972294320033804e-05, "loss": 1.1709, "step": 8946 }, { "epoch": 0.54, "learning_rate": 2.6978330415887005e-05, "loss": 0.9935, "step": 8948 }, { "epoch": 0.54, "learning_rate": 2.6984366511740206e-05, "loss": 1.2507, "step": 8950 }, { "epoch": 0.54, "learning_rate": 2.6990402607593413e-05, "loss": 1.6803, "step": 8952 }, { "epoch": 0.54, "learning_rate": 2.6996438703446614e-05, "loss": 1.09, "step": 8954 }, { "epoch": 0.54, "learning_rate": 2.7002474799299815e-05, "loss": 1.1152, "step": 8956 }, { "epoch": 0.54, "learning_rate": 2.7008510895153016e-05, "loss": 1.2185, "step": 8958 }, { "epoch": 0.54, "learning_rate": 2.7014546991006217e-05, "loss": 1.3791, "step": 8960 }, { "epoch": 0.54, "learning_rate": 2.7020583086859417e-05, "loss": 1.3653, "step": 8962 }, { "epoch": 0.54, "learning_rate": 2.702661918271262e-05, "loss": 1.1937, "step": 8964 }, { "epoch": 0.54, "learning_rate": 2.7032655278565826e-05, "loss": 1.4416, "step": 8966 }, { "epoch": 0.54, "learning_rate": 2.703869137441903e-05, "loss": 1.534, "step": 8968 }, { "epoch": 0.54, "learning_rate": 2.704472747027223e-05, "loss": 1.4078, "step": 8970 }, { "epoch": 0.54, "learning_rate": 2.705076356612543e-05, "loss": 0.9962, "step": 8972 }, { "epoch": 0.54, "learning_rate": 2.7056799661978632e-05, "loss": 1.4502, "step": 8974 }, { "epoch": 0.54, "learning_rate": 2.7062835757831833e-05, "loss": 1.2323, "step": 8976 }, { "epoch": 0.54, "learning_rate": 2.706887185368504e-05, "loss": 1.0193, "step": 8978 }, { "epoch": 0.54, "learning_rate": 2.707490794953824e-05, "loss": 1.1671, "step": 8980 }, { "epoch": 0.54, "learning_rate": 2.7080944045391442e-05, "loss": 0.9845, "step": 8982 }, { "epoch": 0.54, "learning_rate": 2.7086980141244643e-05, "loss": 1.3696, "step": 8984 }, { "epoch": 0.54, "learning_rate": 2.7093016237097844e-05, "loss": 1.2967, "step": 8986 }, { "epoch": 0.54, "learning_rate": 2.709905233295105e-05, "loss": 1.6404, "step": 8988 }, { "epoch": 0.54, "learning_rate": 2.710508842880425e-05, "loss": 1.7631, "step": 8990 }, { "epoch": 0.54, "learning_rate": 2.7111124524657457e-05, "loss": 1.0838, "step": 8992 }, { "epoch": 0.54, "learning_rate": 2.7117160620510658e-05, "loss": 1.1534, "step": 8994 }, { "epoch": 0.54, "learning_rate": 2.712319671636386e-05, "loss": 1.1564, "step": 8996 }, { "epoch": 0.54, "learning_rate": 2.712923281221706e-05, "loss": 1.3137, "step": 8998 }, { "epoch": 0.54, "learning_rate": 2.713526890807026e-05, "loss": 1.4322, "step": 9000 }, { "epoch": 0.54, "learning_rate": 2.714130500392346e-05, "loss": 1.1304, "step": 9002 }, { "epoch": 0.54, "learning_rate": 2.714734109977666e-05, "loss": 1.3109, "step": 9004 }, { "epoch": 0.54, "learning_rate": 2.715337719562987e-05, "loss": 1.1567, "step": 9006 }, { "epoch": 0.54, "learning_rate": 2.715941329148307e-05, "loss": 1.7426, "step": 9008 }, { "epoch": 0.54, "learning_rate": 2.716544938733627e-05, "loss": 1.1477, "step": 9010 }, { "epoch": 0.54, "learning_rate": 2.7171485483189475e-05, "loss": 1.3583, "step": 9012 }, { "epoch": 0.54, "learning_rate": 2.7177521579042676e-05, "loss": 1.2598, "step": 9014 }, { "epoch": 0.54, "learning_rate": 2.7183557674895877e-05, "loss": 1.3181, "step": 9016 }, { "epoch": 0.54, "learning_rate": 2.7189593770749084e-05, "loss": 1.0299, "step": 9018 }, { "epoch": 0.54, "learning_rate": 2.7195629866602285e-05, "loss": 1.5826, "step": 9020 }, { "epoch": 0.54, "learning_rate": 2.7201665962455486e-05, "loss": 1.2566, "step": 9022 }, { "epoch": 0.54, "learning_rate": 2.7207702058308687e-05, "loss": 1.3364, "step": 9024 }, { "epoch": 0.54, "learning_rate": 2.7213738154161888e-05, "loss": 1.1621, "step": 9026 }, { "epoch": 0.54, "learning_rate": 2.721977425001509e-05, "loss": 1.4046, "step": 9028 }, { "epoch": 0.55, "learning_rate": 2.7225810345868293e-05, "loss": 1.3539, "step": 9030 }, { "epoch": 0.55, "learning_rate": 2.7231846441721497e-05, "loss": 1.292, "step": 9032 }, { "epoch": 0.55, "learning_rate": 2.7237882537574698e-05, "loss": 1.194, "step": 9034 }, { "epoch": 0.55, "learning_rate": 2.7243918633427902e-05, "loss": 0.9721, "step": 9036 }, { "epoch": 0.55, "learning_rate": 2.7249954729281103e-05, "loss": 1.1883, "step": 9038 }, { "epoch": 0.55, "learning_rate": 2.7255990825134304e-05, "loss": 0.967, "step": 9040 }, { "epoch": 0.55, "learning_rate": 2.7262026920987504e-05, "loss": 1.2161, "step": 9042 }, { "epoch": 0.55, "learning_rate": 2.7268063016840712e-05, "loss": 1.6846, "step": 9044 }, { "epoch": 0.55, "learning_rate": 2.7274099112693913e-05, "loss": 1.4156, "step": 9046 }, { "epoch": 0.55, "learning_rate": 2.7280135208547114e-05, "loss": 1.2932, "step": 9048 }, { "epoch": 0.55, "learning_rate": 2.7286171304400314e-05, "loss": 1.0009, "step": 9050 }, { "epoch": 0.55, "learning_rate": 2.7292207400253515e-05, "loss": 1.0656, "step": 9052 }, { "epoch": 0.55, "learning_rate": 2.729824349610672e-05, "loss": 1.064, "step": 9054 }, { "epoch": 0.55, "learning_rate": 2.730427959195992e-05, "loss": 1.1954, "step": 9056 }, { "epoch": 0.55, "learning_rate": 2.7310315687813124e-05, "loss": 1.4838, "step": 9058 }, { "epoch": 0.55, "learning_rate": 2.731635178366633e-05, "loss": 1.3192, "step": 9060 }, { "epoch": 0.55, "learning_rate": 2.732238787951953e-05, "loss": 1.3395, "step": 9062 }, { "epoch": 0.55, "learning_rate": 2.732842397537273e-05, "loss": 1.2586, "step": 9064 }, { "epoch": 0.55, "learning_rate": 2.733446007122593e-05, "loss": 1.5488, "step": 9066 }, { "epoch": 0.55, "learning_rate": 2.7340496167079132e-05, "loss": 0.9454, "step": 9068 }, { "epoch": 0.55, "learning_rate": 2.7346532262932333e-05, "loss": 1.1612, "step": 9070 }, { "epoch": 0.55, "learning_rate": 2.735256835878554e-05, "loss": 1.1465, "step": 9072 }, { "epoch": 0.55, "learning_rate": 2.735860445463874e-05, "loss": 1.1471, "step": 9074 }, { "epoch": 0.55, "learning_rate": 2.7364640550491942e-05, "loss": 1.2504, "step": 9076 }, { "epoch": 0.55, "learning_rate": 2.7370676646345146e-05, "loss": 1.4141, "step": 9078 }, { "epoch": 0.55, "learning_rate": 2.7376712742198347e-05, "loss": 1.4735, "step": 9080 }, { "epoch": 0.55, "learning_rate": 2.7382748838051548e-05, "loss": 1.2549, "step": 9082 }, { "epoch": 0.55, "learning_rate": 2.7388784933904755e-05, "loss": 1.2653, "step": 9084 }, { "epoch": 0.55, "learning_rate": 2.7394821029757956e-05, "loss": 1.41, "step": 9086 }, { "epoch": 0.55, "learning_rate": 2.7400857125611157e-05, "loss": 1.1979, "step": 9088 }, { "epoch": 0.55, "learning_rate": 2.7406893221464358e-05, "loss": 1.3202, "step": 9090 }, { "epoch": 0.55, "learning_rate": 2.741292931731756e-05, "loss": 1.0786, "step": 9092 }, { "epoch": 0.55, "learning_rate": 2.741896541317076e-05, "loss": 1.1763, "step": 9094 }, { "epoch": 0.55, "learning_rate": 2.7425001509023964e-05, "loss": 0.8943, "step": 9096 }, { "epoch": 0.55, "learning_rate": 2.7431037604877168e-05, "loss": 1.1707, "step": 9098 }, { "epoch": 0.55, "learning_rate": 2.743707370073037e-05, "loss": 1.1859, "step": 9100 }, { "epoch": 0.55, "learning_rate": 2.7443109796583573e-05, "loss": 1.1298, "step": 9102 }, { "epoch": 0.55, "learning_rate": 2.7449145892436774e-05, "loss": 1.0358, "step": 9104 }, { "epoch": 0.55, "learning_rate": 2.7455181988289975e-05, "loss": 1.2974, "step": 9106 }, { "epoch": 0.55, "learning_rate": 2.7461218084143175e-05, "loss": 1.2473, "step": 9108 }, { "epoch": 0.55, "learning_rate": 2.7467254179996383e-05, "loss": 1.2184, "step": 9110 }, { "epoch": 0.55, "learning_rate": 2.7473290275849584e-05, "loss": 1.0708, "step": 9112 }, { "epoch": 0.55, "learning_rate": 2.7479326371702785e-05, "loss": 1.1177, "step": 9114 }, { "epoch": 0.55, "learning_rate": 2.7485362467555985e-05, "loss": 1.5415, "step": 9116 }, { "epoch": 0.55, "learning_rate": 2.7491398563409186e-05, "loss": 1.3679, "step": 9118 }, { "epoch": 0.55, "learning_rate": 2.749743465926239e-05, "loss": 1.3772, "step": 9120 }, { "epoch": 0.55, "learning_rate": 2.750347075511559e-05, "loss": 1.0277, "step": 9122 }, { "epoch": 0.55, "learning_rate": 2.7509506850968795e-05, "loss": 1.4269, "step": 9124 }, { "epoch": 0.55, "learning_rate": 2.7515542946822e-05, "loss": 1.3946, "step": 9126 }, { "epoch": 0.55, "learning_rate": 2.75215790426752e-05, "loss": 1.1698, "step": 9128 }, { "epoch": 0.55, "learning_rate": 2.75276151385284e-05, "loss": 1.3173, "step": 9130 }, { "epoch": 0.55, "learning_rate": 2.7533651234381602e-05, "loss": 1.6506, "step": 9132 }, { "epoch": 0.55, "learning_rate": 2.7539687330234803e-05, "loss": 1.352, "step": 9134 }, { "epoch": 0.55, "learning_rate": 2.7545723426088004e-05, "loss": 1.3333, "step": 9136 }, { "epoch": 0.55, "learning_rate": 2.755175952194121e-05, "loss": 1.0536, "step": 9138 }, { "epoch": 0.55, "learning_rate": 2.7557795617794412e-05, "loss": 1.6348, "step": 9140 }, { "epoch": 0.55, "learning_rate": 2.7563831713647613e-05, "loss": 1.4235, "step": 9142 }, { "epoch": 0.55, "learning_rate": 2.7569867809500817e-05, "loss": 1.764, "step": 9144 }, { "epoch": 0.55, "learning_rate": 2.7575903905354018e-05, "loss": 1.2449, "step": 9146 }, { "epoch": 0.55, "learning_rate": 2.758194000120722e-05, "loss": 1.101, "step": 9148 }, { "epoch": 0.55, "learning_rate": 2.7587976097060426e-05, "loss": 1.766, "step": 9150 }, { "epoch": 0.55, "learning_rate": 2.7594012192913627e-05, "loss": 1.1699, "step": 9152 }, { "epoch": 0.55, "learning_rate": 2.7600048288766828e-05, "loss": 1.3085, "step": 9154 }, { "epoch": 0.55, "learning_rate": 2.760608438462003e-05, "loss": 1.2836, "step": 9156 }, { "epoch": 0.55, "learning_rate": 2.761212048047323e-05, "loss": 1.4054, "step": 9158 }, { "epoch": 0.55, "learning_rate": 2.761815657632643e-05, "loss": 1.2802, "step": 9160 }, { "epoch": 0.55, "learning_rate": 2.762419267217963e-05, "loss": 1.3318, "step": 9162 }, { "epoch": 0.55, "learning_rate": 2.763022876803284e-05, "loss": 1.2795, "step": 9164 }, { "epoch": 0.55, "learning_rate": 2.763626486388604e-05, "loss": 1.6311, "step": 9166 }, { "epoch": 0.55, "learning_rate": 2.7642300959739244e-05, "loss": 1.2323, "step": 9168 }, { "epoch": 0.55, "learning_rate": 2.7648337055592445e-05, "loss": 1.2635, "step": 9170 }, { "epoch": 0.55, "learning_rate": 2.7654373151445646e-05, "loss": 1.2988, "step": 9172 }, { "epoch": 0.55, "learning_rate": 2.7660409247298846e-05, "loss": 1.612, "step": 9174 }, { "epoch": 0.55, "learning_rate": 2.7666445343152054e-05, "loss": 1.6169, "step": 9176 }, { "epoch": 0.55, "learning_rate": 2.7672481439005255e-05, "loss": 1.0545, "step": 9178 }, { "epoch": 0.55, "learning_rate": 2.7678517534858456e-05, "loss": 1.4323, "step": 9180 }, { "epoch": 0.55, "learning_rate": 2.7684553630711656e-05, "loss": 1.2417, "step": 9182 }, { "epoch": 0.55, "learning_rate": 2.7690589726564857e-05, "loss": 0.8218, "step": 9184 }, { "epoch": 0.55, "learning_rate": 2.7696625822418058e-05, "loss": 1.3139, "step": 9186 }, { "epoch": 0.55, "learning_rate": 2.7702661918271262e-05, "loss": 1.3988, "step": 9188 }, { "epoch": 0.55, "learning_rate": 2.7708698014124467e-05, "loss": 1.4053, "step": 9190 }, { "epoch": 0.55, "learning_rate": 2.771473410997767e-05, "loss": 1.2113, "step": 9192 }, { "epoch": 0.55, "learning_rate": 2.772077020583087e-05, "loss": 1.6619, "step": 9194 }, { "epoch": 0.56, "learning_rate": 2.7726806301684072e-05, "loss": 1.1265, "step": 9196 }, { "epoch": 0.56, "learning_rate": 2.7732842397537273e-05, "loss": 1.6854, "step": 9198 }, { "epoch": 0.56, "learning_rate": 2.7738878493390474e-05, "loss": 1.302, "step": 9200 }, { "epoch": 0.56, "learning_rate": 2.774491458924368e-05, "loss": 1.0748, "step": 9202 }, { "epoch": 0.56, "learning_rate": 2.7750950685096882e-05, "loss": 1.1276, "step": 9204 }, { "epoch": 0.56, "learning_rate": 2.7756986780950083e-05, "loss": 1.1923, "step": 9206 }, { "epoch": 0.56, "learning_rate": 2.7763022876803284e-05, "loss": 1.212, "step": 9208 }, { "epoch": 0.56, "learning_rate": 2.7769058972656485e-05, "loss": 1.268, "step": 9210 }, { "epoch": 0.56, "learning_rate": 2.777509506850969e-05, "loss": 1.4846, "step": 9212 }, { "epoch": 0.56, "learning_rate": 2.778113116436289e-05, "loss": 1.5008, "step": 9214 }, { "epoch": 0.56, "learning_rate": 2.7787167260216098e-05, "loss": 1.1216, "step": 9216 }, { "epoch": 0.56, "learning_rate": 2.77932033560693e-05, "loss": 1.3781, "step": 9218 }, { "epoch": 0.56, "learning_rate": 2.77992394519225e-05, "loss": 1.2631, "step": 9220 }, { "epoch": 0.56, "learning_rate": 2.78052755477757e-05, "loss": 1.0301, "step": 9222 }, { "epoch": 0.56, "learning_rate": 2.78113116436289e-05, "loss": 1.2742, "step": 9224 }, { "epoch": 0.56, "learning_rate": 2.78173477394821e-05, "loss": 1.1533, "step": 9226 }, { "epoch": 0.56, "learning_rate": 2.7823383835335302e-05, "loss": 1.1907, "step": 9228 }, { "epoch": 0.56, "learning_rate": 2.782941993118851e-05, "loss": 0.9372, "step": 9230 }, { "epoch": 0.56, "learning_rate": 2.783545602704171e-05, "loss": 1.2598, "step": 9232 }, { "epoch": 0.56, "learning_rate": 2.784149212289491e-05, "loss": 1.3657, "step": 9234 }, { "epoch": 0.56, "learning_rate": 2.7847528218748116e-05, "loss": 1.0292, "step": 9236 }, { "epoch": 0.56, "learning_rate": 2.7853564314601317e-05, "loss": 1.2997, "step": 9238 }, { "epoch": 0.56, "learning_rate": 2.7859600410454518e-05, "loss": 1.4973, "step": 9240 }, { "epoch": 0.56, "learning_rate": 2.7865636506307725e-05, "loss": 1.0054, "step": 9242 }, { "epoch": 0.56, "learning_rate": 2.7871672602160926e-05, "loss": 1.1209, "step": 9244 }, { "epoch": 0.56, "learning_rate": 2.7877708698014127e-05, "loss": 1.202, "step": 9246 }, { "epoch": 0.56, "learning_rate": 2.7883744793867328e-05, "loss": 0.915, "step": 9248 }, { "epoch": 0.56, "learning_rate": 2.788978088972053e-05, "loss": 1.2724, "step": 9250 }, { "epoch": 0.56, "learning_rate": 2.789581698557373e-05, "loss": 1.2804, "step": 9252 }, { "epoch": 0.56, "learning_rate": 2.7901853081426933e-05, "loss": 1.351, "step": 9254 }, { "epoch": 0.56, "learning_rate": 2.7907889177280138e-05, "loss": 1.3935, "step": 9256 }, { "epoch": 0.56, "learning_rate": 2.791392527313334e-05, "loss": 1.4294, "step": 9258 }, { "epoch": 0.56, "learning_rate": 2.7919961368986543e-05, "loss": 1.4496, "step": 9260 }, { "epoch": 0.56, "learning_rate": 2.7925997464839743e-05, "loss": 1.5576, "step": 9262 }, { "epoch": 0.56, "learning_rate": 2.7932033560692944e-05, "loss": 1.4298, "step": 9264 }, { "epoch": 0.56, "learning_rate": 2.7938069656546145e-05, "loss": 1.1039, "step": 9266 }, { "epoch": 0.56, "learning_rate": 2.7944105752399353e-05, "loss": 1.609, "step": 9268 }, { "epoch": 0.56, "learning_rate": 2.7950141848252554e-05, "loss": 1.8051, "step": 9270 }, { "epoch": 0.56, "learning_rate": 2.7956177944105754e-05, "loss": 1.1773, "step": 9272 }, { "epoch": 0.56, "learning_rate": 2.7962214039958955e-05, "loss": 1.4094, "step": 9274 }, { "epoch": 0.56, "learning_rate": 2.7968250135812156e-05, "loss": 1.0627, "step": 9276 }, { "epoch": 0.56, "learning_rate": 2.797428623166536e-05, "loss": 1.3952, "step": 9278 }, { "epoch": 0.56, "learning_rate": 2.798032232751856e-05, "loss": 1.6501, "step": 9280 }, { "epoch": 0.56, "learning_rate": 2.7986358423371765e-05, "loss": 0.8241, "step": 9282 }, { "epoch": 0.56, "learning_rate": 2.799239451922497e-05, "loss": 1.1899, "step": 9284 }, { "epoch": 0.56, "learning_rate": 2.799843061507817e-05, "loss": 1.3962, "step": 9286 }, { "epoch": 0.56, "learning_rate": 2.800446671093137e-05, "loss": 1.5005, "step": 9288 }, { "epoch": 0.56, "learning_rate": 2.8010502806784572e-05, "loss": 0.8904, "step": 9290 }, { "epoch": 0.56, "learning_rate": 2.8016538902637773e-05, "loss": 1.5182, "step": 9292 }, { "epoch": 0.56, "learning_rate": 2.8022574998490973e-05, "loss": 1.7017, "step": 9294 }, { "epoch": 0.56, "learning_rate": 2.802861109434418e-05, "loss": 1.2818, "step": 9296 }, { "epoch": 0.56, "learning_rate": 2.8034647190197382e-05, "loss": 1.4193, "step": 9298 }, { "epoch": 0.56, "learning_rate": 2.8040683286050583e-05, "loss": 1.3655, "step": 9300 }, { "epoch": 0.56, "learning_rate": 2.8046719381903787e-05, "loss": 1.2354, "step": 9302 }, { "epoch": 0.56, "learning_rate": 2.8052755477756988e-05, "loss": 1.2865, "step": 9304 }, { "epoch": 0.56, "learning_rate": 2.805879157361019e-05, "loss": 1.3301, "step": 9306 }, { "epoch": 0.56, "learning_rate": 2.8064827669463396e-05, "loss": 1.2612, "step": 9308 }, { "epoch": 0.56, "learning_rate": 2.8070863765316597e-05, "loss": 1.3993, "step": 9310 }, { "epoch": 0.56, "learning_rate": 2.8076899861169798e-05, "loss": 1.1496, "step": 9312 }, { "epoch": 0.56, "learning_rate": 2.8082935957023e-05, "loss": 1.3184, "step": 9314 }, { "epoch": 0.56, "learning_rate": 2.80889720528762e-05, "loss": 1.0176, "step": 9316 }, { "epoch": 0.56, "learning_rate": 2.80950081487294e-05, "loss": 1.0801, "step": 9318 }, { "epoch": 0.56, "learning_rate": 2.81010442445826e-05, "loss": 1.2645, "step": 9320 }, { "epoch": 0.56, "learning_rate": 2.810708034043581e-05, "loss": 1.4072, "step": 9322 }, { "epoch": 0.56, "learning_rate": 2.811311643628901e-05, "loss": 1.2413, "step": 9324 }, { "epoch": 0.56, "learning_rate": 2.8119152532142214e-05, "loss": 1.3885, "step": 9326 }, { "epoch": 0.56, "learning_rate": 2.8125188627995415e-05, "loss": 1.4169, "step": 9328 }, { "epoch": 0.56, "learning_rate": 2.8131224723848615e-05, "loss": 1.0838, "step": 9330 }, { "epoch": 0.56, "learning_rate": 2.8137260819701816e-05, "loss": 1.3579, "step": 9332 }, { "epoch": 0.56, "learning_rate": 2.8143296915555024e-05, "loss": 1.3436, "step": 9334 }, { "epoch": 0.56, "learning_rate": 2.8149333011408225e-05, "loss": 1.2633, "step": 9336 }, { "epoch": 0.56, "learning_rate": 2.8155369107261425e-05, "loss": 1.3898, "step": 9338 }, { "epoch": 0.56, "learning_rate": 2.8161405203114626e-05, "loss": 1.0489, "step": 9340 }, { "epoch": 0.56, "learning_rate": 2.8167441298967827e-05, "loss": 1.3052, "step": 9342 }, { "epoch": 0.56, "learning_rate": 2.817347739482103e-05, "loss": 1.1595, "step": 9344 }, { "epoch": 0.56, "learning_rate": 2.8179513490674232e-05, "loss": 1.3074, "step": 9346 }, { "epoch": 0.56, "learning_rate": 2.8185549586527436e-05, "loss": 1.1583, "step": 9348 }, { "epoch": 0.56, "learning_rate": 2.819158568238064e-05, "loss": 1.2457, "step": 9350 }, { "epoch": 0.56, "learning_rate": 2.819762177823384e-05, "loss": 1.1133, "step": 9352 }, { "epoch": 0.56, "learning_rate": 2.8203657874087042e-05, "loss": 1.0796, "step": 9354 }, { "epoch": 0.56, "learning_rate": 2.8209693969940243e-05, "loss": 1.1534, "step": 9356 }, { "epoch": 0.56, "learning_rate": 2.8215730065793444e-05, "loss": 1.025, "step": 9358 }, { "epoch": 0.56, "learning_rate": 2.8221766161646645e-05, "loss": 0.9809, "step": 9360 }, { "epoch": 0.57, "learning_rate": 2.8227802257499852e-05, "loss": 1.2365, "step": 9362 }, { "epoch": 0.57, "learning_rate": 2.8233838353353053e-05, "loss": 1.3707, "step": 9364 }, { "epoch": 0.57, "learning_rate": 2.8239874449206254e-05, "loss": 1.1118, "step": 9366 }, { "epoch": 0.57, "learning_rate": 2.8245910545059458e-05, "loss": 1.5282, "step": 9368 }, { "epoch": 0.57, "learning_rate": 2.825194664091266e-05, "loss": 1.0761, "step": 9370 }, { "epoch": 0.57, "learning_rate": 2.825798273676586e-05, "loss": 1.2586, "step": 9372 }, { "epoch": 0.57, "learning_rate": 2.8264018832619067e-05, "loss": 1.1198, "step": 9374 }, { "epoch": 0.57, "learning_rate": 2.8270054928472268e-05, "loss": 1.1461, "step": 9376 }, { "epoch": 0.57, "learning_rate": 2.827609102432547e-05, "loss": 1.3486, "step": 9378 }, { "epoch": 0.57, "learning_rate": 2.828212712017867e-05, "loss": 1.1453, "step": 9380 }, { "epoch": 0.57, "learning_rate": 2.828816321603187e-05, "loss": 2.1029, "step": 9382 }, { "epoch": 0.57, "learning_rate": 2.829419931188507e-05, "loss": 1.2811, "step": 9384 }, { "epoch": 0.57, "learning_rate": 2.8300235407738272e-05, "loss": 1.1326, "step": 9386 }, { "epoch": 0.57, "learning_rate": 2.830627150359148e-05, "loss": 1.5682, "step": 9388 }, { "epoch": 0.57, "learning_rate": 2.831230759944468e-05, "loss": 1.2824, "step": 9390 }, { "epoch": 0.57, "learning_rate": 2.8318343695297885e-05, "loss": 1.6289, "step": 9392 }, { "epoch": 0.57, "learning_rate": 2.8324379791151086e-05, "loss": 1.2882, "step": 9394 }, { "epoch": 0.57, "learning_rate": 2.8330415887004286e-05, "loss": 1.431, "step": 9396 }, { "epoch": 0.57, "learning_rate": 2.8336451982857487e-05, "loss": 1.4037, "step": 9398 }, { "epoch": 0.57, "learning_rate": 2.8342488078710695e-05, "loss": 1.1904, "step": 9400 }, { "epoch": 0.57, "learning_rate": 2.8348524174563896e-05, "loss": 1.2052, "step": 9402 }, { "epoch": 0.57, "learning_rate": 2.8354560270417096e-05, "loss": 0.9601, "step": 9404 }, { "epoch": 0.57, "learning_rate": 2.8360596366270297e-05, "loss": 1.0833, "step": 9406 }, { "epoch": 0.57, "learning_rate": 2.8366632462123498e-05, "loss": 1.1542, "step": 9408 }, { "epoch": 0.57, "learning_rate": 2.83726685579767e-05, "loss": 1.1028, "step": 9410 }, { "epoch": 0.57, "learning_rate": 2.8378704653829903e-05, "loss": 1.079, "step": 9412 }, { "epoch": 0.57, "learning_rate": 2.8384740749683107e-05, "loss": 1.1756, "step": 9414 }, { "epoch": 0.57, "learning_rate": 2.839077684553631e-05, "loss": 1.2996, "step": 9416 }, { "epoch": 0.57, "learning_rate": 2.8396812941389512e-05, "loss": 1.4445, "step": 9418 }, { "epoch": 0.57, "learning_rate": 2.8402849037242713e-05, "loss": 1.3996, "step": 9420 }, { "epoch": 0.57, "learning_rate": 2.8408885133095914e-05, "loss": 1.3194, "step": 9422 }, { "epoch": 0.57, "learning_rate": 2.8414921228949115e-05, "loss": 1.3099, "step": 9424 }, { "epoch": 0.57, "learning_rate": 2.8420957324802322e-05, "loss": 1.4934, "step": 9426 }, { "epoch": 0.57, "learning_rate": 2.8426993420655523e-05, "loss": 1.4115, "step": 9428 }, { "epoch": 0.57, "learning_rate": 2.8433029516508724e-05, "loss": 1.4935, "step": 9430 }, { "epoch": 0.57, "learning_rate": 2.8439065612361925e-05, "loss": 1.2262, "step": 9432 }, { "epoch": 0.57, "learning_rate": 2.8445101708215126e-05, "loss": 1.3989, "step": 9434 }, { "epoch": 0.57, "learning_rate": 2.845113780406833e-05, "loss": 1.2335, "step": 9436 }, { "epoch": 0.57, "learning_rate": 2.845717389992153e-05, "loss": 1.0248, "step": 9438 }, { "epoch": 0.57, "learning_rate": 2.8463209995774738e-05, "loss": 1.0291, "step": 9440 }, { "epoch": 0.57, "learning_rate": 2.846924609162794e-05, "loss": 1.3979, "step": 9442 }, { "epoch": 0.57, "learning_rate": 2.847528218748114e-05, "loss": 1.0654, "step": 9444 }, { "epoch": 0.57, "learning_rate": 2.848131828333434e-05, "loss": 1.33, "step": 9446 }, { "epoch": 0.57, "learning_rate": 2.848735437918754e-05, "loss": 1.4714, "step": 9448 }, { "epoch": 0.57, "learning_rate": 2.8493390475040742e-05, "loss": 1.5541, "step": 9450 }, { "epoch": 0.57, "learning_rate": 2.8499426570893943e-05, "loss": 1.3678, "step": 9452 }, { "epoch": 0.57, "learning_rate": 2.850546266674715e-05, "loss": 1.8141, "step": 9454 }, { "epoch": 0.57, "learning_rate": 2.851149876260035e-05, "loss": 1.2932, "step": 9456 }, { "epoch": 0.57, "learning_rate": 2.8517534858453552e-05, "loss": 1.2333, "step": 9458 }, { "epoch": 0.57, "learning_rate": 2.8523570954306757e-05, "loss": 1.2171, "step": 9460 }, { "epoch": 0.57, "learning_rate": 2.8529607050159957e-05, "loss": 1.1286, "step": 9462 }, { "epoch": 0.57, "learning_rate": 2.8535643146013158e-05, "loss": 1.2678, "step": 9464 }, { "epoch": 0.57, "learning_rate": 2.8541679241866366e-05, "loss": 1.3444, "step": 9466 }, { "epoch": 0.57, "learning_rate": 2.8547715337719567e-05, "loss": 1.1373, "step": 9468 }, { "epoch": 0.57, "learning_rate": 2.8553751433572767e-05, "loss": 1.5703, "step": 9470 }, { "epoch": 0.57, "learning_rate": 2.855978752942597e-05, "loss": 1.2049, "step": 9472 }, { "epoch": 0.57, "learning_rate": 2.856582362527917e-05, "loss": 1.0727, "step": 9474 }, { "epoch": 0.57, "learning_rate": 2.857185972113237e-05, "loss": 1.4253, "step": 9476 }, { "epoch": 0.57, "learning_rate": 2.8577895816985574e-05, "loss": 1.1239, "step": 9478 }, { "epoch": 0.57, "learning_rate": 2.858393191283878e-05, "loss": 1.4675, "step": 9480 }, { "epoch": 0.57, "learning_rate": 2.858996800869198e-05, "loss": 1.0233, "step": 9482 }, { "epoch": 0.57, "learning_rate": 2.8596004104545183e-05, "loss": 1.3892, "step": 9484 }, { "epoch": 0.57, "learning_rate": 2.8602040200398384e-05, "loss": 0.9986, "step": 9486 }, { "epoch": 0.57, "learning_rate": 2.8608076296251585e-05, "loss": 1.1415, "step": 9488 }, { "epoch": 0.57, "learning_rate": 2.8614112392104786e-05, "loss": 1.1155, "step": 9490 }, { "epoch": 0.57, "learning_rate": 2.8620148487957993e-05, "loss": 1.2146, "step": 9492 }, { "epoch": 0.57, "learning_rate": 2.8626184583811194e-05, "loss": 1.2644, "step": 9494 }, { "epoch": 0.57, "learning_rate": 2.8632220679664395e-05, "loss": 1.1087, "step": 9496 }, { "epoch": 0.57, "learning_rate": 2.8638256775517596e-05, "loss": 1.1342, "step": 9498 }, { "epoch": 0.57, "learning_rate": 2.8644292871370797e-05, "loss": 1.2089, "step": 9500 }, { "epoch": 0.57, "learning_rate": 2.8650328967224e-05, "loss": 1.2816, "step": 9502 }, { "epoch": 0.57, "learning_rate": 2.8656365063077202e-05, "loss": 1.2766, "step": 9504 }, { "epoch": 0.57, "learning_rate": 2.8662401158930406e-05, "loss": 1.1224, "step": 9506 }, { "epoch": 0.57, "learning_rate": 2.866843725478361e-05, "loss": 1.1909, "step": 9508 }, { "epoch": 0.57, "learning_rate": 2.867447335063681e-05, "loss": 1.0297, "step": 9510 }, { "epoch": 0.57, "learning_rate": 2.8680509446490012e-05, "loss": 1.1443, "step": 9512 }, { "epoch": 0.57, "learning_rate": 2.8686545542343213e-05, "loss": 1.4776, "step": 9514 }, { "epoch": 0.57, "learning_rate": 2.8692581638196413e-05, "loss": 1.2447, "step": 9516 }, { "epoch": 0.57, "learning_rate": 2.8698617734049614e-05, "loss": 1.2541, "step": 9518 }, { "epoch": 0.57, "learning_rate": 2.8704653829902822e-05, "loss": 1.2422, "step": 9520 }, { "epoch": 0.57, "learning_rate": 2.8710689925756023e-05, "loss": 0.9862, "step": 9522 }, { "epoch": 0.57, "learning_rate": 2.8716726021609223e-05, "loss": 1.0469, "step": 9524 }, { "epoch": 0.57, "learning_rate": 2.8722762117462428e-05, "loss": 1.3329, "step": 9526 }, { "epoch": 0.58, "learning_rate": 2.872879821331563e-05, "loss": 1.142, "step": 9528 }, { "epoch": 0.58, "learning_rate": 2.873483430916883e-05, "loss": 1.1355, "step": 9530 }, { "epoch": 0.58, "learning_rate": 2.8740870405022037e-05, "loss": 1.1214, "step": 9532 }, { "epoch": 0.58, "learning_rate": 2.8746906500875238e-05, "loss": 1.2893, "step": 9534 }, { "epoch": 0.58, "learning_rate": 2.875294259672844e-05, "loss": 1.2813, "step": 9536 }, { "epoch": 0.58, "learning_rate": 2.875897869258164e-05, "loss": 1.1349, "step": 9538 }, { "epoch": 0.58, "learning_rate": 2.876501478843484e-05, "loss": 1.0862, "step": 9540 }, { "epoch": 0.58, "learning_rate": 2.877105088428804e-05, "loss": 1.3139, "step": 9542 }, { "epoch": 0.58, "learning_rate": 2.8777086980141242e-05, "loss": 1.053, "step": 9544 }, { "epoch": 0.58, "learning_rate": 2.878312307599445e-05, "loss": 1.2805, "step": 9546 }, { "epoch": 0.58, "learning_rate": 2.878915917184765e-05, "loss": 1.0312, "step": 9548 }, { "epoch": 0.58, "learning_rate": 2.8795195267700854e-05, "loss": 1.3212, "step": 9550 }, { "epoch": 0.58, "learning_rate": 2.8801231363554055e-05, "loss": 1.2597, "step": 9552 }, { "epoch": 0.58, "learning_rate": 2.8807267459407256e-05, "loss": 1.2653, "step": 9554 }, { "epoch": 0.58, "learning_rate": 2.8813303555260457e-05, "loss": 1.4116, "step": 9556 }, { "epoch": 0.58, "learning_rate": 2.8819339651113665e-05, "loss": 1.6536, "step": 9558 }, { "epoch": 0.58, "learning_rate": 2.8825375746966865e-05, "loss": 1.0573, "step": 9560 }, { "epoch": 0.58, "learning_rate": 2.8831411842820066e-05, "loss": 1.463, "step": 9562 }, { "epoch": 0.58, "learning_rate": 2.8837447938673267e-05, "loss": 1.3184, "step": 9564 }, { "epoch": 0.58, "learning_rate": 2.8843484034526468e-05, "loss": 1.1131, "step": 9566 }, { "epoch": 0.58, "learning_rate": 2.884952013037967e-05, "loss": 1.4822, "step": 9568 }, { "epoch": 0.58, "learning_rate": 2.8855556226232873e-05, "loss": 1.4288, "step": 9570 }, { "epoch": 0.58, "learning_rate": 2.8861592322086077e-05, "loss": 1.1093, "step": 9572 }, { "epoch": 0.58, "learning_rate": 2.886762841793928e-05, "loss": 1.2067, "step": 9574 }, { "epoch": 0.58, "learning_rate": 2.8873664513792482e-05, "loss": 1.2132, "step": 9576 }, { "epoch": 0.58, "learning_rate": 2.8879700609645683e-05, "loss": 1.1008, "step": 9578 }, { "epoch": 0.58, "learning_rate": 2.8885736705498884e-05, "loss": 1.3051, "step": 9580 }, { "epoch": 0.58, "learning_rate": 2.8891772801352084e-05, "loss": 1.1987, "step": 9582 }, { "epoch": 0.58, "learning_rate": 2.8897808897205285e-05, "loss": 1.5212, "step": 9584 }, { "epoch": 0.58, "learning_rate": 2.8903844993058493e-05, "loss": 1.1194, "step": 9586 }, { "epoch": 0.58, "learning_rate": 2.8909881088911694e-05, "loss": 0.9954, "step": 9588 }, { "epoch": 0.58, "learning_rate": 2.8915917184764895e-05, "loss": 1.5107, "step": 9590 }, { "epoch": 0.58, "learning_rate": 2.89219532806181e-05, "loss": 1.3184, "step": 9592 }, { "epoch": 0.58, "learning_rate": 2.89279893764713e-05, "loss": 1.2782, "step": 9594 }, { "epoch": 0.58, "learning_rate": 2.89340254723245e-05, "loss": 1.2791, "step": 9596 }, { "epoch": 0.58, "learning_rate": 2.8940061568177708e-05, "loss": 1.366, "step": 9598 }, { "epoch": 0.58, "learning_rate": 2.894609766403091e-05, "loss": 1.2807, "step": 9600 }, { "epoch": 0.58, "learning_rate": 2.895213375988411e-05, "loss": 1.2262, "step": 9602 }, { "epoch": 0.58, "learning_rate": 2.895816985573731e-05, "loss": 1.2588, "step": 9604 }, { "epoch": 0.58, "learning_rate": 2.896420595159051e-05, "loss": 1.5502, "step": 9606 }, { "epoch": 0.58, "learning_rate": 2.8970242047443712e-05, "loss": 1.2723, "step": 9608 }, { "epoch": 0.58, "learning_rate": 2.8976278143296913e-05, "loss": 1.4432, "step": 9610 }, { "epoch": 0.58, "learning_rate": 2.898231423915012e-05, "loss": 1.0182, "step": 9612 }, { "epoch": 0.58, "learning_rate": 2.898835033500332e-05, "loss": 1.0076, "step": 9614 }, { "epoch": 0.58, "learning_rate": 2.8994386430856526e-05, "loss": 1.3098, "step": 9616 }, { "epoch": 0.58, "learning_rate": 2.9000422526709726e-05, "loss": 1.8549, "step": 9618 }, { "epoch": 0.58, "learning_rate": 2.9006458622562927e-05, "loss": 1.1401, "step": 9620 }, { "epoch": 0.58, "learning_rate": 2.9012494718416128e-05, "loss": 1.3765, "step": 9622 }, { "epoch": 0.58, "learning_rate": 2.9018530814269336e-05, "loss": 1.3548, "step": 9624 }, { "epoch": 0.58, "learning_rate": 2.9024566910122536e-05, "loss": 1.3177, "step": 9626 }, { "epoch": 0.58, "learning_rate": 2.9030603005975737e-05, "loss": 1.0854, "step": 9628 }, { "epoch": 0.58, "learning_rate": 2.9036639101828938e-05, "loss": 0.9858, "step": 9630 }, { "epoch": 0.58, "learning_rate": 2.904267519768214e-05, "loss": 1.6773, "step": 9632 }, { "epoch": 0.58, "learning_rate": 2.904871129353534e-05, "loss": 1.108, "step": 9634 }, { "epoch": 0.58, "learning_rate": 2.9054747389388544e-05, "loss": 1.338, "step": 9636 }, { "epoch": 0.58, "learning_rate": 2.9060783485241748e-05, "loss": 1.6113, "step": 9638 }, { "epoch": 0.58, "learning_rate": 2.9066819581094952e-05, "loss": 1.7469, "step": 9640 }, { "epoch": 0.58, "learning_rate": 2.9072855676948153e-05, "loss": 1.3383, "step": 9642 }, { "epoch": 0.58, "learning_rate": 2.9078891772801354e-05, "loss": 0.9887, "step": 9644 }, { "epoch": 0.58, "learning_rate": 2.9084927868654555e-05, "loss": 1.2235, "step": 9646 }, { "epoch": 0.58, "learning_rate": 2.9090963964507756e-05, "loss": 1.1572, "step": 9648 }, { "epoch": 0.58, "learning_rate": 2.9097000060360956e-05, "loss": 1.2395, "step": 9650 }, { "epoch": 0.58, "learning_rate": 2.9103036156214164e-05, "loss": 1.4112, "step": 9652 }, { "epoch": 0.58, "learning_rate": 2.9109072252067365e-05, "loss": 1.3332, "step": 9654 }, { "epoch": 0.58, "learning_rate": 2.9115108347920566e-05, "loss": 1.2721, "step": 9656 }, { "epoch": 0.58, "learning_rate": 2.9121144443773766e-05, "loss": 1.3138, "step": 9658 }, { "epoch": 0.58, "learning_rate": 2.912718053962697e-05, "loss": 1.4173, "step": 9660 }, { "epoch": 0.58, "learning_rate": 2.913321663548017e-05, "loss": 1.1882, "step": 9662 }, { "epoch": 0.58, "learning_rate": 2.913925273133338e-05, "loss": 1.4106, "step": 9664 }, { "epoch": 0.58, "learning_rate": 2.914528882718658e-05, "loss": 1.2, "step": 9666 }, { "epoch": 0.58, "learning_rate": 2.915132492303978e-05, "loss": 1.2893, "step": 9668 }, { "epoch": 0.58, "learning_rate": 2.915736101889298e-05, "loss": 1.3756, "step": 9670 }, { "epoch": 0.58, "learning_rate": 2.9163397114746182e-05, "loss": 0.8973, "step": 9672 }, { "epoch": 0.58, "learning_rate": 2.9169433210599383e-05, "loss": 1.1687, "step": 9674 }, { "epoch": 0.58, "learning_rate": 2.9175469306452584e-05, "loss": 1.1051, "step": 9676 }, { "epoch": 0.58, "learning_rate": 2.918150540230579e-05, "loss": 1.264, "step": 9678 }, { "epoch": 0.58, "learning_rate": 2.9187541498158992e-05, "loss": 1.0677, "step": 9680 }, { "epoch": 0.58, "learning_rate": 2.9193577594012193e-05, "loss": 1.3984, "step": 9682 }, { "epoch": 0.58, "learning_rate": 2.9199613689865397e-05, "loss": 1.34, "step": 9684 }, { "epoch": 0.58, "learning_rate": 2.9205649785718598e-05, "loss": 1.28, "step": 9686 }, { "epoch": 0.58, "learning_rate": 2.92116858815718e-05, "loss": 1.2507, "step": 9688 }, { "epoch": 0.58, "learning_rate": 2.9217721977425007e-05, "loss": 1.1156, "step": 9690 }, { "epoch": 0.58, "learning_rate": 2.9223758073278207e-05, "loss": 0.9084, "step": 9692 }, { "epoch": 0.59, "learning_rate": 2.9229794169131408e-05, "loss": 1.4289, "step": 9694 }, { "epoch": 0.59, "learning_rate": 2.923583026498461e-05, "loss": 1.4371, "step": 9696 }, { "epoch": 0.59, "learning_rate": 2.924186636083781e-05, "loss": 1.1198, "step": 9698 }, { "epoch": 0.59, "learning_rate": 2.924790245669101e-05, "loss": 1.1026, "step": 9700 }, { "epoch": 0.59, "learning_rate": 2.9253938552544215e-05, "loss": 1.2877, "step": 9702 }, { "epoch": 0.59, "learning_rate": 2.925997464839742e-05, "loss": 1.4321, "step": 9704 }, { "epoch": 0.59, "learning_rate": 2.926601074425062e-05, "loss": 1.3263, "step": 9706 }, { "epoch": 0.59, "learning_rate": 2.9272046840103824e-05, "loss": 1.3297, "step": 9708 }, { "epoch": 0.59, "learning_rate": 2.9278082935957025e-05, "loss": 1.384, "step": 9710 }, { "epoch": 0.59, "learning_rate": 2.9284119031810226e-05, "loss": 1.2942, "step": 9712 }, { "epoch": 0.59, "learning_rate": 2.9290155127663427e-05, "loss": 1.1963, "step": 9714 }, { "epoch": 0.59, "learning_rate": 2.9296191223516634e-05, "loss": 1.2271, "step": 9716 }, { "epoch": 0.59, "learning_rate": 2.9302227319369835e-05, "loss": 1.1695, "step": 9718 }, { "epoch": 0.59, "learning_rate": 2.9308263415223036e-05, "loss": 1.2454, "step": 9720 }, { "epoch": 0.59, "learning_rate": 2.9314299511076237e-05, "loss": 1.1417, "step": 9722 }, { "epoch": 0.59, "learning_rate": 2.9320335606929437e-05, "loss": 1.4556, "step": 9724 }, { "epoch": 0.59, "learning_rate": 2.932637170278264e-05, "loss": 0.942, "step": 9726 }, { "epoch": 0.59, "learning_rate": 2.9332407798635843e-05, "loss": 1.0748, "step": 9728 }, { "epoch": 0.59, "learning_rate": 2.9338443894489047e-05, "loss": 1.1465, "step": 9730 }, { "epoch": 0.59, "learning_rate": 2.934447999034225e-05, "loss": 0.9966, "step": 9732 }, { "epoch": 0.59, "learning_rate": 2.9350516086195452e-05, "loss": 1.7021, "step": 9734 }, { "epoch": 0.59, "learning_rate": 2.9356552182048653e-05, "loss": 1.1724, "step": 9736 }, { "epoch": 0.59, "learning_rate": 2.9362588277901853e-05, "loss": 1.2461, "step": 9738 }, { "epoch": 0.59, "learning_rate": 2.9368624373755054e-05, "loss": 0.9466, "step": 9740 }, { "epoch": 0.59, "learning_rate": 2.9374660469608255e-05, "loss": 1.2401, "step": 9742 }, { "epoch": 0.59, "learning_rate": 2.9380696565461463e-05, "loss": 1.0191, "step": 9744 }, { "epoch": 0.59, "learning_rate": 2.9386732661314663e-05, "loss": 1.1918, "step": 9746 }, { "epoch": 0.59, "learning_rate": 2.9392768757167864e-05, "loss": 1.4277, "step": 9748 }, { "epoch": 0.59, "learning_rate": 2.939880485302107e-05, "loss": 0.9042, "step": 9750 }, { "epoch": 0.59, "learning_rate": 2.940484094887427e-05, "loss": 1.4277, "step": 9752 }, { "epoch": 0.59, "learning_rate": 2.941087704472747e-05, "loss": 1.6163, "step": 9754 }, { "epoch": 0.59, "learning_rate": 2.9416913140580678e-05, "loss": 1.2735, "step": 9756 }, { "epoch": 0.59, "learning_rate": 2.942294923643388e-05, "loss": 1.2796, "step": 9758 }, { "epoch": 0.59, "learning_rate": 2.942898533228708e-05, "loss": 0.9988, "step": 9760 }, { "epoch": 0.59, "learning_rate": 2.943502142814028e-05, "loss": 1.4496, "step": 9762 }, { "epoch": 0.59, "learning_rate": 2.944105752399348e-05, "loss": 1.7602, "step": 9764 }, { "epoch": 0.59, "learning_rate": 2.9447093619846682e-05, "loss": 1.3054, "step": 9766 }, { "epoch": 0.59, "learning_rate": 2.9453129715699883e-05, "loss": 1.7415, "step": 9768 }, { "epoch": 0.59, "learning_rate": 2.945916581155309e-05, "loss": 1.3362, "step": 9770 }, { "epoch": 0.59, "learning_rate": 2.946520190740629e-05, "loss": 1.2506, "step": 9772 }, { "epoch": 0.59, "learning_rate": 2.9471238003259495e-05, "loss": 1.1945, "step": 9774 }, { "epoch": 0.59, "learning_rate": 2.9477274099112696e-05, "loss": 1.0952, "step": 9776 }, { "epoch": 0.59, "learning_rate": 2.9483310194965897e-05, "loss": 1.3087, "step": 9778 }, { "epoch": 0.59, "learning_rate": 2.9489346290819098e-05, "loss": 1.3447, "step": 9780 }, { "epoch": 0.59, "learning_rate": 2.9495382386672305e-05, "loss": 1.3321, "step": 9782 }, { "epoch": 0.59, "learning_rate": 2.9501418482525506e-05, "loss": 1.1366, "step": 9784 }, { "epoch": 0.59, "learning_rate": 2.9507454578378707e-05, "loss": 1.1941, "step": 9786 }, { "epoch": 0.59, "learning_rate": 2.9513490674231908e-05, "loss": 1.4481, "step": 9788 }, { "epoch": 0.59, "learning_rate": 2.951952677008511e-05, "loss": 1.308, "step": 9790 }, { "epoch": 0.59, "learning_rate": 2.952556286593831e-05, "loss": 1.4826, "step": 9792 }, { "epoch": 0.59, "learning_rate": 2.9531598961791514e-05, "loss": 1.2971, "step": 9794 }, { "epoch": 0.59, "learning_rate": 2.9537635057644718e-05, "loss": 1.3895, "step": 9796 }, { "epoch": 0.59, "learning_rate": 2.9543671153497922e-05, "loss": 1.0554, "step": 9798 }, { "epoch": 0.59, "learning_rate": 2.9549707249351123e-05, "loss": 1.1869, "step": 9800 }, { "epoch": 0.59, "learning_rate": 2.9555743345204324e-05, "loss": 1.3298, "step": 9802 }, { "epoch": 0.59, "learning_rate": 2.9561779441057524e-05, "loss": 1.5327, "step": 9804 }, { "epoch": 0.59, "learning_rate": 2.9567815536910725e-05, "loss": 0.9863, "step": 9806 }, { "epoch": 0.59, "learning_rate": 2.9573851632763926e-05, "loss": 1.0918, "step": 9808 }, { "epoch": 0.59, "learning_rate": 2.9579887728617134e-05, "loss": 1.1366, "step": 9810 }, { "epoch": 0.59, "learning_rate": 2.9585923824470334e-05, "loss": 1.2511, "step": 9812 }, { "epoch": 0.59, "learning_rate": 2.9591959920323535e-05, "loss": 1.8083, "step": 9814 }, { "epoch": 0.59, "learning_rate": 2.9597996016176736e-05, "loss": 1.6828, "step": 9816 }, { "epoch": 0.59, "learning_rate": 2.960403211202994e-05, "loss": 1.4068, "step": 9818 }, { "epoch": 0.59, "learning_rate": 2.961006820788314e-05, "loss": 1.2841, "step": 9820 }, { "epoch": 0.59, "learning_rate": 2.961610430373635e-05, "loss": 1.1202, "step": 9822 }, { "epoch": 0.59, "learning_rate": 2.962214039958955e-05, "loss": 1.1949, "step": 9824 }, { "epoch": 0.59, "learning_rate": 2.962817649544275e-05, "loss": 1.5394, "step": 9826 }, { "epoch": 0.59, "learning_rate": 2.963421259129595e-05, "loss": 1.3339, "step": 9828 }, { "epoch": 0.59, "learning_rate": 2.9640248687149152e-05, "loss": 1.6103, "step": 9830 }, { "epoch": 0.59, "learning_rate": 2.9646284783002353e-05, "loss": 1.9576, "step": 9832 }, { "epoch": 0.59, "learning_rate": 2.9652320878855554e-05, "loss": 1.4149, "step": 9834 }, { "epoch": 0.59, "learning_rate": 2.965835697470876e-05, "loss": 1.3108, "step": 9836 }, { "epoch": 0.59, "learning_rate": 2.9664393070561962e-05, "loss": 1.1546, "step": 9838 }, { "epoch": 0.59, "learning_rate": 2.9670429166415163e-05, "loss": 1.2628, "step": 9840 }, { "epoch": 0.59, "learning_rate": 2.9676465262268367e-05, "loss": 0.901, "step": 9842 }, { "epoch": 0.59, "learning_rate": 2.9682501358121568e-05, "loss": 1.3681, "step": 9844 }, { "epoch": 0.59, "learning_rate": 2.968853745397477e-05, "loss": 1.1305, "step": 9846 }, { "epoch": 0.59, "learning_rate": 2.9694573549827976e-05, "loss": 1.2116, "step": 9848 }, { "epoch": 0.59, "learning_rate": 2.9700609645681177e-05, "loss": 1.023, "step": 9850 }, { "epoch": 0.59, "learning_rate": 2.9706645741534378e-05, "loss": 1.2156, "step": 9852 }, { "epoch": 0.59, "learning_rate": 2.971268183738758e-05, "loss": 1.6463, "step": 9854 }, { "epoch": 0.59, "learning_rate": 2.971871793324078e-05, "loss": 0.952, "step": 9856 }, { "epoch": 0.6, "learning_rate": 2.972475402909398e-05, "loss": 1.2113, "step": 9858 }, { "epoch": 0.6, "learning_rate": 2.9730790124947185e-05, "loss": 1.1594, "step": 9860 }, { "epoch": 0.6, "learning_rate": 2.973682622080039e-05, "loss": 1.0023, "step": 9862 }, { "epoch": 0.6, "learning_rate": 2.9742862316653593e-05, "loss": 1.2879, "step": 9864 }, { "epoch": 0.6, "learning_rate": 2.9748898412506794e-05, "loss": 1.2704, "step": 9866 }, { "epoch": 0.6, "learning_rate": 2.9754934508359995e-05, "loss": 1.674, "step": 9868 }, { "epoch": 0.6, "learning_rate": 2.9760970604213195e-05, "loss": 1.1671, "step": 9870 }, { "epoch": 0.6, "learning_rate": 2.9767006700066396e-05, "loss": 1.3308, "step": 9872 }, { "epoch": 0.6, "learning_rate": 2.9773042795919597e-05, "loss": 1.4087, "step": 9874 }, { "epoch": 0.6, "learning_rate": 2.9779078891772805e-05, "loss": 1.1803, "step": 9876 }, { "epoch": 0.6, "learning_rate": 2.9785114987626006e-05, "loss": 0.9477, "step": 9878 }, { "epoch": 0.6, "learning_rate": 2.9791151083479206e-05, "loss": 1.3548, "step": 9880 }, { "epoch": 0.6, "learning_rate": 2.9797187179332407e-05, "loss": 1.0171, "step": 9882 }, { "epoch": 0.6, "learning_rate": 2.980322327518561e-05, "loss": 1.6668, "step": 9884 }, { "epoch": 0.6, "learning_rate": 2.9809259371038812e-05, "loss": 0.8175, "step": 9886 }, { "epoch": 0.6, "learning_rate": 2.981529546689202e-05, "loss": 0.9463, "step": 9888 }, { "epoch": 0.6, "learning_rate": 2.982133156274522e-05, "loss": 1.2572, "step": 9890 }, { "epoch": 0.6, "learning_rate": 2.982736765859842e-05, "loss": 1.3505, "step": 9892 }, { "epoch": 0.6, "learning_rate": 2.9833403754451622e-05, "loss": 1.2882, "step": 9894 }, { "epoch": 0.6, "learning_rate": 2.9839439850304823e-05, "loss": 1.1082, "step": 9896 }, { "epoch": 0.6, "learning_rate": 2.9845475946158024e-05, "loss": 1.3726, "step": 9898 }, { "epoch": 0.6, "learning_rate": 2.9851512042011225e-05, "loss": 1.3892, "step": 9900 }, { "epoch": 0.6, "learning_rate": 2.9857548137864432e-05, "loss": 0.9195, "step": 9902 }, { "epoch": 0.6, "learning_rate": 2.9863584233717633e-05, "loss": 1.002, "step": 9904 }, { "epoch": 0.6, "learning_rate": 2.9869620329570834e-05, "loss": 1.2325, "step": 9906 }, { "epoch": 0.6, "learning_rate": 2.9875656425424038e-05, "loss": 1.5955, "step": 9908 }, { "epoch": 0.6, "learning_rate": 2.988169252127724e-05, "loss": 1.0898, "step": 9910 }, { "epoch": 0.6, "learning_rate": 2.988772861713044e-05, "loss": 1.1162, "step": 9912 }, { "epoch": 0.6, "learning_rate": 2.9893764712983647e-05, "loss": 1.3042, "step": 9914 }, { "epoch": 0.6, "learning_rate": 2.9899800808836848e-05, "loss": 1.4928, "step": 9916 }, { "epoch": 0.6, "learning_rate": 2.990583690469005e-05, "loss": 1.2442, "step": 9918 }, { "epoch": 0.6, "learning_rate": 2.991187300054325e-05, "loss": 1.3622, "step": 9920 }, { "epoch": 0.6, "learning_rate": 2.991790909639645e-05, "loss": 1.3219, "step": 9922 }, { "epoch": 0.6, "learning_rate": 2.992394519224965e-05, "loss": 1.1589, "step": 9924 }, { "epoch": 0.6, "learning_rate": 2.9929981288102856e-05, "loss": 1.3753, "step": 9926 }, { "epoch": 0.6, "learning_rate": 2.993601738395606e-05, "loss": 1.11, "step": 9928 }, { "epoch": 0.6, "learning_rate": 2.994205347980926e-05, "loss": 1.1474, "step": 9930 }, { "epoch": 0.6, "learning_rate": 2.9948089575662465e-05, "loss": 1.6006, "step": 9932 }, { "epoch": 0.6, "learning_rate": 2.9954125671515666e-05, "loss": 1.0339, "step": 9934 }, { "epoch": 0.6, "learning_rate": 2.9960161767368867e-05, "loss": 0.9349, "step": 9936 }, { "epoch": 0.6, "learning_rate": 2.9966197863222067e-05, "loss": 1.1596, "step": 9938 }, { "epoch": 0.6, "learning_rate": 2.9972233959075275e-05, "loss": 1.3576, "step": 9940 }, { "epoch": 0.6, "learning_rate": 2.9978270054928476e-05, "loss": 1.1987, "step": 9942 }, { "epoch": 0.6, "learning_rate": 2.9984306150781677e-05, "loss": 1.2084, "step": 9944 }, { "epoch": 0.6, "learning_rate": 2.9990342246634877e-05, "loss": 1.1124, "step": 9946 }, { "epoch": 0.6, "learning_rate": 2.9996378342488078e-05, "loss": 1.0854, "step": 9948 }, { "epoch": 0.6, "learning_rate": 3.0002414438341282e-05, "loss": 1.3436, "step": 9950 }, { "epoch": 0.6, "learning_rate": 3.0008450534194483e-05, "loss": 1.1778, "step": 9952 }, { "epoch": 0.6, "learning_rate": 3.0014486630047687e-05, "loss": 0.9577, "step": 9954 }, { "epoch": 0.6, "learning_rate": 3.002052272590089e-05, "loss": 1.3621, "step": 9956 }, { "epoch": 0.6, "learning_rate": 3.0026558821754092e-05, "loss": 1.4141, "step": 9958 }, { "epoch": 0.6, "learning_rate": 3.0032594917607293e-05, "loss": 1.3334, "step": 9960 }, { "epoch": 0.6, "learning_rate": 3.0038631013460494e-05, "loss": 1.4993, "step": 9962 }, { "epoch": 0.6, "learning_rate": 3.0044667109313695e-05, "loss": 1.3688, "step": 9964 }, { "epoch": 0.6, "learning_rate": 3.0050703205166896e-05, "loss": 1.3104, "step": 9966 }, { "epoch": 0.6, "learning_rate": 3.0056739301020103e-05, "loss": 0.9961, "step": 9968 }, { "epoch": 0.6, "learning_rate": 3.0062775396873304e-05, "loss": 1.0634, "step": 9970 }, { "epoch": 0.6, "learning_rate": 3.0068811492726505e-05, "loss": 1.1533, "step": 9972 }, { "epoch": 0.6, "learning_rate": 3.007484758857971e-05, "loss": 1.4776, "step": 9974 }, { "epoch": 0.6, "learning_rate": 3.008088368443291e-05, "loss": 1.2484, "step": 9976 }, { "epoch": 0.6, "learning_rate": 3.008691978028611e-05, "loss": 1.1248, "step": 9978 }, { "epoch": 0.6, "learning_rate": 3.009295587613932e-05, "loss": 1.2487, "step": 9980 }, { "epoch": 0.6, "learning_rate": 3.009899197199252e-05, "loss": 1.0463, "step": 9982 }, { "epoch": 0.6, "learning_rate": 3.010502806784572e-05, "loss": 1.3372, "step": 9984 }, { "epoch": 0.6, "learning_rate": 3.011106416369892e-05, "loss": 1.4497, "step": 9986 }, { "epoch": 0.6, "learning_rate": 3.0117100259552122e-05, "loss": 1.2057, "step": 9988 }, { "epoch": 0.6, "learning_rate": 3.0123136355405323e-05, "loss": 1.332, "step": 9990 }, { "epoch": 0.6, "learning_rate": 3.0129172451258523e-05, "loss": 1.1475, "step": 9992 }, { "epoch": 0.6, "learning_rate": 3.013520854711173e-05, "loss": 1.2625, "step": 9994 }, { "epoch": 0.6, "learning_rate": 3.0141244642964932e-05, "loss": 0.866, "step": 9996 }, { "epoch": 0.6, "learning_rate": 3.0147280738818136e-05, "loss": 1.1639, "step": 9998 }, { "epoch": 0.6, "learning_rate": 3.0153316834671337e-05, "loss": 0.9804, "step": 10000 }, { "epoch": 0.6, "learning_rate": 3.0159352930524538e-05, "loss": 1.4375, "step": 10002 }, { "epoch": 0.6, "learning_rate": 3.016538902637774e-05, "loss": 1.3614, "step": 10004 }, { "epoch": 0.6, "learning_rate": 3.0171425122230946e-05, "loss": 1.2641, "step": 10006 }, { "epoch": 0.6, "learning_rate": 3.0177461218084147e-05, "loss": 1.0772, "step": 10008 }, { "epoch": 0.6, "learning_rate": 3.0183497313937348e-05, "loss": 1.3048, "step": 10010 }, { "epoch": 0.6, "learning_rate": 3.018953340979055e-05, "loss": 1.3451, "step": 10012 }, { "epoch": 0.6, "learning_rate": 3.019556950564375e-05, "loss": 1.0148, "step": 10014 }, { "epoch": 0.6, "learning_rate": 3.020160560149695e-05, "loss": 1.3766, "step": 10016 }, { "epoch": 0.6, "learning_rate": 3.0207641697350154e-05, "loss": 1.2172, "step": 10018 }, { "epoch": 0.6, "learning_rate": 3.021367779320336e-05, "loss": 1.0872, "step": 10020 }, { "epoch": 0.6, "learning_rate": 3.0219713889056563e-05, "loss": 1.1689, "step": 10022 }, { "epoch": 0.61, "learning_rate": 3.0225749984909764e-05, "loss": 1.2059, "step": 10024 }, { "epoch": 0.61, "learning_rate": 3.0231786080762964e-05, "loss": 1.0721, "step": 10026 }, { "epoch": 0.61, "learning_rate": 3.0237822176616165e-05, "loss": 1.0534, "step": 10028 }, { "epoch": 0.61, "learning_rate": 3.0243858272469366e-05, "loss": 1.4003, "step": 10030 }, { "epoch": 0.61, "learning_rate": 3.0249894368322567e-05, "loss": 1.4515, "step": 10032 }, { "epoch": 0.61, "learning_rate": 3.0255930464175774e-05, "loss": 1.1255, "step": 10034 }, { "epoch": 0.61, "learning_rate": 3.0261966560028975e-05, "loss": 1.1868, "step": 10036 }, { "epoch": 0.61, "learning_rate": 3.0268002655882176e-05, "loss": 0.9505, "step": 10038 }, { "epoch": 0.61, "learning_rate": 3.0274038751735377e-05, "loss": 1.7066, "step": 10040 }, { "epoch": 0.61, "learning_rate": 3.028007484758858e-05, "loss": 1.4456, "step": 10042 }, { "epoch": 0.61, "learning_rate": 3.0286110943441782e-05, "loss": 1.3901, "step": 10044 }, { "epoch": 0.61, "learning_rate": 3.029214703929499e-05, "loss": 1.0024, "step": 10046 }, { "epoch": 0.61, "learning_rate": 3.029818313514819e-05, "loss": 1.2106, "step": 10048 }, { "epoch": 0.61, "learning_rate": 3.030421923100139e-05, "loss": 0.8995, "step": 10050 }, { "epoch": 0.61, "learning_rate": 3.0310255326854592e-05, "loss": 1.0868, "step": 10052 }, { "epoch": 0.61, "learning_rate": 3.0316291422707793e-05, "loss": 1.4188, "step": 10054 }, { "epoch": 0.61, "learning_rate": 3.0322327518560994e-05, "loss": 0.9566, "step": 10056 }, { "epoch": 0.61, "learning_rate": 3.0328363614414194e-05, "loss": 1.0256, "step": 10058 }, { "epoch": 0.61, "learning_rate": 3.0334399710267402e-05, "loss": 1.0811, "step": 10060 }, { "epoch": 0.61, "learning_rate": 3.0340435806120603e-05, "loss": 1.1061, "step": 10062 }, { "epoch": 0.61, "learning_rate": 3.0346471901973804e-05, "loss": 0.9923, "step": 10064 }, { "epoch": 0.61, "learning_rate": 3.0352507997827008e-05, "loss": 1.5571, "step": 10066 }, { "epoch": 0.61, "learning_rate": 3.035854409368021e-05, "loss": 1.3597, "step": 10068 }, { "epoch": 0.61, "learning_rate": 3.036458018953341e-05, "loss": 1.2351, "step": 10070 }, { "epoch": 0.61, "learning_rate": 3.0370616285386617e-05, "loss": 1.3042, "step": 10072 }, { "epoch": 0.61, "learning_rate": 3.0376652381239818e-05, "loss": 1.1984, "step": 10074 }, { "epoch": 0.61, "learning_rate": 3.038268847709302e-05, "loss": 1.185, "step": 10076 }, { "epoch": 0.61, "learning_rate": 3.038872457294622e-05, "loss": 1.2223, "step": 10078 }, { "epoch": 0.61, "learning_rate": 3.039476066879942e-05, "loss": 1.0634, "step": 10080 }, { "epoch": 0.61, "learning_rate": 3.040079676465262e-05, "loss": 1.588, "step": 10082 }, { "epoch": 0.61, "learning_rate": 3.0406832860505825e-05, "loss": 1.4076, "step": 10084 }, { "epoch": 0.61, "learning_rate": 3.041286895635903e-05, "loss": 1.1126, "step": 10086 }, { "epoch": 0.61, "learning_rate": 3.041890505221223e-05, "loss": 1.5037, "step": 10088 }, { "epoch": 0.61, "learning_rate": 3.0424941148065435e-05, "loss": 1.2426, "step": 10090 }, { "epoch": 0.61, "learning_rate": 3.0430977243918635e-05, "loss": 2.0939, "step": 10092 }, { "epoch": 0.61, "learning_rate": 3.0437013339771836e-05, "loss": 1.1933, "step": 10094 }, { "epoch": 0.61, "learning_rate": 3.0443049435625037e-05, "loss": 1.1183, "step": 10096 }, { "epoch": 0.61, "learning_rate": 3.0449085531478238e-05, "loss": 1.2104, "step": 10098 }, { "epoch": 0.61, "learning_rate": 3.0455121627331445e-05, "loss": 1.064, "step": 10100 }, { "epoch": 0.61, "learning_rate": 3.0461157723184646e-05, "loss": 1.1619, "step": 10102 }, { "epoch": 0.61, "learning_rate": 3.0467193819037847e-05, "loss": 1.1857, "step": 10104 }, { "epoch": 0.61, "learning_rate": 3.0473229914891048e-05, "loss": 1.1041, "step": 10106 }, { "epoch": 0.61, "learning_rate": 3.0479266010744252e-05, "loss": 1.2911, "step": 10108 }, { "epoch": 0.61, "learning_rate": 3.0485302106597453e-05, "loss": 1.2921, "step": 10110 }, { "epoch": 0.61, "learning_rate": 3.049133820245066e-05, "loss": 1.2565, "step": 10112 }, { "epoch": 0.61, "learning_rate": 3.049737429830386e-05, "loss": 1.1362, "step": 10114 }, { "epoch": 0.61, "learning_rate": 3.0503410394157062e-05, "loss": 1.7419, "step": 10116 }, { "epoch": 0.61, "learning_rate": 3.0509446490010263e-05, "loss": 1.3719, "step": 10118 }, { "epoch": 0.61, "learning_rate": 3.0515482585863464e-05, "loss": 1.0755, "step": 10120 }, { "epoch": 0.61, "learning_rate": 3.0521518681716665e-05, "loss": 1.3796, "step": 10122 }, { "epoch": 0.61, "learning_rate": 3.0527554777569865e-05, "loss": 1.2972, "step": 10124 }, { "epoch": 0.61, "learning_rate": 3.053359087342307e-05, "loss": 1.4757, "step": 10126 }, { "epoch": 0.61, "learning_rate": 3.0539626969276274e-05, "loss": 1.1834, "step": 10128 }, { "epoch": 0.61, "learning_rate": 3.0545663065129475e-05, "loss": 1.0672, "step": 10130 }, { "epoch": 0.61, "learning_rate": 3.0551699160982676e-05, "loss": 1.3369, "step": 10132 }, { "epoch": 0.61, "learning_rate": 3.0557735256835876e-05, "loss": 0.9947, "step": 10134 }, { "epoch": 0.61, "learning_rate": 3.056377135268908e-05, "loss": 1.1386, "step": 10136 }, { "epoch": 0.61, "learning_rate": 3.0569807448542285e-05, "loss": 1.0649, "step": 10138 }, { "epoch": 0.61, "learning_rate": 3.0575843544395486e-05, "loss": 1.2844, "step": 10140 }, { "epoch": 0.61, "learning_rate": 3.0581879640248686e-05, "loss": 1.0077, "step": 10142 }, { "epoch": 0.61, "learning_rate": 3.0587915736101894e-05, "loss": 1.327, "step": 10144 }, { "epoch": 0.61, "learning_rate": 3.0593951831955095e-05, "loss": 1.1316, "step": 10146 }, { "epoch": 0.61, "learning_rate": 3.0599987927808296e-05, "loss": 1.0481, "step": 10148 }, { "epoch": 0.61, "learning_rate": 3.0606024023661496e-05, "loss": 1.2146, "step": 10150 }, { "epoch": 0.61, "learning_rate": 3.0612060119514704e-05, "loss": 1.2908, "step": 10152 }, { "epoch": 0.61, "learning_rate": 3.0618096215367905e-05, "loss": 1.1465, "step": 10154 }, { "epoch": 0.61, "learning_rate": 3.0624132311221106e-05, "loss": 1.2505, "step": 10156 }, { "epoch": 0.61, "learning_rate": 3.0630168407074306e-05, "loss": 1.2466, "step": 10158 }, { "epoch": 0.61, "learning_rate": 3.063620450292751e-05, "loss": 1.8968, "step": 10160 }, { "epoch": 0.61, "learning_rate": 3.064224059878071e-05, "loss": 1.1565, "step": 10162 }, { "epoch": 0.61, "learning_rate": 3.064827669463391e-05, "loss": 1.4108, "step": 10164 }, { "epoch": 0.61, "learning_rate": 3.0654312790487117e-05, "loss": 1.1717, "step": 10166 }, { "epoch": 0.61, "learning_rate": 3.066034888634032e-05, "loss": 1.4638, "step": 10168 }, { "epoch": 0.61, "learning_rate": 3.066638498219352e-05, "loss": 1.0933, "step": 10170 }, { "epoch": 0.61, "learning_rate": 3.067242107804672e-05, "loss": 0.789, "step": 10172 }, { "epoch": 0.61, "learning_rate": 3.067845717389992e-05, "loss": 1.2627, "step": 10174 }, { "epoch": 0.61, "learning_rate": 3.068449326975312e-05, "loss": 1.0278, "step": 10176 }, { "epoch": 0.61, "learning_rate": 3.069052936560633e-05, "loss": 1.2807, "step": 10178 }, { "epoch": 0.61, "learning_rate": 3.069656546145953e-05, "loss": 1.0951, "step": 10180 }, { "epoch": 0.61, "learning_rate": 3.070260155731273e-05, "loss": 1.1999, "step": 10182 }, { "epoch": 0.61, "learning_rate": 3.070863765316593e-05, "loss": 0.9853, "step": 10184 }, { "epoch": 0.61, "learning_rate": 3.071467374901914e-05, "loss": 1.2943, "step": 10186 }, { "epoch": 0.61, "learning_rate": 3.072070984487234e-05, "loss": 1.2608, "step": 10188 }, { "epoch": 0.62, "learning_rate": 3.072674594072554e-05, "loss": 1.2228, "step": 10190 }, { "epoch": 0.62, "learning_rate": 3.073278203657875e-05, "loss": 1.123, "step": 10192 }, { "epoch": 0.62, "learning_rate": 3.073881813243195e-05, "loss": 1.5053, "step": 10194 }, { "epoch": 0.62, "learning_rate": 3.074485422828515e-05, "loss": 0.7513, "step": 10196 }, { "epoch": 0.62, "learning_rate": 3.075089032413835e-05, "loss": 1.8006, "step": 10198 }, { "epoch": 0.62, "learning_rate": 3.075692641999155e-05, "loss": 1.301, "step": 10200 }, { "epoch": 0.62, "learning_rate": 3.076296251584475e-05, "loss": 1.3554, "step": 10202 }, { "epoch": 0.62, "learning_rate": 3.076899861169796e-05, "loss": 3.1234, "step": 10204 }, { "epoch": 0.62, "learning_rate": 3.077503470755116e-05, "loss": 1.2916, "step": 10206 }, { "epoch": 0.62, "learning_rate": 3.078107080340436e-05, "loss": 1.1318, "step": 10208 }, { "epoch": 0.62, "learning_rate": 3.078710689925756e-05, "loss": 0.8543, "step": 10210 }, { "epoch": 0.62, "learning_rate": 3.079314299511076e-05, "loss": 1.5404, "step": 10212 }, { "epoch": 0.62, "learning_rate": 3.079917909096396e-05, "loss": 1.2169, "step": 10214 }, { "epoch": 0.62, "learning_rate": 3.0805215186817164e-05, "loss": 0.9604, "step": 10216 }, { "epoch": 0.62, "learning_rate": 3.081125128267037e-05, "loss": 1.5479, "step": 10218 }, { "epoch": 0.62, "learning_rate": 3.081728737852357e-05, "loss": 1.8314, "step": 10220 }, { "epoch": 0.62, "learning_rate": 3.082332347437677e-05, "loss": 1.1431, "step": 10222 }, { "epoch": 0.62, "learning_rate": 3.0829359570229974e-05, "loss": 1.9453, "step": 10224 }, { "epoch": 0.62, "learning_rate": 3.0835395666083175e-05, "loss": 1.1589, "step": 10226 }, { "epoch": 0.62, "learning_rate": 3.0841431761936376e-05, "loss": 1.3712, "step": 10228 }, { "epoch": 0.62, "learning_rate": 3.0847467857789583e-05, "loss": 1.1982, "step": 10230 }, { "epoch": 0.62, "learning_rate": 3.0853503953642784e-05, "loss": 1.1772, "step": 10232 }, { "epoch": 0.62, "learning_rate": 3.085954004949599e-05, "loss": 1.0315, "step": 10234 }, { "epoch": 0.62, "learning_rate": 3.086557614534919e-05, "loss": 1.2558, "step": 10236 }, { "epoch": 0.62, "learning_rate": 3.0871612241202393e-05, "loss": 1.1628, "step": 10238 }, { "epoch": 0.62, "learning_rate": 3.0877648337055594e-05, "loss": 1.3314, "step": 10240 }, { "epoch": 0.62, "learning_rate": 3.0883684432908795e-05, "loss": 1.3631, "step": 10242 }, { "epoch": 0.62, "learning_rate": 3.0889720528762e-05, "loss": 1.2978, "step": 10244 }, { "epoch": 0.62, "learning_rate": 3.0895756624615203e-05, "loss": 1.2472, "step": 10246 }, { "epoch": 0.62, "learning_rate": 3.0901792720468404e-05, "loss": 1.3205, "step": 10248 }, { "epoch": 0.62, "learning_rate": 3.0907828816321605e-05, "loss": 1.4979, "step": 10250 }, { "epoch": 0.62, "learning_rate": 3.0913864912174806e-05, "loss": 0.9608, "step": 10252 }, { "epoch": 0.62, "learning_rate": 3.091990100802801e-05, "loss": 1.0554, "step": 10254 }, { "epoch": 0.62, "learning_rate": 3.092593710388121e-05, "loss": 1.3129, "step": 10256 }, { "epoch": 0.62, "learning_rate": 3.0931973199734415e-05, "loss": 1.1303, "step": 10258 }, { "epoch": 0.62, "learning_rate": 3.0938009295587616e-05, "loss": 1.1758, "step": 10260 }, { "epoch": 0.62, "learning_rate": 3.094404539144082e-05, "loss": 1.3787, "step": 10262 }, { "epoch": 0.62, "learning_rate": 3.095008148729402e-05, "loss": 1.0984, "step": 10264 }, { "epoch": 0.62, "learning_rate": 3.095611758314722e-05, "loss": 1.0964, "step": 10266 }, { "epoch": 0.62, "learning_rate": 3.096215367900042e-05, "loss": 1.5009, "step": 10268 }, { "epoch": 0.62, "learning_rate": 3.096818977485363e-05, "loss": 1.488, "step": 10270 }, { "epoch": 0.62, "learning_rate": 3.097422587070683e-05, "loss": 1.2599, "step": 10272 }, { "epoch": 0.62, "learning_rate": 3.098026196656003e-05, "loss": 1.3386, "step": 10274 }, { "epoch": 0.62, "learning_rate": 3.098629806241323e-05, "loss": 1.0492, "step": 10276 }, { "epoch": 0.62, "learning_rate": 3.099233415826644e-05, "loss": 1.3145, "step": 10278 }, { "epoch": 0.62, "learning_rate": 3.099837025411964e-05, "loss": 1.4392, "step": 10280 }, { "epoch": 0.62, "learning_rate": 3.100440634997284e-05, "loss": 1.2571, "step": 10282 }, { "epoch": 0.62, "learning_rate": 3.1010442445826046e-05, "loss": 1.2075, "step": 10284 }, { "epoch": 0.62, "learning_rate": 3.101647854167925e-05, "loss": 1.1241, "step": 10286 }, { "epoch": 0.62, "learning_rate": 3.102251463753245e-05, "loss": 1.097, "step": 10288 }, { "epoch": 0.62, "learning_rate": 3.102855073338565e-05, "loss": 1.6328, "step": 10290 }, { "epoch": 0.62, "learning_rate": 3.103458682923885e-05, "loss": 1.3862, "step": 10292 }, { "epoch": 0.62, "learning_rate": 3.104062292509205e-05, "loss": 1.0952, "step": 10294 }, { "epoch": 0.62, "learning_rate": 3.104665902094526e-05, "loss": 0.9464, "step": 10296 }, { "epoch": 0.62, "learning_rate": 3.105269511679846e-05, "loss": 1.3037, "step": 10298 }, { "epoch": 0.62, "learning_rate": 3.105873121265166e-05, "loss": 0.9836, "step": 10300 }, { "epoch": 0.62, "learning_rate": 3.106476730850486e-05, "loss": 1.2736, "step": 10302 }, { "epoch": 0.62, "learning_rate": 3.107080340435806e-05, "loss": 1.1022, "step": 10304 }, { "epoch": 0.62, "learning_rate": 3.107683950021126e-05, "loss": 1.4958, "step": 10306 }, { "epoch": 0.62, "learning_rate": 3.108287559606446e-05, "loss": 1.2906, "step": 10308 }, { "epoch": 0.62, "learning_rate": 3.108891169191767e-05, "loss": 1.3928, "step": 10310 }, { "epoch": 0.62, "learning_rate": 3.109494778777087e-05, "loss": 1.1393, "step": 10312 }, { "epoch": 0.62, "learning_rate": 3.110098388362407e-05, "loss": 1.1317, "step": 10314 }, { "epoch": 0.62, "learning_rate": 3.110701997947727e-05, "loss": 0.9086, "step": 10316 }, { "epoch": 0.62, "learning_rate": 3.1113056075330474e-05, "loss": 1.0577, "step": 10318 }, { "epoch": 0.62, "learning_rate": 3.111909217118368e-05, "loss": 1.2608, "step": 10320 }, { "epoch": 0.62, "learning_rate": 3.112512826703688e-05, "loss": 1.0406, "step": 10322 }, { "epoch": 0.62, "learning_rate": 3.113116436289008e-05, "loss": 1.227, "step": 10324 }, { "epoch": 0.62, "learning_rate": 3.113720045874329e-05, "loss": 1.3119, "step": 10326 }, { "epoch": 0.62, "learning_rate": 3.114323655459649e-05, "loss": 1.0886, "step": 10328 }, { "epoch": 0.62, "learning_rate": 3.114927265044969e-05, "loss": 1.65, "step": 10330 }, { "epoch": 0.62, "learning_rate": 3.115530874630289e-05, "loss": 1.0165, "step": 10332 }, { "epoch": 0.62, "learning_rate": 3.1161344842156094e-05, "loss": 1.1532, "step": 10334 }, { "epoch": 0.62, "learning_rate": 3.11673809380093e-05, "loss": 1.5629, "step": 10336 }, { "epoch": 0.62, "learning_rate": 3.11734170338625e-05, "loss": 1.2337, "step": 10338 }, { "epoch": 0.62, "learning_rate": 3.11794531297157e-05, "loss": 1.6001, "step": 10340 }, { "epoch": 0.62, "learning_rate": 3.1185489225568904e-05, "loss": 0.8365, "step": 10342 }, { "epoch": 0.62, "learning_rate": 3.1191525321422105e-05, "loss": 1.0246, "step": 10344 }, { "epoch": 0.62, "learning_rate": 3.1197561417275305e-05, "loss": 1.2622, "step": 10346 }, { "epoch": 0.62, "learning_rate": 3.1203597513128506e-05, "loss": 1.2061, "step": 10348 }, { "epoch": 0.62, "learning_rate": 3.1209633608981714e-05, "loss": 0.8628, "step": 10350 }, { "epoch": 0.62, "learning_rate": 3.1215669704834915e-05, "loss": 1.3517, "step": 10352 }, { "epoch": 0.62, "learning_rate": 3.1221705800688115e-05, "loss": 1.5439, "step": 10354 }, { "epoch": 0.63, "learning_rate": 3.1227741896541316e-05, "loss": 1.1194, "step": 10356 }, { "epoch": 0.63, "learning_rate": 3.123377799239452e-05, "loss": 1.0191, "step": 10358 }, { "epoch": 0.63, "learning_rate": 3.123981408824772e-05, "loss": 1.4265, "step": 10360 }, { "epoch": 0.63, "learning_rate": 3.1245850184100926e-05, "loss": 1.4396, "step": 10362 }, { "epoch": 0.63, "learning_rate": 3.1251886279954126e-05, "loss": 1.2238, "step": 10364 }, { "epoch": 0.63, "learning_rate": 3.125792237580733e-05, "loss": 1.4104, "step": 10366 }, { "epoch": 0.63, "learning_rate": 3.1263958471660535e-05, "loss": 1.3038, "step": 10368 }, { "epoch": 0.63, "learning_rate": 3.1269994567513736e-05, "loss": 0.8993, "step": 10370 }, { "epoch": 0.63, "learning_rate": 3.1276030663366936e-05, "loss": 1.3474, "step": 10372 }, { "epoch": 0.63, "learning_rate": 3.128206675922014e-05, "loss": 1.1764, "step": 10374 }, { "epoch": 0.63, "learning_rate": 3.1288102855073345e-05, "loss": 1.2016, "step": 10376 }, { "epoch": 0.63, "learning_rate": 3.1294138950926546e-05, "loss": 1.171, "step": 10378 }, { "epoch": 0.63, "learning_rate": 3.1300175046779746e-05, "loss": 1.1373, "step": 10380 }, { "epoch": 0.63, "learning_rate": 3.130621114263295e-05, "loss": 1.4481, "step": 10382 }, { "epoch": 0.63, "learning_rate": 3.131224723848615e-05, "loss": 1.1616, "step": 10384 }, { "epoch": 0.63, "learning_rate": 3.131828333433935e-05, "loss": 1.0588, "step": 10386 }, { "epoch": 0.63, "learning_rate": 3.132431943019255e-05, "loss": 1.2251, "step": 10388 }, { "epoch": 0.63, "learning_rate": 3.133035552604576e-05, "loss": 1.224, "step": 10390 }, { "epoch": 0.63, "learning_rate": 3.133639162189896e-05, "loss": 1.4769, "step": 10392 }, { "epoch": 0.63, "learning_rate": 3.134242771775216e-05, "loss": 1.189, "step": 10394 }, { "epoch": 0.63, "learning_rate": 3.134846381360536e-05, "loss": 0.9841, "step": 10396 }, { "epoch": 0.63, "learning_rate": 3.135449990945856e-05, "loss": 1.209, "step": 10398 }, { "epoch": 0.63, "learning_rate": 3.136053600531176e-05, "loss": 1.0627, "step": 10400 }, { "epoch": 0.63, "learning_rate": 3.136657210116497e-05, "loss": 1.468, "step": 10402 }, { "epoch": 0.63, "learning_rate": 3.137260819701817e-05, "loss": 1.1456, "step": 10404 }, { "epoch": 0.63, "learning_rate": 3.137864429287137e-05, "loss": 1.0625, "step": 10406 }, { "epoch": 0.63, "learning_rate": 3.138468038872457e-05, "loss": 1.6174, "step": 10408 }, { "epoch": 0.63, "learning_rate": 3.139071648457778e-05, "loss": 1.207, "step": 10410 }, { "epoch": 0.63, "learning_rate": 3.139675258043098e-05, "loss": 1.382, "step": 10412 }, { "epoch": 0.63, "learning_rate": 3.140278867628418e-05, "loss": 1.1846, "step": 10414 }, { "epoch": 0.63, "learning_rate": 3.140882477213739e-05, "loss": 1.2769, "step": 10416 }, { "epoch": 0.63, "learning_rate": 3.141486086799059e-05, "loss": 1.1343, "step": 10418 }, { "epoch": 0.63, "learning_rate": 3.142089696384379e-05, "loss": 1.2916, "step": 10420 }, { "epoch": 0.63, "learning_rate": 3.142693305969699e-05, "loss": 1.2652, "step": 10422 }, { "epoch": 0.63, "learning_rate": 3.143296915555019e-05, "loss": 1.0084, "step": 10424 }, { "epoch": 0.63, "learning_rate": 3.143900525140339e-05, "loss": 1.477, "step": 10426 }, { "epoch": 0.63, "learning_rate": 3.14450413472566e-05, "loss": 0.8556, "step": 10428 }, { "epoch": 0.63, "learning_rate": 3.14510774431098e-05, "loss": 1.2272, "step": 10430 }, { "epoch": 0.63, "learning_rate": 3.1457113538963e-05, "loss": 1.3915, "step": 10432 }, { "epoch": 0.63, "learning_rate": 3.14631496348162e-05, "loss": 1.1257, "step": 10434 }, { "epoch": 0.63, "learning_rate": 3.14691857306694e-05, "loss": 1.1321, "step": 10436 }, { "epoch": 0.63, "learning_rate": 3.1475221826522604e-05, "loss": 1.3726, "step": 10438 }, { "epoch": 0.63, "learning_rate": 3.1481257922375805e-05, "loss": 1.175, "step": 10440 }, { "epoch": 0.63, "learning_rate": 3.148729401822901e-05, "loss": 1.4318, "step": 10442 }, { "epoch": 0.63, "learning_rate": 3.149333011408221e-05, "loss": 1.3002, "step": 10444 }, { "epoch": 0.63, "learning_rate": 3.1499366209935414e-05, "loss": 1.3012, "step": 10446 }, { "epoch": 0.63, "learning_rate": 3.1505402305788615e-05, "loss": 1.3186, "step": 10448 }, { "epoch": 0.63, "learning_rate": 3.1511438401641816e-05, "loss": 1.0694, "step": 10450 }, { "epoch": 0.63, "learning_rate": 3.1517474497495017e-05, "loss": 1.3325, "step": 10452 }, { "epoch": 0.63, "learning_rate": 3.1523510593348224e-05, "loss": 1.183, "step": 10454 }, { "epoch": 0.63, "learning_rate": 3.1529546689201425e-05, "loss": 1.1511, "step": 10456 }, { "epoch": 0.63, "learning_rate": 3.153558278505463e-05, "loss": 1.1257, "step": 10458 }, { "epoch": 0.63, "learning_rate": 3.154161888090783e-05, "loss": 1.0708, "step": 10460 }, { "epoch": 0.63, "learning_rate": 3.1547654976761034e-05, "loss": 1.6648, "step": 10462 }, { "epoch": 0.63, "learning_rate": 3.1553691072614235e-05, "loss": 1.0172, "step": 10464 }, { "epoch": 0.63, "learning_rate": 3.1559727168467436e-05, "loss": 1.1145, "step": 10466 }, { "epoch": 0.63, "learning_rate": 3.1565763264320643e-05, "loss": 1.3301, "step": 10468 }, { "epoch": 0.63, "learning_rate": 3.1571799360173844e-05, "loss": 1.0214, "step": 10470 }, { "epoch": 0.63, "learning_rate": 3.1577835456027045e-05, "loss": 1.2265, "step": 10472 }, { "epoch": 0.63, "learning_rate": 3.1583871551880246e-05, "loss": 1.2375, "step": 10474 }, { "epoch": 0.63, "learning_rate": 3.158990764773345e-05, "loss": 1.5329, "step": 10476 }, { "epoch": 0.63, "learning_rate": 3.159594374358665e-05, "loss": 1.1231, "step": 10478 }, { "epoch": 0.63, "learning_rate": 3.160197983943985e-05, "loss": 1.3479, "step": 10480 }, { "epoch": 0.63, "learning_rate": 3.1608015935293056e-05, "loss": 1.3435, "step": 10482 }, { "epoch": 0.63, "learning_rate": 3.161405203114626e-05, "loss": 1.1718, "step": 10484 }, { "epoch": 0.63, "learning_rate": 3.162008812699946e-05, "loss": 1.1772, "step": 10486 }, { "epoch": 0.63, "learning_rate": 3.162612422285266e-05, "loss": 1.4377, "step": 10488 }, { "epoch": 0.63, "learning_rate": 3.163216031870586e-05, "loss": 1.2186, "step": 10490 }, { "epoch": 0.63, "learning_rate": 3.163819641455906e-05, "loss": 1.639, "step": 10492 }, { "epoch": 0.63, "learning_rate": 3.164423251041227e-05, "loss": 1.1042, "step": 10494 }, { "epoch": 0.63, "learning_rate": 3.165026860626547e-05, "loss": 1.2459, "step": 10496 }, { "epoch": 0.63, "learning_rate": 3.165630470211867e-05, "loss": 0.8001, "step": 10498 }, { "epoch": 0.63, "learning_rate": 3.166234079797187e-05, "loss": 0.979, "step": 10500 }, { "epoch": 0.63, "learning_rate": 3.166837689382508e-05, "loss": 1.2093, "step": 10502 }, { "epoch": 0.63, "learning_rate": 3.167441298967828e-05, "loss": 1.3488, "step": 10504 }, { "epoch": 0.63, "learning_rate": 3.168044908553148e-05, "loss": 1.2618, "step": 10506 }, { "epoch": 0.63, "learning_rate": 3.168648518138469e-05, "loss": 0.8705, "step": 10508 }, { "epoch": 0.63, "learning_rate": 3.169252127723789e-05, "loss": 1.1962, "step": 10510 }, { "epoch": 0.63, "learning_rate": 3.169855737309109e-05, "loss": 1.0637, "step": 10512 }, { "epoch": 0.63, "learning_rate": 3.170459346894429e-05, "loss": 1.1694, "step": 10514 }, { "epoch": 0.63, "learning_rate": 3.171062956479749e-05, "loss": 1.1303, "step": 10516 }, { "epoch": 0.63, "learning_rate": 3.171666566065069e-05, "loss": 0.9608, "step": 10518 }, { "epoch": 0.63, "learning_rate": 3.17227017565039e-05, "loss": 1.0889, "step": 10520 }, { "epoch": 0.64, "learning_rate": 3.17287378523571e-05, "loss": 1.1412, "step": 10522 }, { "epoch": 0.64, "learning_rate": 3.17347739482103e-05, "loss": 1.1035, "step": 10524 }, { "epoch": 0.64, "learning_rate": 3.17408100440635e-05, "loss": 1.0562, "step": 10526 }, { "epoch": 0.64, "learning_rate": 3.17468461399167e-05, "loss": 1.082, "step": 10528 }, { "epoch": 0.64, "learning_rate": 3.17528822357699e-05, "loss": 1.4845, "step": 10530 }, { "epoch": 0.64, "learning_rate": 3.1758918331623104e-05, "loss": 1.2352, "step": 10532 }, { "epoch": 0.64, "learning_rate": 3.176495442747631e-05, "loss": 1.1892, "step": 10534 }, { "epoch": 0.64, "learning_rate": 3.177099052332951e-05, "loss": 1.4687, "step": 10536 }, { "epoch": 0.64, "learning_rate": 3.177702661918271e-05, "loss": 2.1911, "step": 10538 }, { "epoch": 0.64, "learning_rate": 3.1783062715035914e-05, "loss": 1.211, "step": 10540 }, { "epoch": 0.64, "learning_rate": 3.1789098810889114e-05, "loss": 1.0657, "step": 10542 }, { "epoch": 0.64, "learning_rate": 3.179513490674232e-05, "loss": 1.5696, "step": 10544 }, { "epoch": 0.64, "learning_rate": 3.180117100259552e-05, "loss": 1.0933, "step": 10546 }, { "epoch": 0.64, "learning_rate": 3.1807207098448724e-05, "loss": 1.3891, "step": 10548 }, { "epoch": 0.64, "learning_rate": 3.181324319430193e-05, "loss": 1.0273, "step": 10550 }, { "epoch": 0.64, "learning_rate": 3.181927929015513e-05, "loss": 1.4552, "step": 10552 }, { "epoch": 0.64, "learning_rate": 3.182531538600833e-05, "loss": 1.3897, "step": 10554 }, { "epoch": 0.64, "learning_rate": 3.1831351481861534e-05, "loss": 0.9594, "step": 10556 }, { "epoch": 0.64, "learning_rate": 3.1837387577714734e-05, "loss": 1.3012, "step": 10558 }, { "epoch": 0.64, "learning_rate": 3.184342367356794e-05, "loss": 1.1401, "step": 10560 }, { "epoch": 0.64, "learning_rate": 3.184945976942114e-05, "loss": 1.0951, "step": 10562 }, { "epoch": 0.64, "learning_rate": 3.1855495865274344e-05, "loss": 1.277, "step": 10564 }, { "epoch": 0.64, "learning_rate": 3.1861531961127545e-05, "loss": 0.8659, "step": 10566 }, { "epoch": 0.64, "learning_rate": 3.1867568056980745e-05, "loss": 1.1589, "step": 10568 }, { "epoch": 0.64, "learning_rate": 3.1873604152833946e-05, "loss": 1.1424, "step": 10570 }, { "epoch": 0.64, "learning_rate": 3.187964024868715e-05, "loss": 1.5726, "step": 10572 }, { "epoch": 0.64, "learning_rate": 3.1885676344540355e-05, "loss": 1.2087, "step": 10574 }, { "epoch": 0.64, "learning_rate": 3.1891712440393555e-05, "loss": 1.0033, "step": 10576 }, { "epoch": 0.64, "learning_rate": 3.1897748536246756e-05, "loss": 1.2501, "step": 10578 }, { "epoch": 0.64, "learning_rate": 3.190378463209996e-05, "loss": 1.2938, "step": 10580 }, { "epoch": 0.64, "learning_rate": 3.190982072795316e-05, "loss": 1.3978, "step": 10582 }, { "epoch": 0.64, "learning_rate": 3.191585682380636e-05, "loss": 1.5488, "step": 10584 }, { "epoch": 0.64, "learning_rate": 3.1921892919659566e-05, "loss": 1.1683, "step": 10586 }, { "epoch": 0.64, "learning_rate": 3.192792901551277e-05, "loss": 1.3262, "step": 10588 }, { "epoch": 0.64, "learning_rate": 3.193396511136597e-05, "loss": 1.3905, "step": 10590 }, { "epoch": 0.64, "learning_rate": 3.1940001207219176e-05, "loss": 1.4812, "step": 10592 }, { "epoch": 0.64, "learning_rate": 3.1946037303072376e-05, "loss": 1.4719, "step": 10594 }, { "epoch": 0.64, "learning_rate": 3.195207339892558e-05, "loss": 1.2292, "step": 10596 }, { "epoch": 0.64, "learning_rate": 3.195810949477878e-05, "loss": 1.2344, "step": 10598 }, { "epoch": 0.64, "learning_rate": 3.1964145590631986e-05, "loss": 1.1795, "step": 10600 }, { "epoch": 0.64, "learning_rate": 3.1970181686485186e-05, "loss": 1.1469, "step": 10602 }, { "epoch": 0.64, "learning_rate": 3.197621778233839e-05, "loss": 1.6566, "step": 10604 }, { "epoch": 0.64, "learning_rate": 3.198225387819159e-05, "loss": 1.5165, "step": 10606 }, { "epoch": 0.64, "learning_rate": 3.198828997404479e-05, "loss": 1.2943, "step": 10608 }, { "epoch": 0.64, "learning_rate": 3.199432606989799e-05, "loss": 1.2202, "step": 10610 }, { "epoch": 0.64, "learning_rate": 3.200036216575119e-05, "loss": 1.2811, "step": 10612 }, { "epoch": 0.64, "learning_rate": 3.20063982616044e-05, "loss": 1.4795, "step": 10614 }, { "epoch": 0.64, "learning_rate": 3.20124343574576e-05, "loss": 1.7175, "step": 10616 }, { "epoch": 0.64, "learning_rate": 3.20184704533108e-05, "loss": 1.4034, "step": 10618 }, { "epoch": 0.64, "learning_rate": 3.2024506549164e-05, "loss": 0.9189, "step": 10620 }, { "epoch": 0.64, "learning_rate": 3.20305426450172e-05, "loss": 1.2642, "step": 10622 }, { "epoch": 0.64, "learning_rate": 3.20365787408704e-05, "loss": 1.49, "step": 10624 }, { "epoch": 0.64, "learning_rate": 3.204261483672361e-05, "loss": 1.4409, "step": 10626 }, { "epoch": 0.64, "learning_rate": 3.204865093257681e-05, "loss": 1.4195, "step": 10628 }, { "epoch": 0.64, "learning_rate": 3.205468702843001e-05, "loss": 1.0553, "step": 10630 }, { "epoch": 0.64, "learning_rate": 3.206072312428321e-05, "loss": 1.1243, "step": 10632 }, { "epoch": 0.64, "learning_rate": 3.206675922013642e-05, "loss": 1.0399, "step": 10634 }, { "epoch": 0.64, "learning_rate": 3.207279531598962e-05, "loss": 1.465, "step": 10636 }, { "epoch": 0.64, "learning_rate": 3.207883141184282e-05, "loss": 1.4496, "step": 10638 }, { "epoch": 0.64, "learning_rate": 3.208486750769603e-05, "loss": 1.4036, "step": 10640 }, { "epoch": 0.64, "learning_rate": 3.209090360354923e-05, "loss": 1.4562, "step": 10642 }, { "epoch": 0.64, "learning_rate": 3.209693969940243e-05, "loss": 1.4822, "step": 10644 }, { "epoch": 0.64, "learning_rate": 3.210297579525563e-05, "loss": 1.1013, "step": 10646 }, { "epoch": 0.64, "learning_rate": 3.210901189110883e-05, "loss": 1.3344, "step": 10648 }, { "epoch": 0.64, "learning_rate": 3.211504798696203e-05, "loss": 1.1964, "step": 10650 }, { "epoch": 0.64, "learning_rate": 3.212108408281524e-05, "loss": 1.0501, "step": 10652 }, { "epoch": 0.64, "learning_rate": 3.212712017866844e-05, "loss": 1.5328, "step": 10654 }, { "epoch": 0.64, "learning_rate": 3.213315627452164e-05, "loss": 1.4901, "step": 10656 }, { "epoch": 0.64, "learning_rate": 3.213919237037484e-05, "loss": 0.9373, "step": 10658 }, { "epoch": 0.64, "learning_rate": 3.2145228466228044e-05, "loss": 1.4722, "step": 10660 }, { "epoch": 0.64, "learning_rate": 3.2151264562081245e-05, "loss": 1.0016, "step": 10662 }, { "epoch": 0.64, "learning_rate": 3.2157300657934446e-05, "loss": 1.3027, "step": 10664 }, { "epoch": 0.64, "learning_rate": 3.216333675378765e-05, "loss": 1.3274, "step": 10666 }, { "epoch": 0.64, "learning_rate": 3.2169372849640854e-05, "loss": 1.3685, "step": 10668 }, { "epoch": 0.64, "learning_rate": 3.2175408945494055e-05, "loss": 0.9806, "step": 10670 }, { "epoch": 0.64, "learning_rate": 3.2181445041347256e-05, "loss": 1.3202, "step": 10672 }, { "epoch": 0.64, "learning_rate": 3.2187481137200456e-05, "loss": 1.1436, "step": 10674 }, { "epoch": 0.64, "learning_rate": 3.219351723305366e-05, "loss": 0.9951, "step": 10676 }, { "epoch": 0.64, "learning_rate": 3.2199553328906865e-05, "loss": 1.0921, "step": 10678 }, { "epoch": 0.64, "learning_rate": 3.2205589424760066e-05, "loss": 1.1107, "step": 10680 }, { "epoch": 0.64, "learning_rate": 3.221162552061327e-05, "loss": 1.4508, "step": 10682 }, { "epoch": 0.64, "learning_rate": 3.2217661616466474e-05, "loss": 1.4209, "step": 10684 }, { "epoch": 0.64, "learning_rate": 3.2223697712319675e-05, "loss": 1.0877, "step": 10686 }, { "epoch": 0.65, "learning_rate": 3.2229733808172876e-05, "loss": 1.2288, "step": 10688 }, { "epoch": 0.65, "learning_rate": 3.2235769904026077e-05, "loss": 0.9374, "step": 10690 }, { "epoch": 0.65, "learning_rate": 3.2241805999879284e-05, "loss": 1.3076, "step": 10692 }, { "epoch": 0.65, "learning_rate": 3.2247842095732485e-05, "loss": 0.9977, "step": 10694 }, { "epoch": 0.65, "learning_rate": 3.2253878191585686e-05, "loss": 1.0693, "step": 10696 }, { "epoch": 0.65, "learning_rate": 3.225991428743889e-05, "loss": 1.1345, "step": 10698 }, { "epoch": 0.65, "learning_rate": 3.226595038329209e-05, "loss": 1.4163, "step": 10700 }, { "epoch": 0.65, "learning_rate": 3.227198647914529e-05, "loss": 1.2307, "step": 10702 }, { "epoch": 0.65, "learning_rate": 3.227802257499849e-05, "loss": 1.6453, "step": 10704 }, { "epoch": 0.65, "learning_rate": 3.22840586708517e-05, "loss": 1.301, "step": 10706 }, { "epoch": 0.65, "learning_rate": 3.22900947667049e-05, "loss": 1.1545, "step": 10708 }, { "epoch": 0.65, "learning_rate": 3.22961308625581e-05, "loss": 1.4502, "step": 10710 }, { "epoch": 0.65, "learning_rate": 3.23021669584113e-05, "loss": 1.0049, "step": 10712 }, { "epoch": 0.65, "learning_rate": 3.23082030542645e-05, "loss": 1.2852, "step": 10714 }, { "epoch": 0.65, "learning_rate": 3.23142391501177e-05, "loss": 1.5061, "step": 10716 }, { "epoch": 0.65, "learning_rate": 3.232027524597091e-05, "loss": 0.9474, "step": 10718 }, { "epoch": 0.65, "learning_rate": 3.232631134182411e-05, "loss": 1.1927, "step": 10720 }, { "epoch": 0.65, "learning_rate": 3.233234743767731e-05, "loss": 0.9377, "step": 10722 }, { "epoch": 0.65, "learning_rate": 3.233838353353051e-05, "loss": 1.0393, "step": 10724 }, { "epoch": 0.65, "learning_rate": 3.234441962938372e-05, "loss": 1.6117, "step": 10726 }, { "epoch": 0.65, "learning_rate": 3.235045572523692e-05, "loss": 1.6028, "step": 10728 }, { "epoch": 0.65, "learning_rate": 3.235649182109012e-05, "loss": 1.422, "step": 10730 }, { "epoch": 0.65, "learning_rate": 3.236252791694333e-05, "loss": 1.4125, "step": 10732 }, { "epoch": 0.65, "learning_rate": 3.236856401279653e-05, "loss": 1.2072, "step": 10734 }, { "epoch": 0.65, "learning_rate": 3.237460010864973e-05, "loss": 1.2128, "step": 10736 }, { "epoch": 0.65, "learning_rate": 3.238063620450293e-05, "loss": 1.5416, "step": 10738 }, { "epoch": 0.65, "learning_rate": 3.238667230035613e-05, "loss": 1.4443, "step": 10740 }, { "epoch": 0.65, "learning_rate": 3.239270839620933e-05, "loss": 1.2571, "step": 10742 }, { "epoch": 0.65, "learning_rate": 3.239874449206254e-05, "loss": 1.1333, "step": 10744 }, { "epoch": 0.65, "learning_rate": 3.240478058791574e-05, "loss": 1.6612, "step": 10746 }, { "epoch": 0.65, "learning_rate": 3.241081668376894e-05, "loss": 1.1127, "step": 10748 }, { "epoch": 0.65, "learning_rate": 3.241685277962214e-05, "loss": 1.1603, "step": 10750 }, { "epoch": 0.65, "learning_rate": 3.242288887547534e-05, "loss": 1.2549, "step": 10752 }, { "epoch": 0.65, "learning_rate": 3.2428924971328543e-05, "loss": 1.1632, "step": 10754 }, { "epoch": 0.65, "learning_rate": 3.2434961067181744e-05, "loss": 1.0479, "step": 10756 }, { "epoch": 0.65, "learning_rate": 3.244099716303495e-05, "loss": 1.3478, "step": 10758 }, { "epoch": 0.65, "learning_rate": 3.244703325888815e-05, "loss": 1.1799, "step": 10760 }, { "epoch": 0.65, "learning_rate": 3.2453069354741354e-05, "loss": 1.6169, "step": 10762 }, { "epoch": 0.65, "learning_rate": 3.2459105450594554e-05, "loss": 1.4788, "step": 10764 }, { "epoch": 0.65, "learning_rate": 3.2465141546447755e-05, "loss": 1.6336, "step": 10766 }, { "epoch": 0.65, "learning_rate": 3.247117764230096e-05, "loss": 1.5856, "step": 10768 }, { "epoch": 0.65, "learning_rate": 3.2477213738154164e-05, "loss": 1.1892, "step": 10770 }, { "epoch": 0.65, "learning_rate": 3.2483249834007364e-05, "loss": 1.4035, "step": 10772 }, { "epoch": 0.65, "learning_rate": 3.248928592986057e-05, "loss": 1.1924, "step": 10774 }, { "epoch": 0.65, "learning_rate": 3.249532202571377e-05, "loss": 1.2549, "step": 10776 }, { "epoch": 0.65, "learning_rate": 3.2501358121566974e-05, "loss": 0.8289, "step": 10778 }, { "epoch": 0.65, "learning_rate": 3.2507394217420174e-05, "loss": 1.0736, "step": 10780 }, { "epoch": 0.65, "learning_rate": 3.2513430313273375e-05, "loss": 1.0092, "step": 10782 }, { "epoch": 0.65, "learning_rate": 3.251946640912658e-05, "loss": 1.1637, "step": 10784 }, { "epoch": 0.65, "learning_rate": 3.2525502504979784e-05, "loss": 1.25, "step": 10786 }, { "epoch": 0.65, "learning_rate": 3.2531538600832984e-05, "loss": 1.1255, "step": 10788 }, { "epoch": 0.65, "learning_rate": 3.2537574696686185e-05, "loss": 1.2586, "step": 10790 }, { "epoch": 0.65, "learning_rate": 3.2543610792539386e-05, "loss": 1.0137, "step": 10792 }, { "epoch": 0.65, "learning_rate": 3.254964688839259e-05, "loss": 1.7199, "step": 10794 }, { "epoch": 0.65, "learning_rate": 3.255568298424579e-05, "loss": 1.3551, "step": 10796 }, { "epoch": 0.65, "learning_rate": 3.2561719080098995e-05, "loss": 1.3855, "step": 10798 }, { "epoch": 0.65, "learning_rate": 3.2567755175952196e-05, "loss": 1.2935, "step": 10800 }, { "epoch": 0.65, "learning_rate": 3.25737912718054e-05, "loss": 1.4283, "step": 10802 }, { "epoch": 0.65, "learning_rate": 3.25798273676586e-05, "loss": 1.5022, "step": 10804 }, { "epoch": 0.65, "learning_rate": 3.25858634635118e-05, "loss": 1.1021, "step": 10806 }, { "epoch": 0.65, "learning_rate": 3.2591899559365e-05, "loss": 1.3604, "step": 10808 }, { "epoch": 0.65, "learning_rate": 3.259793565521821e-05, "loss": 1.3233, "step": 10810 }, { "epoch": 0.65, "learning_rate": 3.260397175107141e-05, "loss": 1.1112, "step": 10812 }, { "epoch": 0.65, "learning_rate": 3.261000784692461e-05, "loss": 1.2656, "step": 10814 }, { "epoch": 0.65, "learning_rate": 3.2616043942777816e-05, "loss": 1.0544, "step": 10816 }, { "epoch": 0.65, "learning_rate": 3.262208003863102e-05, "loss": 1.2327, "step": 10818 }, { "epoch": 0.65, "learning_rate": 3.262811613448422e-05, "loss": 1.4585, "step": 10820 }, { "epoch": 0.65, "learning_rate": 3.263415223033742e-05, "loss": 1.0727, "step": 10822 }, { "epoch": 0.65, "learning_rate": 3.2640188326190626e-05, "loss": 1.3171, "step": 10824 }, { "epoch": 0.65, "learning_rate": 3.264622442204383e-05, "loss": 0.7981, "step": 10826 }, { "epoch": 0.65, "learning_rate": 3.265226051789703e-05, "loss": 1.2583, "step": 10828 }, { "epoch": 0.65, "learning_rate": 3.265829661375023e-05, "loss": 1.0695, "step": 10830 }, { "epoch": 0.65, "learning_rate": 3.266433270960343e-05, "loss": 1.2649, "step": 10832 }, { "epoch": 0.65, "learning_rate": 3.267036880545663e-05, "loss": 0.9776, "step": 10834 }, { "epoch": 0.65, "learning_rate": 3.267640490130983e-05, "loss": 1.2991, "step": 10836 }, { "epoch": 0.65, "learning_rate": 3.268244099716304e-05, "loss": 0.7398, "step": 10838 }, { "epoch": 0.65, "learning_rate": 3.268847709301624e-05, "loss": 1.5376, "step": 10840 }, { "epoch": 0.65, "learning_rate": 3.269451318886944e-05, "loss": 1.2992, "step": 10842 }, { "epoch": 0.65, "learning_rate": 3.270054928472264e-05, "loss": 1.5633, "step": 10844 }, { "epoch": 0.65, "learning_rate": 3.270658538057584e-05, "loss": 1.2116, "step": 10846 }, { "epoch": 0.65, "learning_rate": 3.271262147642904e-05, "loss": 1.3163, "step": 10848 }, { "epoch": 0.65, "learning_rate": 3.271865757228225e-05, "loss": 1.5675, "step": 10850 }, { "epoch": 0.65, "learning_rate": 3.272469366813545e-05, "loss": 1.2293, "step": 10852 }, { "epoch": 0.66, "learning_rate": 3.273072976398865e-05, "loss": 1.3485, "step": 10854 }, { "epoch": 0.66, "learning_rate": 3.273676585984185e-05, "loss": 1.2528, "step": 10856 }, { "epoch": 0.66, "learning_rate": 3.2742801955695054e-05, "loss": 1.107, "step": 10858 }, { "epoch": 0.66, "learning_rate": 3.274883805154826e-05, "loss": 1.2196, "step": 10860 }, { "epoch": 0.66, "learning_rate": 3.275487414740146e-05, "loss": 1.1637, "step": 10862 }, { "epoch": 0.66, "learning_rate": 3.276091024325467e-05, "loss": 1.4819, "step": 10864 }, { "epoch": 0.66, "learning_rate": 3.276694633910787e-05, "loss": 1.5623, "step": 10866 }, { "epoch": 0.66, "learning_rate": 3.277298243496107e-05, "loss": 1.483, "step": 10868 }, { "epoch": 0.66, "learning_rate": 3.277901853081427e-05, "loss": 1.4773, "step": 10870 }, { "epoch": 0.66, "learning_rate": 3.278505462666747e-05, "loss": 1.171, "step": 10872 }, { "epoch": 0.66, "learning_rate": 3.2791090722520674e-05, "loss": 1.3205, "step": 10874 }, { "epoch": 0.66, "learning_rate": 3.279712681837388e-05, "loss": 1.134, "step": 10876 }, { "epoch": 0.66, "learning_rate": 3.280316291422708e-05, "loss": 1.3863, "step": 10878 }, { "epoch": 0.66, "learning_rate": 3.280919901008028e-05, "loss": 1.4575, "step": 10880 }, { "epoch": 0.66, "learning_rate": 3.2815235105933484e-05, "loss": 1.1492, "step": 10882 }, { "epoch": 0.66, "learning_rate": 3.2821271201786685e-05, "loss": 1.2748, "step": 10884 }, { "epoch": 0.66, "learning_rate": 3.2827307297639886e-05, "loss": 1.152, "step": 10886 }, { "epoch": 0.66, "learning_rate": 3.2833343393493086e-05, "loss": 1.3518, "step": 10888 }, { "epoch": 0.66, "learning_rate": 3.2839379489346294e-05, "loss": 1.5715, "step": 10890 }, { "epoch": 0.66, "learning_rate": 3.2845415585199495e-05, "loss": 1.2359, "step": 10892 }, { "epoch": 0.66, "learning_rate": 3.2851451681052696e-05, "loss": 1.323, "step": 10894 }, { "epoch": 0.66, "learning_rate": 3.2857487776905896e-05, "loss": 1.3662, "step": 10896 }, { "epoch": 0.66, "learning_rate": 3.28635238727591e-05, "loss": 1.2257, "step": 10898 }, { "epoch": 0.66, "learning_rate": 3.28695599686123e-05, "loss": 1.6144, "step": 10900 }, { "epoch": 0.66, "learning_rate": 3.2875596064465506e-05, "loss": 0.8497, "step": 10902 }, { "epoch": 0.66, "learning_rate": 3.2881632160318706e-05, "loss": 1.3723, "step": 10904 }, { "epoch": 0.66, "learning_rate": 3.2887668256171914e-05, "loss": 1.2011, "step": 10906 }, { "epoch": 0.66, "learning_rate": 3.2893704352025115e-05, "loss": 1.3718, "step": 10908 }, { "epoch": 0.66, "learning_rate": 3.2899740447878316e-05, "loss": 1.2992, "step": 10910 }, { "epoch": 0.66, "learning_rate": 3.2905776543731517e-05, "loss": 1.3063, "step": 10912 }, { "epoch": 0.66, "learning_rate": 3.291181263958472e-05, "loss": 1.3003, "step": 10914 }, { "epoch": 0.66, "learning_rate": 3.2917848735437925e-05, "loss": 1.2601, "step": 10916 }, { "epoch": 0.66, "learning_rate": 3.2923884831291126e-05, "loss": 1.4477, "step": 10918 }, { "epoch": 0.66, "learning_rate": 3.2929920927144327e-05, "loss": 1.3619, "step": 10920 }, { "epoch": 0.66, "learning_rate": 3.293595702299753e-05, "loss": 1.2368, "step": 10922 }, { "epoch": 0.66, "learning_rate": 3.294199311885073e-05, "loss": 1.3781, "step": 10924 }, { "epoch": 0.66, "learning_rate": 3.294802921470393e-05, "loss": 1.1005, "step": 10926 }, { "epoch": 0.66, "learning_rate": 3.295406531055713e-05, "loss": 1.346, "step": 10928 }, { "epoch": 0.66, "learning_rate": 3.296010140641034e-05, "loss": 1.3191, "step": 10930 }, { "epoch": 0.66, "learning_rate": 3.296613750226354e-05, "loss": 1.5081, "step": 10932 }, { "epoch": 0.66, "learning_rate": 3.297217359811674e-05, "loss": 1.1189, "step": 10934 }, { "epoch": 0.66, "learning_rate": 3.297820969396994e-05, "loss": 1.231, "step": 10936 }, { "epoch": 0.66, "learning_rate": 3.298424578982314e-05, "loss": 1.6189, "step": 10938 }, { "epoch": 0.66, "learning_rate": 3.299028188567634e-05, "loss": 1.0548, "step": 10940 }, { "epoch": 0.66, "learning_rate": 3.299631798152955e-05, "loss": 1.0313, "step": 10942 }, { "epoch": 0.66, "learning_rate": 3.300235407738275e-05, "loss": 1.3655, "step": 10944 }, { "epoch": 0.66, "learning_rate": 3.300839017323595e-05, "loss": 1.1141, "step": 10946 }, { "epoch": 0.66, "learning_rate": 3.301442626908915e-05, "loss": 1.1631, "step": 10948 }, { "epoch": 0.66, "learning_rate": 3.302046236494236e-05, "loss": 1.0838, "step": 10950 }, { "epoch": 0.66, "learning_rate": 3.302649846079556e-05, "loss": 1.1868, "step": 10952 }, { "epoch": 0.66, "learning_rate": 3.303253455664876e-05, "loss": 1.3338, "step": 10954 }, { "epoch": 0.66, "learning_rate": 3.303857065250197e-05, "loss": 1.5821, "step": 10956 }, { "epoch": 0.66, "learning_rate": 3.304460674835517e-05, "loss": 1.2434, "step": 10958 }, { "epoch": 0.66, "learning_rate": 3.305064284420837e-05, "loss": 1.2504, "step": 10960 }, { "epoch": 0.66, "learning_rate": 3.305667894006157e-05, "loss": 1.4405, "step": 10962 }, { "epoch": 0.66, "learning_rate": 3.306271503591477e-05, "loss": 1.169, "step": 10964 }, { "epoch": 0.66, "learning_rate": 3.306875113176797e-05, "loss": 1.1167, "step": 10966 }, { "epoch": 0.66, "learning_rate": 3.307478722762118e-05, "loss": 1.1804, "step": 10968 }, { "epoch": 0.66, "learning_rate": 3.308082332347438e-05, "loss": 1.2231, "step": 10970 }, { "epoch": 0.66, "learning_rate": 3.308685941932758e-05, "loss": 1.5021, "step": 10972 }, { "epoch": 0.66, "learning_rate": 3.309289551518078e-05, "loss": 1.0234, "step": 10974 }, { "epoch": 0.66, "learning_rate": 3.3098931611033983e-05, "loss": 1.2324, "step": 10976 }, { "epoch": 0.66, "learning_rate": 3.3104967706887184e-05, "loss": 1.1959, "step": 10978 }, { "epoch": 0.66, "learning_rate": 3.3111003802740385e-05, "loss": 1.0398, "step": 10980 }, { "epoch": 0.66, "learning_rate": 3.311703989859359e-05, "loss": 1.2737, "step": 10982 }, { "epoch": 0.66, "learning_rate": 3.3123075994446793e-05, "loss": 1.7234, "step": 10984 }, { "epoch": 0.66, "learning_rate": 3.3129112090299994e-05, "loss": 1.1259, "step": 10986 }, { "epoch": 0.66, "learning_rate": 3.3135148186153195e-05, "loss": 1.4394, "step": 10988 }, { "epoch": 0.66, "learning_rate": 3.3141184282006396e-05, "loss": 0.9329, "step": 10990 }, { "epoch": 0.66, "learning_rate": 3.3147220377859603e-05, "loss": 1.3701, "step": 10992 }, { "epoch": 0.66, "learning_rate": 3.3153256473712804e-05, "loss": 1.2073, "step": 10994 }, { "epoch": 0.66, "learning_rate": 3.3159292569566005e-05, "loss": 1.1653, "step": 10996 }, { "epoch": 0.66, "learning_rate": 3.316532866541921e-05, "loss": 1.1751, "step": 10998 }, { "epoch": 0.66, "learning_rate": 3.3171364761272414e-05, "loss": 1.297, "step": 11000 }, { "epoch": 0.66, "learning_rate": 3.3177400857125614e-05, "loss": 1.092, "step": 11002 }, { "epoch": 0.66, "learning_rate": 3.3183436952978815e-05, "loss": 1.177, "step": 11004 }, { "epoch": 0.66, "learning_rate": 3.3189473048832016e-05, "loss": 0.9467, "step": 11006 }, { "epoch": 0.66, "learning_rate": 3.3195509144685224e-05, "loss": 1.1848, "step": 11008 }, { "epoch": 0.66, "learning_rate": 3.3201545240538424e-05, "loss": 1.6589, "step": 11010 }, { "epoch": 0.66, "learning_rate": 3.3207581336391625e-05, "loss": 1.3052, "step": 11012 }, { "epoch": 0.66, "learning_rate": 3.3213617432244826e-05, "loss": 1.1312, "step": 11014 }, { "epoch": 0.66, "learning_rate": 3.321965352809803e-05, "loss": 0.9817, "step": 11016 }, { "epoch": 0.67, "learning_rate": 3.322568962395123e-05, "loss": 1.2175, "step": 11018 }, { "epoch": 0.67, "learning_rate": 3.323172571980443e-05, "loss": 1.4561, "step": 11020 }, { "epoch": 0.67, "learning_rate": 3.3237761815657636e-05, "loss": 1.2656, "step": 11022 }, { "epoch": 0.67, "learning_rate": 3.324379791151084e-05, "loss": 1.2446, "step": 11024 }, { "epoch": 0.67, "learning_rate": 3.324983400736404e-05, "loss": 1.4974, "step": 11026 }, { "epoch": 0.67, "learning_rate": 3.325587010321724e-05, "loss": 0.9855, "step": 11028 }, { "epoch": 0.67, "learning_rate": 3.326190619907044e-05, "loss": 1.1333, "step": 11030 }, { "epoch": 0.67, "learning_rate": 3.326794229492364e-05, "loss": 1.3801, "step": 11032 }, { "epoch": 0.67, "learning_rate": 3.327397839077685e-05, "loss": 0.9388, "step": 11034 }, { "epoch": 0.67, "learning_rate": 3.328001448663005e-05, "loss": 1.1901, "step": 11036 }, { "epoch": 0.67, "learning_rate": 3.328605058248325e-05, "loss": 1.4213, "step": 11038 }, { "epoch": 0.67, "learning_rate": 3.329208667833646e-05, "loss": 1.2505, "step": 11040 }, { "epoch": 0.67, "learning_rate": 3.329812277418966e-05, "loss": 1.1399, "step": 11042 }, { "epoch": 0.67, "learning_rate": 3.330415887004286e-05, "loss": 0.9044, "step": 11044 }, { "epoch": 0.67, "learning_rate": 3.331019496589606e-05, "loss": 1.2802, "step": 11046 }, { "epoch": 0.67, "learning_rate": 3.331623106174927e-05, "loss": 1.2785, "step": 11048 }, { "epoch": 0.67, "learning_rate": 3.332226715760247e-05, "loss": 1.3757, "step": 11050 }, { "epoch": 0.67, "learning_rate": 3.332830325345567e-05, "loss": 1.1583, "step": 11052 }, { "epoch": 0.67, "learning_rate": 3.333433934930887e-05, "loss": 1.1368, "step": 11054 }, { "epoch": 0.67, "learning_rate": 3.334037544516207e-05, "loss": 1.1699, "step": 11056 }, { "epoch": 0.67, "learning_rate": 3.334641154101527e-05, "loss": 1.2864, "step": 11058 }, { "epoch": 0.67, "learning_rate": 3.335244763686847e-05, "loss": 1.1983, "step": 11060 }, { "epoch": 0.67, "learning_rate": 3.335848373272168e-05, "loss": 1.207, "step": 11062 }, { "epoch": 0.67, "learning_rate": 3.336451982857488e-05, "loss": 1.4138, "step": 11064 }, { "epoch": 0.67, "learning_rate": 3.337055592442808e-05, "loss": 1.4917, "step": 11066 }, { "epoch": 0.67, "learning_rate": 3.337659202028128e-05, "loss": 1.1605, "step": 11068 }, { "epoch": 0.67, "learning_rate": 3.338262811613448e-05, "loss": 0.9336, "step": 11070 }, { "epoch": 0.67, "learning_rate": 3.3388664211987684e-05, "loss": 1.1105, "step": 11072 }, { "epoch": 0.67, "learning_rate": 3.339470030784089e-05, "loss": 1.1937, "step": 11074 }, { "epoch": 0.67, "learning_rate": 3.340073640369409e-05, "loss": 1.1661, "step": 11076 }, { "epoch": 0.67, "learning_rate": 3.340677249954729e-05, "loss": 1.1976, "step": 11078 }, { "epoch": 0.67, "learning_rate": 3.3412808595400494e-05, "loss": 1.2997, "step": 11080 }, { "epoch": 0.67, "learning_rate": 3.3418844691253695e-05, "loss": 1.3446, "step": 11082 }, { "epoch": 0.67, "learning_rate": 3.34248807871069e-05, "loss": 1.3129, "step": 11084 }, { "epoch": 0.67, "learning_rate": 3.34309168829601e-05, "loss": 1.1219, "step": 11086 }, { "epoch": 0.67, "learning_rate": 3.343695297881331e-05, "loss": 1.2426, "step": 11088 }, { "epoch": 0.67, "learning_rate": 3.344298907466651e-05, "loss": 0.6249, "step": 11090 }, { "epoch": 0.67, "learning_rate": 3.344902517051971e-05, "loss": 1.0271, "step": 11092 }, { "epoch": 0.67, "learning_rate": 3.345506126637291e-05, "loss": 1.4529, "step": 11094 }, { "epoch": 0.67, "learning_rate": 3.3461097362226114e-05, "loss": 1.3403, "step": 11096 }, { "epoch": 0.67, "learning_rate": 3.3467133458079315e-05, "loss": 1.3584, "step": 11098 }, { "epoch": 0.67, "learning_rate": 3.347316955393252e-05, "loss": 1.4582, "step": 11100 }, { "epoch": 0.67, "learning_rate": 3.347920564978572e-05, "loss": 1.3096, "step": 11102 }, { "epoch": 0.67, "learning_rate": 3.3485241745638924e-05, "loss": 1.281, "step": 11104 }, { "epoch": 0.67, "learning_rate": 3.3491277841492125e-05, "loss": 0.9157, "step": 11106 }, { "epoch": 0.67, "learning_rate": 3.3497313937345326e-05, "loss": 1.2277, "step": 11108 }, { "epoch": 0.67, "learning_rate": 3.3503350033198526e-05, "loss": 1.395, "step": 11110 }, { "epoch": 0.67, "learning_rate": 3.350938612905173e-05, "loss": 1.3703, "step": 11112 }, { "epoch": 0.67, "learning_rate": 3.3515422224904935e-05, "loss": 1.2435, "step": 11114 }, { "epoch": 0.67, "learning_rate": 3.3521458320758136e-05, "loss": 1.268, "step": 11116 }, { "epoch": 0.67, "learning_rate": 3.3527494416611336e-05, "loss": 1.4155, "step": 11118 }, { "epoch": 0.67, "learning_rate": 3.353353051246454e-05, "loss": 1.4757, "step": 11120 }, { "epoch": 0.67, "learning_rate": 3.353956660831774e-05, "loss": 1.1927, "step": 11122 }, { "epoch": 0.67, "learning_rate": 3.354560270417094e-05, "loss": 1.5664, "step": 11124 }, { "epoch": 0.67, "learning_rate": 3.3551638800024146e-05, "loss": 1.3521, "step": 11126 }, { "epoch": 0.67, "learning_rate": 3.355767489587735e-05, "loss": 1.2932, "step": 11128 }, { "epoch": 0.67, "learning_rate": 3.356371099173055e-05, "loss": 1.5329, "step": 11130 }, { "epoch": 0.67, "learning_rate": 3.3569747087583756e-05, "loss": 1.436, "step": 11132 }, { "epoch": 0.67, "learning_rate": 3.3575783183436956e-05, "loss": 1.2496, "step": 11134 }, { "epoch": 0.67, "learning_rate": 3.358181927929016e-05, "loss": 1.5465, "step": 11136 }, { "epoch": 0.67, "learning_rate": 3.358785537514336e-05, "loss": 1.3644, "step": 11138 }, { "epoch": 0.67, "learning_rate": 3.3593891470996566e-05, "loss": 1.2782, "step": 11140 }, { "epoch": 0.67, "learning_rate": 3.3599927566849767e-05, "loss": 1.1052, "step": 11142 }, { "epoch": 0.67, "learning_rate": 3.360596366270297e-05, "loss": 1.4646, "step": 11144 }, { "epoch": 0.67, "learning_rate": 3.361199975855617e-05, "loss": 1.2558, "step": 11146 }, { "epoch": 0.67, "learning_rate": 3.361803585440937e-05, "loss": 1.5202, "step": 11148 }, { "epoch": 0.67, "learning_rate": 3.362407195026257e-05, "loss": 1.0464, "step": 11150 }, { "epoch": 0.67, "learning_rate": 3.363010804611577e-05, "loss": 1.2941, "step": 11152 }, { "epoch": 0.67, "learning_rate": 3.363614414196898e-05, "loss": 0.9315, "step": 11154 }, { "epoch": 0.67, "learning_rate": 3.364218023782218e-05, "loss": 1.4848, "step": 11156 }, { "epoch": 0.67, "learning_rate": 3.364821633367538e-05, "loss": 1.1067, "step": 11158 }, { "epoch": 0.67, "learning_rate": 3.365425242952858e-05, "loss": 1.5144, "step": 11160 }, { "epoch": 0.67, "learning_rate": 3.366028852538178e-05, "loss": 1.2259, "step": 11162 }, { "epoch": 0.67, "learning_rate": 3.366632462123498e-05, "loss": 1.2743, "step": 11164 }, { "epoch": 0.67, "learning_rate": 3.367236071708819e-05, "loss": 1.1084, "step": 11166 }, { "epoch": 0.67, "learning_rate": 3.367839681294139e-05, "loss": 1.4242, "step": 11168 }, { "epoch": 0.67, "learning_rate": 3.368443290879459e-05, "loss": 1.308, "step": 11170 }, { "epoch": 0.67, "learning_rate": 3.369046900464779e-05, "loss": 1.0573, "step": 11172 }, { "epoch": 0.67, "learning_rate": 3.3696505100501e-05, "loss": 0.9465, "step": 11174 }, { "epoch": 0.67, "learning_rate": 3.37025411963542e-05, "loss": 1.2203, "step": 11176 }, { "epoch": 0.67, "learning_rate": 3.37085772922074e-05, "loss": 1.3402, "step": 11178 }, { "epoch": 0.67, "learning_rate": 3.371461338806061e-05, "loss": 1.493, "step": 11180 }, { "epoch": 0.67, "learning_rate": 3.372064948391381e-05, "loss": 1.7573, "step": 11182 }, { "epoch": 0.68, "learning_rate": 3.372668557976701e-05, "loss": 1.3641, "step": 11184 }, { "epoch": 0.68, "learning_rate": 3.373272167562021e-05, "loss": 1.1296, "step": 11186 }, { "epoch": 0.68, "learning_rate": 3.373875777147341e-05, "loss": 1.3092, "step": 11188 }, { "epoch": 0.68, "learning_rate": 3.374479386732661e-05, "loss": 1.0176, "step": 11190 }, { "epoch": 0.68, "learning_rate": 3.3750829963179814e-05, "loss": 1.4362, "step": 11192 }, { "epoch": 0.68, "learning_rate": 3.375686605903302e-05, "loss": 1.1204, "step": 11194 }, { "epoch": 0.68, "learning_rate": 3.376290215488622e-05, "loss": 1.0339, "step": 11196 }, { "epoch": 0.68, "learning_rate": 3.376893825073942e-05, "loss": 1.3072, "step": 11198 }, { "epoch": 0.68, "learning_rate": 3.3774974346592624e-05, "loss": 1.361, "step": 11200 }, { "epoch": 0.68, "learning_rate": 3.3781010442445825e-05, "loss": 1.0324, "step": 11202 }, { "epoch": 0.68, "learning_rate": 3.3787046538299026e-05, "loss": 1.2939, "step": 11204 }, { "epoch": 0.68, "learning_rate": 3.379308263415223e-05, "loss": 1.0811, "step": 11206 }, { "epoch": 0.68, "learning_rate": 3.3799118730005434e-05, "loss": 1.2522, "step": 11208 }, { "epoch": 0.68, "learning_rate": 3.3805154825858635e-05, "loss": 1.3585, "step": 11210 }, { "epoch": 0.68, "learning_rate": 3.3811190921711836e-05, "loss": 1.2969, "step": 11212 }, { "epoch": 0.68, "learning_rate": 3.381722701756504e-05, "loss": 1.5364, "step": 11214 }, { "epoch": 0.68, "learning_rate": 3.3823263113418244e-05, "loss": 1.4153, "step": 11216 }, { "epoch": 0.68, "learning_rate": 3.3829299209271445e-05, "loss": 1.2336, "step": 11218 }, { "epoch": 0.68, "learning_rate": 3.3835335305124646e-05, "loss": 1.0873, "step": 11220 }, { "epoch": 0.68, "learning_rate": 3.3841371400977853e-05, "loss": 1.1905, "step": 11222 }, { "epoch": 0.68, "learning_rate": 3.3847407496831054e-05, "loss": 1.0709, "step": 11224 }, { "epoch": 0.68, "learning_rate": 3.3853443592684255e-05, "loss": 1.309, "step": 11226 }, { "epoch": 0.68, "learning_rate": 3.3859479688537456e-05, "loss": 1.5071, "step": 11228 }, { "epoch": 0.68, "learning_rate": 3.386551578439066e-05, "loss": 1.1954, "step": 11230 }, { "epoch": 0.68, "learning_rate": 3.3871551880243864e-05, "loss": 1.3918, "step": 11232 }, { "epoch": 0.68, "learning_rate": 3.3877587976097065e-05, "loss": 1.3368, "step": 11234 }, { "epoch": 0.68, "learning_rate": 3.3883624071950266e-05, "loss": 1.1568, "step": 11236 }, { "epoch": 0.68, "learning_rate": 3.388966016780347e-05, "loss": 1.0097, "step": 11238 }, { "epoch": 0.68, "learning_rate": 3.389569626365667e-05, "loss": 1.5288, "step": 11240 }, { "epoch": 0.68, "learning_rate": 3.390173235950987e-05, "loss": 1.425, "step": 11242 }, { "epoch": 0.68, "learning_rate": 3.390776845536307e-05, "loss": 1.1336, "step": 11244 }, { "epoch": 0.68, "learning_rate": 3.391380455121628e-05, "loss": 1.1582, "step": 11246 }, { "epoch": 0.68, "learning_rate": 3.391984064706948e-05, "loss": 1.1265, "step": 11248 }, { "epoch": 0.68, "learning_rate": 3.392587674292268e-05, "loss": 1.3481, "step": 11250 }, { "epoch": 0.68, "learning_rate": 3.393191283877588e-05, "loss": 1.4281, "step": 11252 }, { "epoch": 0.68, "learning_rate": 3.393794893462908e-05, "loss": 1.3745, "step": 11254 }, { "epoch": 0.68, "learning_rate": 3.394398503048228e-05, "loss": 1.1844, "step": 11256 }, { "epoch": 0.68, "learning_rate": 3.395002112633549e-05, "loss": 0.9945, "step": 11258 }, { "epoch": 0.68, "learning_rate": 3.395605722218869e-05, "loss": 1.0442, "step": 11260 }, { "epoch": 0.68, "learning_rate": 3.396209331804189e-05, "loss": 1.0712, "step": 11262 }, { "epoch": 0.68, "learning_rate": 3.39681294138951e-05, "loss": 1.085, "step": 11264 }, { "epoch": 0.68, "learning_rate": 3.39741655097483e-05, "loss": 1.3398, "step": 11266 }, { "epoch": 0.68, "learning_rate": 3.39802016056015e-05, "loss": 0.8819, "step": 11268 }, { "epoch": 0.68, "learning_rate": 3.39862377014547e-05, "loss": 1.4649, "step": 11270 }, { "epoch": 0.68, "learning_rate": 3.399227379730791e-05, "loss": 1.1804, "step": 11272 }, { "epoch": 0.68, "learning_rate": 3.399830989316111e-05, "loss": 1.1599, "step": 11274 }, { "epoch": 0.68, "learning_rate": 3.400434598901431e-05, "loss": 1.4223, "step": 11276 }, { "epoch": 0.68, "learning_rate": 3.401038208486751e-05, "loss": 1.2278, "step": 11278 }, { "epoch": 0.68, "learning_rate": 3.401641818072071e-05, "loss": 1.072, "step": 11280 }, { "epoch": 0.68, "learning_rate": 3.402245427657391e-05, "loss": 1.1114, "step": 11282 }, { "epoch": 0.68, "learning_rate": 3.402849037242711e-05, "loss": 1.052, "step": 11284 }, { "epoch": 0.68, "learning_rate": 3.403452646828032e-05, "loss": 1.0585, "step": 11286 }, { "epoch": 0.68, "learning_rate": 3.404056256413352e-05, "loss": 1.1935, "step": 11288 }, { "epoch": 0.68, "learning_rate": 3.404659865998672e-05, "loss": 1.0503, "step": 11290 }, { "epoch": 0.68, "learning_rate": 3.405263475583992e-05, "loss": 1.2323, "step": 11292 }, { "epoch": 0.68, "learning_rate": 3.4058670851693124e-05, "loss": 1.384, "step": 11294 }, { "epoch": 0.68, "learning_rate": 3.4064706947546324e-05, "loss": 1.1598, "step": 11296 }, { "epoch": 0.68, "learning_rate": 3.407074304339953e-05, "loss": 1.0029, "step": 11298 }, { "epoch": 0.68, "learning_rate": 3.407677913925273e-05, "loss": 1.6223, "step": 11300 }, { "epoch": 0.68, "learning_rate": 3.4082815235105934e-05, "loss": 0.9347, "step": 11302 }, { "epoch": 0.68, "learning_rate": 3.4088851330959134e-05, "loss": 0.8229, "step": 11304 }, { "epoch": 0.68, "learning_rate": 3.4094887426812335e-05, "loss": 1.2523, "step": 11306 }, { "epoch": 0.68, "learning_rate": 3.410092352266554e-05, "loss": 1.6258, "step": 11308 }, { "epoch": 0.68, "learning_rate": 3.4106959618518744e-05, "loss": 1.5343, "step": 11310 }, { "epoch": 0.68, "learning_rate": 3.411299571437195e-05, "loss": 0.9443, "step": 11312 }, { "epoch": 0.68, "learning_rate": 3.411903181022515e-05, "loss": 1.0712, "step": 11314 }, { "epoch": 0.68, "learning_rate": 3.412506790607835e-05, "loss": 1.2332, "step": 11316 }, { "epoch": 0.68, "learning_rate": 3.4131104001931554e-05, "loss": 1.7254, "step": 11318 }, { "epoch": 0.68, "learning_rate": 3.4137140097784755e-05, "loss": 1.145, "step": 11320 }, { "epoch": 0.68, "learning_rate": 3.4143176193637955e-05, "loss": 1.2236, "step": 11322 }, { "epoch": 0.68, "learning_rate": 3.414921228949116e-05, "loss": 0.92, "step": 11324 }, { "epoch": 0.68, "learning_rate": 3.4155248385344364e-05, "loss": 1.1832, "step": 11326 }, { "epoch": 0.68, "learning_rate": 3.4161284481197565e-05, "loss": 1.0775, "step": 11328 }, { "epoch": 0.68, "learning_rate": 3.4167320577050765e-05, "loss": 1.2272, "step": 11330 }, { "epoch": 0.68, "learning_rate": 3.4173356672903966e-05, "loss": 0.9393, "step": 11332 }, { "epoch": 0.68, "learning_rate": 3.417939276875717e-05, "loss": 1.4344, "step": 11334 }, { "epoch": 0.68, "learning_rate": 3.418542886461037e-05, "loss": 1.2458, "step": 11336 }, { "epoch": 0.68, "learning_rate": 3.4191464960463576e-05, "loss": 1.0442, "step": 11338 }, { "epoch": 0.68, "learning_rate": 3.4197501056316776e-05, "loss": 1.3651, "step": 11340 }, { "epoch": 0.68, "learning_rate": 3.420353715216998e-05, "loss": 1.4242, "step": 11342 }, { "epoch": 0.68, "learning_rate": 3.420957324802318e-05, "loss": 1.0289, "step": 11344 }, { "epoch": 0.68, "learning_rate": 3.421560934387638e-05, "loss": 1.002, "step": 11346 }, { "epoch": 0.68, "learning_rate": 3.422164543972958e-05, "loss": 1.1812, "step": 11348 }, { "epoch": 0.69, "learning_rate": 3.422768153558279e-05, "loss": 1.8739, "step": 11350 }, { "epoch": 0.69, "learning_rate": 3.423371763143599e-05, "loss": 1.5065, "step": 11352 }, { "epoch": 0.69, "learning_rate": 3.423975372728919e-05, "loss": 1.0951, "step": 11354 }, { "epoch": 0.69, "learning_rate": 3.4245789823142396e-05, "loss": 1.1147, "step": 11356 }, { "epoch": 0.69, "learning_rate": 3.42518259189956e-05, "loss": 1.0682, "step": 11358 }, { "epoch": 0.69, "learning_rate": 3.42578620148488e-05, "loss": 1.001, "step": 11360 }, { "epoch": 0.69, "learning_rate": 3.4263898110702e-05, "loss": 1.212, "step": 11362 }, { "epoch": 0.69, "learning_rate": 3.4269934206555206e-05, "loss": 1.253, "step": 11364 }, { "epoch": 0.69, "learning_rate": 3.427597030240841e-05, "loss": 1.5139, "step": 11366 }, { "epoch": 0.69, "learning_rate": 3.428200639826161e-05, "loss": 1.1311, "step": 11368 }, { "epoch": 0.69, "learning_rate": 3.428804249411481e-05, "loss": 0.8684, "step": 11370 }, { "epoch": 0.69, "learning_rate": 3.429407858996801e-05, "loss": 1.2448, "step": 11372 }, { "epoch": 0.69, "learning_rate": 3.430011468582121e-05, "loss": 1.2466, "step": 11374 }, { "epoch": 0.69, "learning_rate": 3.430615078167441e-05, "loss": 1.3954, "step": 11376 }, { "epoch": 0.69, "learning_rate": 3.431218687752762e-05, "loss": 1.2104, "step": 11378 }, { "epoch": 0.69, "learning_rate": 3.431822297338082e-05, "loss": 1.1774, "step": 11380 }, { "epoch": 0.69, "learning_rate": 3.432425906923402e-05, "loss": 1.2797, "step": 11382 }, { "epoch": 0.69, "learning_rate": 3.433029516508722e-05, "loss": 1.536, "step": 11384 }, { "epoch": 0.69, "learning_rate": 3.433633126094042e-05, "loss": 1.2042, "step": 11386 }, { "epoch": 0.69, "learning_rate": 3.434236735679362e-05, "loss": 1.2621, "step": 11388 }, { "epoch": 0.69, "learning_rate": 3.434840345264683e-05, "loss": 1.3038, "step": 11390 }, { "epoch": 0.69, "learning_rate": 3.435443954850003e-05, "loss": 1.1685, "step": 11392 }, { "epoch": 0.69, "learning_rate": 3.436047564435323e-05, "loss": 1.3424, "step": 11394 }, { "epoch": 0.69, "learning_rate": 3.436651174020643e-05, "loss": 1.1587, "step": 11396 }, { "epoch": 0.69, "learning_rate": 3.437254783605964e-05, "loss": 0.9884, "step": 11398 }, { "epoch": 0.69, "learning_rate": 3.437858393191284e-05, "loss": 1.1299, "step": 11400 }, { "epoch": 0.69, "learning_rate": 3.438462002776604e-05, "loss": 1.4354, "step": 11402 }, { "epoch": 0.69, "learning_rate": 3.439065612361925e-05, "loss": 1.3183, "step": 11404 }, { "epoch": 0.69, "learning_rate": 3.439669221947245e-05, "loss": 1.1285, "step": 11406 }, { "epoch": 0.69, "learning_rate": 3.440272831532565e-05, "loss": 1.5321, "step": 11408 }, { "epoch": 0.69, "learning_rate": 3.440876441117885e-05, "loss": 1.1454, "step": 11410 }, { "epoch": 0.69, "learning_rate": 3.441480050703205e-05, "loss": 1.2911, "step": 11412 }, { "epoch": 0.69, "learning_rate": 3.4420836602885254e-05, "loss": 1.1712, "step": 11414 }, { "epoch": 0.69, "learning_rate": 3.4426872698738455e-05, "loss": 1.559, "step": 11416 }, { "epoch": 0.69, "learning_rate": 3.443290879459166e-05, "loss": 1.037, "step": 11418 }, { "epoch": 0.69, "learning_rate": 3.443894489044486e-05, "loss": 1.1389, "step": 11420 }, { "epoch": 0.69, "learning_rate": 3.4444980986298064e-05, "loss": 1.079, "step": 11422 }, { "epoch": 0.69, "learning_rate": 3.4451017082151265e-05, "loss": 1.0563, "step": 11424 }, { "epoch": 0.69, "learning_rate": 3.4457053178004466e-05, "loss": 1.2882, "step": 11426 }, { "epoch": 0.69, "learning_rate": 3.4463089273857667e-05, "loss": 1.3009, "step": 11428 }, { "epoch": 0.69, "learning_rate": 3.4469125369710874e-05, "loss": 1.3186, "step": 11430 }, { "epoch": 0.69, "learning_rate": 3.4475161465564075e-05, "loss": 1.3322, "step": 11432 }, { "epoch": 0.69, "learning_rate": 3.4481197561417276e-05, "loss": 1.4625, "step": 11434 }, { "epoch": 0.69, "learning_rate": 3.4487233657270477e-05, "loss": 0.986, "step": 11436 }, { "epoch": 0.69, "learning_rate": 3.449326975312368e-05, "loss": 1.4437, "step": 11438 }, { "epoch": 0.69, "learning_rate": 3.4499305848976885e-05, "loss": 1.2312, "step": 11440 }, { "epoch": 0.69, "learning_rate": 3.4505341944830086e-05, "loss": 0.9179, "step": 11442 }, { "epoch": 0.69, "learning_rate": 3.451137804068329e-05, "loss": 1.0168, "step": 11444 }, { "epoch": 0.69, "learning_rate": 3.4517414136536494e-05, "loss": 1.2639, "step": 11446 }, { "epoch": 0.69, "learning_rate": 3.4523450232389695e-05, "loss": 0.9148, "step": 11448 }, { "epoch": 0.69, "learning_rate": 3.4529486328242896e-05, "loss": 0.877, "step": 11450 }, { "epoch": 0.69, "learning_rate": 3.45355224240961e-05, "loss": 1.2375, "step": 11452 }, { "epoch": 0.69, "learning_rate": 3.45415585199493e-05, "loss": 1.8172, "step": 11454 }, { "epoch": 0.69, "learning_rate": 3.4547594615802505e-05, "loss": 1.468, "step": 11456 }, { "epoch": 0.69, "learning_rate": 3.4553630711655706e-05, "loss": 1.5168, "step": 11458 }, { "epoch": 0.69, "learning_rate": 3.455966680750891e-05, "loss": 1.3458, "step": 11460 }, { "epoch": 0.69, "learning_rate": 3.456570290336211e-05, "loss": 0.9261, "step": 11462 }, { "epoch": 0.69, "learning_rate": 3.457173899921531e-05, "loss": 1.4531, "step": 11464 }, { "epoch": 0.69, "learning_rate": 3.457777509506851e-05, "loss": 1.2298, "step": 11466 }, { "epoch": 0.69, "learning_rate": 3.458381119092171e-05, "loss": 1.2558, "step": 11468 }, { "epoch": 0.69, "learning_rate": 3.458984728677492e-05, "loss": 1.5991, "step": 11470 }, { "epoch": 0.69, "learning_rate": 3.459588338262812e-05, "loss": 1.31, "step": 11472 }, { "epoch": 0.69, "learning_rate": 3.460191947848132e-05, "loss": 1.5408, "step": 11474 }, { "epoch": 0.69, "learning_rate": 3.460795557433452e-05, "loss": 1.3749, "step": 11476 }, { "epoch": 0.69, "learning_rate": 3.461399167018772e-05, "loss": 1.5059, "step": 11478 }, { "epoch": 0.69, "learning_rate": 3.462002776604092e-05, "loss": 1.0524, "step": 11480 }, { "epoch": 0.69, "learning_rate": 3.462606386189412e-05, "loss": 1.42, "step": 11482 }, { "epoch": 0.69, "learning_rate": 3.463209995774733e-05, "loss": 1.4095, "step": 11484 }, { "epoch": 0.69, "learning_rate": 3.463813605360053e-05, "loss": 0.9823, "step": 11486 }, { "epoch": 0.69, "learning_rate": 3.464417214945374e-05, "loss": 1.7687, "step": 11488 }, { "epoch": 0.69, "learning_rate": 3.465020824530694e-05, "loss": 0.9182, "step": 11490 }, { "epoch": 0.69, "learning_rate": 3.465624434116014e-05, "loss": 1.6372, "step": 11492 }, { "epoch": 0.69, "learning_rate": 3.466228043701334e-05, "loss": 1.1203, "step": 11494 }, { "epoch": 0.69, "learning_rate": 3.466831653286655e-05, "loss": 1.3349, "step": 11496 }, { "epoch": 0.69, "learning_rate": 3.467435262871975e-05, "loss": 1.1371, "step": 11498 }, { "epoch": 0.69, "learning_rate": 3.468038872457295e-05, "loss": 1.3249, "step": 11500 }, { "epoch": 0.69, "learning_rate": 3.468642482042615e-05, "loss": 1.6688, "step": 11502 }, { "epoch": 0.69, "learning_rate": 3.469246091627935e-05, "loss": 1.0351, "step": 11504 }, { "epoch": 0.69, "learning_rate": 3.469849701213255e-05, "loss": 1.0854, "step": 11506 }, { "epoch": 0.69, "learning_rate": 3.4704533107985754e-05, "loss": 1.537, "step": 11508 }, { "epoch": 0.69, "learning_rate": 3.471056920383896e-05, "loss": 1.2055, "step": 11510 }, { "epoch": 0.69, "learning_rate": 3.471660529969216e-05, "loss": 1.3505, "step": 11512 }, { "epoch": 0.69, "learning_rate": 3.472264139554536e-05, "loss": 1.4518, "step": 11514 }, { "epoch": 0.7, "learning_rate": 3.4728677491398564e-05, "loss": 1.27, "step": 11516 }, { "epoch": 0.7, "learning_rate": 3.4734713587251764e-05, "loss": 1.4428, "step": 11518 }, { "epoch": 0.7, "learning_rate": 3.4740749683104965e-05, "loss": 1.457, "step": 11520 }, { "epoch": 0.7, "learning_rate": 3.474678577895817e-05, "loss": 1.1671, "step": 11522 }, { "epoch": 0.7, "learning_rate": 3.4752821874811374e-05, "loss": 1.0956, "step": 11524 }, { "epoch": 0.7, "learning_rate": 3.4758857970664574e-05, "loss": 1.1165, "step": 11526 }, { "epoch": 0.7, "learning_rate": 3.4764894066517775e-05, "loss": 1.1608, "step": 11528 }, { "epoch": 0.7, "learning_rate": 3.4770930162370976e-05, "loss": 1.1212, "step": 11530 }, { "epoch": 0.7, "learning_rate": 3.4776966258224184e-05, "loss": 1.3266, "step": 11532 }, { "epoch": 0.7, "learning_rate": 3.4783002354077384e-05, "loss": 1.2563, "step": 11534 }, { "epoch": 0.7, "learning_rate": 3.478903844993059e-05, "loss": 1.361, "step": 11536 }, { "epoch": 0.7, "learning_rate": 3.479507454578379e-05, "loss": 1.3977, "step": 11538 }, { "epoch": 0.7, "learning_rate": 3.4801110641636994e-05, "loss": 1.293, "step": 11540 }, { "epoch": 0.7, "learning_rate": 3.4807146737490195e-05, "loss": 1.1604, "step": 11542 }, { "epoch": 0.7, "learning_rate": 3.4813182833343395e-05, "loss": 1.0028, "step": 11544 }, { "epoch": 0.7, "learning_rate": 3.4819218929196596e-05, "loss": 1.384, "step": 11546 }, { "epoch": 0.7, "learning_rate": 3.4825255025049804e-05, "loss": 1.0921, "step": 11548 }, { "epoch": 0.7, "learning_rate": 3.4831291120903005e-05, "loss": 1.3637, "step": 11550 }, { "epoch": 0.7, "learning_rate": 3.4837327216756205e-05, "loss": 1.2963, "step": 11552 }, { "epoch": 0.7, "learning_rate": 3.4843363312609406e-05, "loss": 1.1923, "step": 11554 }, { "epoch": 0.7, "learning_rate": 3.484939940846261e-05, "loss": 1.2664, "step": 11556 }, { "epoch": 0.7, "learning_rate": 3.485543550431581e-05, "loss": 1.4907, "step": 11558 }, { "epoch": 0.7, "learning_rate": 3.486147160016901e-05, "loss": 1.2672, "step": 11560 }, { "epoch": 0.7, "learning_rate": 3.4867507696022216e-05, "loss": 1.5977, "step": 11562 }, { "epoch": 0.7, "learning_rate": 3.487354379187542e-05, "loss": 1.186, "step": 11564 }, { "epoch": 0.7, "learning_rate": 3.487957988772862e-05, "loss": 1.2164, "step": 11566 }, { "epoch": 0.7, "learning_rate": 3.488561598358182e-05, "loss": 1.5787, "step": 11568 }, { "epoch": 0.7, "learning_rate": 3.489165207943502e-05, "loss": 1.1523, "step": 11570 }, { "epoch": 0.7, "learning_rate": 3.489768817528822e-05, "loss": 1.3033, "step": 11572 }, { "epoch": 0.7, "learning_rate": 3.490372427114143e-05, "loss": 1.394, "step": 11574 }, { "epoch": 0.7, "learning_rate": 3.490976036699463e-05, "loss": 1.1447, "step": 11576 }, { "epoch": 0.7, "learning_rate": 3.491579646284783e-05, "loss": 0.9656, "step": 11578 }, { "epoch": 0.7, "learning_rate": 3.492183255870104e-05, "loss": 1.336, "step": 11580 }, { "epoch": 0.7, "learning_rate": 3.492786865455424e-05, "loss": 1.418, "step": 11582 }, { "epoch": 0.7, "learning_rate": 3.493390475040744e-05, "loss": 1.1621, "step": 11584 }, { "epoch": 0.7, "learning_rate": 3.493994084626064e-05, "loss": 1.1036, "step": 11586 }, { "epoch": 0.7, "learning_rate": 3.494597694211385e-05, "loss": 1.4773, "step": 11588 }, { "epoch": 0.7, "learning_rate": 3.495201303796705e-05, "loss": 1.0316, "step": 11590 }, { "epoch": 0.7, "learning_rate": 3.495804913382025e-05, "loss": 1.1634, "step": 11592 }, { "epoch": 0.7, "learning_rate": 3.496408522967345e-05, "loss": 1.0781, "step": 11594 }, { "epoch": 0.7, "learning_rate": 3.497012132552665e-05, "loss": 1.1894, "step": 11596 }, { "epoch": 0.7, "learning_rate": 3.497615742137985e-05, "loss": 1.019, "step": 11598 }, { "epoch": 0.7, "learning_rate": 3.498219351723305e-05, "loss": 1.3577, "step": 11600 }, { "epoch": 0.7, "learning_rate": 3.498822961308626e-05, "loss": 1.2512, "step": 11602 }, { "epoch": 0.7, "learning_rate": 3.499426570893946e-05, "loss": 1.5325, "step": 11604 }, { "epoch": 0.7, "learning_rate": 3.500030180479266e-05, "loss": 1.2573, "step": 11606 }, { "epoch": 0.7, "learning_rate": 3.500633790064586e-05, "loss": 1.293, "step": 11608 }, { "epoch": 0.7, "learning_rate": 3.501237399649906e-05, "loss": 1.2537, "step": 11610 }, { "epoch": 0.7, "learning_rate": 3.5018410092352264e-05, "loss": 1.2172, "step": 11612 }, { "epoch": 0.7, "learning_rate": 3.502444618820547e-05, "loss": 1.3535, "step": 11614 }, { "epoch": 0.7, "learning_rate": 3.503048228405867e-05, "loss": 1.2211, "step": 11616 }, { "epoch": 0.7, "learning_rate": 3.503651837991187e-05, "loss": 1.0886, "step": 11618 }, { "epoch": 0.7, "learning_rate": 3.5042554475765074e-05, "loss": 1.2131, "step": 11620 }, { "epoch": 0.7, "learning_rate": 3.504859057161828e-05, "loss": 1.3971, "step": 11622 }, { "epoch": 0.7, "learning_rate": 3.505462666747148e-05, "loss": 1.0598, "step": 11624 }, { "epoch": 0.7, "learning_rate": 3.506066276332468e-05, "loss": 0.9608, "step": 11626 }, { "epoch": 0.7, "learning_rate": 3.506669885917789e-05, "loss": 1.1056, "step": 11628 }, { "epoch": 0.7, "learning_rate": 3.507273495503109e-05, "loss": 1.2918, "step": 11630 }, { "epoch": 0.7, "learning_rate": 3.507877105088429e-05, "loss": 1.861, "step": 11632 }, { "epoch": 0.7, "learning_rate": 3.508480714673749e-05, "loss": 1.4855, "step": 11634 }, { "epoch": 0.7, "learning_rate": 3.5090843242590694e-05, "loss": 1.3918, "step": 11636 }, { "epoch": 0.7, "learning_rate": 3.5096879338443895e-05, "loss": 1.3993, "step": 11638 }, { "epoch": 0.7, "learning_rate": 3.5102915434297096e-05, "loss": 1.3036, "step": 11640 }, { "epoch": 0.7, "learning_rate": 3.51089515301503e-05, "loss": 1.7096, "step": 11642 }, { "epoch": 0.7, "learning_rate": 3.5114987626003504e-05, "loss": 1.1873, "step": 11644 }, { "epoch": 0.7, "learning_rate": 3.5121023721856705e-05, "loss": 1.2448, "step": 11646 }, { "epoch": 0.7, "learning_rate": 3.5127059817709906e-05, "loss": 1.442, "step": 11648 }, { "epoch": 0.7, "learning_rate": 3.5133095913563106e-05, "loss": 1.1446, "step": 11650 }, { "epoch": 0.7, "learning_rate": 3.513913200941631e-05, "loss": 1.1627, "step": 11652 }, { "epoch": 0.7, "learning_rate": 3.5145168105269515e-05, "loss": 1.1318, "step": 11654 }, { "epoch": 0.7, "learning_rate": 3.5151204201122716e-05, "loss": 1.6125, "step": 11656 }, { "epoch": 0.7, "learning_rate": 3.5157240296975917e-05, "loss": 1.2968, "step": 11658 }, { "epoch": 0.7, "learning_rate": 3.516327639282912e-05, "loss": 1.2985, "step": 11660 }, { "epoch": 0.7, "learning_rate": 3.516931248868232e-05, "loss": 1.2896, "step": 11662 }, { "epoch": 0.7, "learning_rate": 3.5175348584535526e-05, "loss": 1.3535, "step": 11664 }, { "epoch": 0.7, "learning_rate": 3.5181384680388727e-05, "loss": 1.3272, "step": 11666 }, { "epoch": 0.7, "learning_rate": 3.518742077624193e-05, "loss": 1.2748, "step": 11668 }, { "epoch": 0.7, "learning_rate": 3.5193456872095135e-05, "loss": 1.3437, "step": 11670 }, { "epoch": 0.7, "learning_rate": 3.5199492967948336e-05, "loss": 1.404, "step": 11672 }, { "epoch": 0.7, "learning_rate": 3.520552906380154e-05, "loss": 1.3528, "step": 11674 }, { "epoch": 0.7, "learning_rate": 3.521156515965474e-05, "loss": 0.966, "step": 11676 }, { "epoch": 0.7, "learning_rate": 3.521760125550794e-05, "loss": 1.1967, "step": 11678 }, { "epoch": 0.7, "learning_rate": 3.5223637351361146e-05, "loss": 0.9767, "step": 11680 }, { "epoch": 0.71, "learning_rate": 3.522967344721435e-05, "loss": 1.3925, "step": 11682 }, { "epoch": 0.71, "learning_rate": 3.523570954306755e-05, "loss": 1.4329, "step": 11684 }, { "epoch": 0.71, "learning_rate": 3.524174563892075e-05, "loss": 1.12, "step": 11686 }, { "epoch": 0.71, "learning_rate": 3.524778173477395e-05, "loss": 1.3716, "step": 11688 }, { "epoch": 0.71, "learning_rate": 3.525381783062715e-05, "loss": 1.3924, "step": 11690 }, { "epoch": 0.71, "learning_rate": 3.525985392648035e-05, "loss": 1.403, "step": 11692 }, { "epoch": 0.71, "learning_rate": 3.526589002233356e-05, "loss": 1.3098, "step": 11694 }, { "epoch": 0.71, "learning_rate": 3.527192611818676e-05, "loss": 1.5795, "step": 11696 }, { "epoch": 0.71, "learning_rate": 3.527796221403996e-05, "loss": 1.1869, "step": 11698 }, { "epoch": 0.71, "learning_rate": 3.528399830989316e-05, "loss": 1.1251, "step": 11700 }, { "epoch": 0.71, "learning_rate": 3.529003440574636e-05, "loss": 1.3264, "step": 11702 }, { "epoch": 0.71, "learning_rate": 3.529607050159956e-05, "loss": 1.3188, "step": 11704 }, { "epoch": 0.71, "learning_rate": 3.530210659745276e-05, "loss": 1.2665, "step": 11706 }, { "epoch": 0.71, "learning_rate": 3.530814269330597e-05, "loss": 1.4889, "step": 11708 }, { "epoch": 0.71, "learning_rate": 3.531417878915917e-05, "loss": 1.0579, "step": 11710 }, { "epoch": 0.71, "learning_rate": 3.532021488501238e-05, "loss": 1.434, "step": 11712 }, { "epoch": 0.71, "learning_rate": 3.532625098086558e-05, "loss": 1.4321, "step": 11714 }, { "epoch": 0.71, "learning_rate": 3.533228707671878e-05, "loss": 1.6648, "step": 11716 }, { "epoch": 0.71, "learning_rate": 3.533832317257198e-05, "loss": 1.4927, "step": 11718 }, { "epoch": 0.71, "learning_rate": 3.534435926842519e-05, "loss": 1.2709, "step": 11720 }, { "epoch": 0.71, "learning_rate": 3.535039536427839e-05, "loss": 1.474, "step": 11722 }, { "epoch": 0.71, "learning_rate": 3.535643146013159e-05, "loss": 1.3116, "step": 11724 }, { "epoch": 0.71, "learning_rate": 3.536246755598479e-05, "loss": 1.3981, "step": 11726 }, { "epoch": 0.71, "learning_rate": 3.536850365183799e-05, "loss": 1.3905, "step": 11728 }, { "epoch": 0.71, "learning_rate": 3.5374539747691193e-05, "loss": 1.2055, "step": 11730 }, { "epoch": 0.71, "learning_rate": 3.5380575843544394e-05, "loss": 1.0393, "step": 11732 }, { "epoch": 0.71, "learning_rate": 3.53866119393976e-05, "loss": 1.3427, "step": 11734 }, { "epoch": 0.71, "learning_rate": 3.53926480352508e-05, "loss": 1.255, "step": 11736 }, { "epoch": 0.71, "learning_rate": 3.5398684131104003e-05, "loss": 1.3453, "step": 11738 }, { "epoch": 0.71, "learning_rate": 3.5404720226957204e-05, "loss": 1.1823, "step": 11740 }, { "epoch": 0.71, "learning_rate": 3.5410756322810405e-05, "loss": 1.4172, "step": 11742 }, { "epoch": 0.71, "learning_rate": 3.5416792418663606e-05, "loss": 1.2904, "step": 11744 }, { "epoch": 0.71, "learning_rate": 3.5422828514516814e-05, "loss": 1.0356, "step": 11746 }, { "epoch": 0.71, "learning_rate": 3.5428864610370014e-05, "loss": 1.1832, "step": 11748 }, { "epoch": 0.71, "learning_rate": 3.5434900706223215e-05, "loss": 1.3257, "step": 11750 }, { "epoch": 0.71, "learning_rate": 3.5440936802076416e-05, "loss": 0.8573, "step": 11752 }, { "epoch": 0.71, "learning_rate": 3.544697289792962e-05, "loss": 1.0134, "step": 11754 }, { "epoch": 0.71, "learning_rate": 3.5453008993782824e-05, "loss": 1.4286, "step": 11756 }, { "epoch": 0.71, "learning_rate": 3.5459045089636025e-05, "loss": 1.5108, "step": 11758 }, { "epoch": 0.71, "learning_rate": 3.546508118548923e-05, "loss": 1.5943, "step": 11760 }, { "epoch": 0.71, "learning_rate": 3.5471117281342434e-05, "loss": 1.2973, "step": 11762 }, { "epoch": 0.71, "learning_rate": 3.5477153377195634e-05, "loss": 1.1804, "step": 11764 }, { "epoch": 0.71, "learning_rate": 3.5483189473048835e-05, "loss": 1.3551, "step": 11766 }, { "epoch": 0.71, "learning_rate": 3.5489225568902036e-05, "loss": 1.4133, "step": 11768 }, { "epoch": 0.71, "learning_rate": 3.549526166475524e-05, "loss": 1.204, "step": 11770 }, { "epoch": 0.71, "learning_rate": 3.5501297760608445e-05, "loss": 1.2363, "step": 11772 }, { "epoch": 0.71, "learning_rate": 3.5507333856461645e-05, "loss": 1.3018, "step": 11774 }, { "epoch": 0.71, "learning_rate": 3.5513369952314846e-05, "loss": 1.4522, "step": 11776 }, { "epoch": 0.71, "learning_rate": 3.551940604816805e-05, "loss": 1.2319, "step": 11778 }, { "epoch": 0.71, "learning_rate": 3.552544214402125e-05, "loss": 1.1347, "step": 11780 }, { "epoch": 0.71, "learning_rate": 3.553147823987445e-05, "loss": 1.3178, "step": 11782 }, { "epoch": 0.71, "learning_rate": 3.553751433572765e-05, "loss": 1.1576, "step": 11784 }, { "epoch": 0.71, "learning_rate": 3.554355043158086e-05, "loss": 1.2017, "step": 11786 }, { "epoch": 0.71, "learning_rate": 3.554958652743406e-05, "loss": 1.7035, "step": 11788 }, { "epoch": 0.71, "learning_rate": 3.555562262328726e-05, "loss": 1.4441, "step": 11790 }, { "epoch": 0.71, "learning_rate": 3.556165871914046e-05, "loss": 1.4507, "step": 11792 }, { "epoch": 0.71, "learning_rate": 3.556769481499366e-05, "loss": 1.3427, "step": 11794 }, { "epoch": 0.71, "learning_rate": 3.557373091084686e-05, "loss": 1.3836, "step": 11796 }, { "epoch": 0.71, "learning_rate": 3.557976700670007e-05, "loss": 1.2663, "step": 11798 }, { "epoch": 0.71, "learning_rate": 3.558580310255327e-05, "loss": 1.5763, "step": 11800 }, { "epoch": 0.71, "learning_rate": 3.559183919840647e-05, "loss": 1.302, "step": 11802 }, { "epoch": 0.71, "learning_rate": 3.559787529425968e-05, "loss": 1.4391, "step": 11804 }, { "epoch": 0.71, "learning_rate": 3.560391139011288e-05, "loss": 1.1897, "step": 11806 }, { "epoch": 0.71, "learning_rate": 3.560994748596608e-05, "loss": 1.2733, "step": 11808 }, { "epoch": 0.71, "learning_rate": 3.561598358181928e-05, "loss": 1.3513, "step": 11810 }, { "epoch": 0.71, "learning_rate": 3.562201967767249e-05, "loss": 1.0412, "step": 11812 }, { "epoch": 0.71, "learning_rate": 3.562805577352569e-05, "loss": 1.1494, "step": 11814 }, { "epoch": 0.71, "learning_rate": 3.563409186937889e-05, "loss": 1.3229, "step": 11816 }, { "epoch": 0.71, "learning_rate": 3.564012796523209e-05, "loss": 1.5662, "step": 11818 }, { "epoch": 0.71, "learning_rate": 3.564616406108529e-05, "loss": 1.1785, "step": 11820 }, { "epoch": 0.71, "learning_rate": 3.565220015693849e-05, "loss": 1.5536, "step": 11822 }, { "epoch": 0.71, "learning_rate": 3.565823625279169e-05, "loss": 0.98, "step": 11824 }, { "epoch": 0.71, "learning_rate": 3.56642723486449e-05, "loss": 1.0193, "step": 11826 }, { "epoch": 0.71, "learning_rate": 3.56703084444981e-05, "loss": 1.2221, "step": 11828 }, { "epoch": 0.71, "learning_rate": 3.56763445403513e-05, "loss": 1.0828, "step": 11830 }, { "epoch": 0.71, "learning_rate": 3.56823806362045e-05, "loss": 1.1499, "step": 11832 }, { "epoch": 0.71, "learning_rate": 3.5688416732057704e-05, "loss": 0.8838, "step": 11834 }, { "epoch": 0.71, "learning_rate": 3.5694452827910905e-05, "loss": 1.1053, "step": 11836 }, { "epoch": 0.71, "learning_rate": 3.570048892376411e-05, "loss": 1.0842, "step": 11838 }, { "epoch": 0.71, "learning_rate": 3.570652501961731e-05, "loss": 1.4237, "step": 11840 }, { "epoch": 0.71, "learning_rate": 3.5712561115470514e-05, "loss": 1.3061, "step": 11842 }, { "epoch": 0.71, "learning_rate": 3.5718597211323715e-05, "loss": 1.4009, "step": 11844 }, { "epoch": 0.71, "learning_rate": 3.572463330717692e-05, "loss": 1.2579, "step": 11846 }, { "epoch": 0.72, "learning_rate": 3.573066940303012e-05, "loss": 1.3992, "step": 11848 }, { "epoch": 0.72, "learning_rate": 3.5736705498883324e-05, "loss": 1.2876, "step": 11850 }, { "epoch": 0.72, "learning_rate": 3.574274159473653e-05, "loss": 1.275, "step": 11852 }, { "epoch": 0.72, "learning_rate": 3.574877769058973e-05, "loss": 1.3978, "step": 11854 }, { "epoch": 0.72, "learning_rate": 3.575481378644293e-05, "loss": 1.6108, "step": 11856 }, { "epoch": 0.72, "learning_rate": 3.5760849882296134e-05, "loss": 1.4004, "step": 11858 }, { "epoch": 0.72, "learning_rate": 3.5766885978149335e-05, "loss": 1.2348, "step": 11860 }, { "epoch": 0.72, "learning_rate": 3.5772922074002536e-05, "loss": 1.4315, "step": 11862 }, { "epoch": 0.72, "learning_rate": 3.5778958169855736e-05, "loss": 1.1377, "step": 11864 }, { "epoch": 0.72, "learning_rate": 3.5784994265708944e-05, "loss": 1.1947, "step": 11866 }, { "epoch": 0.72, "learning_rate": 3.5791030361562145e-05, "loss": 1.0472, "step": 11868 }, { "epoch": 0.72, "learning_rate": 3.5797066457415346e-05, "loss": 1.1341, "step": 11870 }, { "epoch": 0.72, "learning_rate": 3.5803102553268546e-05, "loss": 1.0989, "step": 11872 }, { "epoch": 0.72, "learning_rate": 3.580913864912175e-05, "loss": 1.1362, "step": 11874 }, { "epoch": 0.72, "learning_rate": 3.581517474497495e-05, "loss": 1.0642, "step": 11876 }, { "epoch": 0.72, "learning_rate": 3.5821210840828156e-05, "loss": 1.1081, "step": 11878 }, { "epoch": 0.72, "learning_rate": 3.5827246936681356e-05, "loss": 1.6867, "step": 11880 }, { "epoch": 0.72, "learning_rate": 3.583328303253456e-05, "loss": 1.0183, "step": 11882 }, { "epoch": 0.72, "learning_rate": 3.583931912838776e-05, "loss": 0.8275, "step": 11884 }, { "epoch": 0.72, "learning_rate": 3.584535522424096e-05, "loss": 0.9958, "step": 11886 }, { "epoch": 0.72, "learning_rate": 3.585139132009416e-05, "loss": 1.7557, "step": 11888 }, { "epoch": 0.72, "learning_rate": 3.585742741594737e-05, "loss": 1.1337, "step": 11890 }, { "epoch": 0.72, "learning_rate": 3.586346351180057e-05, "loss": 0.927, "step": 11892 }, { "epoch": 0.72, "learning_rate": 3.5869499607653776e-05, "loss": 1.7492, "step": 11894 }, { "epoch": 0.72, "learning_rate": 3.5875535703506977e-05, "loss": 1.0892, "step": 11896 }, { "epoch": 0.72, "learning_rate": 3.588157179936018e-05, "loss": 1.4405, "step": 11898 }, { "epoch": 0.72, "learning_rate": 3.588760789521338e-05, "loss": 1.1989, "step": 11900 }, { "epoch": 0.72, "learning_rate": 3.589364399106658e-05, "loss": 1.6173, "step": 11902 }, { "epoch": 0.72, "learning_rate": 3.589968008691979e-05, "loss": 1.2201, "step": 11904 }, { "epoch": 0.72, "learning_rate": 3.590571618277299e-05, "loss": 1.3148, "step": 11906 }, { "epoch": 0.72, "learning_rate": 3.591175227862619e-05, "loss": 1.4833, "step": 11908 }, { "epoch": 0.72, "learning_rate": 3.591778837447939e-05, "loss": 1.4613, "step": 11910 }, { "epoch": 0.72, "learning_rate": 3.592382447033259e-05, "loss": 0.994, "step": 11912 }, { "epoch": 0.72, "learning_rate": 3.592986056618579e-05, "loss": 1.2688, "step": 11914 }, { "epoch": 0.72, "learning_rate": 3.593589666203899e-05, "loss": 1.2261, "step": 11916 }, { "epoch": 0.72, "learning_rate": 3.59419327578922e-05, "loss": 1.228, "step": 11918 }, { "epoch": 0.72, "learning_rate": 3.59479688537454e-05, "loss": 1.3917, "step": 11920 }, { "epoch": 0.72, "learning_rate": 3.59540049495986e-05, "loss": 1.3398, "step": 11922 }, { "epoch": 0.72, "learning_rate": 3.59600410454518e-05, "loss": 0.9399, "step": 11924 }, { "epoch": 0.72, "learning_rate": 3.5966077141305e-05, "loss": 1.6061, "step": 11926 }, { "epoch": 0.72, "learning_rate": 3.59721132371582e-05, "loss": 1.2982, "step": 11928 }, { "epoch": 0.72, "learning_rate": 3.5978149333011404e-05, "loss": 1.326, "step": 11930 }, { "epoch": 0.72, "learning_rate": 3.598418542886461e-05, "loss": 1.3489, "step": 11932 }, { "epoch": 0.72, "learning_rate": 3.599022152471781e-05, "loss": 1.4184, "step": 11934 }, { "epoch": 0.72, "learning_rate": 3.599625762057102e-05, "loss": 1.099, "step": 11936 }, { "epoch": 0.72, "learning_rate": 3.600229371642422e-05, "loss": 1.2254, "step": 11938 }, { "epoch": 0.72, "learning_rate": 3.600832981227742e-05, "loss": 1.6703, "step": 11940 }, { "epoch": 0.72, "learning_rate": 3.601436590813062e-05, "loss": 1.0966, "step": 11942 }, { "epoch": 0.72, "learning_rate": 3.602040200398383e-05, "loss": 1.2266, "step": 11944 }, { "epoch": 0.72, "learning_rate": 3.602643809983703e-05, "loss": 1.0219, "step": 11946 }, { "epoch": 0.72, "learning_rate": 3.603247419569023e-05, "loss": 1.5074, "step": 11948 }, { "epoch": 0.72, "learning_rate": 3.603851029154343e-05, "loss": 1.2575, "step": 11950 }, { "epoch": 0.72, "learning_rate": 3.6044546387396633e-05, "loss": 1.2795, "step": 11952 }, { "epoch": 0.72, "learning_rate": 3.6050582483249834e-05, "loss": 1.3343, "step": 11954 }, { "epoch": 0.72, "learning_rate": 3.6056618579103035e-05, "loss": 1.1809, "step": 11956 }, { "epoch": 0.72, "learning_rate": 3.606265467495624e-05, "loss": 1.1993, "step": 11958 }, { "epoch": 0.72, "learning_rate": 3.6068690770809443e-05, "loss": 1.0986, "step": 11960 }, { "epoch": 0.72, "learning_rate": 3.6074726866662644e-05, "loss": 1.7269, "step": 11962 }, { "epoch": 0.72, "learning_rate": 3.6080762962515845e-05, "loss": 1.3088, "step": 11964 }, { "epoch": 0.72, "learning_rate": 3.6086799058369046e-05, "loss": 1.1934, "step": 11966 }, { "epoch": 0.72, "learning_rate": 3.609283515422225e-05, "loss": 1.3308, "step": 11968 }, { "epoch": 0.72, "learning_rate": 3.6098871250075454e-05, "loss": 1.3353, "step": 11970 }, { "epoch": 0.72, "learning_rate": 3.6104907345928655e-05, "loss": 1.1421, "step": 11972 }, { "epoch": 0.72, "learning_rate": 3.6110943441781856e-05, "loss": 1.3219, "step": 11974 }, { "epoch": 0.72, "learning_rate": 3.611697953763506e-05, "loss": 1.4603, "step": 11976 }, { "epoch": 0.72, "learning_rate": 3.612301563348826e-05, "loss": 1.4491, "step": 11978 }, { "epoch": 0.72, "learning_rate": 3.6129051729341465e-05, "loss": 1.2435, "step": 11980 }, { "epoch": 0.72, "learning_rate": 3.6135087825194666e-05, "loss": 1.4017, "step": 11982 }, { "epoch": 0.72, "learning_rate": 3.6141123921047874e-05, "loss": 0.9946, "step": 11984 }, { "epoch": 0.72, "learning_rate": 3.6147160016901074e-05, "loss": 1.7692, "step": 11986 }, { "epoch": 0.72, "learning_rate": 3.6153196112754275e-05, "loss": 1.0865, "step": 11988 }, { "epoch": 0.72, "learning_rate": 3.6159232208607476e-05, "loss": 1.5352, "step": 11990 }, { "epoch": 0.72, "learning_rate": 3.616526830446068e-05, "loss": 1.1525, "step": 11992 }, { "epoch": 0.72, "learning_rate": 3.617130440031388e-05, "loss": 2.5097, "step": 11994 }, { "epoch": 0.72, "learning_rate": 3.617734049616708e-05, "loss": 1.3371, "step": 11996 }, { "epoch": 0.72, "learning_rate": 3.6183376592020286e-05, "loss": 1.1823, "step": 11998 }, { "epoch": 0.72, "learning_rate": 3.618941268787349e-05, "loss": 1.6029, "step": 12000 }, { "epoch": 0.72, "learning_rate": 3.619544878372669e-05, "loss": 1.5309, "step": 12002 }, { "epoch": 0.72, "learning_rate": 3.620148487957989e-05, "loss": 1.2895, "step": 12004 }, { "epoch": 0.72, "learning_rate": 3.620752097543309e-05, "loss": 1.211, "step": 12006 }, { "epoch": 0.72, "learning_rate": 3.621355707128629e-05, "loss": 1.1515, "step": 12008 }, { "epoch": 0.72, "learning_rate": 3.62195931671395e-05, "loss": 1.4603, "step": 12010 }, { "epoch": 0.73, "learning_rate": 3.62256292629927e-05, "loss": 0.9958, "step": 12012 }, { "epoch": 0.73, "learning_rate": 3.62316653588459e-05, "loss": 1.0136, "step": 12014 }, { "epoch": 0.73, "learning_rate": 3.62377014546991e-05, "loss": 1.2785, "step": 12016 }, { "epoch": 0.73, "learning_rate": 3.62437375505523e-05, "loss": 1.3633, "step": 12018 }, { "epoch": 0.73, "learning_rate": 3.62497736464055e-05, "loss": 1.0401, "step": 12020 }, { "epoch": 0.73, "learning_rate": 3.625580974225871e-05, "loss": 1.2251, "step": 12022 }, { "epoch": 0.73, "learning_rate": 3.626184583811191e-05, "loss": 1.4207, "step": 12024 }, { "epoch": 0.73, "learning_rate": 3.626788193396511e-05, "loss": 1.2812, "step": 12026 }, { "epoch": 0.73, "learning_rate": 3.627391802981832e-05, "loss": 1.1165, "step": 12028 }, { "epoch": 0.73, "learning_rate": 3.627995412567152e-05, "loss": 1.076, "step": 12030 }, { "epoch": 0.73, "learning_rate": 3.628599022152472e-05, "loss": 1.3345, "step": 12032 }, { "epoch": 0.73, "learning_rate": 3.629202631737792e-05, "loss": 1.1752, "step": 12034 }, { "epoch": 0.73, "learning_rate": 3.629806241323113e-05, "loss": 1.2482, "step": 12036 }, { "epoch": 0.73, "learning_rate": 3.630409850908433e-05, "loss": 1.2849, "step": 12038 }, { "epoch": 0.73, "learning_rate": 3.631013460493753e-05, "loss": 0.8987, "step": 12040 }, { "epoch": 0.73, "learning_rate": 3.631617070079073e-05, "loss": 1.2688, "step": 12042 }, { "epoch": 0.73, "learning_rate": 3.632220679664393e-05, "loss": 1.3428, "step": 12044 }, { "epoch": 0.73, "learning_rate": 3.632824289249713e-05, "loss": 1.6669, "step": 12046 }, { "epoch": 0.73, "learning_rate": 3.6334278988350334e-05, "loss": 1.3553, "step": 12048 }, { "epoch": 0.73, "learning_rate": 3.634031508420354e-05, "loss": 1.0721, "step": 12050 }, { "epoch": 0.73, "learning_rate": 3.634635118005674e-05, "loss": 1.2219, "step": 12052 }, { "epoch": 0.73, "learning_rate": 3.635238727590994e-05, "loss": 0.8577, "step": 12054 }, { "epoch": 0.73, "learning_rate": 3.6358423371763144e-05, "loss": 1.1215, "step": 12056 }, { "epoch": 0.73, "learning_rate": 3.6364459467616345e-05, "loss": 0.8272, "step": 12058 }, { "epoch": 0.73, "learning_rate": 3.6370495563469545e-05, "loss": 1.3166, "step": 12060 }, { "epoch": 0.73, "learning_rate": 3.637653165932275e-05, "loss": 0.9792, "step": 12062 }, { "epoch": 0.73, "learning_rate": 3.6382567755175954e-05, "loss": 1.7708, "step": 12064 }, { "epoch": 0.73, "learning_rate": 3.6388603851029155e-05, "loss": 1.5351, "step": 12066 }, { "epoch": 0.73, "learning_rate": 3.6394639946882355e-05, "loss": 1.3512, "step": 12068 }, { "epoch": 0.73, "learning_rate": 3.640067604273556e-05, "loss": 0.8867, "step": 12070 }, { "epoch": 0.73, "learning_rate": 3.6406712138588764e-05, "loss": 1.2712, "step": 12072 }, { "epoch": 0.73, "learning_rate": 3.6412748234441965e-05, "loss": 1.1548, "step": 12074 }, { "epoch": 0.73, "learning_rate": 3.641878433029517e-05, "loss": 1.1237, "step": 12076 }, { "epoch": 0.73, "learning_rate": 3.642482042614837e-05, "loss": 1.6201, "step": 12078 }, { "epoch": 0.73, "learning_rate": 3.6430856522001574e-05, "loss": 1.1147, "step": 12080 }, { "epoch": 0.73, "learning_rate": 3.6436892617854775e-05, "loss": 1.5444, "step": 12082 }, { "epoch": 0.73, "learning_rate": 3.6442928713707976e-05, "loss": 1.4133, "step": 12084 }, { "epoch": 0.73, "learning_rate": 3.6448964809561176e-05, "loss": 0.8358, "step": 12086 }, { "epoch": 0.73, "learning_rate": 3.645500090541438e-05, "loss": 1.235, "step": 12088 }, { "epoch": 0.73, "learning_rate": 3.6461037001267585e-05, "loss": 1.1302, "step": 12090 }, { "epoch": 0.73, "learning_rate": 3.6467073097120786e-05, "loss": 1.3525, "step": 12092 }, { "epoch": 0.73, "learning_rate": 3.6473109192973986e-05, "loss": 1.4532, "step": 12094 }, { "epoch": 0.73, "learning_rate": 3.647914528882719e-05, "loss": 1.1583, "step": 12096 }, { "epoch": 0.73, "learning_rate": 3.648518138468039e-05, "loss": 1.1889, "step": 12098 }, { "epoch": 0.73, "learning_rate": 3.649121748053359e-05, "loss": 1.3404, "step": 12100 }, { "epoch": 0.73, "learning_rate": 3.6497253576386796e-05, "loss": 1.0121, "step": 12102 }, { "epoch": 0.73, "learning_rate": 3.650328967224e-05, "loss": 1.3939, "step": 12104 }, { "epoch": 0.73, "learning_rate": 3.65093257680932e-05, "loss": 1.066, "step": 12106 }, { "epoch": 0.73, "learning_rate": 3.65153618639464e-05, "loss": 1.3374, "step": 12108 }, { "epoch": 0.73, "learning_rate": 3.65213979597996e-05, "loss": 1.0763, "step": 12110 }, { "epoch": 0.73, "learning_rate": 3.65274340556528e-05, "loss": 1.0619, "step": 12112 }, { "epoch": 0.73, "learning_rate": 3.653347015150601e-05, "loss": 1.3444, "step": 12114 }, { "epoch": 0.73, "learning_rate": 3.653950624735921e-05, "loss": 1.193, "step": 12116 }, { "epoch": 0.73, "learning_rate": 3.6545542343212417e-05, "loss": 1.3279, "step": 12118 }, { "epoch": 0.73, "learning_rate": 3.655157843906562e-05, "loss": 1.3582, "step": 12120 }, { "epoch": 0.73, "learning_rate": 3.655761453491882e-05, "loss": 1.3381, "step": 12122 }, { "epoch": 0.73, "learning_rate": 3.656365063077202e-05, "loss": 1.149, "step": 12124 }, { "epoch": 0.73, "learning_rate": 3.656968672662522e-05, "loss": 1.1863, "step": 12126 }, { "epoch": 0.73, "learning_rate": 3.657572282247843e-05, "loss": 1.0922, "step": 12128 }, { "epoch": 0.73, "learning_rate": 3.658175891833163e-05, "loss": 0.8657, "step": 12130 }, { "epoch": 0.73, "learning_rate": 3.658779501418483e-05, "loss": 0.9825, "step": 12132 }, { "epoch": 0.73, "learning_rate": 3.659383111003803e-05, "loss": 1.1641, "step": 12134 }, { "epoch": 0.73, "learning_rate": 3.659986720589123e-05, "loss": 1.1665, "step": 12136 }, { "epoch": 0.73, "learning_rate": 3.660590330174443e-05, "loss": 1.0595, "step": 12138 }, { "epoch": 0.73, "learning_rate": 3.661193939759763e-05, "loss": 1.5565, "step": 12140 }, { "epoch": 0.73, "learning_rate": 3.661797549345084e-05, "loss": 1.0758, "step": 12142 }, { "epoch": 0.73, "learning_rate": 3.662401158930404e-05, "loss": 1.1497, "step": 12144 }, { "epoch": 0.73, "learning_rate": 3.663004768515724e-05, "loss": 0.9653, "step": 12146 }, { "epoch": 0.73, "learning_rate": 3.663608378101044e-05, "loss": 1.2221, "step": 12148 }, { "epoch": 0.73, "learning_rate": 3.664211987686364e-05, "loss": 1.0227, "step": 12150 }, { "epoch": 0.73, "learning_rate": 3.6648155972716844e-05, "loss": 1.2877, "step": 12152 }, { "epoch": 0.73, "learning_rate": 3.6654192068570045e-05, "loss": 1.2667, "step": 12154 }, { "epoch": 0.73, "learning_rate": 3.666022816442325e-05, "loss": 1.2997, "step": 12156 }, { "epoch": 0.73, "learning_rate": 3.666626426027645e-05, "loss": 1.0234, "step": 12158 }, { "epoch": 0.73, "learning_rate": 3.6672300356129654e-05, "loss": 1.3348, "step": 12160 }, { "epoch": 0.73, "learning_rate": 3.667833645198286e-05, "loss": 1.3529, "step": 12162 }, { "epoch": 0.73, "learning_rate": 3.668437254783606e-05, "loss": 1.267, "step": 12164 }, { "epoch": 0.73, "learning_rate": 3.669040864368926e-05, "loss": 1.2516, "step": 12166 }, { "epoch": 0.73, "learning_rate": 3.669644473954247e-05, "loss": 0.9894, "step": 12168 }, { "epoch": 0.73, "learning_rate": 3.670248083539567e-05, "loss": 1.1833, "step": 12170 }, { "epoch": 0.73, "learning_rate": 3.670851693124887e-05, "loss": 0.906, "step": 12172 }, { "epoch": 0.73, "learning_rate": 3.671455302710207e-05, "loss": 1.0363, "step": 12174 }, { "epoch": 0.73, "learning_rate": 3.6720589122955274e-05, "loss": 1.5923, "step": 12176 }, { "epoch": 0.74, "learning_rate": 3.6726625218808475e-05, "loss": 1.0495, "step": 12178 }, { "epoch": 0.74, "learning_rate": 3.6732661314661676e-05, "loss": 1.2949, "step": 12180 }, { "epoch": 0.74, "learning_rate": 3.673869741051488e-05, "loss": 1.2638, "step": 12182 }, { "epoch": 0.74, "learning_rate": 3.6744733506368084e-05, "loss": 1.3905, "step": 12184 }, { "epoch": 0.74, "learning_rate": 3.6750769602221285e-05, "loss": 1.3259, "step": 12186 }, { "epoch": 0.74, "learning_rate": 3.6756805698074486e-05, "loss": 1.1466, "step": 12188 }, { "epoch": 0.74, "learning_rate": 3.676284179392769e-05, "loss": 1.1037, "step": 12190 }, { "epoch": 0.74, "learning_rate": 3.676887788978089e-05, "loss": 1.3196, "step": 12192 }, { "epoch": 0.74, "learning_rate": 3.6774913985634095e-05, "loss": 1.3927, "step": 12194 }, { "epoch": 0.74, "learning_rate": 3.6780950081487296e-05, "loss": 1.136, "step": 12196 }, { "epoch": 0.74, "learning_rate": 3.67869861773405e-05, "loss": 1.1033, "step": 12198 }, { "epoch": 0.74, "learning_rate": 3.67930222731937e-05, "loss": 1.3956, "step": 12200 }, { "epoch": 0.74, "learning_rate": 3.67990583690469e-05, "loss": 1.4769, "step": 12202 }, { "epoch": 0.74, "learning_rate": 3.6805094464900106e-05, "loss": 1.3465, "step": 12204 }, { "epoch": 0.74, "learning_rate": 3.681113056075331e-05, "loss": 1.3624, "step": 12206 }, { "epoch": 0.74, "learning_rate": 3.6817166656606514e-05, "loss": 1.2437, "step": 12208 }, { "epoch": 0.74, "learning_rate": 3.6823202752459715e-05, "loss": 1.4411, "step": 12210 }, { "epoch": 0.74, "learning_rate": 3.6829238848312916e-05, "loss": 1.2222, "step": 12212 }, { "epoch": 0.74, "learning_rate": 3.683527494416612e-05, "loss": 1.226, "step": 12214 }, { "epoch": 0.74, "learning_rate": 3.684131104001932e-05, "loss": 1.4278, "step": 12216 }, { "epoch": 0.74, "learning_rate": 3.684734713587252e-05, "loss": 1.4276, "step": 12218 }, { "epoch": 0.74, "learning_rate": 3.685338323172572e-05, "loss": 1.1686, "step": 12220 }, { "epoch": 0.74, "learning_rate": 3.685941932757893e-05, "loss": 1.2996, "step": 12222 }, { "epoch": 0.74, "learning_rate": 3.686545542343213e-05, "loss": 1.0682, "step": 12224 }, { "epoch": 0.74, "learning_rate": 3.687149151928533e-05, "loss": 0.9387, "step": 12226 }, { "epoch": 0.74, "learning_rate": 3.687752761513853e-05, "loss": 1.1171, "step": 12228 }, { "epoch": 0.74, "learning_rate": 3.688356371099173e-05, "loss": 1.2182, "step": 12230 }, { "epoch": 0.74, "learning_rate": 3.688959980684493e-05, "loss": 0.7428, "step": 12232 }, { "epoch": 0.74, "learning_rate": 3.689563590269814e-05, "loss": 1.2633, "step": 12234 }, { "epoch": 0.74, "learning_rate": 3.690167199855134e-05, "loss": 1.4151, "step": 12236 }, { "epoch": 0.74, "learning_rate": 3.690770809440454e-05, "loss": 1.1874, "step": 12238 }, { "epoch": 0.74, "learning_rate": 3.691374419025774e-05, "loss": 1.6999, "step": 12240 }, { "epoch": 0.74, "learning_rate": 3.691978028611094e-05, "loss": 1.3791, "step": 12242 }, { "epoch": 0.74, "learning_rate": 3.692581638196414e-05, "loss": 1.2161, "step": 12244 }, { "epoch": 0.74, "learning_rate": 3.693185247781735e-05, "loss": 1.4397, "step": 12246 }, { "epoch": 0.74, "learning_rate": 3.693788857367055e-05, "loss": 1.5069, "step": 12248 }, { "epoch": 0.74, "learning_rate": 3.694392466952375e-05, "loss": 0.964, "step": 12250 }, { "epoch": 0.74, "learning_rate": 3.694996076537696e-05, "loss": 1.3181, "step": 12252 }, { "epoch": 0.74, "learning_rate": 3.695599686123016e-05, "loss": 0.9083, "step": 12254 }, { "epoch": 0.74, "learning_rate": 3.696203295708336e-05, "loss": 1.2466, "step": 12256 }, { "epoch": 0.74, "learning_rate": 3.696806905293656e-05, "loss": 1.2323, "step": 12258 }, { "epoch": 0.74, "learning_rate": 3.697410514878977e-05, "loss": 1.4114, "step": 12260 }, { "epoch": 0.74, "learning_rate": 3.698014124464297e-05, "loss": 1.3165, "step": 12262 }, { "epoch": 0.74, "learning_rate": 3.698617734049617e-05, "loss": 1.5406, "step": 12264 }, { "epoch": 0.74, "learning_rate": 3.699221343634937e-05, "loss": 1.1474, "step": 12266 }, { "epoch": 0.74, "learning_rate": 3.699824953220257e-05, "loss": 1.4153, "step": 12268 }, { "epoch": 0.74, "learning_rate": 3.7004285628055774e-05, "loss": 0.9203, "step": 12270 }, { "epoch": 0.74, "learning_rate": 3.7010321723908974e-05, "loss": 1.9791, "step": 12272 }, { "epoch": 0.74, "learning_rate": 3.701635781976218e-05, "loss": 1.435, "step": 12274 }, { "epoch": 0.74, "learning_rate": 3.702239391561538e-05, "loss": 1.5022, "step": 12276 }, { "epoch": 0.74, "learning_rate": 3.7028430011468584e-05, "loss": 1.6655, "step": 12278 }, { "epoch": 0.74, "learning_rate": 3.7034466107321784e-05, "loss": 0.9604, "step": 12280 }, { "epoch": 0.74, "learning_rate": 3.7040502203174985e-05, "loss": 1.5417, "step": 12282 }, { "epoch": 0.74, "learning_rate": 3.7046538299028186e-05, "loss": 1.0007, "step": 12284 }, { "epoch": 0.74, "learning_rate": 3.7052574394881394e-05, "loss": 1.3976, "step": 12286 }, { "epoch": 0.74, "learning_rate": 3.7058610490734595e-05, "loss": 1.2529, "step": 12288 }, { "epoch": 0.74, "learning_rate": 3.7064646586587795e-05, "loss": 1.4419, "step": 12290 }, { "epoch": 0.74, "learning_rate": 3.7070682682440996e-05, "loss": 0.963, "step": 12292 }, { "epoch": 0.74, "learning_rate": 3.7076718778294204e-05, "loss": 1.1061, "step": 12294 }, { "epoch": 0.74, "learning_rate": 3.7082754874147405e-05, "loss": 1.5413, "step": 12296 }, { "epoch": 0.74, "learning_rate": 3.7088790970000605e-05, "loss": 1.2823, "step": 12298 }, { "epoch": 0.74, "learning_rate": 3.709482706585381e-05, "loss": 1.1456, "step": 12300 }, { "epoch": 0.74, "learning_rate": 3.7100863161707014e-05, "loss": 1.5164, "step": 12302 }, { "epoch": 0.74, "learning_rate": 3.7106899257560215e-05, "loss": 1.5908, "step": 12304 }, { "epoch": 0.74, "learning_rate": 3.7112935353413415e-05, "loss": 1.5128, "step": 12306 }, { "epoch": 0.74, "learning_rate": 3.7118971449266616e-05, "loss": 1.3295, "step": 12308 }, { "epoch": 0.74, "learning_rate": 3.712500754511982e-05, "loss": 0.8757, "step": 12310 }, { "epoch": 0.74, "learning_rate": 3.713104364097302e-05, "loss": 1.2508, "step": 12312 }, { "epoch": 0.74, "learning_rate": 3.7137079736826225e-05, "loss": 1.5606, "step": 12314 }, { "epoch": 0.74, "learning_rate": 3.7143115832679426e-05, "loss": 1.2746, "step": 12316 }, { "epoch": 0.74, "learning_rate": 3.714915192853263e-05, "loss": 1.6777, "step": 12318 }, { "epoch": 0.74, "learning_rate": 3.715518802438583e-05, "loss": 1.2954, "step": 12320 }, { "epoch": 0.74, "learning_rate": 3.716122412023903e-05, "loss": 1.3337, "step": 12322 }, { "epoch": 0.74, "learning_rate": 3.716726021609223e-05, "loss": 1.0368, "step": 12324 }, { "epoch": 0.74, "learning_rate": 3.717329631194544e-05, "loss": 1.3341, "step": 12326 }, { "epoch": 0.74, "learning_rate": 3.717933240779864e-05, "loss": 1.1166, "step": 12328 }, { "epoch": 0.74, "learning_rate": 3.718536850365184e-05, "loss": 0.9616, "step": 12330 }, { "epoch": 0.74, "learning_rate": 3.719140459950504e-05, "loss": 0.8925, "step": 12332 }, { "epoch": 0.74, "learning_rate": 3.719744069535824e-05, "loss": 1.0364, "step": 12334 }, { "epoch": 0.74, "learning_rate": 3.720347679121144e-05, "loss": 1.4146, "step": 12336 }, { "epoch": 0.74, "learning_rate": 3.720951288706465e-05, "loss": 1.3791, "step": 12338 }, { "epoch": 0.74, "learning_rate": 3.721554898291785e-05, "loss": 1.7004, "step": 12340 }, { "epoch": 0.74, "learning_rate": 3.722158507877106e-05, "loss": 1.6028, "step": 12342 }, { "epoch": 0.75, "learning_rate": 3.722762117462426e-05, "loss": 0.986, "step": 12344 }, { "epoch": 0.75, "learning_rate": 3.723365727047746e-05, "loss": 1.3578, "step": 12346 }, { "epoch": 0.75, "learning_rate": 3.723969336633066e-05, "loss": 1.4383, "step": 12348 }, { "epoch": 0.75, "learning_rate": 3.724572946218386e-05, "loss": 1.5064, "step": 12350 }, { "epoch": 0.75, "learning_rate": 3.725176555803707e-05, "loss": 1.2277, "step": 12352 }, { "epoch": 0.75, "learning_rate": 3.725780165389027e-05, "loss": 1.3575, "step": 12354 }, { "epoch": 0.75, "learning_rate": 3.726383774974347e-05, "loss": 1.0437, "step": 12356 }, { "epoch": 0.75, "learning_rate": 3.726987384559667e-05, "loss": 1.4989, "step": 12358 }, { "epoch": 0.75, "learning_rate": 3.727590994144987e-05, "loss": 0.9482, "step": 12360 }, { "epoch": 0.75, "learning_rate": 3.728194603730307e-05, "loss": 1.0738, "step": 12362 }, { "epoch": 0.75, "learning_rate": 3.728798213315627e-05, "loss": 1.4753, "step": 12364 }, { "epoch": 0.75, "learning_rate": 3.729401822900948e-05, "loss": 1.5026, "step": 12366 }, { "epoch": 0.75, "learning_rate": 3.730005432486268e-05, "loss": 1.028, "step": 12368 }, { "epoch": 0.75, "learning_rate": 3.730609042071588e-05, "loss": 1.4118, "step": 12370 }, { "epoch": 0.75, "learning_rate": 3.731212651656908e-05, "loss": 1.1935, "step": 12372 }, { "epoch": 0.75, "learning_rate": 3.7318162612422284e-05, "loss": 1.316, "step": 12374 }, { "epoch": 0.75, "learning_rate": 3.7324198708275485e-05, "loss": 1.1689, "step": 12376 }, { "epoch": 0.75, "learning_rate": 3.7330234804128686e-05, "loss": 1.3768, "step": 12378 }, { "epoch": 0.75, "learning_rate": 3.733627089998189e-05, "loss": 1.1543, "step": 12380 }, { "epoch": 0.75, "learning_rate": 3.7342306995835094e-05, "loss": 1.097, "step": 12382 }, { "epoch": 0.75, "learning_rate": 3.7348343091688295e-05, "loss": 1.1968, "step": 12384 }, { "epoch": 0.75, "learning_rate": 3.73543791875415e-05, "loss": 1.3402, "step": 12386 }, { "epoch": 0.75, "learning_rate": 3.73604152833947e-05, "loss": 1.2707, "step": 12388 }, { "epoch": 0.75, "learning_rate": 3.7366451379247904e-05, "loss": 1.421, "step": 12390 }, { "epoch": 0.75, "learning_rate": 3.737248747510111e-05, "loss": 1.3596, "step": 12392 }, { "epoch": 0.75, "learning_rate": 3.737852357095431e-05, "loss": 1.5214, "step": 12394 }, { "epoch": 0.75, "learning_rate": 3.738455966680751e-05, "loss": 1.5059, "step": 12396 }, { "epoch": 0.75, "learning_rate": 3.7390595762660714e-05, "loss": 0.989, "step": 12398 }, { "epoch": 0.75, "learning_rate": 3.7396631858513915e-05, "loss": 1.3903, "step": 12400 }, { "epoch": 0.75, "learning_rate": 3.7402667954367116e-05, "loss": 0.9868, "step": 12402 }, { "epoch": 0.75, "learning_rate": 3.7408704050220317e-05, "loss": 1.0558, "step": 12404 }, { "epoch": 0.75, "learning_rate": 3.7414740146073524e-05, "loss": 1.3531, "step": 12406 }, { "epoch": 0.75, "learning_rate": 3.7420776241926725e-05, "loss": 1.1777, "step": 12408 }, { "epoch": 0.75, "learning_rate": 3.7426812337779926e-05, "loss": 1.3986, "step": 12410 }, { "epoch": 0.75, "learning_rate": 3.7432848433633127e-05, "loss": 0.9375, "step": 12412 }, { "epoch": 0.75, "learning_rate": 3.743888452948633e-05, "loss": 1.2646, "step": 12414 }, { "epoch": 0.75, "learning_rate": 3.744492062533953e-05, "loss": 1.4622, "step": 12416 }, { "epoch": 0.75, "learning_rate": 3.7450956721192736e-05, "loss": 1.116, "step": 12418 }, { "epoch": 0.75, "learning_rate": 3.745699281704594e-05, "loss": 1.0797, "step": 12420 }, { "epoch": 0.75, "learning_rate": 3.746302891289914e-05, "loss": 1.2429, "step": 12422 }, { "epoch": 0.75, "learning_rate": 3.746906500875234e-05, "loss": 0.9216, "step": 12424 }, { "epoch": 0.75, "learning_rate": 3.747510110460554e-05, "loss": 1.2212, "step": 12426 }, { "epoch": 0.75, "learning_rate": 3.748113720045875e-05, "loss": 1.2513, "step": 12428 }, { "epoch": 0.75, "learning_rate": 3.748717329631195e-05, "loss": 1.2086, "step": 12430 }, { "epoch": 0.75, "learning_rate": 3.749320939216515e-05, "loss": 1.3695, "step": 12432 }, { "epoch": 0.75, "learning_rate": 3.7499245488018356e-05, "loss": 0.8212, "step": 12434 }, { "epoch": 0.75, "learning_rate": 3.750528158387156e-05, "loss": 1.2188, "step": 12436 }, { "epoch": 0.75, "learning_rate": 3.751131767972476e-05, "loss": 1.2611, "step": 12438 }, { "epoch": 0.75, "learning_rate": 3.751735377557796e-05, "loss": 1.3863, "step": 12440 }, { "epoch": 0.75, "learning_rate": 3.752338987143116e-05, "loss": 1.1882, "step": 12442 }, { "epoch": 0.75, "learning_rate": 3.752942596728436e-05, "loss": 0.9952, "step": 12444 }, { "epoch": 0.75, "learning_rate": 3.753546206313757e-05, "loss": 1.5219, "step": 12446 }, { "epoch": 0.75, "learning_rate": 3.754149815899077e-05, "loss": 0.9839, "step": 12448 }, { "epoch": 0.75, "learning_rate": 3.754753425484397e-05, "loss": 1.3504, "step": 12450 }, { "epoch": 0.75, "learning_rate": 3.755357035069717e-05, "loss": 1.0655, "step": 12452 }, { "epoch": 0.75, "learning_rate": 3.755960644655037e-05, "loss": 1.6522, "step": 12454 }, { "epoch": 0.75, "learning_rate": 3.756564254240357e-05, "loss": 1.5482, "step": 12456 }, { "epoch": 0.75, "learning_rate": 3.757167863825678e-05, "loss": 1.2303, "step": 12458 }, { "epoch": 0.75, "learning_rate": 3.757771473410998e-05, "loss": 0.9227, "step": 12460 }, { "epoch": 0.75, "learning_rate": 3.758375082996318e-05, "loss": 1.3927, "step": 12462 }, { "epoch": 0.75, "learning_rate": 3.758978692581638e-05, "loss": 1.4458, "step": 12464 }, { "epoch": 0.75, "learning_rate": 3.759582302166958e-05, "loss": 1.1127, "step": 12466 }, { "epoch": 0.75, "learning_rate": 3.7601859117522783e-05, "loss": 1.2989, "step": 12468 }, { "epoch": 0.75, "learning_rate": 3.760789521337599e-05, "loss": 1.2226, "step": 12470 }, { "epoch": 0.75, "learning_rate": 3.761393130922919e-05, "loss": 0.9911, "step": 12472 }, { "epoch": 0.75, "learning_rate": 3.761996740508239e-05, "loss": 1.2898, "step": 12474 }, { "epoch": 0.75, "learning_rate": 3.76260035009356e-05, "loss": 1.3007, "step": 12476 }, { "epoch": 0.75, "learning_rate": 3.76320395967888e-05, "loss": 0.8839, "step": 12478 }, { "epoch": 0.75, "learning_rate": 3.7638075692642e-05, "loss": 1.2365, "step": 12480 }, { "epoch": 0.75, "learning_rate": 3.76441117884952e-05, "loss": 0.9269, "step": 12482 }, { "epoch": 0.75, "learning_rate": 3.765014788434841e-05, "loss": 1.2521, "step": 12484 }, { "epoch": 0.75, "learning_rate": 3.765618398020161e-05, "loss": 1.4132, "step": 12486 }, { "epoch": 0.75, "learning_rate": 3.766222007605481e-05, "loss": 1.1696, "step": 12488 }, { "epoch": 0.75, "learning_rate": 3.766825617190801e-05, "loss": 1.2919, "step": 12490 }, { "epoch": 0.75, "learning_rate": 3.7674292267761214e-05, "loss": 1.4306, "step": 12492 }, { "epoch": 0.75, "learning_rate": 3.7680328363614414e-05, "loss": 1.5386, "step": 12494 }, { "epoch": 0.75, "learning_rate": 3.7686364459467615e-05, "loss": 1.1328, "step": 12496 }, { "epoch": 0.75, "learning_rate": 3.769240055532082e-05, "loss": 1.6174, "step": 12498 }, { "epoch": 0.75, "learning_rate": 3.7698436651174024e-05, "loss": 1.2577, "step": 12500 }, { "epoch": 0.75, "learning_rate": 3.7704472747027224e-05, "loss": 1.2836, "step": 12502 }, { "epoch": 0.75, "learning_rate": 3.7710508842880425e-05, "loss": 1.3464, "step": 12504 }, { "epoch": 0.75, "learning_rate": 3.7716544938733626e-05, "loss": 1.0007, "step": 12506 }, { "epoch": 0.75, "learning_rate": 3.772258103458683e-05, "loss": 1.2016, "step": 12508 }, { "epoch": 0.76, "learning_rate": 3.772861713044003e-05, "loss": 1.3416, "step": 12510 }, { "epoch": 0.76, "learning_rate": 3.7734653226293235e-05, "loss": 1.315, "step": 12512 }, { "epoch": 0.76, "learning_rate": 3.7740689322146436e-05, "loss": 1.3196, "step": 12514 }, { "epoch": 0.76, "learning_rate": 3.774672541799964e-05, "loss": 1.5562, "step": 12516 }, { "epoch": 0.76, "learning_rate": 3.7752761513852845e-05, "loss": 1.5188, "step": 12518 }, { "epoch": 0.76, "learning_rate": 3.7758797609706045e-05, "loss": 0.8749, "step": 12520 }, { "epoch": 0.76, "learning_rate": 3.7764833705559246e-05, "loss": 1.5562, "step": 12522 }, { "epoch": 0.76, "learning_rate": 3.7770869801412454e-05, "loss": 1.1403, "step": 12524 }, { "epoch": 0.76, "learning_rate": 3.7776905897265655e-05, "loss": 1.5027, "step": 12526 }, { "epoch": 0.76, "learning_rate": 3.7782941993118855e-05, "loss": 1.2437, "step": 12528 }, { "epoch": 0.76, "learning_rate": 3.7788978088972056e-05, "loss": 0.8921, "step": 12530 }, { "epoch": 0.76, "learning_rate": 3.779501418482526e-05, "loss": 1.1041, "step": 12532 }, { "epoch": 0.76, "learning_rate": 3.780105028067846e-05, "loss": 1.129, "step": 12534 }, { "epoch": 0.76, "learning_rate": 3.780708637653166e-05, "loss": 0.9839, "step": 12536 }, { "epoch": 0.76, "learning_rate": 3.7813122472384866e-05, "loss": 1.0064, "step": 12538 }, { "epoch": 0.76, "learning_rate": 3.781915856823807e-05, "loss": 1.4026, "step": 12540 }, { "epoch": 0.76, "learning_rate": 3.782519466409127e-05, "loss": 1.2796, "step": 12542 }, { "epoch": 0.76, "learning_rate": 3.783123075994447e-05, "loss": 1.2441, "step": 12544 }, { "epoch": 0.76, "learning_rate": 3.783726685579767e-05, "loss": 1.2259, "step": 12546 }, { "epoch": 0.76, "learning_rate": 3.784330295165087e-05, "loss": 1.2548, "step": 12548 }, { "epoch": 0.76, "learning_rate": 3.784933904750408e-05, "loss": 1.2964, "step": 12550 }, { "epoch": 0.76, "learning_rate": 3.785537514335728e-05, "loss": 1.1299, "step": 12552 }, { "epoch": 0.76, "learning_rate": 3.786141123921048e-05, "loss": 1.1345, "step": 12554 }, { "epoch": 0.76, "learning_rate": 3.786744733506368e-05, "loss": 1.0399, "step": 12556 }, { "epoch": 0.76, "learning_rate": 3.787348343091688e-05, "loss": 1.6864, "step": 12558 }, { "epoch": 0.76, "learning_rate": 3.787951952677008e-05, "loss": 1.1472, "step": 12560 }, { "epoch": 0.76, "learning_rate": 3.788555562262329e-05, "loss": 1.5012, "step": 12562 }, { "epoch": 0.76, "learning_rate": 3.789159171847649e-05, "loss": 1.2764, "step": 12564 }, { "epoch": 0.76, "learning_rate": 3.78976278143297e-05, "loss": 1.3629, "step": 12566 }, { "epoch": 0.76, "learning_rate": 3.79036639101829e-05, "loss": 1.44, "step": 12568 }, { "epoch": 0.76, "learning_rate": 3.79097000060361e-05, "loss": 1.4943, "step": 12570 }, { "epoch": 0.76, "learning_rate": 3.79157361018893e-05, "loss": 0.9939, "step": 12572 }, { "epoch": 0.76, "learning_rate": 3.79217721977425e-05, "loss": 0.9479, "step": 12574 }, { "epoch": 0.76, "learning_rate": 3.792780829359571e-05, "loss": 1.2352, "step": 12576 }, { "epoch": 0.76, "learning_rate": 3.793384438944891e-05, "loss": 1.1965, "step": 12578 }, { "epoch": 0.76, "learning_rate": 3.793988048530211e-05, "loss": 1.0875, "step": 12580 }, { "epoch": 0.76, "learning_rate": 3.794591658115531e-05, "loss": 1.1751, "step": 12582 }, { "epoch": 0.76, "learning_rate": 3.795195267700851e-05, "loss": 1.2332, "step": 12584 }, { "epoch": 0.76, "learning_rate": 3.795798877286171e-05, "loss": 1.6283, "step": 12586 }, { "epoch": 0.76, "learning_rate": 3.7964024868714914e-05, "loss": 1.3261, "step": 12588 }, { "epoch": 0.76, "learning_rate": 3.797006096456812e-05, "loss": 1.2486, "step": 12590 }, { "epoch": 0.76, "learning_rate": 3.797609706042132e-05, "loss": 1.4696, "step": 12592 }, { "epoch": 0.76, "learning_rate": 3.798213315627452e-05, "loss": 1.1396, "step": 12594 }, { "epoch": 0.76, "learning_rate": 3.7988169252127724e-05, "loss": 1.0024, "step": 12596 }, { "epoch": 0.76, "learning_rate": 3.7994205347980925e-05, "loss": 1.2835, "step": 12598 }, { "epoch": 0.76, "learning_rate": 3.8000241443834126e-05, "loss": 1.3227, "step": 12600 }, { "epoch": 0.76, "learning_rate": 3.8006277539687326e-05, "loss": 1.3466, "step": 12602 }, { "epoch": 0.76, "learning_rate": 3.8012313635540534e-05, "loss": 1.0169, "step": 12604 }, { "epoch": 0.76, "learning_rate": 3.8018349731393735e-05, "loss": 1.0839, "step": 12606 }, { "epoch": 0.76, "learning_rate": 3.8024385827246936e-05, "loss": 1.4881, "step": 12608 }, { "epoch": 0.76, "learning_rate": 3.803042192310014e-05, "loss": 1.3637, "step": 12610 }, { "epoch": 0.76, "learning_rate": 3.8036458018953344e-05, "loss": 1.2701, "step": 12612 }, { "epoch": 0.76, "learning_rate": 3.8042494114806545e-05, "loss": 0.9176, "step": 12614 }, { "epoch": 0.76, "learning_rate": 3.804853021065975e-05, "loss": 1.1267, "step": 12616 }, { "epoch": 0.76, "learning_rate": 3.805456630651295e-05, "loss": 1.4413, "step": 12618 }, { "epoch": 0.76, "learning_rate": 3.8060602402366154e-05, "loss": 1.356, "step": 12620 }, { "epoch": 0.76, "learning_rate": 3.8066638498219355e-05, "loss": 0.9981, "step": 12622 }, { "epoch": 0.76, "learning_rate": 3.8072674594072556e-05, "loss": 1.166, "step": 12624 }, { "epoch": 0.76, "learning_rate": 3.8078710689925756e-05, "loss": 1.6198, "step": 12626 }, { "epoch": 0.76, "learning_rate": 3.808474678577896e-05, "loss": 1.2907, "step": 12628 }, { "epoch": 0.76, "learning_rate": 3.8090782881632165e-05, "loss": 1.4174, "step": 12630 }, { "epoch": 0.76, "learning_rate": 3.8096818977485366e-05, "loss": 1.0019, "step": 12632 }, { "epoch": 0.76, "learning_rate": 3.8102855073338567e-05, "loss": 1.3328, "step": 12634 }, { "epoch": 0.76, "learning_rate": 3.810889116919177e-05, "loss": 1.1976, "step": 12636 }, { "epoch": 0.76, "learning_rate": 3.811492726504497e-05, "loss": 1.0605, "step": 12638 }, { "epoch": 0.76, "learning_rate": 3.812096336089817e-05, "loss": 1.1106, "step": 12640 }, { "epoch": 0.76, "learning_rate": 3.8126999456751377e-05, "loss": 1.1611, "step": 12642 }, { "epoch": 0.76, "learning_rate": 3.813303555260458e-05, "loss": 1.187, "step": 12644 }, { "epoch": 0.76, "learning_rate": 3.813907164845778e-05, "loss": 1.6621, "step": 12646 }, { "epoch": 0.76, "learning_rate": 3.814510774431098e-05, "loss": 1.0152, "step": 12648 }, { "epoch": 0.76, "learning_rate": 3.815114384016418e-05, "loss": 1.3406, "step": 12650 }, { "epoch": 0.76, "learning_rate": 3.815717993601739e-05, "loss": 1.2266, "step": 12652 }, { "epoch": 0.76, "learning_rate": 3.816321603187059e-05, "loss": 1.3836, "step": 12654 }, { "epoch": 0.76, "learning_rate": 3.816925212772379e-05, "loss": 1.1409, "step": 12656 }, { "epoch": 0.76, "learning_rate": 3.8175288223577e-05, "loss": 1.1376, "step": 12658 }, { "epoch": 0.76, "learning_rate": 3.81813243194302e-05, "loss": 1.4353, "step": 12660 }, { "epoch": 0.76, "learning_rate": 3.81873604152834e-05, "loss": 1.3588, "step": 12662 }, { "epoch": 0.76, "learning_rate": 3.81933965111366e-05, "loss": 1.2766, "step": 12664 }, { "epoch": 0.76, "learning_rate": 3.81994326069898e-05, "loss": 1.3465, "step": 12666 }, { "epoch": 0.76, "learning_rate": 3.8205468702843e-05, "loss": 1.3702, "step": 12668 }, { "epoch": 0.76, "learning_rate": 3.821150479869621e-05, "loss": 1.1131, "step": 12670 }, { "epoch": 0.76, "learning_rate": 3.821754089454941e-05, "loss": 1.1536, "step": 12672 }, { "epoch": 0.76, "learning_rate": 3.822357699040261e-05, "loss": 1.2787, "step": 12674 }, { "epoch": 0.77, "learning_rate": 3.822961308625581e-05, "loss": 1.3099, "step": 12676 }, { "epoch": 0.77, "learning_rate": 3.823564918210901e-05, "loss": 1.2178, "step": 12678 }, { "epoch": 0.77, "learning_rate": 3.824168527796221e-05, "loss": 0.9549, "step": 12680 }, { "epoch": 0.77, "learning_rate": 3.824772137381542e-05, "loss": 1.4368, "step": 12682 }, { "epoch": 0.77, "learning_rate": 3.825375746966862e-05, "loss": 1.3866, "step": 12684 }, { "epoch": 0.77, "learning_rate": 3.825979356552182e-05, "loss": 1.4957, "step": 12686 }, { "epoch": 0.77, "learning_rate": 3.826582966137502e-05, "loss": 1.0505, "step": 12688 }, { "epoch": 0.77, "learning_rate": 3.827186575722822e-05, "loss": 1.0443, "step": 12690 }, { "epoch": 0.77, "learning_rate": 3.8277901853081424e-05, "loss": 1.1901, "step": 12692 }, { "epoch": 0.77, "learning_rate": 3.828393794893463e-05, "loss": 1.0527, "step": 12694 }, { "epoch": 0.77, "learning_rate": 3.828997404478783e-05, "loss": 1.1611, "step": 12696 }, { "epoch": 0.77, "learning_rate": 3.8296010140641033e-05, "loss": 1.0455, "step": 12698 }, { "epoch": 0.77, "learning_rate": 3.830204623649424e-05, "loss": 1.0551, "step": 12700 }, { "epoch": 0.77, "learning_rate": 3.830808233234744e-05, "loss": 1.1979, "step": 12702 }, { "epoch": 0.77, "learning_rate": 3.831411842820064e-05, "loss": 1.3271, "step": 12704 }, { "epoch": 0.77, "learning_rate": 3.8320154524053843e-05, "loss": 1.3422, "step": 12706 }, { "epoch": 0.77, "learning_rate": 3.832619061990705e-05, "loss": 1.1703, "step": 12708 }, { "epoch": 0.77, "learning_rate": 3.833222671576025e-05, "loss": 1.3637, "step": 12710 }, { "epoch": 0.77, "learning_rate": 3.833826281161345e-05, "loss": 1.1875, "step": 12712 }, { "epoch": 0.77, "learning_rate": 3.8344298907466653e-05, "loss": 1.3605, "step": 12714 }, { "epoch": 0.77, "learning_rate": 3.8350335003319854e-05, "loss": 0.83, "step": 12716 }, { "epoch": 0.77, "learning_rate": 3.8356371099173055e-05, "loss": 1.5648, "step": 12718 }, { "epoch": 0.77, "learning_rate": 3.8362407195026256e-05, "loss": 0.938, "step": 12720 }, { "epoch": 0.77, "learning_rate": 3.8368443290879464e-05, "loss": 1.4444, "step": 12722 }, { "epoch": 0.77, "learning_rate": 3.8374479386732664e-05, "loss": 1.4936, "step": 12724 }, { "epoch": 0.77, "learning_rate": 3.8380515482585865e-05, "loss": 0.9493, "step": 12726 }, { "epoch": 0.77, "learning_rate": 3.8386551578439066e-05, "loss": 1.1297, "step": 12728 }, { "epoch": 0.77, "learning_rate": 3.839258767429227e-05, "loss": 1.0533, "step": 12730 }, { "epoch": 0.77, "learning_rate": 3.839862377014547e-05, "loss": 1.1693, "step": 12732 }, { "epoch": 0.77, "learning_rate": 3.840465986599867e-05, "loss": 1.324, "step": 12734 }, { "epoch": 0.77, "learning_rate": 3.8410695961851876e-05, "loss": 1.4099, "step": 12736 }, { "epoch": 0.77, "learning_rate": 3.841673205770508e-05, "loss": 1.1786, "step": 12738 }, { "epoch": 0.77, "learning_rate": 3.842276815355828e-05, "loss": 1.1839, "step": 12740 }, { "epoch": 0.77, "learning_rate": 3.8428804249411485e-05, "loss": 1.4328, "step": 12742 }, { "epoch": 0.77, "learning_rate": 3.8434840345264686e-05, "loss": 1.0292, "step": 12744 }, { "epoch": 0.77, "learning_rate": 3.844087644111789e-05, "loss": 1.0574, "step": 12746 }, { "epoch": 0.77, "learning_rate": 3.8446912536971095e-05, "loss": 1.1857, "step": 12748 }, { "epoch": 0.77, "learning_rate": 3.8452948632824295e-05, "loss": 1.1364, "step": 12750 }, { "epoch": 0.77, "learning_rate": 3.8458984728677496e-05, "loss": 1.2949, "step": 12752 }, { "epoch": 0.77, "learning_rate": 3.84650208245307e-05, "loss": 0.8893, "step": 12754 }, { "epoch": 0.77, "learning_rate": 3.84710569203839e-05, "loss": 1.3312, "step": 12756 }, { "epoch": 0.77, "learning_rate": 3.84770930162371e-05, "loss": 1.3589, "step": 12758 }, { "epoch": 0.77, "learning_rate": 3.84831291120903e-05, "loss": 1.3883, "step": 12760 }, { "epoch": 0.77, "learning_rate": 3.848916520794351e-05, "loss": 1.5475, "step": 12762 }, { "epoch": 0.77, "learning_rate": 3.849520130379671e-05, "loss": 1.2102, "step": 12764 }, { "epoch": 0.77, "learning_rate": 3.850123739964991e-05, "loss": 1.0596, "step": 12766 }, { "epoch": 0.77, "learning_rate": 3.850727349550311e-05, "loss": 1.3118, "step": 12768 }, { "epoch": 0.77, "learning_rate": 3.851330959135631e-05, "loss": 1.2884, "step": 12770 }, { "epoch": 0.77, "learning_rate": 3.851934568720951e-05, "loss": 1.1292, "step": 12772 }, { "epoch": 0.77, "learning_rate": 3.852538178306272e-05, "loss": 1.4722, "step": 12774 }, { "epoch": 0.77, "learning_rate": 3.853141787891592e-05, "loss": 1.3863, "step": 12776 }, { "epoch": 0.77, "learning_rate": 3.853745397476912e-05, "loss": 1.1056, "step": 12778 }, { "epoch": 0.77, "learning_rate": 3.854349007062232e-05, "loss": 0.9537, "step": 12780 }, { "epoch": 0.77, "learning_rate": 3.854952616647552e-05, "loss": 1.6818, "step": 12782 }, { "epoch": 0.77, "learning_rate": 3.855556226232872e-05, "loss": 0.9664, "step": 12784 }, { "epoch": 0.77, "learning_rate": 3.856159835818193e-05, "loss": 1.069, "step": 12786 }, { "epoch": 0.77, "learning_rate": 3.856763445403513e-05, "loss": 0.9336, "step": 12788 }, { "epoch": 0.77, "learning_rate": 3.857367054988834e-05, "loss": 1.5728, "step": 12790 }, { "epoch": 0.77, "learning_rate": 3.857970664574154e-05, "loss": 1.1247, "step": 12792 }, { "epoch": 0.77, "learning_rate": 3.858574274159474e-05, "loss": 1.1489, "step": 12794 }, { "epoch": 0.77, "learning_rate": 3.859177883744794e-05, "loss": 1.2289, "step": 12796 }, { "epoch": 0.77, "learning_rate": 3.859781493330114e-05, "loss": 1.1896, "step": 12798 }, { "epoch": 0.77, "learning_rate": 3.860385102915435e-05, "loss": 1.3935, "step": 12800 }, { "epoch": 0.77, "learning_rate": 3.860988712500755e-05, "loss": 1.4571, "step": 12802 }, { "epoch": 0.77, "learning_rate": 3.861592322086075e-05, "loss": 1.2232, "step": 12804 }, { "epoch": 0.77, "learning_rate": 3.862195931671395e-05, "loss": 0.9778, "step": 12806 }, { "epoch": 0.77, "learning_rate": 3.862799541256715e-05, "loss": 1.1797, "step": 12808 }, { "epoch": 0.77, "learning_rate": 3.8634031508420354e-05, "loss": 1.4802, "step": 12810 }, { "epoch": 0.77, "learning_rate": 3.8640067604273555e-05, "loss": 1.2501, "step": 12812 }, { "epoch": 0.77, "learning_rate": 3.864610370012676e-05, "loss": 1.4367, "step": 12814 }, { "epoch": 0.77, "learning_rate": 3.865213979597996e-05, "loss": 1.7658, "step": 12816 }, { "epoch": 0.77, "learning_rate": 3.8658175891833164e-05, "loss": 1.277, "step": 12818 }, { "epoch": 0.77, "learning_rate": 3.8664211987686365e-05, "loss": 1.0831, "step": 12820 }, { "epoch": 0.77, "learning_rate": 3.8670248083539565e-05, "loss": 1.1001, "step": 12822 }, { "epoch": 0.77, "learning_rate": 3.8676284179392766e-05, "loss": 1.2978, "step": 12824 }, { "epoch": 0.77, "learning_rate": 3.868232027524597e-05, "loss": 1.1818, "step": 12826 }, { "epoch": 0.77, "learning_rate": 3.8688356371099175e-05, "loss": 1.0119, "step": 12828 }, { "epoch": 0.77, "learning_rate": 3.8694392466952376e-05, "loss": 1.0716, "step": 12830 }, { "epoch": 0.77, "learning_rate": 3.8700428562805576e-05, "loss": 0.994, "step": 12832 }, { "epoch": 0.77, "learning_rate": 3.8706464658658784e-05, "loss": 1.455, "step": 12834 }, { "epoch": 0.77, "learning_rate": 3.8712500754511985e-05, "loss": 1.3367, "step": 12836 }, { "epoch": 0.77, "learning_rate": 3.8718536850365186e-05, "loss": 1.2245, "step": 12838 }, { "epoch": 0.77, "learning_rate": 3.872457294621839e-05, "loss": 1.6711, "step": 12840 }, { "epoch": 0.78, "learning_rate": 3.8730609042071594e-05, "loss": 1.1812, "step": 12842 }, { "epoch": 0.78, "learning_rate": 3.8736645137924795e-05, "loss": 1.3159, "step": 12844 }, { "epoch": 0.78, "learning_rate": 3.8742681233777996e-05, "loss": 1.0503, "step": 12846 }, { "epoch": 0.78, "learning_rate": 3.8748717329631196e-05, "loss": 1.5054, "step": 12848 }, { "epoch": 0.78, "learning_rate": 3.87547534254844e-05, "loss": 1.4275, "step": 12850 }, { "epoch": 0.78, "learning_rate": 3.87607895213376e-05, "loss": 1.1933, "step": 12852 }, { "epoch": 0.78, "learning_rate": 3.8766825617190806e-05, "loss": 1.3139, "step": 12854 }, { "epoch": 0.78, "learning_rate": 3.8772861713044006e-05, "loss": 1.3515, "step": 12856 }, { "epoch": 0.78, "learning_rate": 3.877889780889721e-05, "loss": 1.0863, "step": 12858 }, { "epoch": 0.78, "learning_rate": 3.878493390475041e-05, "loss": 1.3179, "step": 12860 }, { "epoch": 0.78, "learning_rate": 3.879097000060361e-05, "loss": 1.1643, "step": 12862 }, { "epoch": 0.78, "learning_rate": 3.879700609645681e-05, "loss": 1.2919, "step": 12864 }, { "epoch": 0.78, "learning_rate": 3.880304219231002e-05, "loss": 1.2027, "step": 12866 }, { "epoch": 0.78, "learning_rate": 3.880907828816322e-05, "loss": 1.1867, "step": 12868 }, { "epoch": 0.78, "learning_rate": 3.881511438401642e-05, "loss": 1.3971, "step": 12870 }, { "epoch": 0.78, "learning_rate": 3.882115047986962e-05, "loss": 1.5498, "step": 12872 }, { "epoch": 0.78, "learning_rate": 3.882718657572282e-05, "loss": 0.9742, "step": 12874 }, { "epoch": 0.78, "learning_rate": 3.883322267157603e-05, "loss": 1.1674, "step": 12876 }, { "epoch": 0.78, "learning_rate": 3.883925876742923e-05, "loss": 1.0619, "step": 12878 }, { "epoch": 0.78, "learning_rate": 3.884529486328243e-05, "loss": 1.3809, "step": 12880 }, { "epoch": 0.78, "learning_rate": 3.885133095913564e-05, "loss": 1.0287, "step": 12882 }, { "epoch": 0.78, "learning_rate": 3.885736705498884e-05, "loss": 1.5072, "step": 12884 }, { "epoch": 0.78, "learning_rate": 3.886340315084204e-05, "loss": 1.2799, "step": 12886 }, { "epoch": 0.78, "learning_rate": 3.886943924669524e-05, "loss": 1.2153, "step": 12888 }, { "epoch": 0.78, "learning_rate": 3.887547534254844e-05, "loss": 1.1828, "step": 12890 }, { "epoch": 0.78, "learning_rate": 3.888151143840164e-05, "loss": 1.4735, "step": 12892 }, { "epoch": 0.78, "learning_rate": 3.888754753425485e-05, "loss": 1.284, "step": 12894 }, { "epoch": 0.78, "learning_rate": 3.889358363010805e-05, "loss": 1.4224, "step": 12896 }, { "epoch": 0.78, "learning_rate": 3.889961972596125e-05, "loss": 1.1527, "step": 12898 }, { "epoch": 0.78, "learning_rate": 3.890565582181445e-05, "loss": 1.284, "step": 12900 }, { "epoch": 0.78, "learning_rate": 3.891169191766765e-05, "loss": 1.2865, "step": 12902 }, { "epoch": 0.78, "learning_rate": 3.891772801352085e-05, "loss": 1.4684, "step": 12904 }, { "epoch": 0.78, "learning_rate": 3.892376410937406e-05, "loss": 1.5055, "step": 12906 }, { "epoch": 0.78, "learning_rate": 3.892980020522726e-05, "loss": 1.1072, "step": 12908 }, { "epoch": 0.78, "learning_rate": 3.893583630108046e-05, "loss": 1.0269, "step": 12910 }, { "epoch": 0.78, "learning_rate": 3.894187239693366e-05, "loss": 1.3631, "step": 12912 }, { "epoch": 0.78, "learning_rate": 3.8947908492786864e-05, "loss": 1.2601, "step": 12914 }, { "epoch": 0.78, "learning_rate": 3.8953944588640065e-05, "loss": 1.221, "step": 12916 }, { "epoch": 0.78, "learning_rate": 3.8959980684493266e-05, "loss": 1.2358, "step": 12918 }, { "epoch": 0.78, "learning_rate": 3.896601678034647e-05, "loss": 1.4048, "step": 12920 }, { "epoch": 0.78, "learning_rate": 3.8972052876199674e-05, "loss": 1.2969, "step": 12922 }, { "epoch": 0.78, "learning_rate": 3.897808897205288e-05, "loss": 1.2622, "step": 12924 }, { "epoch": 0.78, "learning_rate": 3.898412506790608e-05, "loss": 1.0442, "step": 12926 }, { "epoch": 0.78, "learning_rate": 3.899016116375928e-05, "loss": 1.4157, "step": 12928 }, { "epoch": 0.78, "learning_rate": 3.8996197259612484e-05, "loss": 1.3294, "step": 12930 }, { "epoch": 0.78, "learning_rate": 3.900223335546569e-05, "loss": 1.1039, "step": 12932 }, { "epoch": 0.78, "learning_rate": 3.900826945131889e-05, "loss": 1.1067, "step": 12934 }, { "epoch": 0.78, "learning_rate": 3.9014305547172093e-05, "loss": 1.1041, "step": 12936 }, { "epoch": 0.78, "learning_rate": 3.9020341643025294e-05, "loss": 1.1301, "step": 12938 }, { "epoch": 0.78, "learning_rate": 3.9026377738878495e-05, "loss": 1.0821, "step": 12940 }, { "epoch": 0.78, "learning_rate": 3.9032413834731696e-05, "loss": 1.7091, "step": 12942 }, { "epoch": 0.78, "learning_rate": 3.90384499305849e-05, "loss": 1.5474, "step": 12944 }, { "epoch": 0.78, "learning_rate": 3.9044486026438104e-05, "loss": 1.3042, "step": 12946 }, { "epoch": 0.78, "learning_rate": 3.9050522122291305e-05, "loss": 0.9232, "step": 12948 }, { "epoch": 0.78, "learning_rate": 3.9056558218144506e-05, "loss": 1.2751, "step": 12950 }, { "epoch": 0.78, "learning_rate": 3.906259431399771e-05, "loss": 0.8502, "step": 12952 }, { "epoch": 0.78, "learning_rate": 3.906863040985091e-05, "loss": 0.9607, "step": 12954 }, { "epoch": 0.78, "learning_rate": 3.907466650570411e-05, "loss": 1.7876, "step": 12956 }, { "epoch": 0.78, "learning_rate": 3.908070260155731e-05, "loss": 1.0737, "step": 12958 }, { "epoch": 0.78, "learning_rate": 3.908673869741052e-05, "loss": 1.2639, "step": 12960 }, { "epoch": 0.78, "learning_rate": 3.909277479326372e-05, "loss": 1.1626, "step": 12962 }, { "epoch": 0.78, "learning_rate": 3.909881088911692e-05, "loss": 1.2621, "step": 12964 }, { "epoch": 0.78, "learning_rate": 3.9104846984970126e-05, "loss": 1.1014, "step": 12966 }, { "epoch": 0.78, "learning_rate": 3.911088308082333e-05, "loss": 1.343, "step": 12968 }, { "epoch": 0.78, "learning_rate": 3.911691917667653e-05, "loss": 1.3814, "step": 12970 }, { "epoch": 0.78, "learning_rate": 3.9122955272529735e-05, "loss": 1.0097, "step": 12972 }, { "epoch": 0.78, "learning_rate": 3.9128991368382936e-05, "loss": 1.1205, "step": 12974 }, { "epoch": 0.78, "learning_rate": 3.913502746423614e-05, "loss": 1.0278, "step": 12976 }, { "epoch": 0.78, "learning_rate": 3.914106356008934e-05, "loss": 1.3275, "step": 12978 }, { "epoch": 0.78, "learning_rate": 3.914709965594254e-05, "loss": 1.1683, "step": 12980 }, { "epoch": 0.78, "learning_rate": 3.915313575179574e-05, "loss": 1.4359, "step": 12982 }, { "epoch": 0.78, "learning_rate": 3.915917184764894e-05, "loss": 0.9821, "step": 12984 }, { "epoch": 0.78, "learning_rate": 3.916520794350215e-05, "loss": 1.2471, "step": 12986 }, { "epoch": 0.78, "learning_rate": 3.917124403935535e-05, "loss": 1.1889, "step": 12988 }, { "epoch": 0.78, "learning_rate": 3.917728013520855e-05, "loss": 1.3917, "step": 12990 }, { "epoch": 0.78, "learning_rate": 3.918331623106175e-05, "loss": 1.2833, "step": 12992 }, { "epoch": 0.78, "learning_rate": 3.918935232691495e-05, "loss": 1.1723, "step": 12994 }, { "epoch": 0.78, "learning_rate": 3.919538842276815e-05, "loss": 1.2242, "step": 12996 }, { "epoch": 0.78, "learning_rate": 3.920142451862136e-05, "loss": 0.9884, "step": 12998 }, { "epoch": 0.78, "learning_rate": 3.920746061447456e-05, "loss": 1.4157, "step": 13000 }, { "epoch": 0.78, "learning_rate": 3.921349671032776e-05, "loss": 1.2768, "step": 13002 }, { "epoch": 0.78, "learning_rate": 3.921953280618096e-05, "loss": 1.0133, "step": 13004 }, { "epoch": 0.79, "learning_rate": 3.922556890203416e-05, "loss": 1.168, "step": 13006 }, { "epoch": 0.79, "learning_rate": 3.9231604997887364e-05, "loss": 1.2656, "step": 13008 }, { "epoch": 0.79, "learning_rate": 3.923764109374057e-05, "loss": 1.3397, "step": 13010 }, { "epoch": 0.79, "learning_rate": 3.924367718959377e-05, "loss": 1.1595, "step": 13012 }, { "epoch": 0.79, "learning_rate": 3.924971328544698e-05, "loss": 0.9205, "step": 13014 }, { "epoch": 0.79, "learning_rate": 3.925574938130018e-05, "loss": 1.1547, "step": 13016 }, { "epoch": 0.79, "learning_rate": 3.926178547715338e-05, "loss": 1.103, "step": 13018 }, { "epoch": 0.79, "learning_rate": 3.926782157300658e-05, "loss": 1.4979, "step": 13020 }, { "epoch": 0.79, "learning_rate": 3.927385766885978e-05, "loss": 1.4939, "step": 13022 }, { "epoch": 0.79, "learning_rate": 3.9279893764712984e-05, "loss": 1.4543, "step": 13024 }, { "epoch": 0.79, "learning_rate": 3.928592986056619e-05, "loss": 1.5196, "step": 13026 }, { "epoch": 0.79, "learning_rate": 3.929196595641939e-05, "loss": 1.0195, "step": 13028 }, { "epoch": 0.79, "learning_rate": 3.929800205227259e-05, "loss": 1.1556, "step": 13030 }, { "epoch": 0.79, "learning_rate": 3.9304038148125794e-05, "loss": 1.3122, "step": 13032 }, { "epoch": 0.79, "learning_rate": 3.9310074243978995e-05, "loss": 0.8892, "step": 13034 }, { "epoch": 0.79, "learning_rate": 3.9316110339832195e-05, "loss": 1.4309, "step": 13036 }, { "epoch": 0.79, "learning_rate": 3.93221464356854e-05, "loss": 0.9464, "step": 13038 }, { "epoch": 0.79, "learning_rate": 3.9328182531538604e-05, "loss": 1.4233, "step": 13040 }, { "epoch": 0.79, "learning_rate": 3.9334218627391805e-05, "loss": 1.4368, "step": 13042 }, { "epoch": 0.79, "learning_rate": 3.9340254723245005e-05, "loss": 1.4586, "step": 13044 }, { "epoch": 0.79, "learning_rate": 3.9346290819098206e-05, "loss": 1.0417, "step": 13046 }, { "epoch": 0.79, "learning_rate": 3.935232691495141e-05, "loss": 1.5024, "step": 13048 }, { "epoch": 0.79, "learning_rate": 3.935836301080461e-05, "loss": 0.9366, "step": 13050 }, { "epoch": 0.79, "learning_rate": 3.9364399106657815e-05, "loss": 1.1313, "step": 13052 }, { "epoch": 0.79, "learning_rate": 3.9370435202511016e-05, "loss": 1.3719, "step": 13054 }, { "epoch": 0.79, "learning_rate": 3.937647129836422e-05, "loss": 0.9689, "step": 13056 }, { "epoch": 0.79, "learning_rate": 3.9382507394217425e-05, "loss": 1.5281, "step": 13058 }, { "epoch": 0.79, "learning_rate": 3.9388543490070626e-05, "loss": 1.5339, "step": 13060 }, { "epoch": 0.79, "learning_rate": 3.9394579585923826e-05, "loss": 1.5128, "step": 13062 }, { "epoch": 0.79, "learning_rate": 3.9400615681777034e-05, "loss": 1.506, "step": 13064 }, { "epoch": 0.79, "learning_rate": 3.9406651777630235e-05, "loss": 1.5639, "step": 13066 }, { "epoch": 0.79, "learning_rate": 3.9412687873483436e-05, "loss": 1.836, "step": 13068 }, { "epoch": 0.79, "learning_rate": 3.9418723969336636e-05, "loss": 1.8157, "step": 13070 }, { "epoch": 0.79, "learning_rate": 3.942476006518984e-05, "loss": 1.3364, "step": 13072 }, { "epoch": 0.79, "learning_rate": 3.943079616104304e-05, "loss": 1.1565, "step": 13074 }, { "epoch": 0.79, "learning_rate": 3.943683225689624e-05, "loss": 1.2082, "step": 13076 }, { "epoch": 0.79, "learning_rate": 3.9442868352749446e-05, "loss": 0.9899, "step": 13078 }, { "epoch": 0.79, "learning_rate": 3.944890444860265e-05, "loss": 1.0565, "step": 13080 }, { "epoch": 0.79, "learning_rate": 3.945494054445585e-05, "loss": 1.2123, "step": 13082 }, { "epoch": 0.79, "learning_rate": 3.946097664030905e-05, "loss": 1.3244, "step": 13084 }, { "epoch": 0.79, "learning_rate": 3.946701273616225e-05, "loss": 1.3506, "step": 13086 }, { "epoch": 0.79, "learning_rate": 3.947304883201545e-05, "loss": 1.6058, "step": 13088 }, { "epoch": 0.79, "learning_rate": 3.947908492786866e-05, "loss": 1.2629, "step": 13090 }, { "epoch": 0.79, "learning_rate": 3.948512102372186e-05, "loss": 1.1267, "step": 13092 }, { "epoch": 0.79, "learning_rate": 3.949115711957506e-05, "loss": 1.0497, "step": 13094 }, { "epoch": 0.79, "learning_rate": 3.949719321542826e-05, "loss": 1.3628, "step": 13096 }, { "epoch": 0.79, "learning_rate": 3.950322931128146e-05, "loss": 1.4278, "step": 13098 }, { "epoch": 0.79, "learning_rate": 3.950926540713467e-05, "loss": 0.7913, "step": 13100 }, { "epoch": 0.79, "learning_rate": 3.951530150298787e-05, "loss": 0.8095, "step": 13102 }, { "epoch": 0.79, "learning_rate": 3.952133759884107e-05, "loss": 1.1569, "step": 13104 }, { "epoch": 0.79, "learning_rate": 3.952737369469428e-05, "loss": 1.1903, "step": 13106 }, { "epoch": 0.79, "learning_rate": 3.953340979054748e-05, "loss": 1.2832, "step": 13108 }, { "epoch": 0.79, "learning_rate": 3.953944588640068e-05, "loss": 1.9353, "step": 13110 }, { "epoch": 0.79, "learning_rate": 3.954548198225388e-05, "loss": 1.6348, "step": 13112 }, { "epoch": 0.79, "learning_rate": 3.955151807810708e-05, "loss": 0.9524, "step": 13114 }, { "epoch": 0.79, "learning_rate": 3.955755417396028e-05, "loss": 1.1662, "step": 13116 }, { "epoch": 0.79, "learning_rate": 3.956359026981349e-05, "loss": 0.8499, "step": 13118 }, { "epoch": 0.79, "learning_rate": 3.956962636566669e-05, "loss": 1.2927, "step": 13120 }, { "epoch": 0.79, "learning_rate": 3.957566246151989e-05, "loss": 1.0748, "step": 13122 }, { "epoch": 0.79, "learning_rate": 3.958169855737309e-05, "loss": 1.337, "step": 13124 }, { "epoch": 0.79, "learning_rate": 3.958773465322629e-05, "loss": 1.1935, "step": 13126 }, { "epoch": 0.79, "learning_rate": 3.9593770749079494e-05, "loss": 1.5866, "step": 13128 }, { "epoch": 0.79, "learning_rate": 3.95998068449327e-05, "loss": 1.5578, "step": 13130 }, { "epoch": 0.79, "learning_rate": 3.96058429407859e-05, "loss": 1.1955, "step": 13132 }, { "epoch": 0.79, "learning_rate": 3.96118790366391e-05, "loss": 0.9014, "step": 13134 }, { "epoch": 0.79, "learning_rate": 3.9617915132492304e-05, "loss": 1.1155, "step": 13136 }, { "epoch": 0.79, "learning_rate": 3.9623951228345505e-05, "loss": 1.4509, "step": 13138 }, { "epoch": 0.79, "learning_rate": 3.9629987324198706e-05, "loss": 1.6336, "step": 13140 }, { "epoch": 0.79, "learning_rate": 3.9636023420051906e-05, "loss": 1.3241, "step": 13142 }, { "epoch": 0.79, "learning_rate": 3.9642059515905114e-05, "loss": 1.229, "step": 13144 }, { "epoch": 0.79, "learning_rate": 3.9648095611758315e-05, "loss": 0.9484, "step": 13146 }, { "epoch": 0.79, "learning_rate": 3.965413170761152e-05, "loss": 1.1645, "step": 13148 }, { "epoch": 0.79, "learning_rate": 3.966016780346472e-05, "loss": 1.1162, "step": 13150 }, { "epoch": 0.79, "learning_rate": 3.9666203899317924e-05, "loss": 0.9721, "step": 13152 }, { "epoch": 0.79, "learning_rate": 3.9672239995171125e-05, "loss": 1.2496, "step": 13154 }, { "epoch": 0.79, "learning_rate": 3.967827609102433e-05, "loss": 0.9973, "step": 13156 }, { "epoch": 0.79, "learning_rate": 3.968431218687753e-05, "loss": 1.4084, "step": 13158 }, { "epoch": 0.79, "learning_rate": 3.9690348282730734e-05, "loss": 0.9724, "step": 13160 }, { "epoch": 0.79, "learning_rate": 3.9696384378583935e-05, "loss": 1.2194, "step": 13162 }, { "epoch": 0.79, "learning_rate": 3.9702420474437136e-05, "loss": 1.3483, "step": 13164 }, { "epoch": 0.79, "learning_rate": 3.970845657029034e-05, "loss": 1.4739, "step": 13166 }, { "epoch": 0.79, "learning_rate": 3.971449266614354e-05, "loss": 1.1367, "step": 13168 }, { "epoch": 0.79, "learning_rate": 3.9720528761996745e-05, "loss": 1.137, "step": 13170 }, { "epoch": 0.8, "learning_rate": 3.9726564857849946e-05, "loss": 1.2865, "step": 13172 }, { "epoch": 0.8, "learning_rate": 3.973260095370315e-05, "loss": 1.4738, "step": 13174 }, { "epoch": 0.8, "learning_rate": 3.973863704955635e-05, "loss": 0.9742, "step": 13176 }, { "epoch": 0.8, "learning_rate": 3.974467314540955e-05, "loss": 1.2814, "step": 13178 }, { "epoch": 0.8, "learning_rate": 3.975070924126275e-05, "loss": 1.4577, "step": 13180 }, { "epoch": 0.8, "learning_rate": 3.975674533711595e-05, "loss": 1.0102, "step": 13182 }, { "epoch": 0.8, "learning_rate": 3.976278143296916e-05, "loss": 1.1609, "step": 13184 }, { "epoch": 0.8, "learning_rate": 3.976881752882236e-05, "loss": 0.914, "step": 13186 }, { "epoch": 0.8, "learning_rate": 3.977485362467556e-05, "loss": 1.1926, "step": 13188 }, { "epoch": 0.8, "learning_rate": 3.978088972052876e-05, "loss": 1.04, "step": 13190 }, { "epoch": 0.8, "learning_rate": 3.978692581638197e-05, "loss": 1.5901, "step": 13192 }, { "epoch": 0.8, "learning_rate": 3.979296191223517e-05, "loss": 1.2194, "step": 13194 }, { "epoch": 0.8, "learning_rate": 3.9798998008088376e-05, "loss": 1.3892, "step": 13196 }, { "epoch": 0.8, "learning_rate": 3.980503410394158e-05, "loss": 1.1449, "step": 13198 }, { "epoch": 0.8, "learning_rate": 3.981107019979478e-05, "loss": 1.2288, "step": 13200 }, { "epoch": 0.8, "learning_rate": 3.981710629564798e-05, "loss": 0.9651, "step": 13202 }, { "epoch": 0.8, "learning_rate": 3.982314239150118e-05, "loss": 1.2174, "step": 13204 }, { "epoch": 0.8, "learning_rate": 3.982917848735438e-05, "loss": 0.9547, "step": 13206 }, { "epoch": 0.8, "learning_rate": 3.983521458320758e-05, "loss": 1.5511, "step": 13208 }, { "epoch": 0.8, "learning_rate": 3.984125067906079e-05, "loss": 1.1703, "step": 13210 }, { "epoch": 0.8, "learning_rate": 3.984728677491399e-05, "loss": 0.9408, "step": 13212 }, { "epoch": 0.8, "learning_rate": 3.985332287076719e-05, "loss": 1.4123, "step": 13214 }, { "epoch": 0.8, "learning_rate": 3.985935896662039e-05, "loss": 1.3852, "step": 13216 }, { "epoch": 0.8, "learning_rate": 3.986539506247359e-05, "loss": 1.3875, "step": 13218 }, { "epoch": 0.8, "learning_rate": 3.987143115832679e-05, "loss": 1.2918, "step": 13220 }, { "epoch": 0.8, "learning_rate": 3.987746725418e-05, "loss": 0.9553, "step": 13222 }, { "epoch": 0.8, "learning_rate": 3.98835033500332e-05, "loss": 1.4355, "step": 13224 }, { "epoch": 0.8, "learning_rate": 3.98895394458864e-05, "loss": 1.2888, "step": 13226 }, { "epoch": 0.8, "learning_rate": 3.98955755417396e-05, "loss": 1.039, "step": 13228 }, { "epoch": 0.8, "learning_rate": 3.9901611637592804e-05, "loss": 1.2602, "step": 13230 }, { "epoch": 0.8, "learning_rate": 3.9907647733446004e-05, "loss": 1.0526, "step": 13232 }, { "epoch": 0.8, "learning_rate": 3.991368382929921e-05, "loss": 1.3736, "step": 13234 }, { "epoch": 0.8, "learning_rate": 3.991971992515241e-05, "loss": 1.0877, "step": 13236 }, { "epoch": 0.8, "learning_rate": 3.992575602100562e-05, "loss": 1.3679, "step": 13238 }, { "epoch": 0.8, "learning_rate": 3.993179211685882e-05, "loss": 1.395, "step": 13240 }, { "epoch": 0.8, "learning_rate": 3.993782821271202e-05, "loss": 1.3932, "step": 13242 }, { "epoch": 0.8, "learning_rate": 3.994386430856522e-05, "loss": 1.2744, "step": 13244 }, { "epoch": 0.8, "learning_rate": 3.9949900404418424e-05, "loss": 1.2789, "step": 13246 }, { "epoch": 0.8, "learning_rate": 3.9955936500271624e-05, "loss": 1.3906, "step": 13248 }, { "epoch": 0.8, "learning_rate": 3.996197259612483e-05, "loss": 1.3598, "step": 13250 }, { "epoch": 0.8, "learning_rate": 3.996800869197803e-05, "loss": 1.3476, "step": 13252 }, { "epoch": 0.8, "learning_rate": 3.9974044787831234e-05, "loss": 1.2551, "step": 13254 }, { "epoch": 0.8, "learning_rate": 3.9980080883684434e-05, "loss": 1.0407, "step": 13256 }, { "epoch": 0.8, "learning_rate": 3.9986116979537635e-05, "loss": 1.2777, "step": 13258 }, { "epoch": 0.8, "learning_rate": 3.9992153075390836e-05, "loss": 1.027, "step": 13260 }, { "epoch": 0.8, "learning_rate": 3.9998189171244044e-05, "loss": 1.508, "step": 13262 }, { "epoch": 0.8, "learning_rate": 4.0004225267097245e-05, "loss": 1.0335, "step": 13264 }, { "epoch": 0.8, "learning_rate": 4.0010261362950445e-05, "loss": 1.3498, "step": 13266 }, { "epoch": 0.8, "learning_rate": 4.0016297458803646e-05, "loss": 1.2353, "step": 13268 }, { "epoch": 0.8, "learning_rate": 4.002233355465685e-05, "loss": 1.5689, "step": 13270 }, { "epoch": 0.8, "learning_rate": 4.002836965051005e-05, "loss": 1.4445, "step": 13272 }, { "epoch": 0.8, "learning_rate": 4.003440574636325e-05, "loss": 1.0541, "step": 13274 }, { "epoch": 0.8, "learning_rate": 4.0040441842216456e-05, "loss": 1.3711, "step": 13276 }, { "epoch": 0.8, "learning_rate": 4.004647793806966e-05, "loss": 0.9596, "step": 13278 }, { "epoch": 0.8, "learning_rate": 4.005251403392286e-05, "loss": 0.8725, "step": 13280 }, { "epoch": 0.8, "learning_rate": 4.0058550129776065e-05, "loss": 1.0902, "step": 13282 }, { "epoch": 0.8, "learning_rate": 4.0064586225629266e-05, "loss": 1.4067, "step": 13284 }, { "epoch": 0.8, "learning_rate": 4.007062232148247e-05, "loss": 1.332, "step": 13286 }, { "epoch": 0.8, "learning_rate": 4.0076658417335675e-05, "loss": 1.4131, "step": 13288 }, { "epoch": 0.8, "learning_rate": 4.0082694513188875e-05, "loss": 1.2987, "step": 13290 }, { "epoch": 0.8, "learning_rate": 4.0088730609042076e-05, "loss": 1.0937, "step": 13292 }, { "epoch": 0.8, "learning_rate": 4.009476670489528e-05, "loss": 1.6482, "step": 13294 }, { "epoch": 0.8, "learning_rate": 4.010080280074848e-05, "loss": 1.3433, "step": 13296 }, { "epoch": 0.8, "learning_rate": 4.010683889660168e-05, "loss": 1.1812, "step": 13298 }, { "epoch": 0.8, "learning_rate": 4.011287499245488e-05, "loss": 0.9601, "step": 13300 }, { "epoch": 0.8, "learning_rate": 4.011891108830809e-05, "loss": 1.3206, "step": 13302 }, { "epoch": 0.8, "learning_rate": 4.012494718416129e-05, "loss": 1.5304, "step": 13304 }, { "epoch": 0.8, "learning_rate": 4.013098328001449e-05, "loss": 1.0978, "step": 13306 }, { "epoch": 0.8, "learning_rate": 4.013701937586769e-05, "loss": 1.9699, "step": 13308 }, { "epoch": 0.8, "learning_rate": 4.014305547172089e-05, "loss": 0.9693, "step": 13310 }, { "epoch": 0.8, "learning_rate": 4.014909156757409e-05, "loss": 1.3793, "step": 13312 }, { "epoch": 0.8, "learning_rate": 4.015512766342729e-05, "loss": 0.8979, "step": 13314 }, { "epoch": 0.8, "learning_rate": 4.01611637592805e-05, "loss": 1.4893, "step": 13316 }, { "epoch": 0.8, "learning_rate": 4.01671998551337e-05, "loss": 0.8219, "step": 13318 }, { "epoch": 0.8, "learning_rate": 4.01732359509869e-05, "loss": 1.1336, "step": 13320 }, { "epoch": 0.8, "learning_rate": 4.01792720468401e-05, "loss": 1.0303, "step": 13322 }, { "epoch": 0.8, "learning_rate": 4.018530814269331e-05, "loss": 1.2017, "step": 13324 }, { "epoch": 0.8, "learning_rate": 4.019134423854651e-05, "loss": 1.5938, "step": 13326 }, { "epoch": 0.8, "learning_rate": 4.019738033439971e-05, "loss": 1.5642, "step": 13328 }, { "epoch": 0.8, "learning_rate": 4.020341643025292e-05, "loss": 1.7101, "step": 13330 }, { "epoch": 0.8, "learning_rate": 4.020945252610612e-05, "loss": 1.1658, "step": 13332 }, { "epoch": 0.8, "learning_rate": 4.021548862195932e-05, "loss": 1.015, "step": 13334 }, { "epoch": 0.8, "learning_rate": 4.022152471781252e-05, "loss": 1.0734, "step": 13336 }, { "epoch": 0.81, "learning_rate": 4.022756081366572e-05, "loss": 1.2523, "step": 13338 }, { "epoch": 0.81, "learning_rate": 4.023359690951892e-05, "loss": 1.1894, "step": 13340 }, { "epoch": 0.81, "learning_rate": 4.023963300537213e-05, "loss": 1.4494, "step": 13342 }, { "epoch": 0.81, "learning_rate": 4.024566910122533e-05, "loss": 1.2027, "step": 13344 }, { "epoch": 0.81, "learning_rate": 4.025170519707853e-05, "loss": 1.849, "step": 13346 }, { "epoch": 0.81, "learning_rate": 4.025774129293173e-05, "loss": 1.2279, "step": 13348 }, { "epoch": 0.81, "learning_rate": 4.0263777388784934e-05, "loss": 2.3053, "step": 13350 }, { "epoch": 0.81, "learning_rate": 4.0269813484638135e-05, "loss": 1.2859, "step": 13352 }, { "epoch": 0.81, "learning_rate": 4.027584958049134e-05, "loss": 1.3865, "step": 13354 }, { "epoch": 0.81, "learning_rate": 4.028188567634454e-05, "loss": 1.2624, "step": 13356 }, { "epoch": 0.81, "learning_rate": 4.0287921772197744e-05, "loss": 1.4046, "step": 13358 }, { "epoch": 0.81, "learning_rate": 4.0293957868050945e-05, "loss": 1.0409, "step": 13360 }, { "epoch": 0.81, "learning_rate": 4.0299993963904146e-05, "loss": 1.2555, "step": 13362 }, { "epoch": 0.81, "learning_rate": 4.0306030059757346e-05, "loss": 1.4289, "step": 13364 }, { "epoch": 0.81, "learning_rate": 4.031206615561055e-05, "loss": 1.3436, "step": 13366 }, { "epoch": 0.81, "learning_rate": 4.0318102251463755e-05, "loss": 1.6777, "step": 13368 }, { "epoch": 0.81, "learning_rate": 4.0324138347316956e-05, "loss": 1.313, "step": 13370 }, { "epoch": 0.81, "learning_rate": 4.033017444317016e-05, "loss": 1.5603, "step": 13372 }, { "epoch": 0.81, "learning_rate": 4.0336210539023364e-05, "loss": 1.0608, "step": 13374 }, { "epoch": 0.81, "learning_rate": 4.0342246634876565e-05, "loss": 1.2134, "step": 13376 }, { "epoch": 0.81, "learning_rate": 4.0348282730729766e-05, "loss": 1.6464, "step": 13378 }, { "epoch": 0.81, "learning_rate": 4.035431882658297e-05, "loss": 1.3685, "step": 13380 }, { "epoch": 0.81, "learning_rate": 4.0360354922436174e-05, "loss": 1.054, "step": 13382 }, { "epoch": 0.81, "learning_rate": 4.0366391018289375e-05, "loss": 0.8763, "step": 13384 }, { "epoch": 0.81, "learning_rate": 4.0372427114142576e-05, "loss": 1.261, "step": 13386 }, { "epoch": 0.81, "learning_rate": 4.0378463209995777e-05, "loss": 1.2824, "step": 13388 }, { "epoch": 0.81, "learning_rate": 4.038449930584898e-05, "loss": 1.1567, "step": 13390 }, { "epoch": 0.81, "learning_rate": 4.039053540170218e-05, "loss": 0.8585, "step": 13392 }, { "epoch": 0.81, "learning_rate": 4.0396571497555386e-05, "loss": 0.8821, "step": 13394 }, { "epoch": 0.81, "learning_rate": 4.040260759340859e-05, "loss": 1.3032, "step": 13396 }, { "epoch": 0.81, "learning_rate": 4.040864368926179e-05, "loss": 1.2255, "step": 13398 }, { "epoch": 0.81, "learning_rate": 4.041467978511499e-05, "loss": 1.1256, "step": 13400 }, { "epoch": 0.81, "learning_rate": 4.042071588096819e-05, "loss": 1.1049, "step": 13402 }, { "epoch": 0.81, "learning_rate": 4.042675197682139e-05, "loss": 1.1553, "step": 13404 }, { "epoch": 0.81, "learning_rate": 4.043278807267459e-05, "loss": 1.346, "step": 13406 }, { "epoch": 0.81, "learning_rate": 4.04388241685278e-05, "loss": 1.5689, "step": 13408 }, { "epoch": 0.81, "learning_rate": 4.0444860264381e-05, "loss": 1.2603, "step": 13410 }, { "epoch": 0.81, "learning_rate": 4.04508963602342e-05, "loss": 1.4148, "step": 13412 }, { "epoch": 0.81, "learning_rate": 4.04569324560874e-05, "loss": 1.7918, "step": 13414 }, { "epoch": 0.81, "learning_rate": 4.046296855194061e-05, "loss": 1.1238, "step": 13416 }, { "epoch": 0.81, "learning_rate": 4.046900464779381e-05, "loss": 1.0865, "step": 13418 }, { "epoch": 0.81, "learning_rate": 4.047504074364702e-05, "loss": 1.3601, "step": 13420 }, { "epoch": 0.81, "learning_rate": 4.048107683950022e-05, "loss": 1.3667, "step": 13422 }, { "epoch": 0.81, "learning_rate": 4.048711293535342e-05, "loss": 1.2504, "step": 13424 }, { "epoch": 0.81, "learning_rate": 4.049314903120662e-05, "loss": 1.0849, "step": 13426 }, { "epoch": 0.81, "learning_rate": 4.049918512705982e-05, "loss": 1.0201, "step": 13428 }, { "epoch": 0.81, "learning_rate": 4.050522122291302e-05, "loss": 1.4344, "step": 13430 }, { "epoch": 0.81, "learning_rate": 4.051125731876622e-05, "loss": 0.8898, "step": 13432 }, { "epoch": 0.81, "learning_rate": 4.051729341461943e-05, "loss": 1.1056, "step": 13434 }, { "epoch": 0.81, "learning_rate": 4.052332951047263e-05, "loss": 1.3745, "step": 13436 }, { "epoch": 0.81, "learning_rate": 4.052936560632583e-05, "loss": 1.1561, "step": 13438 }, { "epoch": 0.81, "learning_rate": 4.053540170217903e-05, "loss": 0.979, "step": 13440 }, { "epoch": 0.81, "learning_rate": 4.054143779803223e-05, "loss": 1.1644, "step": 13442 }, { "epoch": 0.81, "learning_rate": 4.0547473893885433e-05, "loss": 1.2456, "step": 13444 }, { "epoch": 0.81, "learning_rate": 4.055350998973864e-05, "loss": 1.0339, "step": 13446 }, { "epoch": 0.81, "learning_rate": 4.055954608559184e-05, "loss": 1.0408, "step": 13448 }, { "epoch": 0.81, "learning_rate": 4.056558218144504e-05, "loss": 1.1593, "step": 13450 }, { "epoch": 0.81, "learning_rate": 4.0571618277298243e-05, "loss": 1.0964, "step": 13452 }, { "epoch": 0.81, "learning_rate": 4.0577654373151444e-05, "loss": 1.1417, "step": 13454 }, { "epoch": 0.81, "learning_rate": 4.0583690469004645e-05, "loss": 1.2536, "step": 13456 }, { "epoch": 0.81, "learning_rate": 4.058972656485785e-05, "loss": 1.2373, "step": 13458 }, { "epoch": 0.81, "learning_rate": 4.0595762660711053e-05, "loss": 1.2758, "step": 13460 }, { "epoch": 0.81, "learning_rate": 4.0601798756564254e-05, "loss": 1.1479, "step": 13462 }, { "epoch": 0.81, "learning_rate": 4.060783485241746e-05, "loss": 1.1848, "step": 13464 }, { "epoch": 0.81, "learning_rate": 4.061387094827066e-05, "loss": 1.3325, "step": 13466 }, { "epoch": 0.81, "learning_rate": 4.0619907044123864e-05, "loss": 1.1458, "step": 13468 }, { "epoch": 0.81, "learning_rate": 4.0625943139977064e-05, "loss": 1.2801, "step": 13470 }, { "epoch": 0.81, "learning_rate": 4.0631979235830265e-05, "loss": 1.5064, "step": 13472 }, { "epoch": 0.81, "learning_rate": 4.063801533168347e-05, "loss": 1.4272, "step": 13474 }, { "epoch": 0.81, "learning_rate": 4.0644051427536674e-05, "loss": 1.7506, "step": 13476 }, { "epoch": 0.81, "learning_rate": 4.0650087523389874e-05, "loss": 0.9046, "step": 13478 }, { "epoch": 0.81, "learning_rate": 4.0656123619243075e-05, "loss": 1.4447, "step": 13480 }, { "epoch": 0.81, "learning_rate": 4.0662159715096276e-05, "loss": 1.0871, "step": 13482 }, { "epoch": 0.81, "learning_rate": 4.066819581094948e-05, "loss": 1.0645, "step": 13484 }, { "epoch": 0.81, "learning_rate": 4.0674231906802684e-05, "loss": 1.2262, "step": 13486 }, { "epoch": 0.81, "learning_rate": 4.0680268002655885e-05, "loss": 0.965, "step": 13488 }, { "epoch": 0.81, "learning_rate": 4.0686304098509086e-05, "loss": 1.1271, "step": 13490 }, { "epoch": 0.81, "learning_rate": 4.069234019436229e-05, "loss": 1.1241, "step": 13492 }, { "epoch": 0.81, "learning_rate": 4.069837629021549e-05, "loss": 1.0496, "step": 13494 }, { "epoch": 0.81, "learning_rate": 4.070441238606869e-05, "loss": 1.5637, "step": 13496 }, { "epoch": 0.81, "learning_rate": 4.071044848192189e-05, "loss": 1.155, "step": 13498 }, { "epoch": 0.81, "learning_rate": 4.07164845777751e-05, "loss": 1.1406, "step": 13500 }, { "epoch": 0.81, "learning_rate": 4.07225206736283e-05, "loss": 0.9896, "step": 13502 }, { "epoch": 0.82, "learning_rate": 4.07285567694815e-05, "loss": 1.3972, "step": 13504 }, { "epoch": 0.82, "learning_rate": 4.0734592865334706e-05, "loss": 1.0893, "step": 13506 }, { "epoch": 0.82, "learning_rate": 4.074062896118791e-05, "loss": 1.2485, "step": 13508 }, { "epoch": 0.82, "learning_rate": 4.074666505704111e-05, "loss": 1.3541, "step": 13510 }, { "epoch": 0.82, "learning_rate": 4.0752701152894315e-05, "loss": 1.2324, "step": 13512 }, { "epoch": 0.82, "learning_rate": 4.0758737248747516e-05, "loss": 1.083, "step": 13514 }, { "epoch": 0.82, "learning_rate": 4.076477334460072e-05, "loss": 1.4253, "step": 13516 }, { "epoch": 0.82, "learning_rate": 4.077080944045392e-05, "loss": 1.2626, "step": 13518 }, { "epoch": 0.82, "learning_rate": 4.077684553630712e-05, "loss": 0.9598, "step": 13520 }, { "epoch": 0.82, "learning_rate": 4.078288163216032e-05, "loss": 1.2657, "step": 13522 }, { "epoch": 0.82, "learning_rate": 4.078891772801352e-05, "loss": 1.0862, "step": 13524 }, { "epoch": 0.82, "learning_rate": 4.079495382386673e-05, "loss": 1.31, "step": 13526 }, { "epoch": 0.82, "learning_rate": 4.080098991971993e-05, "loss": 1.2, "step": 13528 }, { "epoch": 0.82, "learning_rate": 4.080702601557313e-05, "loss": 1.5181, "step": 13530 }, { "epoch": 0.82, "learning_rate": 4.081306211142633e-05, "loss": 1.8191, "step": 13532 }, { "epoch": 0.82, "learning_rate": 4.081909820727953e-05, "loss": 1.2368, "step": 13534 }, { "epoch": 0.82, "learning_rate": 4.082513430313273e-05, "loss": 0.9385, "step": 13536 }, { "epoch": 0.82, "learning_rate": 4.083117039898593e-05, "loss": 1.2792, "step": 13538 }, { "epoch": 0.82, "learning_rate": 4.083720649483914e-05, "loss": 1.1372, "step": 13540 }, { "epoch": 0.82, "learning_rate": 4.084324259069234e-05, "loss": 0.9194, "step": 13542 }, { "epoch": 0.82, "learning_rate": 4.084927868654554e-05, "loss": 1.21, "step": 13544 }, { "epoch": 0.82, "learning_rate": 4.085531478239874e-05, "loss": 1.4541, "step": 13546 }, { "epoch": 0.82, "learning_rate": 4.086135087825195e-05, "loss": 1.4885, "step": 13548 }, { "epoch": 0.82, "learning_rate": 4.086738697410515e-05, "loss": 1.3056, "step": 13550 }, { "epoch": 0.82, "learning_rate": 4.087342306995835e-05, "loss": 1.1371, "step": 13552 }, { "epoch": 0.82, "learning_rate": 4.087945916581156e-05, "loss": 1.168, "step": 13554 }, { "epoch": 0.82, "learning_rate": 4.088549526166476e-05, "loss": 1.3254, "step": 13556 }, { "epoch": 0.82, "learning_rate": 4.089153135751796e-05, "loss": 1.2163, "step": 13558 }, { "epoch": 0.82, "learning_rate": 4.089756745337116e-05, "loss": 1.1343, "step": 13560 }, { "epoch": 0.82, "learning_rate": 4.090360354922436e-05, "loss": 1.4792, "step": 13562 }, { "epoch": 0.82, "learning_rate": 4.0909639645077564e-05, "loss": 1.4143, "step": 13564 }, { "epoch": 0.82, "learning_rate": 4.091567574093077e-05, "loss": 1.1589, "step": 13566 }, { "epoch": 0.82, "learning_rate": 4.092171183678397e-05, "loss": 1.2569, "step": 13568 }, { "epoch": 0.82, "learning_rate": 4.092774793263717e-05, "loss": 1.1951, "step": 13570 }, { "epoch": 0.82, "learning_rate": 4.0933784028490374e-05, "loss": 1.6423, "step": 13572 }, { "epoch": 0.82, "learning_rate": 4.0939820124343575e-05, "loss": 1.0607, "step": 13574 }, { "epoch": 0.82, "learning_rate": 4.0945856220196776e-05, "loss": 1.1316, "step": 13576 }, { "epoch": 0.82, "learning_rate": 4.095189231604998e-05, "loss": 1.3642, "step": 13578 }, { "epoch": 0.82, "learning_rate": 4.0957928411903184e-05, "loss": 1.0632, "step": 13580 }, { "epoch": 0.82, "learning_rate": 4.0963964507756385e-05, "loss": 1.78, "step": 13582 }, { "epoch": 0.82, "learning_rate": 4.0970000603609586e-05, "loss": 1.3205, "step": 13584 }, { "epoch": 0.82, "learning_rate": 4.0976036699462786e-05, "loss": 1.5236, "step": 13586 }, { "epoch": 0.82, "learning_rate": 4.098207279531599e-05, "loss": 1.3508, "step": 13588 }, { "epoch": 0.82, "learning_rate": 4.098810889116919e-05, "loss": 1.184, "step": 13590 }, { "epoch": 0.82, "learning_rate": 4.0994144987022396e-05, "loss": 1.2177, "step": 13592 }, { "epoch": 0.82, "learning_rate": 4.1000181082875596e-05, "loss": 1.3223, "step": 13594 }, { "epoch": 0.82, "learning_rate": 4.1006217178728804e-05, "loss": 0.9953, "step": 13596 }, { "epoch": 0.82, "learning_rate": 4.1012253274582005e-05, "loss": 1.0162, "step": 13598 }, { "epoch": 0.82, "learning_rate": 4.1018289370435206e-05, "loss": 1.2572, "step": 13600 }, { "epoch": 0.82, "learning_rate": 4.1024325466288406e-05, "loss": 1.4964, "step": 13602 }, { "epoch": 0.82, "learning_rate": 4.1030361562141614e-05, "loss": 1.8525, "step": 13604 }, { "epoch": 0.82, "learning_rate": 4.1036397657994815e-05, "loss": 1.1982, "step": 13606 }, { "epoch": 0.82, "learning_rate": 4.1042433753848016e-05, "loss": 1.4071, "step": 13608 }, { "epoch": 0.82, "learning_rate": 4.1048469849701217e-05, "loss": 1.4157, "step": 13610 }, { "epoch": 0.82, "learning_rate": 4.105450594555442e-05, "loss": 1.2464, "step": 13612 }, { "epoch": 0.82, "learning_rate": 4.106054204140762e-05, "loss": 1.1733, "step": 13614 }, { "epoch": 0.82, "learning_rate": 4.106657813726082e-05, "loss": 1.4116, "step": 13616 }, { "epoch": 0.82, "learning_rate": 4.1072614233114027e-05, "loss": 1.3186, "step": 13618 }, { "epoch": 0.82, "learning_rate": 4.107865032896723e-05, "loss": 1.2847, "step": 13620 }, { "epoch": 0.82, "learning_rate": 4.108468642482043e-05, "loss": 1.2377, "step": 13622 }, { "epoch": 0.82, "learning_rate": 4.109072252067363e-05, "loss": 1.1062, "step": 13624 }, { "epoch": 0.82, "learning_rate": 4.109675861652683e-05, "loss": 1.4437, "step": 13626 }, { "epoch": 0.82, "learning_rate": 4.110279471238003e-05, "loss": 1.4753, "step": 13628 }, { "epoch": 0.82, "learning_rate": 4.110883080823323e-05, "loss": 1.341, "step": 13630 }, { "epoch": 0.82, "learning_rate": 4.111486690408644e-05, "loss": 1.2759, "step": 13632 }, { "epoch": 0.82, "learning_rate": 4.112090299993964e-05, "loss": 1.6727, "step": 13634 }, { "epoch": 0.82, "learning_rate": 4.112693909579284e-05, "loss": 1.2144, "step": 13636 }, { "epoch": 0.82, "learning_rate": 4.113297519164604e-05, "loss": 1.2881, "step": 13638 }, { "epoch": 0.82, "learning_rate": 4.113901128749925e-05, "loss": 0.7549, "step": 13640 }, { "epoch": 0.82, "learning_rate": 4.114504738335245e-05, "loss": 1.0069, "step": 13642 }, { "epoch": 0.82, "learning_rate": 4.115108347920566e-05, "loss": 1.3384, "step": 13644 }, { "epoch": 0.82, "learning_rate": 4.115711957505886e-05, "loss": 1.0434, "step": 13646 }, { "epoch": 0.82, "learning_rate": 4.116315567091206e-05, "loss": 1.3434, "step": 13648 }, { "epoch": 0.82, "learning_rate": 4.116919176676526e-05, "loss": 1.1978, "step": 13650 }, { "epoch": 0.82, "learning_rate": 4.117522786261846e-05, "loss": 1.0315, "step": 13652 }, { "epoch": 0.82, "learning_rate": 4.118126395847166e-05, "loss": 1.2012, "step": 13654 }, { "epoch": 0.82, "learning_rate": 4.118730005432486e-05, "loss": 1.2143, "step": 13656 }, { "epoch": 0.82, "learning_rate": 4.119333615017807e-05, "loss": 1.5529, "step": 13658 }, { "epoch": 0.82, "learning_rate": 4.119937224603127e-05, "loss": 1.3848, "step": 13660 }, { "epoch": 0.82, "learning_rate": 4.120540834188447e-05, "loss": 1.3409, "step": 13662 }, { "epoch": 0.82, "learning_rate": 4.121144443773767e-05, "loss": 1.2998, "step": 13664 }, { "epoch": 0.82, "learning_rate": 4.121748053359087e-05, "loss": 1.0525, "step": 13666 }, { "epoch": 0.82, "learning_rate": 4.1223516629444074e-05, "loss": 1.3584, "step": 13668 }, { "epoch": 0.83, "learning_rate": 4.122955272529728e-05, "loss": 1.087, "step": 13670 }, { "epoch": 0.83, "learning_rate": 4.123558882115048e-05, "loss": 1.4926, "step": 13672 }, { "epoch": 0.83, "learning_rate": 4.124162491700368e-05, "loss": 1.115, "step": 13674 }, { "epoch": 0.83, "learning_rate": 4.1247661012856884e-05, "loss": 1.1455, "step": 13676 }, { "epoch": 0.83, "learning_rate": 4.1253697108710085e-05, "loss": 1.539, "step": 13678 }, { "epoch": 0.83, "learning_rate": 4.1259733204563286e-05, "loss": 1.6885, "step": 13680 }, { "epoch": 0.83, "learning_rate": 4.1265769300416493e-05, "loss": 1.0459, "step": 13682 }, { "epoch": 0.83, "learning_rate": 4.1271805396269694e-05, "loss": 1.0079, "step": 13684 }, { "epoch": 0.83, "learning_rate": 4.1277841492122895e-05, "loss": 1.196, "step": 13686 }, { "epoch": 0.83, "learning_rate": 4.12838775879761e-05, "loss": 0.9856, "step": 13688 }, { "epoch": 0.83, "learning_rate": 4.1289913683829303e-05, "loss": 1.3179, "step": 13690 }, { "epoch": 0.83, "learning_rate": 4.1295949779682504e-05, "loss": 1.2826, "step": 13692 }, { "epoch": 0.83, "learning_rate": 4.1301985875535705e-05, "loss": 1.1228, "step": 13694 }, { "epoch": 0.83, "learning_rate": 4.1308021971388906e-05, "loss": 1.1305, "step": 13696 }, { "epoch": 0.83, "learning_rate": 4.1314058067242114e-05, "loss": 1.2583, "step": 13698 }, { "epoch": 0.83, "learning_rate": 4.1320094163095314e-05, "loss": 1.3854, "step": 13700 }, { "epoch": 0.83, "learning_rate": 4.1326130258948515e-05, "loss": 1.6032, "step": 13702 }, { "epoch": 0.83, "learning_rate": 4.1332166354801716e-05, "loss": 1.2366, "step": 13704 }, { "epoch": 0.83, "learning_rate": 4.133820245065492e-05, "loss": 1.1129, "step": 13706 }, { "epoch": 0.83, "learning_rate": 4.134423854650812e-05, "loss": 0.9658, "step": 13708 }, { "epoch": 0.83, "learning_rate": 4.1350274642361325e-05, "loss": 1.3281, "step": 13710 }, { "epoch": 0.83, "learning_rate": 4.1356310738214526e-05, "loss": 1.3398, "step": 13712 }, { "epoch": 0.83, "learning_rate": 4.136234683406773e-05, "loss": 1.1233, "step": 13714 }, { "epoch": 0.83, "learning_rate": 4.136838292992093e-05, "loss": 1.1379, "step": 13716 }, { "epoch": 0.83, "learning_rate": 4.137441902577413e-05, "loss": 1.3821, "step": 13718 }, { "epoch": 0.83, "learning_rate": 4.138045512162733e-05, "loss": 1.0074, "step": 13720 }, { "epoch": 0.83, "learning_rate": 4.138649121748053e-05, "loss": 1.3178, "step": 13722 }, { "epoch": 0.83, "learning_rate": 4.139252731333374e-05, "loss": 1.0621, "step": 13724 }, { "epoch": 0.83, "learning_rate": 4.139856340918694e-05, "loss": 1.2369, "step": 13726 }, { "epoch": 0.83, "learning_rate": 4.140459950504014e-05, "loss": 0.9267, "step": 13728 }, { "epoch": 0.83, "learning_rate": 4.141063560089335e-05, "loss": 1.2336, "step": 13730 }, { "epoch": 0.83, "learning_rate": 4.141667169674655e-05, "loss": 1.2564, "step": 13732 }, { "epoch": 0.83, "learning_rate": 4.142270779259975e-05, "loss": 1.2895, "step": 13734 }, { "epoch": 0.83, "learning_rate": 4.1428743888452956e-05, "loss": 1.1759, "step": 13736 }, { "epoch": 0.83, "learning_rate": 4.143477998430616e-05, "loss": 1.1351, "step": 13738 }, { "epoch": 0.83, "learning_rate": 4.144081608015936e-05, "loss": 1.0582, "step": 13740 }, { "epoch": 0.83, "learning_rate": 4.144685217601256e-05, "loss": 1.6905, "step": 13742 }, { "epoch": 0.83, "learning_rate": 4.145288827186576e-05, "loss": 1.1303, "step": 13744 }, { "epoch": 0.83, "learning_rate": 4.145892436771896e-05, "loss": 1.4927, "step": 13746 }, { "epoch": 0.83, "learning_rate": 4.146496046357216e-05, "loss": 1.1028, "step": 13748 }, { "epoch": 0.83, "learning_rate": 4.147099655942537e-05, "loss": 1.4445, "step": 13750 }, { "epoch": 0.83, "learning_rate": 4.147703265527857e-05, "loss": 1.2008, "step": 13752 }, { "epoch": 0.83, "learning_rate": 4.148306875113177e-05, "loss": 0.9074, "step": 13754 }, { "epoch": 0.83, "learning_rate": 4.148910484698497e-05, "loss": 1.4577, "step": 13756 }, { "epoch": 0.83, "learning_rate": 4.149514094283817e-05, "loss": 1.1012, "step": 13758 }, { "epoch": 0.83, "learning_rate": 4.150117703869137e-05, "loss": 2.0929, "step": 13760 }, { "epoch": 0.83, "learning_rate": 4.1507213134544574e-05, "loss": 1.1381, "step": 13762 }, { "epoch": 0.83, "learning_rate": 4.151324923039778e-05, "loss": 1.2773, "step": 13764 }, { "epoch": 0.83, "learning_rate": 4.151928532625098e-05, "loss": 0.7356, "step": 13766 }, { "epoch": 0.83, "learning_rate": 4.152532142210418e-05, "loss": 1.219, "step": 13768 }, { "epoch": 0.83, "learning_rate": 4.1531357517957384e-05, "loss": 1.223, "step": 13770 }, { "epoch": 0.83, "learning_rate": 4.153739361381059e-05, "loss": 1.3221, "step": 13772 }, { "epoch": 0.83, "learning_rate": 4.154342970966379e-05, "loss": 1.2373, "step": 13774 }, { "epoch": 0.83, "learning_rate": 4.154946580551699e-05, "loss": 1.4074, "step": 13776 }, { "epoch": 0.83, "learning_rate": 4.15555019013702e-05, "loss": 1.4361, "step": 13778 }, { "epoch": 0.83, "learning_rate": 4.15615379972234e-05, "loss": 1.2831, "step": 13780 }, { "epoch": 0.83, "learning_rate": 4.15675740930766e-05, "loss": 1.0548, "step": 13782 }, { "epoch": 0.83, "learning_rate": 4.15736101889298e-05, "loss": 1.276, "step": 13784 }, { "epoch": 0.83, "learning_rate": 4.1579646284783004e-05, "loss": 1.0374, "step": 13786 }, { "epoch": 0.83, "learning_rate": 4.1585682380636205e-05, "loss": 0.9931, "step": 13788 }, { "epoch": 0.83, "learning_rate": 4.159171847648941e-05, "loss": 1.0617, "step": 13790 }, { "epoch": 0.83, "learning_rate": 4.159775457234261e-05, "loss": 0.9524, "step": 13792 }, { "epoch": 0.83, "learning_rate": 4.1603790668195814e-05, "loss": 1.1123, "step": 13794 }, { "epoch": 0.83, "learning_rate": 4.1609826764049015e-05, "loss": 1.4408, "step": 13796 }, { "epoch": 0.83, "learning_rate": 4.1615862859902215e-05, "loss": 1.2127, "step": 13798 }, { "epoch": 0.83, "learning_rate": 4.1621898955755416e-05, "loss": 0.8216, "step": 13800 }, { "epoch": 0.83, "learning_rate": 4.1627935051608624e-05, "loss": 0.8767, "step": 13802 }, { "epoch": 0.83, "learning_rate": 4.1633971147461825e-05, "loss": 1.4003, "step": 13804 }, { "epoch": 0.83, "learning_rate": 4.1640007243315026e-05, "loss": 1.5166, "step": 13806 }, { "epoch": 0.83, "learning_rate": 4.1646043339168226e-05, "loss": 1.306, "step": 13808 }, { "epoch": 0.83, "learning_rate": 4.165207943502143e-05, "loss": 1.6163, "step": 13810 }, { "epoch": 0.83, "learning_rate": 4.165811553087463e-05, "loss": 1.1598, "step": 13812 }, { "epoch": 0.83, "learning_rate": 4.166415162672783e-05, "loss": 0.9813, "step": 13814 }, { "epoch": 0.83, "learning_rate": 4.1670187722581036e-05, "loss": 1.2978, "step": 13816 }, { "epoch": 0.83, "learning_rate": 4.167622381843424e-05, "loss": 1.0805, "step": 13818 }, { "epoch": 0.83, "learning_rate": 4.1682259914287445e-05, "loss": 1.4214, "step": 13820 }, { "epoch": 0.83, "learning_rate": 4.1688296010140646e-05, "loss": 1.3254, "step": 13822 }, { "epoch": 0.83, "learning_rate": 4.1694332105993846e-05, "loss": 1.4641, "step": 13824 }, { "epoch": 0.83, "learning_rate": 4.170036820184705e-05, "loss": 1.3862, "step": 13826 }, { "epoch": 0.83, "learning_rate": 4.170640429770025e-05, "loss": 1.0582, "step": 13828 }, { "epoch": 0.83, "learning_rate": 4.1712440393553456e-05, "loss": 1.1204, "step": 13830 }, { "epoch": 0.83, "learning_rate": 4.1718476489406656e-05, "loss": 1.546, "step": 13832 }, { "epoch": 0.83, "learning_rate": 4.172451258525986e-05, "loss": 0.9372, "step": 13834 }, { "epoch": 0.84, "learning_rate": 4.173054868111306e-05, "loss": 1.3972, "step": 13836 }, { "epoch": 0.84, "learning_rate": 4.173658477696626e-05, "loss": 1.5539, "step": 13838 }, { "epoch": 0.84, "learning_rate": 4.174262087281946e-05, "loss": 1.093, "step": 13840 }, { "epoch": 0.84, "learning_rate": 4.174865696867267e-05, "loss": 1.4777, "step": 13842 }, { "epoch": 0.84, "learning_rate": 4.175469306452587e-05, "loss": 1.5446, "step": 13844 }, { "epoch": 0.84, "learning_rate": 4.176072916037907e-05, "loss": 1.0606, "step": 13846 }, { "epoch": 0.84, "learning_rate": 4.176676525623227e-05, "loss": 1.356, "step": 13848 }, { "epoch": 0.84, "learning_rate": 4.177280135208547e-05, "loss": 1.2163, "step": 13850 }, { "epoch": 0.84, "learning_rate": 4.177883744793867e-05, "loss": 1.6329, "step": 13852 }, { "epoch": 0.84, "learning_rate": 4.178487354379187e-05, "loss": 1.3735, "step": 13854 }, { "epoch": 0.84, "learning_rate": 4.179090963964508e-05, "loss": 1.3014, "step": 13856 }, { "epoch": 0.84, "learning_rate": 4.179694573549828e-05, "loss": 1.3242, "step": 13858 }, { "epoch": 0.84, "learning_rate": 4.180298183135148e-05, "loss": 1.2384, "step": 13860 }, { "epoch": 0.84, "learning_rate": 4.180901792720468e-05, "loss": 0.9514, "step": 13862 }, { "epoch": 0.84, "learning_rate": 4.181505402305789e-05, "loss": 1.1358, "step": 13864 }, { "epoch": 0.84, "learning_rate": 4.182109011891109e-05, "loss": 1.182, "step": 13866 }, { "epoch": 0.84, "learning_rate": 4.18271262147643e-05, "loss": 1.451, "step": 13868 }, { "epoch": 0.84, "learning_rate": 4.18331623106175e-05, "loss": 1.1038, "step": 13870 }, { "epoch": 0.84, "learning_rate": 4.18391984064707e-05, "loss": 1.2399, "step": 13872 }, { "epoch": 0.84, "learning_rate": 4.18452345023239e-05, "loss": 1.0685, "step": 13874 }, { "epoch": 0.84, "learning_rate": 4.18512705981771e-05, "loss": 0.8434, "step": 13876 }, { "epoch": 0.84, "learning_rate": 4.18573066940303e-05, "loss": 1.1806, "step": 13878 }, { "epoch": 0.84, "learning_rate": 4.18633427898835e-05, "loss": 1.2513, "step": 13880 }, { "epoch": 0.84, "learning_rate": 4.186937888573671e-05, "loss": 1.3607, "step": 13882 }, { "epoch": 0.84, "learning_rate": 4.187541498158991e-05, "loss": 1.1439, "step": 13884 }, { "epoch": 0.84, "learning_rate": 4.188145107744311e-05, "loss": 1.5093, "step": 13886 }, { "epoch": 0.84, "learning_rate": 4.188748717329631e-05, "loss": 1.2112, "step": 13888 }, { "epoch": 0.84, "learning_rate": 4.1893523269149514e-05, "loss": 1.4915, "step": 13890 }, { "epoch": 0.84, "learning_rate": 4.1899559365002715e-05, "loss": 0.8235, "step": 13892 }, { "epoch": 0.84, "learning_rate": 4.190559546085592e-05, "loss": 1.4176, "step": 13894 }, { "epoch": 0.84, "learning_rate": 4.191163155670912e-05, "loss": 1.6202, "step": 13896 }, { "epoch": 0.84, "learning_rate": 4.1917667652562324e-05, "loss": 1.33, "step": 13898 }, { "epoch": 0.84, "learning_rate": 4.1923703748415525e-05, "loss": 1.4715, "step": 13900 }, { "epoch": 0.84, "learning_rate": 4.1929739844268726e-05, "loss": 1.6508, "step": 13902 }, { "epoch": 0.84, "learning_rate": 4.1935775940121927e-05, "loss": 1.2394, "step": 13904 }, { "epoch": 0.84, "learning_rate": 4.1941812035975134e-05, "loss": 1.1478, "step": 13906 }, { "epoch": 0.84, "learning_rate": 4.1947848131828335e-05, "loss": 1.2748, "step": 13908 }, { "epoch": 0.84, "learning_rate": 4.1953884227681536e-05, "loss": 1.1931, "step": 13910 }, { "epoch": 0.84, "learning_rate": 4.1959920323534743e-05, "loss": 1.7752, "step": 13912 }, { "epoch": 0.84, "learning_rate": 4.1965956419387944e-05, "loss": 1.4281, "step": 13914 }, { "epoch": 0.84, "learning_rate": 4.1971992515241145e-05, "loss": 1.2965, "step": 13916 }, { "epoch": 0.84, "learning_rate": 4.1978028611094346e-05, "loss": 1.1202, "step": 13918 }, { "epoch": 0.84, "learning_rate": 4.198406470694755e-05, "loss": 1.4028, "step": 13920 }, { "epoch": 0.84, "learning_rate": 4.1990100802800754e-05, "loss": 1.2548, "step": 13922 }, { "epoch": 0.84, "learning_rate": 4.1996136898653955e-05, "loss": 0.8577, "step": 13924 }, { "epoch": 0.84, "learning_rate": 4.2002172994507156e-05, "loss": 1.2401, "step": 13926 }, { "epoch": 0.84, "learning_rate": 4.200820909036036e-05, "loss": 1.5746, "step": 13928 }, { "epoch": 0.84, "learning_rate": 4.201424518621356e-05, "loss": 1.3049, "step": 13930 }, { "epoch": 0.84, "learning_rate": 4.202028128206676e-05, "loss": 1.1369, "step": 13932 }, { "epoch": 0.84, "learning_rate": 4.2026317377919966e-05, "loss": 1.4136, "step": 13934 }, { "epoch": 0.84, "learning_rate": 4.203235347377317e-05, "loss": 0.7439, "step": 13936 }, { "epoch": 0.84, "learning_rate": 4.203838956962637e-05, "loss": 1.1197, "step": 13938 }, { "epoch": 0.84, "learning_rate": 4.204442566547957e-05, "loss": 1.243, "step": 13940 }, { "epoch": 0.84, "learning_rate": 4.205046176133277e-05, "loss": 1.7531, "step": 13942 }, { "epoch": 0.84, "learning_rate": 4.205649785718597e-05, "loss": 1.3912, "step": 13944 }, { "epoch": 0.84, "learning_rate": 4.206253395303917e-05, "loss": 1.4669, "step": 13946 }, { "epoch": 0.84, "learning_rate": 4.206857004889238e-05, "loss": 1.083, "step": 13948 }, { "epoch": 0.84, "learning_rate": 4.207460614474558e-05, "loss": 1.2359, "step": 13950 }, { "epoch": 0.84, "learning_rate": 4.208064224059878e-05, "loss": 1.303, "step": 13952 }, { "epoch": 0.84, "learning_rate": 4.208667833645199e-05, "loss": 1.1728, "step": 13954 }, { "epoch": 0.84, "learning_rate": 4.209271443230519e-05, "loss": 1.2406, "step": 13956 }, { "epoch": 0.84, "learning_rate": 4.209875052815839e-05, "loss": 1.1424, "step": 13958 }, { "epoch": 0.84, "learning_rate": 4.21047866240116e-05, "loss": 1.1822, "step": 13960 }, { "epoch": 0.84, "learning_rate": 4.21108227198648e-05, "loss": 1.5545, "step": 13962 }, { "epoch": 0.84, "learning_rate": 4.2116858815718e-05, "loss": 1.3914, "step": 13964 }, { "epoch": 0.84, "learning_rate": 4.21228949115712e-05, "loss": 1.2909, "step": 13966 }, { "epoch": 0.84, "learning_rate": 4.21289310074244e-05, "loss": 1.1833, "step": 13968 }, { "epoch": 0.84, "learning_rate": 4.21349671032776e-05, "loss": 1.1135, "step": 13970 }, { "epoch": 0.84, "learning_rate": 4.21410031991308e-05, "loss": 1.2769, "step": 13972 }, { "epoch": 0.84, "learning_rate": 4.214703929498401e-05, "loss": 1.3653, "step": 13974 }, { "epoch": 0.84, "learning_rate": 4.215307539083721e-05, "loss": 1.3225, "step": 13976 }, { "epoch": 0.84, "learning_rate": 4.215911148669041e-05, "loss": 0.9983, "step": 13978 }, { "epoch": 0.84, "learning_rate": 4.216514758254361e-05, "loss": 1.6004, "step": 13980 }, { "epoch": 0.84, "learning_rate": 4.217118367839681e-05, "loss": 1.5129, "step": 13982 }, { "epoch": 0.84, "learning_rate": 4.2177219774250014e-05, "loss": 1.2986, "step": 13984 }, { "epoch": 0.84, "learning_rate": 4.2183255870103214e-05, "loss": 1.3863, "step": 13986 }, { "epoch": 0.84, "learning_rate": 4.218929196595642e-05, "loss": 1.298, "step": 13988 }, { "epoch": 0.84, "learning_rate": 4.219532806180962e-05, "loss": 1.4134, "step": 13990 }, { "epoch": 0.84, "learning_rate": 4.2201364157662824e-05, "loss": 0.9982, "step": 13992 }, { "epoch": 0.84, "learning_rate": 4.2207400253516024e-05, "loss": 1.2318, "step": 13994 }, { "epoch": 0.84, "learning_rate": 4.221343634936923e-05, "loss": 1.1375, "step": 13996 }, { "epoch": 0.84, "learning_rate": 4.221947244522243e-05, "loss": 1.1564, "step": 13998 }, { "epoch": 0.85, "learning_rate": 4.2225508541075634e-05, "loss": 0.9811, "step": 14000 }, { "epoch": 0.85, "learning_rate": 4.223154463692884e-05, "loss": 1.1875, "step": 14002 }, { "epoch": 0.85, "learning_rate": 4.223758073278204e-05, "loss": 1.5966, "step": 14004 }, { "epoch": 0.85, "learning_rate": 4.224361682863524e-05, "loss": 0.921, "step": 14006 }, { "epoch": 0.85, "learning_rate": 4.2249652924488444e-05, "loss": 1.2714, "step": 14008 }, { "epoch": 0.85, "learning_rate": 4.2255689020341645e-05, "loss": 1.0787, "step": 14010 }, { "epoch": 0.85, "learning_rate": 4.2261725116194845e-05, "loss": 1.0807, "step": 14012 }, { "epoch": 0.85, "learning_rate": 4.226776121204805e-05, "loss": 1.2646, "step": 14014 }, { "epoch": 0.85, "learning_rate": 4.2273797307901254e-05, "loss": 1.0109, "step": 14016 }, { "epoch": 0.85, "learning_rate": 4.2279833403754455e-05, "loss": 1.6054, "step": 14018 }, { "epoch": 0.85, "learning_rate": 4.2285869499607655e-05, "loss": 1.2701, "step": 14020 }, { "epoch": 0.85, "learning_rate": 4.2291905595460856e-05, "loss": 1.1941, "step": 14022 }, { "epoch": 0.85, "learning_rate": 4.229794169131406e-05, "loss": 1.488, "step": 14024 }, { "epoch": 0.85, "learning_rate": 4.2303977787167265e-05, "loss": 1.4211, "step": 14026 }, { "epoch": 0.85, "learning_rate": 4.2310013883020465e-05, "loss": 1.5784, "step": 14028 }, { "epoch": 0.85, "learning_rate": 4.2316049978873666e-05, "loss": 0.8939, "step": 14030 }, { "epoch": 0.85, "learning_rate": 4.232208607472687e-05, "loss": 1.1157, "step": 14032 }, { "epoch": 0.85, "learning_rate": 4.232812217058007e-05, "loss": 1.3404, "step": 14034 }, { "epoch": 0.85, "learning_rate": 4.233415826643327e-05, "loss": 1.1443, "step": 14036 }, { "epoch": 0.85, "learning_rate": 4.234019436228647e-05, "loss": 1.1942, "step": 14038 }, { "epoch": 0.85, "learning_rate": 4.234623045813968e-05, "loss": 1.8843, "step": 14040 }, { "epoch": 0.85, "learning_rate": 4.235226655399288e-05, "loss": 1.0323, "step": 14042 }, { "epoch": 0.85, "learning_rate": 4.2358302649846086e-05, "loss": 1.1179, "step": 14044 }, { "epoch": 0.85, "learning_rate": 4.2364338745699286e-05, "loss": 1.0097, "step": 14046 }, { "epoch": 0.85, "learning_rate": 4.237037484155249e-05, "loss": 1.3158, "step": 14048 }, { "epoch": 0.85, "learning_rate": 4.237641093740569e-05, "loss": 1.5509, "step": 14050 }, { "epoch": 0.85, "learning_rate": 4.238244703325889e-05, "loss": 1.2934, "step": 14052 }, { "epoch": 0.85, "learning_rate": 4.2388483129112096e-05, "loss": 0.8633, "step": 14054 }, { "epoch": 0.85, "learning_rate": 4.23945192249653e-05, "loss": 1.1918, "step": 14056 }, { "epoch": 0.85, "learning_rate": 4.24005553208185e-05, "loss": 1.5497, "step": 14058 }, { "epoch": 0.85, "learning_rate": 4.24065914166717e-05, "loss": 1.2928, "step": 14060 }, { "epoch": 0.85, "learning_rate": 4.24126275125249e-05, "loss": 1.1208, "step": 14062 }, { "epoch": 0.85, "learning_rate": 4.24186636083781e-05, "loss": 1.3533, "step": 14064 }, { "epoch": 0.85, "learning_rate": 4.242469970423131e-05, "loss": 1.2176, "step": 14066 }, { "epoch": 0.85, "learning_rate": 4.243073580008451e-05, "loss": 1.4217, "step": 14068 }, { "epoch": 0.85, "learning_rate": 4.243677189593771e-05, "loss": 1.1643, "step": 14070 }, { "epoch": 0.85, "learning_rate": 4.244280799179091e-05, "loss": 1.0317, "step": 14072 }, { "epoch": 0.85, "learning_rate": 4.244884408764411e-05, "loss": 1.519, "step": 14074 }, { "epoch": 0.85, "learning_rate": 4.245488018349731e-05, "loss": 1.4167, "step": 14076 }, { "epoch": 0.85, "learning_rate": 4.246091627935051e-05, "loss": 1.2667, "step": 14078 }, { "epoch": 0.85, "learning_rate": 4.246695237520372e-05, "loss": 1.2688, "step": 14080 }, { "epoch": 0.85, "learning_rate": 4.247298847105692e-05, "loss": 1.5529, "step": 14082 }, { "epoch": 0.85, "learning_rate": 4.247902456691012e-05, "loss": 1.3074, "step": 14084 }, { "epoch": 0.85, "learning_rate": 4.248506066276332e-05, "loss": 1.4773, "step": 14086 }, { "epoch": 0.85, "learning_rate": 4.249109675861653e-05, "loss": 0.96, "step": 14088 }, { "epoch": 0.85, "learning_rate": 4.249713285446973e-05, "loss": 1.4075, "step": 14090 }, { "epoch": 0.85, "learning_rate": 4.250316895032294e-05, "loss": 1.3694, "step": 14092 }, { "epoch": 0.85, "learning_rate": 4.250920504617614e-05, "loss": 1.479, "step": 14094 }, { "epoch": 0.85, "learning_rate": 4.251524114202934e-05, "loss": 1.2166, "step": 14096 }, { "epoch": 0.85, "learning_rate": 4.252127723788254e-05, "loss": 1.449, "step": 14098 }, { "epoch": 0.85, "learning_rate": 4.252731333373574e-05, "loss": 1.4403, "step": 14100 }, { "epoch": 0.85, "learning_rate": 4.253334942958894e-05, "loss": 1.2535, "step": 14102 }, { "epoch": 0.85, "learning_rate": 4.2539385525442144e-05, "loss": 1.4518, "step": 14104 }, { "epoch": 0.85, "learning_rate": 4.254542162129535e-05, "loss": 1.113, "step": 14106 }, { "epoch": 0.85, "learning_rate": 4.255145771714855e-05, "loss": 1.1845, "step": 14108 }, { "epoch": 0.85, "learning_rate": 4.255749381300175e-05, "loss": 1.3913, "step": 14110 }, { "epoch": 0.85, "learning_rate": 4.2563529908854954e-05, "loss": 1.1037, "step": 14112 }, { "epoch": 0.85, "learning_rate": 4.2569566004708155e-05, "loss": 1.382, "step": 14114 }, { "epoch": 0.85, "learning_rate": 4.2575602100561356e-05, "loss": 1.0035, "step": 14116 }, { "epoch": 0.85, "learning_rate": 4.258163819641456e-05, "loss": 1.4998, "step": 14118 }, { "epoch": 0.85, "learning_rate": 4.2587674292267764e-05, "loss": 1.4285, "step": 14120 }, { "epoch": 0.85, "learning_rate": 4.2593710388120965e-05, "loss": 1.0755, "step": 14122 }, { "epoch": 0.85, "learning_rate": 4.2599746483974166e-05, "loss": 1.3309, "step": 14124 }, { "epoch": 0.85, "learning_rate": 4.2605782579827367e-05, "loss": 0.9308, "step": 14126 }, { "epoch": 0.85, "learning_rate": 4.261181867568057e-05, "loss": 1.2082, "step": 14128 }, { "epoch": 0.85, "learning_rate": 4.2617854771533775e-05, "loss": 1.1236, "step": 14130 }, { "epoch": 0.85, "learning_rate": 4.2623890867386976e-05, "loss": 1.141, "step": 14132 }, { "epoch": 0.85, "learning_rate": 4.2629926963240177e-05, "loss": 1.5697, "step": 14134 }, { "epoch": 0.85, "learning_rate": 4.2635963059093384e-05, "loss": 1.2619, "step": 14136 }, { "epoch": 0.85, "learning_rate": 4.2641999154946585e-05, "loss": 1.4058, "step": 14138 }, { "epoch": 0.85, "learning_rate": 4.2648035250799786e-05, "loss": 1.3348, "step": 14140 }, { "epoch": 0.85, "learning_rate": 4.265407134665299e-05, "loss": 1.0408, "step": 14142 }, { "epoch": 0.85, "learning_rate": 4.266010744250619e-05, "loss": 1.1171, "step": 14144 }, { "epoch": 0.85, "learning_rate": 4.2666143538359395e-05, "loss": 1.1382, "step": 14146 }, { "epoch": 0.85, "learning_rate": 4.2672179634212596e-05, "loss": 1.2664, "step": 14148 }, { "epoch": 0.85, "learning_rate": 4.26782157300658e-05, "loss": 0.8509, "step": 14150 }, { "epoch": 0.85, "learning_rate": 4.2684251825919e-05, "loss": 0.737, "step": 14152 }, { "epoch": 0.85, "learning_rate": 4.26902879217722e-05, "loss": 1.0825, "step": 14154 }, { "epoch": 0.85, "learning_rate": 4.26963240176254e-05, "loss": 1.2972, "step": 14156 }, { "epoch": 0.85, "learning_rate": 4.270236011347861e-05, "loss": 1.1651, "step": 14158 }, { "epoch": 0.85, "learning_rate": 4.270839620933181e-05, "loss": 0.8946, "step": 14160 }, { "epoch": 0.85, "learning_rate": 4.271443230518501e-05, "loss": 1.406, "step": 14162 }, { "epoch": 0.85, "learning_rate": 4.272046840103821e-05, "loss": 1.4879, "step": 14164 }, { "epoch": 0.86, "learning_rate": 4.272650449689141e-05, "loss": 0.9768, "step": 14166 }, { "epoch": 0.86, "learning_rate": 4.273254059274461e-05, "loss": 1.0313, "step": 14168 }, { "epoch": 0.86, "learning_rate": 4.273857668859781e-05, "loss": 1.1481, "step": 14170 }, { "epoch": 0.86, "learning_rate": 4.274461278445102e-05, "loss": 1.3418, "step": 14172 }, { "epoch": 0.86, "learning_rate": 4.275064888030422e-05, "loss": 1.3338, "step": 14174 }, { "epoch": 0.86, "learning_rate": 4.275668497615742e-05, "loss": 1.1822, "step": 14176 }, { "epoch": 0.86, "learning_rate": 4.276272107201063e-05, "loss": 0.956, "step": 14178 }, { "epoch": 0.86, "learning_rate": 4.276875716786383e-05, "loss": 1.219, "step": 14180 }, { "epoch": 0.86, "learning_rate": 4.277479326371703e-05, "loss": 1.5229, "step": 14182 }, { "epoch": 0.86, "learning_rate": 4.278082935957024e-05, "loss": 1.1941, "step": 14184 }, { "epoch": 0.86, "learning_rate": 4.278686545542344e-05, "loss": 1.109, "step": 14186 }, { "epoch": 0.86, "learning_rate": 4.279290155127664e-05, "loss": 1.1967, "step": 14188 }, { "epoch": 0.86, "learning_rate": 4.279893764712984e-05, "loss": 1.7484, "step": 14190 }, { "epoch": 0.86, "learning_rate": 4.280497374298304e-05, "loss": 1.2449, "step": 14192 }, { "epoch": 0.86, "learning_rate": 4.281100983883624e-05, "loss": 1.4024, "step": 14194 }, { "epoch": 0.86, "learning_rate": 4.281704593468944e-05, "loss": 0.8833, "step": 14196 }, { "epoch": 0.86, "learning_rate": 4.282308203054265e-05, "loss": 0.9956, "step": 14198 }, { "epoch": 0.86, "learning_rate": 4.282911812639585e-05, "loss": 1.441, "step": 14200 }, { "epoch": 0.86, "learning_rate": 4.283515422224905e-05, "loss": 1.18, "step": 14202 }, { "epoch": 0.86, "learning_rate": 4.284119031810225e-05, "loss": 1.5261, "step": 14204 }, { "epoch": 0.86, "learning_rate": 4.2847226413955453e-05, "loss": 1.0268, "step": 14206 }, { "epoch": 0.86, "learning_rate": 4.2853262509808654e-05, "loss": 1.0146, "step": 14208 }, { "epoch": 0.86, "learning_rate": 4.2859298605661855e-05, "loss": 1.3095, "step": 14210 }, { "epoch": 0.86, "learning_rate": 4.286533470151506e-05, "loss": 1.2714, "step": 14212 }, { "epoch": 0.86, "learning_rate": 4.2871370797368264e-05, "loss": 1.1949, "step": 14214 }, { "epoch": 0.86, "learning_rate": 4.2877406893221464e-05, "loss": 1.0305, "step": 14216 }, { "epoch": 0.86, "learning_rate": 4.2883442989074665e-05, "loss": 1.518, "step": 14218 }, { "epoch": 0.86, "learning_rate": 4.2889479084927866e-05, "loss": 1.1948, "step": 14220 }, { "epoch": 0.86, "learning_rate": 4.2895515180781074e-05, "loss": 1.0162, "step": 14222 }, { "epoch": 0.86, "learning_rate": 4.2901551276634274e-05, "loss": 1.3237, "step": 14224 }, { "epoch": 0.86, "learning_rate": 4.290758737248748e-05, "loss": 1.3988, "step": 14226 }, { "epoch": 0.86, "learning_rate": 4.291362346834068e-05, "loss": 1.0972, "step": 14228 }, { "epoch": 0.86, "learning_rate": 4.2919659564193884e-05, "loss": 1.4379, "step": 14230 }, { "epoch": 0.86, "learning_rate": 4.2925695660047084e-05, "loss": 1.2241, "step": 14232 }, { "epoch": 0.86, "learning_rate": 4.2931731755900285e-05, "loss": 1.6036, "step": 14234 }, { "epoch": 0.86, "learning_rate": 4.2937767851753486e-05, "loss": 1.0955, "step": 14236 }, { "epoch": 0.86, "learning_rate": 4.2943803947606694e-05, "loss": 1.3593, "step": 14238 }, { "epoch": 0.86, "learning_rate": 4.2949840043459895e-05, "loss": 1.4205, "step": 14240 }, { "epoch": 0.86, "learning_rate": 4.2955876139313095e-05, "loss": 1.2766, "step": 14242 }, { "epoch": 0.86, "learning_rate": 4.2961912235166296e-05, "loss": 1.5052, "step": 14244 }, { "epoch": 0.86, "learning_rate": 4.29679483310195e-05, "loss": 1.1574, "step": 14246 }, { "epoch": 0.86, "learning_rate": 4.29739844268727e-05, "loss": 1.2253, "step": 14248 }, { "epoch": 0.86, "learning_rate": 4.2980020522725905e-05, "loss": 1.1479, "step": 14250 }, { "epoch": 0.86, "learning_rate": 4.2986056618579106e-05, "loss": 1.2946, "step": 14252 }, { "epoch": 0.86, "learning_rate": 4.299209271443231e-05, "loss": 1.0961, "step": 14254 }, { "epoch": 0.86, "learning_rate": 4.299812881028551e-05, "loss": 1.2814, "step": 14256 }, { "epoch": 0.86, "learning_rate": 4.300416490613871e-05, "loss": 1.587, "step": 14258 }, { "epoch": 0.86, "learning_rate": 4.301020100199191e-05, "loss": 1.286, "step": 14260 }, { "epoch": 0.86, "learning_rate": 4.301623709784511e-05, "loss": 1.4939, "step": 14262 }, { "epoch": 0.86, "learning_rate": 4.302227319369832e-05, "loss": 1.4491, "step": 14264 }, { "epoch": 0.86, "learning_rate": 4.302830928955152e-05, "loss": 1.4294, "step": 14266 }, { "epoch": 0.86, "learning_rate": 4.3034345385404726e-05, "loss": 1.5529, "step": 14268 }, { "epoch": 0.86, "learning_rate": 4.304038148125793e-05, "loss": 1.0903, "step": 14270 }, { "epoch": 0.86, "learning_rate": 4.304641757711113e-05, "loss": 1.7291, "step": 14272 }, { "epoch": 0.86, "learning_rate": 4.305245367296433e-05, "loss": 1.4774, "step": 14274 }, { "epoch": 0.86, "learning_rate": 4.305848976881753e-05, "loss": 1.0686, "step": 14276 }, { "epoch": 0.86, "learning_rate": 4.306452586467074e-05, "loss": 1.2822, "step": 14278 }, { "epoch": 0.86, "learning_rate": 4.307056196052394e-05, "loss": 1.297, "step": 14280 }, { "epoch": 0.86, "learning_rate": 4.307659805637714e-05, "loss": 1.3425, "step": 14282 }, { "epoch": 0.86, "learning_rate": 4.308263415223034e-05, "loss": 1.115, "step": 14284 }, { "epoch": 0.86, "learning_rate": 4.308867024808354e-05, "loss": 0.9895, "step": 14286 }, { "epoch": 0.86, "learning_rate": 4.309470634393674e-05, "loss": 1.1698, "step": 14288 }, { "epoch": 0.86, "learning_rate": 4.310074243978995e-05, "loss": 1.3353, "step": 14290 }, { "epoch": 0.86, "learning_rate": 4.310677853564315e-05, "loss": 0.9882, "step": 14292 }, { "epoch": 0.86, "learning_rate": 4.311281463149635e-05, "loss": 1.3129, "step": 14294 }, { "epoch": 0.86, "learning_rate": 4.311885072734955e-05, "loss": 1.4311, "step": 14296 }, { "epoch": 0.86, "learning_rate": 4.312488682320275e-05, "loss": 1.5269, "step": 14298 }, { "epoch": 0.86, "learning_rate": 4.313092291905595e-05, "loss": 1.1949, "step": 14300 }, { "epoch": 0.86, "learning_rate": 4.3136959014909154e-05, "loss": 1.4181, "step": 14302 }, { "epoch": 0.86, "learning_rate": 4.314299511076236e-05, "loss": 1.4926, "step": 14304 }, { "epoch": 0.86, "learning_rate": 4.314903120661556e-05, "loss": 1.0765, "step": 14306 }, { "epoch": 0.86, "learning_rate": 4.315506730246876e-05, "loss": 1.1028, "step": 14308 }, { "epoch": 0.86, "learning_rate": 4.3161103398321964e-05, "loss": 1.0937, "step": 14310 }, { "epoch": 0.86, "learning_rate": 4.316713949417517e-05, "loss": 1.0132, "step": 14312 }, { "epoch": 0.86, "learning_rate": 4.317317559002837e-05, "loss": 1.1808, "step": 14314 }, { "epoch": 0.86, "learning_rate": 4.317921168588158e-05, "loss": 1.265, "step": 14316 }, { "epoch": 0.86, "learning_rate": 4.318524778173478e-05, "loss": 1.3154, "step": 14318 }, { "epoch": 0.86, "learning_rate": 4.319128387758798e-05, "loss": 1.7109, "step": 14320 }, { "epoch": 0.86, "learning_rate": 4.319731997344118e-05, "loss": 1.3988, "step": 14322 }, { "epoch": 0.86, "learning_rate": 4.320335606929438e-05, "loss": 1.1041, "step": 14324 }, { "epoch": 0.86, "learning_rate": 4.3209392165147584e-05, "loss": 1.1837, "step": 14326 }, { "epoch": 0.86, "learning_rate": 4.3215428261000785e-05, "loss": 1.4281, "step": 14328 }, { "epoch": 0.86, "learning_rate": 4.322146435685399e-05, "loss": 1.274, "step": 14330 }, { "epoch": 0.87, "learning_rate": 4.322750045270719e-05, "loss": 1.0441, "step": 14332 }, { "epoch": 0.87, "learning_rate": 4.3233536548560394e-05, "loss": 1.0542, "step": 14334 }, { "epoch": 0.87, "learning_rate": 4.3239572644413595e-05, "loss": 0.7008, "step": 14336 }, { "epoch": 0.87, "learning_rate": 4.3245608740266796e-05, "loss": 1.2211, "step": 14338 }, { "epoch": 0.87, "learning_rate": 4.3251644836119996e-05, "loss": 1.1756, "step": 14340 }, { "epoch": 0.87, "learning_rate": 4.32576809319732e-05, "loss": 1.3048, "step": 14342 }, { "epoch": 0.87, "learning_rate": 4.3263717027826405e-05, "loss": 1.4576, "step": 14344 }, { "epoch": 0.87, "learning_rate": 4.3269753123679606e-05, "loss": 1.2605, "step": 14346 }, { "epoch": 0.87, "learning_rate": 4.3275789219532806e-05, "loss": 1.4504, "step": 14348 }, { "epoch": 0.87, "learning_rate": 4.328182531538601e-05, "loss": 1.2126, "step": 14350 }, { "epoch": 0.87, "learning_rate": 4.328786141123921e-05, "loss": 1.1723, "step": 14352 }, { "epoch": 0.87, "learning_rate": 4.3293897507092416e-05, "loss": 0.972, "step": 14354 }, { "epoch": 0.87, "learning_rate": 4.3299933602945617e-05, "loss": 1.0433, "step": 14356 }, { "epoch": 0.87, "learning_rate": 4.330596969879882e-05, "loss": 1.1842, "step": 14358 }, { "epoch": 0.87, "learning_rate": 4.3312005794652025e-05, "loss": 1.1844, "step": 14360 }, { "epoch": 0.87, "learning_rate": 4.3318041890505226e-05, "loss": 1.3882, "step": 14362 }, { "epoch": 0.87, "learning_rate": 4.3324077986358427e-05, "loss": 1.3978, "step": 14364 }, { "epoch": 0.87, "learning_rate": 4.333011408221163e-05, "loss": 1.4502, "step": 14366 }, { "epoch": 0.87, "learning_rate": 4.333615017806483e-05, "loss": 1.6245, "step": 14368 }, { "epoch": 0.87, "learning_rate": 4.3342186273918036e-05, "loss": 1.2828, "step": 14370 }, { "epoch": 0.87, "learning_rate": 4.334822236977124e-05, "loss": 1.35, "step": 14372 }, { "epoch": 0.87, "learning_rate": 4.335425846562444e-05, "loss": 1.0023, "step": 14374 }, { "epoch": 0.87, "learning_rate": 4.336029456147764e-05, "loss": 1.3004, "step": 14376 }, { "epoch": 0.87, "learning_rate": 4.336633065733084e-05, "loss": 1.1845, "step": 14378 }, { "epoch": 0.87, "learning_rate": 4.337236675318404e-05, "loss": 1.3926, "step": 14380 }, { "epoch": 0.87, "learning_rate": 4.337840284903725e-05, "loss": 1.7206, "step": 14382 }, { "epoch": 0.87, "learning_rate": 4.338443894489045e-05, "loss": 1.5255, "step": 14384 }, { "epoch": 0.87, "learning_rate": 4.339047504074365e-05, "loss": 1.2863, "step": 14386 }, { "epoch": 0.87, "learning_rate": 4.339651113659685e-05, "loss": 1.3341, "step": 14388 }, { "epoch": 0.87, "learning_rate": 4.340254723245005e-05, "loss": 1.6185, "step": 14390 }, { "epoch": 0.87, "learning_rate": 4.340858332830325e-05, "loss": 1.1731, "step": 14392 }, { "epoch": 0.87, "learning_rate": 4.341461942415645e-05, "loss": 1.5041, "step": 14394 }, { "epoch": 0.87, "learning_rate": 4.342065552000966e-05, "loss": 1.2661, "step": 14396 }, { "epoch": 0.87, "learning_rate": 4.342669161586286e-05, "loss": 1.3558, "step": 14398 }, { "epoch": 0.87, "learning_rate": 4.343272771171606e-05, "loss": 1.3556, "step": 14400 }, { "epoch": 0.87, "learning_rate": 4.343876380756927e-05, "loss": 1.3363, "step": 14402 }, { "epoch": 0.87, "learning_rate": 4.344479990342247e-05, "loss": 1.544, "step": 14404 }, { "epoch": 0.87, "learning_rate": 4.345083599927567e-05, "loss": 1.1282, "step": 14406 }, { "epoch": 0.87, "learning_rate": 4.345687209512888e-05, "loss": 1.3419, "step": 14408 }, { "epoch": 0.87, "learning_rate": 4.346290819098208e-05, "loss": 1.0916, "step": 14410 }, { "epoch": 0.87, "learning_rate": 4.346894428683528e-05, "loss": 1.131, "step": 14412 }, { "epoch": 0.87, "learning_rate": 4.347498038268848e-05, "loss": 1.097, "step": 14414 }, { "epoch": 0.87, "learning_rate": 4.348101647854168e-05, "loss": 0.9746, "step": 14416 }, { "epoch": 0.87, "learning_rate": 4.348705257439488e-05, "loss": 1.297, "step": 14418 }, { "epoch": 0.87, "learning_rate": 4.349308867024808e-05, "loss": 0.9693, "step": 14420 }, { "epoch": 0.87, "learning_rate": 4.349912476610129e-05, "loss": 1.5666, "step": 14422 }, { "epoch": 0.87, "learning_rate": 4.350516086195449e-05, "loss": 1.5468, "step": 14424 }, { "epoch": 0.87, "learning_rate": 4.351119695780769e-05, "loss": 1.3524, "step": 14426 }, { "epoch": 0.87, "learning_rate": 4.3517233053660893e-05, "loss": 1.1052, "step": 14428 }, { "epoch": 0.87, "learning_rate": 4.3523269149514094e-05, "loss": 1.1544, "step": 14430 }, { "epoch": 0.87, "learning_rate": 4.3529305245367295e-05, "loss": 0.7354, "step": 14432 }, { "epoch": 0.87, "learning_rate": 4.3535341341220496e-05, "loss": 1.3799, "step": 14434 }, { "epoch": 0.87, "learning_rate": 4.3541377437073703e-05, "loss": 1.2448, "step": 14436 }, { "epoch": 0.87, "learning_rate": 4.3547413532926904e-05, "loss": 1.2059, "step": 14438 }, { "epoch": 0.87, "learning_rate": 4.3553449628780105e-05, "loss": 1.2446, "step": 14440 }, { "epoch": 0.87, "learning_rate": 4.3559485724633306e-05, "loss": 1.5729, "step": 14442 }, { "epoch": 0.87, "learning_rate": 4.356552182048651e-05, "loss": 1.2233, "step": 14444 }, { "epoch": 0.87, "learning_rate": 4.3571557916339714e-05, "loss": 1.2116, "step": 14446 }, { "epoch": 0.87, "learning_rate": 4.3577594012192915e-05, "loss": 1.0995, "step": 14448 }, { "epoch": 0.87, "learning_rate": 4.358363010804612e-05, "loss": 0.9879, "step": 14450 }, { "epoch": 0.87, "learning_rate": 4.3589666203899324e-05, "loss": 1.0657, "step": 14452 }, { "epoch": 0.87, "learning_rate": 4.3595702299752524e-05, "loss": 1.567, "step": 14454 }, { "epoch": 0.87, "learning_rate": 4.3601738395605725e-05, "loss": 1.1389, "step": 14456 }, { "epoch": 0.87, "learning_rate": 4.3607774491458926e-05, "loss": 0.9562, "step": 14458 }, { "epoch": 0.87, "learning_rate": 4.361381058731213e-05, "loss": 1.0452, "step": 14460 }, { "epoch": 0.87, "learning_rate": 4.3619846683165334e-05, "loss": 1.4343, "step": 14462 }, { "epoch": 0.87, "learning_rate": 4.3625882779018535e-05, "loss": 1.3068, "step": 14464 }, { "epoch": 0.87, "learning_rate": 4.3631918874871736e-05, "loss": 1.2284, "step": 14466 }, { "epoch": 0.87, "learning_rate": 4.363795497072494e-05, "loss": 1.349, "step": 14468 }, { "epoch": 0.87, "learning_rate": 4.364399106657814e-05, "loss": 1.2115, "step": 14470 }, { "epoch": 0.87, "learning_rate": 4.365002716243134e-05, "loss": 0.9565, "step": 14472 }, { "epoch": 0.87, "learning_rate": 4.3656063258284546e-05, "loss": 1.4917, "step": 14474 }, { "epoch": 0.87, "learning_rate": 4.366209935413775e-05, "loss": 0.9781, "step": 14476 }, { "epoch": 0.87, "learning_rate": 4.366813544999095e-05, "loss": 0.8375, "step": 14478 }, { "epoch": 0.87, "learning_rate": 4.367417154584415e-05, "loss": 1.3624, "step": 14480 }, { "epoch": 0.87, "learning_rate": 4.368020764169735e-05, "loss": 1.5095, "step": 14482 }, { "epoch": 0.87, "learning_rate": 4.368624373755055e-05, "loss": 1.6301, "step": 14484 }, { "epoch": 0.87, "learning_rate": 4.369227983340375e-05, "loss": 1.0389, "step": 14486 }, { "epoch": 0.87, "learning_rate": 4.369831592925696e-05, "loss": 1.2477, "step": 14488 }, { "epoch": 0.87, "learning_rate": 4.370435202511016e-05, "loss": 1.3136, "step": 14490 }, { "epoch": 0.87, "learning_rate": 4.371038812096336e-05, "loss": 1.4748, "step": 14492 }, { "epoch": 0.87, "learning_rate": 4.371642421681657e-05, "loss": 1.1889, "step": 14494 }, { "epoch": 0.87, "learning_rate": 4.372246031266977e-05, "loss": 1.3277, "step": 14496 }, { "epoch": 0.88, "learning_rate": 4.372849640852297e-05, "loss": 1.1303, "step": 14498 }, { "epoch": 0.88, "learning_rate": 4.373453250437617e-05, "loss": 1.1017, "step": 14500 }, { "epoch": 0.88, "learning_rate": 4.374056860022938e-05, "loss": 1.1388, "step": 14502 }, { "epoch": 0.88, "learning_rate": 4.374660469608258e-05, "loss": 1.247, "step": 14504 }, { "epoch": 0.88, "learning_rate": 4.375264079193578e-05, "loss": 1.1591, "step": 14506 }, { "epoch": 0.88, "learning_rate": 4.375867688778898e-05, "loss": 1.1526, "step": 14508 }, { "epoch": 0.88, "learning_rate": 4.376471298364218e-05, "loss": 1.0246, "step": 14510 }, { "epoch": 0.88, "learning_rate": 4.377074907949538e-05, "loss": 1.3047, "step": 14512 }, { "epoch": 0.88, "learning_rate": 4.377678517534859e-05, "loss": 1.1557, "step": 14514 }, { "epoch": 0.88, "learning_rate": 4.378282127120179e-05, "loss": 1.0185, "step": 14516 }, { "epoch": 0.88, "learning_rate": 4.378885736705499e-05, "loss": 1.1139, "step": 14518 }, { "epoch": 0.88, "learning_rate": 4.379489346290819e-05, "loss": 1.4955, "step": 14520 }, { "epoch": 0.88, "learning_rate": 4.380092955876139e-05, "loss": 1.1378, "step": 14522 }, { "epoch": 0.88, "learning_rate": 4.3806965654614594e-05, "loss": 1.3134, "step": 14524 }, { "epoch": 0.88, "learning_rate": 4.3813001750467795e-05, "loss": 1.2835, "step": 14526 }, { "epoch": 0.88, "learning_rate": 4.3819037846321e-05, "loss": 1.0472, "step": 14528 }, { "epoch": 0.88, "learning_rate": 4.38250739421742e-05, "loss": 1.0798, "step": 14530 }, { "epoch": 0.88, "learning_rate": 4.3831110038027404e-05, "loss": 1.3094, "step": 14532 }, { "epoch": 0.88, "learning_rate": 4.3837146133880605e-05, "loss": 1.4749, "step": 14534 }, { "epoch": 0.88, "learning_rate": 4.384318222973381e-05, "loss": 1.0506, "step": 14536 }, { "epoch": 0.88, "learning_rate": 4.384921832558701e-05, "loss": 0.929, "step": 14538 }, { "epoch": 0.88, "learning_rate": 4.385525442144022e-05, "loss": 1.2427, "step": 14540 }, { "epoch": 0.88, "learning_rate": 4.386129051729342e-05, "loss": 1.1312, "step": 14542 }, { "epoch": 0.88, "learning_rate": 4.386732661314662e-05, "loss": 1.3738, "step": 14544 }, { "epoch": 0.88, "learning_rate": 4.387336270899982e-05, "loss": 0.8672, "step": 14546 }, { "epoch": 0.88, "learning_rate": 4.3879398804853024e-05, "loss": 1.4484, "step": 14548 }, { "epoch": 0.88, "learning_rate": 4.3885434900706225e-05, "loss": 1.0428, "step": 14550 }, { "epoch": 0.88, "learning_rate": 4.3891470996559426e-05, "loss": 1.1764, "step": 14552 }, { "epoch": 0.88, "learning_rate": 4.389750709241263e-05, "loss": 1.3682, "step": 14554 }, { "epoch": 0.88, "learning_rate": 4.3903543188265834e-05, "loss": 1.3708, "step": 14556 }, { "epoch": 0.88, "learning_rate": 4.3909579284119035e-05, "loss": 1.2693, "step": 14558 }, { "epoch": 0.88, "learning_rate": 4.3915615379972236e-05, "loss": 1.5891, "step": 14560 }, { "epoch": 0.88, "learning_rate": 4.3921651475825436e-05, "loss": 1.1902, "step": 14562 }, { "epoch": 0.88, "learning_rate": 4.392768757167864e-05, "loss": 1.4454, "step": 14564 }, { "epoch": 0.88, "learning_rate": 4.393372366753184e-05, "loss": 1.4258, "step": 14566 }, { "epoch": 0.88, "learning_rate": 4.3939759763385046e-05, "loss": 1.6103, "step": 14568 }, { "epoch": 0.88, "learning_rate": 4.3945795859238246e-05, "loss": 1.2596, "step": 14570 }, { "epoch": 0.88, "learning_rate": 4.395183195509145e-05, "loss": 1.1832, "step": 14572 }, { "epoch": 0.88, "learning_rate": 4.395786805094465e-05, "loss": 1.1248, "step": 14574 }, { "epoch": 0.88, "learning_rate": 4.396390414679785e-05, "loss": 1.2222, "step": 14576 }, { "epoch": 0.88, "learning_rate": 4.3969940242651056e-05, "loss": 1.2497, "step": 14578 }, { "epoch": 0.88, "learning_rate": 4.397597633850426e-05, "loss": 1.3516, "step": 14580 }, { "epoch": 0.88, "learning_rate": 4.398201243435746e-05, "loss": 1.3004, "step": 14582 }, { "epoch": 0.88, "learning_rate": 4.3988048530210666e-05, "loss": 1.1555, "step": 14584 }, { "epoch": 0.88, "learning_rate": 4.3994084626063867e-05, "loss": 1.3549, "step": 14586 }, { "epoch": 0.88, "learning_rate": 4.400012072191707e-05, "loss": 1.0574, "step": 14588 }, { "epoch": 0.88, "learning_rate": 4.400615681777027e-05, "loss": 1.3482, "step": 14590 }, { "epoch": 0.88, "learning_rate": 4.401219291362347e-05, "loss": 1.2339, "step": 14592 }, { "epoch": 0.88, "learning_rate": 4.4018229009476677e-05, "loss": 1.6984, "step": 14594 }, { "epoch": 0.88, "learning_rate": 4.402426510532988e-05, "loss": 1.4691, "step": 14596 }, { "epoch": 0.88, "learning_rate": 4.403030120118308e-05, "loss": 1.0231, "step": 14598 }, { "epoch": 0.88, "learning_rate": 4.403633729703628e-05, "loss": 0.886, "step": 14600 }, { "epoch": 0.88, "learning_rate": 4.404237339288948e-05, "loss": 1.1583, "step": 14602 }, { "epoch": 0.88, "learning_rate": 4.404840948874268e-05, "loss": 1.496, "step": 14604 }, { "epoch": 0.88, "learning_rate": 4.405444558459589e-05, "loss": 1.5821, "step": 14606 }, { "epoch": 0.88, "learning_rate": 4.406048168044909e-05, "loss": 1.4166, "step": 14608 }, { "epoch": 0.88, "learning_rate": 4.406651777630229e-05, "loss": 1.2533, "step": 14610 }, { "epoch": 0.88, "learning_rate": 4.407255387215549e-05, "loss": 1.5013, "step": 14612 }, { "epoch": 0.88, "learning_rate": 4.407858996800869e-05, "loss": 1.4132, "step": 14614 }, { "epoch": 0.88, "learning_rate": 4.408462606386189e-05, "loss": 0.9925, "step": 14616 }, { "epoch": 0.88, "learning_rate": 4.409066215971509e-05, "loss": 1.6461, "step": 14618 }, { "epoch": 0.88, "learning_rate": 4.40966982555683e-05, "loss": 1.2277, "step": 14620 }, { "epoch": 0.88, "learning_rate": 4.41027343514215e-05, "loss": 1.2843, "step": 14622 }, { "epoch": 0.88, "learning_rate": 4.41087704472747e-05, "loss": 1.1396, "step": 14624 }, { "epoch": 0.88, "learning_rate": 4.411480654312791e-05, "loss": 1.0203, "step": 14626 }, { "epoch": 0.88, "learning_rate": 4.412084263898111e-05, "loss": 1.9731, "step": 14628 }, { "epoch": 0.88, "learning_rate": 4.412687873483431e-05, "loss": 1.421, "step": 14630 }, { "epoch": 0.88, "learning_rate": 4.413291483068752e-05, "loss": 1.2741, "step": 14632 }, { "epoch": 0.88, "learning_rate": 4.413895092654072e-05, "loss": 1.3038, "step": 14634 }, { "epoch": 0.88, "learning_rate": 4.414498702239392e-05, "loss": 1.249, "step": 14636 }, { "epoch": 0.88, "learning_rate": 4.415102311824712e-05, "loss": 1.1217, "step": 14638 }, { "epoch": 0.88, "learning_rate": 4.415705921410032e-05, "loss": 1.0119, "step": 14640 }, { "epoch": 0.88, "learning_rate": 4.416309530995352e-05, "loss": 1.0847, "step": 14642 }, { "epoch": 0.88, "learning_rate": 4.4169131405806724e-05, "loss": 1.1651, "step": 14644 }, { "epoch": 0.88, "learning_rate": 4.417516750165993e-05, "loss": 1.538, "step": 14646 }, { "epoch": 0.88, "learning_rate": 4.418120359751313e-05, "loss": 1.3983, "step": 14648 }, { "epoch": 0.88, "learning_rate": 4.418723969336633e-05, "loss": 1.2895, "step": 14650 }, { "epoch": 0.88, "learning_rate": 4.4193275789219534e-05, "loss": 1.2463, "step": 14652 }, { "epoch": 0.88, "learning_rate": 4.4199311885072735e-05, "loss": 1.1953, "step": 14654 }, { "epoch": 0.88, "learning_rate": 4.4205347980925936e-05, "loss": 1.4388, "step": 14656 }, { "epoch": 0.88, "learning_rate": 4.421138407677914e-05, "loss": 1.5726, "step": 14658 }, { "epoch": 0.88, "learning_rate": 4.4217420172632344e-05, "loss": 1.0525, "step": 14660 }, { "epoch": 0.88, "learning_rate": 4.4223456268485545e-05, "loss": 1.0716, "step": 14662 }, { "epoch": 0.89, "learning_rate": 4.4229492364338746e-05, "loss": 1.1515, "step": 14664 }, { "epoch": 0.89, "learning_rate": 4.423552846019195e-05, "loss": 0.9159, "step": 14666 }, { "epoch": 0.89, "learning_rate": 4.424156455604515e-05, "loss": 1.1476, "step": 14668 }, { "epoch": 0.89, "learning_rate": 4.4247600651898355e-05, "loss": 0.9737, "step": 14670 }, { "epoch": 0.89, "learning_rate": 4.4253636747751556e-05, "loss": 0.8208, "step": 14672 }, { "epoch": 0.89, "learning_rate": 4.4259672843604764e-05, "loss": 1.6165, "step": 14674 }, { "epoch": 0.89, "learning_rate": 4.4265708939457964e-05, "loss": 1.3028, "step": 14676 }, { "epoch": 0.89, "learning_rate": 4.4271745035311165e-05, "loss": 0.942, "step": 14678 }, { "epoch": 0.89, "learning_rate": 4.4277781131164366e-05, "loss": 1.5782, "step": 14680 }, { "epoch": 0.89, "learning_rate": 4.428381722701757e-05, "loss": 1.4276, "step": 14682 }, { "epoch": 0.89, "learning_rate": 4.428985332287077e-05, "loss": 1.3794, "step": 14684 }, { "epoch": 0.89, "learning_rate": 4.4295889418723975e-05, "loss": 1.0974, "step": 14686 }, { "epoch": 0.89, "learning_rate": 4.4301925514577176e-05, "loss": 1.7625, "step": 14688 }, { "epoch": 0.89, "learning_rate": 4.430796161043038e-05, "loss": 1.3552, "step": 14690 }, { "epoch": 0.89, "learning_rate": 4.431399770628358e-05, "loss": 0.9257, "step": 14692 }, { "epoch": 0.89, "learning_rate": 4.432003380213678e-05, "loss": 1.0524, "step": 14694 }, { "epoch": 0.89, "learning_rate": 4.432606989798998e-05, "loss": 1.7567, "step": 14696 }, { "epoch": 0.89, "learning_rate": 4.433210599384319e-05, "loss": 1.0214, "step": 14698 }, { "epoch": 0.89, "learning_rate": 4.433814208969639e-05, "loss": 0.9474, "step": 14700 }, { "epoch": 0.89, "learning_rate": 4.434417818554959e-05, "loss": 1.1932, "step": 14702 }, { "epoch": 0.89, "learning_rate": 4.435021428140279e-05, "loss": 1.2147, "step": 14704 }, { "epoch": 0.89, "learning_rate": 4.435625037725599e-05, "loss": 1.1132, "step": 14706 }, { "epoch": 0.89, "learning_rate": 4.436228647310919e-05, "loss": 1.5461, "step": 14708 }, { "epoch": 0.89, "learning_rate": 4.436832256896239e-05, "loss": 1.0301, "step": 14710 }, { "epoch": 0.89, "learning_rate": 4.43743586648156e-05, "loss": 1.1531, "step": 14712 }, { "epoch": 0.89, "learning_rate": 4.43803947606688e-05, "loss": 1.2877, "step": 14714 }, { "epoch": 0.89, "learning_rate": 4.4386430856522e-05, "loss": 1.5131, "step": 14716 }, { "epoch": 0.89, "learning_rate": 4.439246695237521e-05, "loss": 1.1142, "step": 14718 }, { "epoch": 0.89, "learning_rate": 4.439850304822841e-05, "loss": 1.2667, "step": 14720 }, { "epoch": 0.89, "learning_rate": 4.440453914408161e-05, "loss": 1.0418, "step": 14722 }, { "epoch": 0.89, "learning_rate": 4.441057523993481e-05, "loss": 1.091, "step": 14724 }, { "epoch": 0.89, "learning_rate": 4.441661133578802e-05, "loss": 1.1321, "step": 14726 }, { "epoch": 0.89, "learning_rate": 4.442264743164122e-05, "loss": 0.8792, "step": 14728 }, { "epoch": 0.89, "learning_rate": 4.442868352749442e-05, "loss": 1.171, "step": 14730 }, { "epoch": 0.89, "learning_rate": 4.443471962334762e-05, "loss": 1.0932, "step": 14732 }, { "epoch": 0.89, "learning_rate": 4.444075571920082e-05, "loss": 1.4672, "step": 14734 }, { "epoch": 0.89, "learning_rate": 4.444679181505402e-05, "loss": 1.3575, "step": 14736 }, { "epoch": 0.89, "learning_rate": 4.445282791090723e-05, "loss": 1.4156, "step": 14738 }, { "epoch": 0.89, "learning_rate": 4.445886400676043e-05, "loss": 1.4942, "step": 14740 }, { "epoch": 0.89, "learning_rate": 4.446490010261363e-05, "loss": 1.0414, "step": 14742 }, { "epoch": 0.89, "learning_rate": 4.447093619846683e-05, "loss": 1.0132, "step": 14744 }, { "epoch": 0.89, "learning_rate": 4.4476972294320034e-05, "loss": 1.2814, "step": 14746 }, { "epoch": 0.89, "learning_rate": 4.4483008390173234e-05, "loss": 0.9789, "step": 14748 }, { "epoch": 0.89, "learning_rate": 4.4489044486026435e-05, "loss": 0.9761, "step": 14750 }, { "epoch": 0.89, "learning_rate": 4.449508058187964e-05, "loss": 1.4171, "step": 14752 }, { "epoch": 0.89, "learning_rate": 4.4501116677732844e-05, "loss": 1.2694, "step": 14754 }, { "epoch": 0.89, "learning_rate": 4.4507152773586045e-05, "loss": 1.0166, "step": 14756 }, { "epoch": 0.89, "learning_rate": 4.4513188869439245e-05, "loss": 1.1708, "step": 14758 }, { "epoch": 0.89, "learning_rate": 4.451922496529245e-05, "loss": 1.2409, "step": 14760 }, { "epoch": 0.89, "learning_rate": 4.4525261061145654e-05, "loss": 1.1998, "step": 14762 }, { "epoch": 0.89, "learning_rate": 4.4531297156998855e-05, "loss": 1.2896, "step": 14764 }, { "epoch": 0.89, "learning_rate": 4.453733325285206e-05, "loss": 1.3764, "step": 14766 }, { "epoch": 0.89, "learning_rate": 4.454336934870526e-05, "loss": 0.8903, "step": 14768 }, { "epoch": 0.89, "learning_rate": 4.4549405444558464e-05, "loss": 0.946, "step": 14770 }, { "epoch": 0.89, "learning_rate": 4.4555441540411665e-05, "loss": 1.2312, "step": 14772 }, { "epoch": 0.89, "learning_rate": 4.4561477636264865e-05, "loss": 1.3628, "step": 14774 }, { "epoch": 0.89, "learning_rate": 4.4567513732118066e-05, "loss": 1.2281, "step": 14776 }, { "epoch": 0.89, "learning_rate": 4.4573549827971274e-05, "loss": 1.2857, "step": 14778 }, { "epoch": 0.89, "learning_rate": 4.4579585923824475e-05, "loss": 1.0672, "step": 14780 }, { "epoch": 0.89, "learning_rate": 4.4585622019677675e-05, "loss": 1.1387, "step": 14782 }, { "epoch": 0.89, "learning_rate": 4.4591658115530876e-05, "loss": 0.9963, "step": 14784 }, { "epoch": 0.89, "learning_rate": 4.459769421138408e-05, "loss": 1.4398, "step": 14786 }, { "epoch": 0.89, "learning_rate": 4.460373030723728e-05, "loss": 1.001, "step": 14788 }, { "epoch": 0.89, "learning_rate": 4.460976640309048e-05, "loss": 1.3138, "step": 14790 }, { "epoch": 0.89, "learning_rate": 4.4615802498943686e-05, "loss": 1.3185, "step": 14792 }, { "epoch": 0.89, "learning_rate": 4.462183859479689e-05, "loss": 1.2061, "step": 14794 }, { "epoch": 0.89, "learning_rate": 4.462787469065009e-05, "loss": 1.1048, "step": 14796 }, { "epoch": 0.89, "learning_rate": 4.463391078650329e-05, "loss": 1.4264, "step": 14798 }, { "epoch": 0.89, "learning_rate": 4.463994688235649e-05, "loss": 0.9076, "step": 14800 }, { "epoch": 0.89, "learning_rate": 4.46459829782097e-05, "loss": 1.2369, "step": 14802 }, { "epoch": 0.89, "learning_rate": 4.46520190740629e-05, "loss": 1.1905, "step": 14804 }, { "epoch": 0.89, "learning_rate": 4.46580551699161e-05, "loss": 1.2079, "step": 14806 }, { "epoch": 0.89, "learning_rate": 4.4664091265769306e-05, "loss": 1.2915, "step": 14808 }, { "epoch": 0.89, "learning_rate": 4.467012736162251e-05, "loss": 1.3832, "step": 14810 }, { "epoch": 0.89, "learning_rate": 4.467616345747571e-05, "loss": 0.7916, "step": 14812 }, { "epoch": 0.89, "learning_rate": 4.468219955332891e-05, "loss": 1.2447, "step": 14814 }, { "epoch": 0.89, "learning_rate": 4.468823564918211e-05, "loss": 1.0759, "step": 14816 }, { "epoch": 0.89, "learning_rate": 4.469427174503532e-05, "loss": 1.5476, "step": 14818 }, { "epoch": 0.89, "learning_rate": 4.470030784088852e-05, "loss": 1.2104, "step": 14820 }, { "epoch": 0.89, "learning_rate": 4.470634393674172e-05, "loss": 0.9985, "step": 14822 }, { "epoch": 0.89, "learning_rate": 4.471238003259492e-05, "loss": 0.9807, "step": 14824 }, { "epoch": 0.89, "learning_rate": 4.471841612844812e-05, "loss": 1.2142, "step": 14826 }, { "epoch": 0.89, "learning_rate": 4.472445222430132e-05, "loss": 1.1459, "step": 14828 }, { "epoch": 0.9, "learning_rate": 4.473048832015453e-05, "loss": 1.1704, "step": 14830 }, { "epoch": 0.9, "learning_rate": 4.473652441600773e-05, "loss": 1.2967, "step": 14832 }, { "epoch": 0.9, "learning_rate": 4.474256051186093e-05, "loss": 1.1299, "step": 14834 }, { "epoch": 0.9, "learning_rate": 4.474859660771413e-05, "loss": 1.1768, "step": 14836 }, { "epoch": 0.9, "learning_rate": 4.475463270356733e-05, "loss": 0.8793, "step": 14838 }, { "epoch": 0.9, "learning_rate": 4.476066879942053e-05, "loss": 1.1052, "step": 14840 }, { "epoch": 0.9, "learning_rate": 4.4766704895273734e-05, "loss": 1.3958, "step": 14842 }, { "epoch": 0.9, "learning_rate": 4.477274099112694e-05, "loss": 1.2773, "step": 14844 }, { "epoch": 0.9, "learning_rate": 4.477877708698014e-05, "loss": 0.9468, "step": 14846 }, { "epoch": 0.9, "learning_rate": 4.478481318283334e-05, "loss": 1.3387, "step": 14848 }, { "epoch": 0.9, "learning_rate": 4.479084927868655e-05, "loss": 1.1697, "step": 14850 }, { "epoch": 0.9, "learning_rate": 4.479688537453975e-05, "loss": 1.4469, "step": 14852 }, { "epoch": 0.9, "learning_rate": 4.480292147039295e-05, "loss": 1.2609, "step": 14854 }, { "epoch": 0.9, "learning_rate": 4.480895756624615e-05, "loss": 1.2742, "step": 14856 }, { "epoch": 0.9, "learning_rate": 4.481499366209936e-05, "loss": 1.0265, "step": 14858 }, { "epoch": 0.9, "learning_rate": 4.482102975795256e-05, "loss": 1.3745, "step": 14860 }, { "epoch": 0.9, "learning_rate": 4.482706585380576e-05, "loss": 1.3172, "step": 14862 }, { "epoch": 0.9, "learning_rate": 4.483310194965896e-05, "loss": 1.0121, "step": 14864 }, { "epoch": 0.9, "learning_rate": 4.4839138045512164e-05, "loss": 1.3996, "step": 14866 }, { "epoch": 0.9, "learning_rate": 4.4845174141365365e-05, "loss": 1.3267, "step": 14868 }, { "epoch": 0.9, "learning_rate": 4.485121023721857e-05, "loss": 1.2139, "step": 14870 }, { "epoch": 0.9, "learning_rate": 4.485724633307177e-05, "loss": 1.2647, "step": 14872 }, { "epoch": 0.9, "learning_rate": 4.4863282428924974e-05, "loss": 1.3683, "step": 14874 }, { "epoch": 0.9, "learning_rate": 4.4869318524778175e-05, "loss": 0.9394, "step": 14876 }, { "epoch": 0.9, "learning_rate": 4.4875354620631376e-05, "loss": 1.3189, "step": 14878 }, { "epoch": 0.9, "learning_rate": 4.4881390716484577e-05, "loss": 1.2319, "step": 14880 }, { "epoch": 0.9, "learning_rate": 4.488742681233778e-05, "loss": 1.2893, "step": 14882 }, { "epoch": 0.9, "learning_rate": 4.4893462908190985e-05, "loss": 1.4508, "step": 14884 }, { "epoch": 0.9, "learning_rate": 4.4899499004044186e-05, "loss": 1.1601, "step": 14886 }, { "epoch": 0.9, "learning_rate": 4.490553509989739e-05, "loss": 0.7803, "step": 14888 }, { "epoch": 0.9, "learning_rate": 4.491157119575059e-05, "loss": 1.3601, "step": 14890 }, { "epoch": 0.9, "learning_rate": 4.491760729160379e-05, "loss": 1.3294, "step": 14892 }, { "epoch": 0.9, "learning_rate": 4.4923643387456996e-05, "loss": 1.194, "step": 14894 }, { "epoch": 0.9, "learning_rate": 4.49296794833102e-05, "loss": 1.6458, "step": 14896 }, { "epoch": 0.9, "learning_rate": 4.4935715579163404e-05, "loss": 1.5518, "step": 14898 }, { "epoch": 0.9, "learning_rate": 4.4941751675016605e-05, "loss": 0.8457, "step": 14900 }, { "epoch": 0.9, "learning_rate": 4.4947787770869806e-05, "loss": 1.3045, "step": 14902 }, { "epoch": 0.9, "learning_rate": 4.495382386672301e-05, "loss": 1.1204, "step": 14904 }, { "epoch": 0.9, "learning_rate": 4.495985996257621e-05, "loss": 1.3273, "step": 14906 }, { "epoch": 0.9, "learning_rate": 4.496589605842941e-05, "loss": 1.3436, "step": 14908 }, { "epoch": 0.9, "learning_rate": 4.4971932154282616e-05, "loss": 1.2964, "step": 14910 }, { "epoch": 0.9, "learning_rate": 4.497796825013582e-05, "loss": 1.3664, "step": 14912 }, { "epoch": 0.9, "learning_rate": 4.498400434598902e-05, "loss": 0.8966, "step": 14914 }, { "epoch": 0.9, "learning_rate": 4.499004044184222e-05, "loss": 1.1771, "step": 14916 }, { "epoch": 0.9, "learning_rate": 4.499607653769542e-05, "loss": 0.9784, "step": 14918 }, { "epoch": 0.9, "learning_rate": 4.500211263354862e-05, "loss": 1.372, "step": 14920 }, { "epoch": 0.9, "learning_rate": 4.500814872940183e-05, "loss": 1.3343, "step": 14922 }, { "epoch": 0.9, "learning_rate": 4.501418482525503e-05, "loss": 1.0905, "step": 14924 }, { "epoch": 0.9, "learning_rate": 4.502022092110823e-05, "loss": 1.1556, "step": 14926 }, { "epoch": 0.9, "learning_rate": 4.502625701696143e-05, "loss": 1.3244, "step": 14928 }, { "epoch": 0.9, "learning_rate": 4.503229311281463e-05, "loss": 0.9927, "step": 14930 }, { "epoch": 0.9, "learning_rate": 4.503832920866783e-05, "loss": 1.3645, "step": 14932 }, { "epoch": 0.9, "learning_rate": 4.504436530452103e-05, "loss": 1.4091, "step": 14934 }, { "epoch": 0.9, "learning_rate": 4.505040140037424e-05, "loss": 1.1038, "step": 14936 }, { "epoch": 0.9, "learning_rate": 4.505643749622744e-05, "loss": 1.3573, "step": 14938 }, { "epoch": 0.9, "learning_rate": 4.506247359208064e-05, "loss": 1.2975, "step": 14940 }, { "epoch": 0.9, "learning_rate": 4.506850968793385e-05, "loss": 1.3397, "step": 14942 }, { "epoch": 0.9, "learning_rate": 4.507454578378705e-05, "loss": 1.3493, "step": 14944 }, { "epoch": 0.9, "learning_rate": 4.508058187964025e-05, "loss": 1.3159, "step": 14946 }, { "epoch": 0.9, "learning_rate": 4.508661797549345e-05, "loss": 1.5255, "step": 14948 }, { "epoch": 0.9, "learning_rate": 4.509265407134666e-05, "loss": 1.2787, "step": 14950 }, { "epoch": 0.9, "learning_rate": 4.509869016719986e-05, "loss": 1.3967, "step": 14952 }, { "epoch": 0.9, "learning_rate": 4.510472626305306e-05, "loss": 1.3856, "step": 14954 }, { "epoch": 0.9, "learning_rate": 4.511076235890626e-05, "loss": 1.1108, "step": 14956 }, { "epoch": 0.9, "learning_rate": 4.511679845475946e-05, "loss": 1.2289, "step": 14958 }, { "epoch": 0.9, "learning_rate": 4.5122834550612664e-05, "loss": 1.109, "step": 14960 }, { "epoch": 0.9, "learning_rate": 4.512887064646587e-05, "loss": 1.1845, "step": 14962 }, { "epoch": 0.9, "learning_rate": 4.513490674231907e-05, "loss": 1.1464, "step": 14964 }, { "epoch": 0.9, "learning_rate": 4.514094283817227e-05, "loss": 0.7486, "step": 14966 }, { "epoch": 0.9, "learning_rate": 4.5146978934025474e-05, "loss": 1.4037, "step": 14968 }, { "epoch": 0.9, "learning_rate": 4.5153015029878674e-05, "loss": 1.1102, "step": 14970 }, { "epoch": 0.9, "learning_rate": 4.5159051125731875e-05, "loss": 1.0331, "step": 14972 }, { "epoch": 0.9, "learning_rate": 4.5165087221585076e-05, "loss": 1.5578, "step": 14974 }, { "epoch": 0.9, "learning_rate": 4.5171123317438284e-05, "loss": 1.3602, "step": 14976 }, { "epoch": 0.9, "learning_rate": 4.5177159413291484e-05, "loss": 1.1432, "step": 14978 }, { "epoch": 0.9, "learning_rate": 4.5183195509144685e-05, "loss": 1.0926, "step": 14980 }, { "epoch": 0.9, "learning_rate": 4.5189231604997886e-05, "loss": 1.3506, "step": 14982 }, { "epoch": 0.9, "learning_rate": 4.5195267700851094e-05, "loss": 1.3194, "step": 14984 }, { "epoch": 0.9, "learning_rate": 4.5201303796704295e-05, "loss": 1.2992, "step": 14986 }, { "epoch": 0.9, "learning_rate": 4.5207339892557495e-05, "loss": 1.6094, "step": 14988 }, { "epoch": 0.9, "learning_rate": 4.52133759884107e-05, "loss": 1.2449, "step": 14990 }, { "epoch": 0.9, "learning_rate": 4.5219412084263904e-05, "loss": 1.0833, "step": 14992 }, { "epoch": 0.9, "learning_rate": 4.5225448180117105e-05, "loss": 1.0115, "step": 14994 }, { "epoch": 0.91, "learning_rate": 4.5231484275970305e-05, "loss": 1.275, "step": 14996 }, { "epoch": 0.91, "learning_rate": 4.5237520371823506e-05, "loss": 1.433, "step": 14998 }, { "epoch": 0.91, "learning_rate": 4.524355646767671e-05, "loss": 1.6457, "step": 15000 }, { "epoch": 0.91, "learning_rate": 4.5249592563529915e-05, "loss": 1.0327, "step": 15002 }, { "epoch": 0.91, "learning_rate": 4.5255628659383115e-05, "loss": 1.3339, "step": 15004 }, { "epoch": 0.91, "learning_rate": 4.5261664755236316e-05, "loss": 1.6551, "step": 15006 }, { "epoch": 0.91, "learning_rate": 4.526770085108952e-05, "loss": 1.2391, "step": 15008 }, { "epoch": 0.91, "learning_rate": 4.527373694694272e-05, "loss": 1.4587, "step": 15010 }, { "epoch": 0.91, "learning_rate": 4.527977304279592e-05, "loss": 1.6848, "step": 15012 }, { "epoch": 0.91, "learning_rate": 4.528580913864912e-05, "loss": 1.1078, "step": 15014 }, { "epoch": 0.91, "learning_rate": 4.529184523450233e-05, "loss": 1.3292, "step": 15016 }, { "epoch": 0.91, "learning_rate": 4.529788133035553e-05, "loss": 1.4496, "step": 15018 }, { "epoch": 0.91, "learning_rate": 4.530391742620873e-05, "loss": 1.4784, "step": 15020 }, { "epoch": 0.91, "learning_rate": 4.530995352206193e-05, "loss": 0.9828, "step": 15022 }, { "epoch": 0.91, "learning_rate": 4.531598961791513e-05, "loss": 0.9783, "step": 15024 }, { "epoch": 0.91, "learning_rate": 4.532202571376834e-05, "loss": 1.1474, "step": 15026 }, { "epoch": 0.91, "learning_rate": 4.532806180962154e-05, "loss": 1.1827, "step": 15028 }, { "epoch": 0.91, "learning_rate": 4.533409790547474e-05, "loss": 1.0396, "step": 15030 }, { "epoch": 0.91, "learning_rate": 4.534013400132795e-05, "loss": 0.9405, "step": 15032 }, { "epoch": 0.91, "learning_rate": 4.534617009718115e-05, "loss": 1.3847, "step": 15034 }, { "epoch": 0.91, "learning_rate": 4.535220619303435e-05, "loss": 1.5188, "step": 15036 }, { "epoch": 0.91, "learning_rate": 4.535824228888755e-05, "loss": 1.6779, "step": 15038 }, { "epoch": 0.91, "learning_rate": 4.536427838474075e-05, "loss": 1.5714, "step": 15040 }, { "epoch": 0.91, "learning_rate": 4.537031448059396e-05, "loss": 1.2689, "step": 15042 }, { "epoch": 0.91, "learning_rate": 4.537635057644716e-05, "loss": 1.2932, "step": 15044 }, { "epoch": 0.91, "learning_rate": 4.538238667230036e-05, "loss": 1.0596, "step": 15046 }, { "epoch": 0.91, "learning_rate": 4.538842276815356e-05, "loss": 1.1758, "step": 15048 }, { "epoch": 0.91, "learning_rate": 4.539445886400676e-05, "loss": 1.0056, "step": 15050 }, { "epoch": 0.91, "learning_rate": 4.540049495985996e-05, "loss": 1.0318, "step": 15052 }, { "epoch": 0.91, "learning_rate": 4.540653105571317e-05, "loss": 1.3767, "step": 15054 }, { "epoch": 0.91, "learning_rate": 4.541256715156637e-05, "loss": 1.012, "step": 15056 }, { "epoch": 0.91, "learning_rate": 4.541860324741957e-05, "loss": 1.6831, "step": 15058 }, { "epoch": 0.91, "learning_rate": 4.542463934327277e-05, "loss": 1.5522, "step": 15060 }, { "epoch": 0.91, "learning_rate": 4.543067543912597e-05, "loss": 1.1256, "step": 15062 }, { "epoch": 0.91, "learning_rate": 4.5436711534979174e-05, "loss": 1.2089, "step": 15064 }, { "epoch": 0.91, "learning_rate": 4.5442747630832375e-05, "loss": 1.2265, "step": 15066 }, { "epoch": 0.91, "learning_rate": 4.544878372668558e-05, "loss": 1.242, "step": 15068 }, { "epoch": 0.91, "learning_rate": 4.545481982253878e-05, "loss": 1.5381, "step": 15070 }, { "epoch": 0.91, "learning_rate": 4.5460855918391984e-05, "loss": 1.2807, "step": 15072 }, { "epoch": 0.91, "learning_rate": 4.546689201424519e-05, "loss": 1.2815, "step": 15074 }, { "epoch": 0.91, "learning_rate": 4.547292811009839e-05, "loss": 1.6556, "step": 15076 }, { "epoch": 0.91, "learning_rate": 4.547896420595159e-05, "loss": 0.9691, "step": 15078 }, { "epoch": 0.91, "learning_rate": 4.5485000301804794e-05, "loss": 0.9762, "step": 15080 }, { "epoch": 0.91, "learning_rate": 4.5491036397658e-05, "loss": 1.1503, "step": 15082 }, { "epoch": 0.91, "learning_rate": 4.54970724935112e-05, "loss": 1.2796, "step": 15084 }, { "epoch": 0.91, "learning_rate": 4.55031085893644e-05, "loss": 1.227, "step": 15086 }, { "epoch": 0.91, "learning_rate": 4.5509144685217604e-05, "loss": 1.147, "step": 15088 }, { "epoch": 0.91, "learning_rate": 4.5515180781070805e-05, "loss": 1.2626, "step": 15090 }, { "epoch": 0.91, "learning_rate": 4.5521216876924006e-05, "loss": 1.2718, "step": 15092 }, { "epoch": 0.91, "learning_rate": 4.552725297277721e-05, "loss": 0.9068, "step": 15094 }, { "epoch": 0.91, "learning_rate": 4.5533289068630414e-05, "loss": 1.3615, "step": 15096 }, { "epoch": 0.91, "learning_rate": 4.5539325164483615e-05, "loss": 1.2094, "step": 15098 }, { "epoch": 0.91, "learning_rate": 4.5545361260336816e-05, "loss": 1.5002, "step": 15100 }, { "epoch": 0.91, "learning_rate": 4.5551397356190017e-05, "loss": 0.9883, "step": 15102 }, { "epoch": 0.91, "learning_rate": 4.555743345204322e-05, "loss": 1.0221, "step": 15104 }, { "epoch": 0.91, "learning_rate": 4.556346954789642e-05, "loss": 1.1752, "step": 15106 }, { "epoch": 0.91, "learning_rate": 4.5569505643749626e-05, "loss": 1.6904, "step": 15108 }, { "epoch": 0.91, "learning_rate": 4.5575541739602827e-05, "loss": 1.589, "step": 15110 }, { "epoch": 0.91, "learning_rate": 4.558157783545603e-05, "loss": 1.0086, "step": 15112 }, { "epoch": 0.91, "learning_rate": 4.558761393130923e-05, "loss": 1.3438, "step": 15114 }, { "epoch": 0.91, "learning_rate": 4.559365002716243e-05, "loss": 1.1327, "step": 15116 }, { "epoch": 0.91, "learning_rate": 4.559968612301564e-05, "loss": 1.1529, "step": 15118 }, { "epoch": 0.91, "learning_rate": 4.560572221886884e-05, "loss": 1.2138, "step": 15120 }, { "epoch": 0.91, "learning_rate": 4.5611758314722045e-05, "loss": 1.2566, "step": 15122 }, { "epoch": 0.91, "learning_rate": 4.5617794410575246e-05, "loss": 1.2168, "step": 15124 }, { "epoch": 0.91, "learning_rate": 4.562383050642845e-05, "loss": 1.4003, "step": 15126 }, { "epoch": 0.91, "learning_rate": 4.562986660228165e-05, "loss": 1.2908, "step": 15128 }, { "epoch": 0.91, "learning_rate": 4.563590269813485e-05, "loss": 1.2137, "step": 15130 }, { "epoch": 0.91, "learning_rate": 4.564193879398805e-05, "loss": 1.1264, "step": 15132 }, { "epoch": 0.91, "learning_rate": 4.564797488984126e-05, "loss": 0.9866, "step": 15134 }, { "epoch": 0.91, "learning_rate": 4.565401098569446e-05, "loss": 1.3139, "step": 15136 }, { "epoch": 0.91, "learning_rate": 4.566004708154766e-05, "loss": 1.0219, "step": 15138 }, { "epoch": 0.91, "learning_rate": 4.566608317740086e-05, "loss": 1.0757, "step": 15140 }, { "epoch": 0.91, "learning_rate": 4.567211927325406e-05, "loss": 1.1923, "step": 15142 }, { "epoch": 0.91, "learning_rate": 4.567815536910726e-05, "loss": 1.3161, "step": 15144 }, { "epoch": 0.91, "learning_rate": 4.568419146496047e-05, "loss": 1.2987, "step": 15146 }, { "epoch": 0.91, "learning_rate": 4.569022756081367e-05, "loss": 1.417, "step": 15148 }, { "epoch": 0.91, "learning_rate": 4.569626365666687e-05, "loss": 1.0054, "step": 15150 }, { "epoch": 0.91, "learning_rate": 4.570229975252007e-05, "loss": 1.3033, "step": 15152 }, { "epoch": 0.91, "learning_rate": 4.570833584837327e-05, "loss": 1.1674, "step": 15154 }, { "epoch": 0.91, "learning_rate": 4.571437194422647e-05, "loss": 0.8635, "step": 15156 }, { "epoch": 0.91, "learning_rate": 4.572040804007967e-05, "loss": 1.1758, "step": 15158 }, { "epoch": 0.92, "learning_rate": 4.572644413593288e-05, "loss": 1.1518, "step": 15160 }, { "epoch": 0.92, "learning_rate": 4.573248023178608e-05, "loss": 1.4237, "step": 15162 }, { "epoch": 0.92, "learning_rate": 4.573851632763928e-05, "loss": 1.2402, "step": 15164 }, { "epoch": 0.92, "learning_rate": 4.574455242349249e-05, "loss": 1.1666, "step": 15166 }, { "epoch": 0.92, "learning_rate": 4.575058851934569e-05, "loss": 1.4405, "step": 15168 }, { "epoch": 0.92, "learning_rate": 4.575662461519889e-05, "loss": 1.0138, "step": 15170 }, { "epoch": 0.92, "learning_rate": 4.576266071105209e-05, "loss": 1.5018, "step": 15172 }, { "epoch": 0.92, "learning_rate": 4.57686968069053e-05, "loss": 1.0008, "step": 15174 }, { "epoch": 0.92, "learning_rate": 4.57747329027585e-05, "loss": 1.8146, "step": 15176 }, { "epoch": 0.92, "learning_rate": 4.57807689986117e-05, "loss": 1.3076, "step": 15178 }, { "epoch": 0.92, "learning_rate": 4.57868050944649e-05, "loss": 1.0257, "step": 15180 }, { "epoch": 0.92, "learning_rate": 4.5792841190318103e-05, "loss": 1.3457, "step": 15182 }, { "epoch": 0.92, "learning_rate": 4.5798877286171304e-05, "loss": 1.2422, "step": 15184 }, { "epoch": 0.92, "learning_rate": 4.580491338202451e-05, "loss": 0.8442, "step": 15186 }, { "epoch": 0.92, "learning_rate": 4.581094947787771e-05, "loss": 1.4254, "step": 15188 }, { "epoch": 0.92, "learning_rate": 4.5816985573730914e-05, "loss": 1.2066, "step": 15190 }, { "epoch": 0.92, "learning_rate": 4.5823021669584114e-05, "loss": 1.4379, "step": 15192 }, { "epoch": 0.92, "learning_rate": 4.5829057765437315e-05, "loss": 1.4931, "step": 15194 }, { "epoch": 0.92, "learning_rate": 4.5835093861290516e-05, "loss": 1.1304, "step": 15196 }, { "epoch": 0.92, "learning_rate": 4.584112995714372e-05, "loss": 1.6024, "step": 15198 }, { "epoch": 0.92, "learning_rate": 4.5847166052996924e-05, "loss": 1.3631, "step": 15200 }, { "epoch": 0.92, "learning_rate": 4.5853202148850125e-05, "loss": 1.4457, "step": 15202 }, { "epoch": 0.92, "learning_rate": 4.5859238244703326e-05, "loss": 1.4971, "step": 15204 }, { "epoch": 0.92, "learning_rate": 4.586527434055653e-05, "loss": 1.6124, "step": 15206 }, { "epoch": 0.92, "learning_rate": 4.5871310436409734e-05, "loss": 1.286, "step": 15208 }, { "epoch": 0.92, "learning_rate": 4.5877346532262935e-05, "loss": 1.3693, "step": 15210 }, { "epoch": 0.92, "learning_rate": 4.5883382628116136e-05, "loss": 0.9702, "step": 15212 }, { "epoch": 0.92, "learning_rate": 4.5889418723969344e-05, "loss": 1.2147, "step": 15214 }, { "epoch": 0.92, "learning_rate": 4.5895454819822545e-05, "loss": 0.8691, "step": 15216 }, { "epoch": 0.92, "learning_rate": 4.5901490915675745e-05, "loss": 1.6689, "step": 15218 }, { "epoch": 0.92, "learning_rate": 4.5907527011528946e-05, "loss": 1.5196, "step": 15220 }, { "epoch": 0.92, "learning_rate": 4.591356310738215e-05, "loss": 1.3406, "step": 15222 }, { "epoch": 0.92, "learning_rate": 4.591959920323535e-05, "loss": 1.1905, "step": 15224 }, { "epoch": 0.92, "learning_rate": 4.5925635299088555e-05, "loss": 1.4087, "step": 15226 }, { "epoch": 0.92, "learning_rate": 4.5931671394941756e-05, "loss": 1.3293, "step": 15228 }, { "epoch": 0.92, "learning_rate": 4.593770749079496e-05, "loss": 1.3388, "step": 15230 }, { "epoch": 0.92, "learning_rate": 4.594374358664816e-05, "loss": 1.6387, "step": 15232 }, { "epoch": 0.92, "learning_rate": 4.594977968250136e-05, "loss": 1.7308, "step": 15234 }, { "epoch": 0.92, "learning_rate": 4.595581577835456e-05, "loss": 0.7928, "step": 15236 }, { "epoch": 0.92, "learning_rate": 4.596185187420776e-05, "loss": 1.4224, "step": 15238 }, { "epoch": 0.92, "learning_rate": 4.596788797006097e-05, "loss": 1.2988, "step": 15240 }, { "epoch": 0.92, "learning_rate": 4.597392406591417e-05, "loss": 1.1486, "step": 15242 }, { "epoch": 0.92, "learning_rate": 4.597996016176737e-05, "loss": 1.4811, "step": 15244 }, { "epoch": 0.92, "learning_rate": 4.598599625762057e-05, "loss": 1.3857, "step": 15246 }, { "epoch": 0.92, "learning_rate": 4.599203235347377e-05, "loss": 1.0382, "step": 15248 }, { "epoch": 0.92, "learning_rate": 4.599806844932697e-05, "loss": 1.0442, "step": 15250 }, { "epoch": 0.92, "learning_rate": 4.600410454518018e-05, "loss": 1.1654, "step": 15252 }, { "epoch": 0.92, "learning_rate": 4.601014064103338e-05, "loss": 1.0223, "step": 15254 }, { "epoch": 0.92, "learning_rate": 4.601617673688659e-05, "loss": 1.1689, "step": 15256 }, { "epoch": 0.92, "learning_rate": 4.602221283273979e-05, "loss": 1.1944, "step": 15258 }, { "epoch": 0.92, "learning_rate": 4.602824892859299e-05, "loss": 1.7358, "step": 15260 }, { "epoch": 0.92, "learning_rate": 4.603428502444619e-05, "loss": 1.0527, "step": 15262 }, { "epoch": 0.92, "learning_rate": 4.604032112029939e-05, "loss": 1.3812, "step": 15264 }, { "epoch": 0.92, "learning_rate": 4.60463572161526e-05, "loss": 1.2655, "step": 15266 }, { "epoch": 0.92, "learning_rate": 4.60523933120058e-05, "loss": 1.2294, "step": 15268 }, { "epoch": 0.92, "learning_rate": 4.6058429407859e-05, "loss": 1.3391, "step": 15270 }, { "epoch": 0.92, "learning_rate": 4.60644655037122e-05, "loss": 1.1264, "step": 15272 }, { "epoch": 0.92, "learning_rate": 4.60705015995654e-05, "loss": 0.9063, "step": 15274 }, { "epoch": 0.92, "learning_rate": 4.60765376954186e-05, "loss": 1.1336, "step": 15276 }, { "epoch": 0.92, "learning_rate": 4.608257379127181e-05, "loss": 1.0671, "step": 15278 }, { "epoch": 0.92, "learning_rate": 4.608860988712501e-05, "loss": 1.0672, "step": 15280 }, { "epoch": 0.92, "learning_rate": 4.609464598297821e-05, "loss": 1.1446, "step": 15282 }, { "epoch": 0.92, "learning_rate": 4.610068207883141e-05, "loss": 1.1829, "step": 15284 }, { "epoch": 0.92, "learning_rate": 4.6106718174684614e-05, "loss": 1.102, "step": 15286 }, { "epoch": 0.92, "learning_rate": 4.6112754270537815e-05, "loss": 1.1522, "step": 15288 }, { "epoch": 0.92, "learning_rate": 4.6118790366391015e-05, "loss": 1.2754, "step": 15290 }, { "epoch": 0.92, "learning_rate": 4.612482646224422e-05, "loss": 1.1034, "step": 15292 }, { "epoch": 0.92, "learning_rate": 4.6130862558097424e-05, "loss": 1.3396, "step": 15294 }, { "epoch": 0.92, "learning_rate": 4.6136898653950625e-05, "loss": 1.1897, "step": 15296 }, { "epoch": 0.92, "learning_rate": 4.614293474980383e-05, "loss": 1.6213, "step": 15298 }, { "epoch": 0.92, "learning_rate": 4.614897084565703e-05, "loss": 1.0074, "step": 15300 }, { "epoch": 0.92, "learning_rate": 4.6155006941510234e-05, "loss": 1.1271, "step": 15302 }, { "epoch": 0.92, "learning_rate": 4.6161043037363435e-05, "loss": 1.4466, "step": 15304 }, { "epoch": 0.92, "learning_rate": 4.616707913321664e-05, "loss": 1.179, "step": 15306 }, { "epoch": 0.92, "learning_rate": 4.617311522906984e-05, "loss": 1.6445, "step": 15308 }, { "epoch": 0.92, "learning_rate": 4.6179151324923044e-05, "loss": 1.5523, "step": 15310 }, { "epoch": 0.92, "learning_rate": 4.6185187420776245e-05, "loss": 1.3029, "step": 15312 }, { "epoch": 0.92, "learning_rate": 4.6191223516629446e-05, "loss": 1.1917, "step": 15314 }, { "epoch": 0.92, "learning_rate": 4.6197259612482646e-05, "loss": 1.0832, "step": 15316 }, { "epoch": 0.92, "learning_rate": 4.6203295708335854e-05, "loss": 1.0562, "step": 15318 }, { "epoch": 0.92, "learning_rate": 4.6209331804189055e-05, "loss": 1.5868, "step": 15320 }, { "epoch": 0.92, "learning_rate": 4.6215367900042256e-05, "loss": 1.1721, "step": 15322 }, { "epoch": 0.92, "learning_rate": 4.6221403995895456e-05, "loss": 1.2496, "step": 15324 }, { "epoch": 0.93, "learning_rate": 4.622744009174866e-05, "loss": 1.4704, "step": 15326 }, { "epoch": 0.93, "learning_rate": 4.623347618760186e-05, "loss": 1.4719, "step": 15328 }, { "epoch": 0.93, "learning_rate": 4.623951228345506e-05, "loss": 1.2055, "step": 15330 }, { "epoch": 0.93, "learning_rate": 4.6245548379308267e-05, "loss": 0.937, "step": 15332 }, { "epoch": 0.93, "learning_rate": 4.625158447516147e-05, "loss": 1.1642, "step": 15334 }, { "epoch": 0.93, "learning_rate": 4.625762057101467e-05, "loss": 1.3381, "step": 15336 }, { "epoch": 0.93, "learning_rate": 4.626365666686787e-05, "loss": 1.6353, "step": 15338 }, { "epoch": 0.93, "learning_rate": 4.626969276272107e-05, "loss": 0.9658, "step": 15340 }, { "epoch": 0.93, "learning_rate": 4.627572885857428e-05, "loss": 1.343, "step": 15342 }, { "epoch": 0.93, "learning_rate": 4.628176495442748e-05, "loss": 1.8664, "step": 15344 }, { "epoch": 0.93, "learning_rate": 4.6287801050280686e-05, "loss": 1.3483, "step": 15346 }, { "epoch": 0.93, "learning_rate": 4.629383714613389e-05, "loss": 1.4456, "step": 15348 }, { "epoch": 0.93, "learning_rate": 4.629987324198709e-05, "loss": 1.0356, "step": 15350 }, { "epoch": 0.93, "learning_rate": 4.630590933784029e-05, "loss": 1.3392, "step": 15352 }, { "epoch": 0.93, "learning_rate": 4.631194543369349e-05, "loss": 1.3167, "step": 15354 }, { "epoch": 0.93, "learning_rate": 4.631798152954669e-05, "loss": 1.3287, "step": 15356 }, { "epoch": 0.93, "learning_rate": 4.63240176253999e-05, "loss": 1.263, "step": 15358 }, { "epoch": 0.93, "learning_rate": 4.63300537212531e-05, "loss": 0.9722, "step": 15360 }, { "epoch": 0.93, "learning_rate": 4.63360898171063e-05, "loss": 1.3083, "step": 15362 }, { "epoch": 0.93, "learning_rate": 4.63421259129595e-05, "loss": 1.3703, "step": 15364 }, { "epoch": 0.93, "learning_rate": 4.63481620088127e-05, "loss": 1.1344, "step": 15366 }, { "epoch": 0.93, "learning_rate": 4.63541981046659e-05, "loss": 1.1152, "step": 15368 }, { "epoch": 0.93, "learning_rate": 4.63602342005191e-05, "loss": 1.4749, "step": 15370 }, { "epoch": 0.93, "learning_rate": 4.636627029637231e-05, "loss": 1.2169, "step": 15372 }, { "epoch": 0.93, "learning_rate": 4.637230639222551e-05, "loss": 1.2721, "step": 15374 }, { "epoch": 0.93, "learning_rate": 4.637834248807871e-05, "loss": 1.2926, "step": 15376 }, { "epoch": 0.93, "learning_rate": 4.638437858393191e-05, "loss": 1.3857, "step": 15378 }, { "epoch": 0.93, "learning_rate": 4.639041467978511e-05, "loss": 1.548, "step": 15380 }, { "epoch": 0.93, "learning_rate": 4.6396450775638314e-05, "loss": 0.9408, "step": 15382 }, { "epoch": 0.93, "learning_rate": 4.640248687149152e-05, "loss": 1.3974, "step": 15384 }, { "epoch": 0.93, "learning_rate": 4.640852296734472e-05, "loss": 1.502, "step": 15386 }, { "epoch": 0.93, "learning_rate": 4.641455906319792e-05, "loss": 1.1411, "step": 15388 }, { "epoch": 0.93, "learning_rate": 4.642059515905113e-05, "loss": 1.1644, "step": 15390 }, { "epoch": 0.93, "learning_rate": 4.642663125490433e-05, "loss": 1.2266, "step": 15392 }, { "epoch": 0.93, "learning_rate": 4.643266735075753e-05, "loss": 1.3799, "step": 15394 }, { "epoch": 0.93, "learning_rate": 4.643870344661073e-05, "loss": 0.9671, "step": 15396 }, { "epoch": 0.93, "learning_rate": 4.644473954246394e-05, "loss": 1.4045, "step": 15398 }, { "epoch": 0.93, "learning_rate": 4.645077563831714e-05, "loss": 1.6132, "step": 15400 }, { "epoch": 0.93, "learning_rate": 4.645681173417034e-05, "loss": 1.0083, "step": 15402 }, { "epoch": 0.93, "learning_rate": 4.6462847830023543e-05, "loss": 1.3952, "step": 15404 }, { "epoch": 0.93, "learning_rate": 4.6468883925876744e-05, "loss": 1.0841, "step": 15406 }, { "epoch": 0.93, "learning_rate": 4.6474920021729945e-05, "loss": 1.2739, "step": 15408 }, { "epoch": 0.93, "learning_rate": 4.648095611758315e-05, "loss": 0.8136, "step": 15410 }, { "epoch": 0.93, "learning_rate": 4.6486992213436353e-05, "loss": 0.9162, "step": 15412 }, { "epoch": 0.93, "learning_rate": 4.6493028309289554e-05, "loss": 1.5469, "step": 15414 }, { "epoch": 0.93, "learning_rate": 4.6499064405142755e-05, "loss": 1.3432, "step": 15416 }, { "epoch": 0.93, "learning_rate": 4.6505100500995956e-05, "loss": 1.539, "step": 15418 }, { "epoch": 0.93, "learning_rate": 4.651113659684916e-05, "loss": 1.69, "step": 15420 }, { "epoch": 0.93, "learning_rate": 4.651717269270236e-05, "loss": 1.1602, "step": 15422 }, { "epoch": 0.93, "learning_rate": 4.6523208788555565e-05, "loss": 1.7087, "step": 15424 }, { "epoch": 0.93, "learning_rate": 4.6529244884408766e-05, "loss": 1.0034, "step": 15426 }, { "epoch": 0.93, "learning_rate": 4.653528098026197e-05, "loss": 1.2954, "step": 15428 }, { "epoch": 0.93, "learning_rate": 4.654131707611517e-05, "loss": 1.2536, "step": 15430 }, { "epoch": 0.93, "learning_rate": 4.6547353171968375e-05, "loss": 1.198, "step": 15432 }, { "epoch": 0.93, "learning_rate": 4.6553389267821576e-05, "loss": 0.9248, "step": 15434 }, { "epoch": 0.93, "learning_rate": 4.655942536367478e-05, "loss": 1.5745, "step": 15436 }, { "epoch": 0.93, "learning_rate": 4.6565461459527984e-05, "loss": 1.3976, "step": 15438 }, { "epoch": 0.93, "learning_rate": 4.6571497555381185e-05, "loss": 1.2362, "step": 15440 }, { "epoch": 0.93, "learning_rate": 4.6577533651234386e-05, "loss": 1.7053, "step": 15442 }, { "epoch": 0.93, "learning_rate": 4.658356974708759e-05, "loss": 1.2077, "step": 15444 }, { "epoch": 0.93, "learning_rate": 4.658960584294079e-05, "loss": 1.1075, "step": 15446 }, { "epoch": 0.93, "learning_rate": 4.659564193879399e-05, "loss": 1.1989, "step": 15448 }, { "epoch": 0.93, "learning_rate": 4.6601678034647196e-05, "loss": 1.2168, "step": 15450 }, { "epoch": 0.93, "learning_rate": 4.66077141305004e-05, "loss": 0.9189, "step": 15452 }, { "epoch": 0.93, "learning_rate": 4.66137502263536e-05, "loss": 1.2444, "step": 15454 }, { "epoch": 0.93, "learning_rate": 4.66197863222068e-05, "loss": 1.0438, "step": 15456 }, { "epoch": 0.93, "learning_rate": 4.662582241806e-05, "loss": 1.3974, "step": 15458 }, { "epoch": 0.93, "learning_rate": 4.66318585139132e-05, "loss": 1.1436, "step": 15460 }, { "epoch": 0.93, "learning_rate": 4.66378946097664e-05, "loss": 1.3585, "step": 15462 }, { "epoch": 0.93, "learning_rate": 4.664393070561961e-05, "loss": 1.2585, "step": 15464 }, { "epoch": 0.93, "learning_rate": 4.664996680147281e-05, "loss": 0.8359, "step": 15466 }, { "epoch": 0.93, "learning_rate": 4.665600289732601e-05, "loss": 1.2866, "step": 15468 }, { "epoch": 0.93, "learning_rate": 4.666203899317921e-05, "loss": 1.2817, "step": 15470 }, { "epoch": 0.93, "learning_rate": 4.666807508903241e-05, "loss": 1.4141, "step": 15472 }, { "epoch": 0.93, "learning_rate": 4.667411118488561e-05, "loss": 1.4891, "step": 15474 }, { "epoch": 0.93, "learning_rate": 4.668014728073882e-05, "loss": 1.3103, "step": 15476 }, { "epoch": 0.93, "learning_rate": 4.668618337659202e-05, "loss": 1.4383, "step": 15478 }, { "epoch": 0.93, "learning_rate": 4.669221947244523e-05, "loss": 1.3091, "step": 15480 }, { "epoch": 0.93, "learning_rate": 4.669825556829843e-05, "loss": 1.5409, "step": 15482 }, { "epoch": 0.93, "learning_rate": 4.670429166415163e-05, "loss": 1.293, "step": 15484 }, { "epoch": 0.93, "learning_rate": 4.671032776000483e-05, "loss": 1.1292, "step": 15486 }, { "epoch": 0.93, "learning_rate": 4.671636385585803e-05, "loss": 1.5241, "step": 15488 }, { "epoch": 0.93, "learning_rate": 4.672239995171124e-05, "loss": 0.9651, "step": 15490 }, { "epoch": 0.94, "learning_rate": 4.672843604756444e-05, "loss": 1.6779, "step": 15492 }, { "epoch": 0.94, "learning_rate": 4.673447214341764e-05, "loss": 1.3931, "step": 15494 }, { "epoch": 0.94, "learning_rate": 4.674050823927084e-05, "loss": 1.0886, "step": 15496 }, { "epoch": 0.94, "learning_rate": 4.674654433512404e-05, "loss": 1.0902, "step": 15498 }, { "epoch": 0.94, "learning_rate": 4.6752580430977244e-05, "loss": 1.2478, "step": 15500 }, { "epoch": 0.94, "learning_rate": 4.675861652683045e-05, "loss": 0.9894, "step": 15502 }, { "epoch": 0.94, "learning_rate": 4.676465262268365e-05, "loss": 1.1637, "step": 15504 }, { "epoch": 0.94, "learning_rate": 4.677068871853685e-05, "loss": 1.1794, "step": 15506 }, { "epoch": 0.94, "learning_rate": 4.6776724814390054e-05, "loss": 1.263, "step": 15508 }, { "epoch": 0.94, "learning_rate": 4.6782760910243255e-05, "loss": 1.0926, "step": 15510 }, { "epoch": 0.94, "learning_rate": 4.6788797006096455e-05, "loss": 1.3738, "step": 15512 }, { "epoch": 0.94, "learning_rate": 4.6794833101949656e-05, "loss": 1.6, "step": 15514 }, { "epoch": 0.94, "learning_rate": 4.6800869197802864e-05, "loss": 1.2617, "step": 15516 }, { "epoch": 0.94, "learning_rate": 4.6806905293656065e-05, "loss": 1.4666, "step": 15518 }, { "epoch": 0.94, "learning_rate": 4.6812941389509265e-05, "loss": 1.0603, "step": 15520 }, { "epoch": 0.94, "learning_rate": 4.6818977485362466e-05, "loss": 1.0924, "step": 15522 }, { "epoch": 0.94, "learning_rate": 4.6825013581215674e-05, "loss": 1.3513, "step": 15524 }, { "epoch": 0.94, "learning_rate": 4.6831049677068875e-05, "loss": 1.2794, "step": 15526 }, { "epoch": 0.94, "learning_rate": 4.6837085772922075e-05, "loss": 1.0705, "step": 15528 }, { "epoch": 0.94, "learning_rate": 4.684312186877528e-05, "loss": 0.9262, "step": 15530 }, { "epoch": 0.94, "learning_rate": 4.6849157964628484e-05, "loss": 1.5524, "step": 15532 }, { "epoch": 0.94, "learning_rate": 4.6855194060481685e-05, "loss": 1.3389, "step": 15534 }, { "epoch": 0.94, "learning_rate": 4.6861230156334886e-05, "loss": 1.2451, "step": 15536 }, { "epoch": 0.94, "learning_rate": 4.6867266252188086e-05, "loss": 1.7934, "step": 15538 }, { "epoch": 0.94, "learning_rate": 4.687330234804129e-05, "loss": 1.1069, "step": 15540 }, { "epoch": 0.94, "learning_rate": 4.6879338443894495e-05, "loss": 1.266, "step": 15542 }, { "epoch": 0.94, "learning_rate": 4.6885374539747696e-05, "loss": 1.8176, "step": 15544 }, { "epoch": 0.94, "learning_rate": 4.6891410635600896e-05, "loss": 1.7317, "step": 15546 }, { "epoch": 0.94, "learning_rate": 4.68974467314541e-05, "loss": 1.5169, "step": 15548 }, { "epoch": 0.94, "learning_rate": 4.69034828273073e-05, "loss": 1.8356, "step": 15550 }, { "epoch": 0.94, "learning_rate": 4.69095189231605e-05, "loss": 1.0207, "step": 15552 }, { "epoch": 0.94, "learning_rate": 4.69155550190137e-05, "loss": 1.0253, "step": 15554 }, { "epoch": 0.94, "learning_rate": 4.692159111486691e-05, "loss": 1.0399, "step": 15556 }, { "epoch": 0.94, "learning_rate": 4.692762721072011e-05, "loss": 1.2656, "step": 15558 }, { "epoch": 0.94, "learning_rate": 4.693366330657331e-05, "loss": 0.9931, "step": 15560 }, { "epoch": 0.94, "learning_rate": 4.693969940242651e-05, "loss": 1.2545, "step": 15562 }, { "epoch": 0.94, "learning_rate": 4.694573549827971e-05, "loss": 1.3857, "step": 15564 }, { "epoch": 0.94, "learning_rate": 4.695177159413292e-05, "loss": 1.2834, "step": 15566 }, { "epoch": 0.94, "learning_rate": 4.695780768998612e-05, "loss": 1.2614, "step": 15568 }, { "epoch": 0.94, "learning_rate": 4.6963843785839327e-05, "loss": 0.9835, "step": 15570 }, { "epoch": 0.94, "learning_rate": 4.696987988169253e-05, "loss": 1.6381, "step": 15572 }, { "epoch": 0.94, "learning_rate": 4.697591597754573e-05, "loss": 0.9605, "step": 15574 }, { "epoch": 0.94, "learning_rate": 4.698195207339893e-05, "loss": 1.0743, "step": 15576 }, { "epoch": 0.94, "learning_rate": 4.698798816925213e-05, "loss": 1.3589, "step": 15578 }, { "epoch": 0.94, "learning_rate": 4.699402426510533e-05, "loss": 1.3847, "step": 15580 }, { "epoch": 0.94, "learning_rate": 4.700006036095854e-05, "loss": 1.1853, "step": 15582 }, { "epoch": 0.94, "learning_rate": 4.700609645681174e-05, "loss": 1.7451, "step": 15584 }, { "epoch": 0.94, "learning_rate": 4.701213255266494e-05, "loss": 1.7316, "step": 15586 }, { "epoch": 0.94, "learning_rate": 4.701816864851814e-05, "loss": 1.1978, "step": 15588 }, { "epoch": 0.94, "learning_rate": 4.702420474437134e-05, "loss": 0.9412, "step": 15590 }, { "epoch": 0.94, "learning_rate": 4.703024084022454e-05, "loss": 1.2321, "step": 15592 }, { "epoch": 0.94, "learning_rate": 4.703627693607774e-05, "loss": 1.0384, "step": 15594 }, { "epoch": 0.94, "learning_rate": 4.704231303193095e-05, "loss": 1.1484, "step": 15596 }, { "epoch": 0.94, "learning_rate": 4.704834912778415e-05, "loss": 0.9842, "step": 15598 }, { "epoch": 0.94, "learning_rate": 4.705438522363735e-05, "loss": 0.9687, "step": 15600 }, { "epoch": 0.94, "learning_rate": 4.706042131949055e-05, "loss": 1.412, "step": 15602 }, { "epoch": 0.94, "learning_rate": 4.7066457415343754e-05, "loss": 1.0669, "step": 15604 }, { "epoch": 0.94, "learning_rate": 4.7072493511196955e-05, "loss": 1.1841, "step": 15606 }, { "epoch": 0.94, "learning_rate": 4.707852960705016e-05, "loss": 1.4573, "step": 15608 }, { "epoch": 0.94, "learning_rate": 4.708456570290336e-05, "loss": 1.6131, "step": 15610 }, { "epoch": 0.94, "learning_rate": 4.7090601798756564e-05, "loss": 1.2417, "step": 15612 }, { "epoch": 0.94, "learning_rate": 4.709663789460977e-05, "loss": 1.2899, "step": 15614 }, { "epoch": 0.94, "learning_rate": 4.710267399046297e-05, "loss": 1.3714, "step": 15616 }, { "epoch": 0.94, "learning_rate": 4.710871008631617e-05, "loss": 1.2953, "step": 15618 }, { "epoch": 0.94, "learning_rate": 4.7114746182169374e-05, "loss": 1.4289, "step": 15620 }, { "epoch": 0.94, "learning_rate": 4.712078227802258e-05, "loss": 1.3257, "step": 15622 }, { "epoch": 0.94, "learning_rate": 4.712681837387578e-05, "loss": 1.3281, "step": 15624 }, { "epoch": 0.94, "learning_rate": 4.713285446972898e-05, "loss": 1.5069, "step": 15626 }, { "epoch": 0.94, "learning_rate": 4.7138890565582184e-05, "loss": 1.5817, "step": 15628 }, { "epoch": 0.94, "learning_rate": 4.7144926661435385e-05, "loss": 1.2873, "step": 15630 }, { "epoch": 0.94, "learning_rate": 4.7150962757288586e-05, "loss": 2.024, "step": 15632 }, { "epoch": 0.94, "learning_rate": 4.7156998853141793e-05, "loss": 1.1056, "step": 15634 }, { "epoch": 0.94, "learning_rate": 4.7163034948994994e-05, "loss": 1.0457, "step": 15636 }, { "epoch": 0.94, "learning_rate": 4.7169071044848195e-05, "loss": 1.5365, "step": 15638 }, { "epoch": 0.94, "learning_rate": 4.7175107140701396e-05, "loss": 1.3915, "step": 15640 }, { "epoch": 0.94, "learning_rate": 4.71811432365546e-05, "loss": 1.4612, "step": 15642 }, { "epoch": 0.94, "learning_rate": 4.71871793324078e-05, "loss": 1.1983, "step": 15644 }, { "epoch": 0.94, "learning_rate": 4.7193215428261e-05, "loss": 1.4542, "step": 15646 }, { "epoch": 0.94, "learning_rate": 4.7199251524114206e-05, "loss": 1.5854, "step": 15648 }, { "epoch": 0.94, "learning_rate": 4.720528761996741e-05, "loss": 1.2893, "step": 15650 }, { "epoch": 0.94, "learning_rate": 4.721132371582061e-05, "loss": 1.1246, "step": 15652 }, { "epoch": 0.94, "learning_rate": 4.721735981167381e-05, "loss": 1.589, "step": 15654 }, { "epoch": 0.94, "learning_rate": 4.7223395907527016e-05, "loss": 1.1522, "step": 15656 }, { "epoch": 0.95, "learning_rate": 4.722943200338022e-05, "loss": 1.4737, "step": 15658 }, { "epoch": 0.95, "learning_rate": 4.723546809923342e-05, "loss": 1.2532, "step": 15660 }, { "epoch": 0.95, "learning_rate": 4.7241504195086625e-05, "loss": 1.3427, "step": 15662 }, { "epoch": 0.95, "learning_rate": 4.7247540290939826e-05, "loss": 1.2429, "step": 15664 }, { "epoch": 0.95, "learning_rate": 4.725357638679303e-05, "loss": 1.3008, "step": 15666 }, { "epoch": 0.95, "learning_rate": 4.725961248264623e-05, "loss": 1.2758, "step": 15668 }, { "epoch": 0.95, "learning_rate": 4.726564857849943e-05, "loss": 1.32, "step": 15670 }, { "epoch": 0.95, "learning_rate": 4.727168467435263e-05, "loss": 1.004, "step": 15672 }, { "epoch": 0.95, "learning_rate": 4.727772077020584e-05, "loss": 1.3968, "step": 15674 }, { "epoch": 0.95, "learning_rate": 4.728375686605904e-05, "loss": 1.2462, "step": 15676 }, { "epoch": 0.95, "learning_rate": 4.728979296191224e-05, "loss": 1.0606, "step": 15678 }, { "epoch": 0.95, "learning_rate": 4.729582905776544e-05, "loss": 1.6797, "step": 15680 }, { "epoch": 0.95, "learning_rate": 4.730186515361864e-05, "loss": 1.3272, "step": 15682 }, { "epoch": 0.95, "learning_rate": 4.730790124947184e-05, "loss": 1.1336, "step": 15684 }, { "epoch": 0.95, "learning_rate": 4.731393734532504e-05, "loss": 0.963, "step": 15686 }, { "epoch": 0.95, "learning_rate": 4.731997344117825e-05, "loss": 1.7251, "step": 15688 }, { "epoch": 0.95, "learning_rate": 4.732600953703145e-05, "loss": 1.278, "step": 15690 }, { "epoch": 0.95, "learning_rate": 4.733204563288465e-05, "loss": 1.2573, "step": 15692 }, { "epoch": 0.95, "learning_rate": 4.733808172873785e-05, "loss": 1.0335, "step": 15694 }, { "epoch": 0.95, "learning_rate": 4.734411782459105e-05, "loss": 1.42, "step": 15696 }, { "epoch": 0.95, "learning_rate": 4.7350153920444253e-05, "loss": 1.1811, "step": 15698 }, { "epoch": 0.95, "learning_rate": 4.735619001629746e-05, "loss": 1.3918, "step": 15700 }, { "epoch": 0.95, "learning_rate": 4.736222611215066e-05, "loss": 1.5479, "step": 15702 }, { "epoch": 0.95, "learning_rate": 4.736826220800387e-05, "loss": 1.1587, "step": 15704 }, { "epoch": 0.95, "learning_rate": 4.737429830385707e-05, "loss": 1.1026, "step": 15706 }, { "epoch": 0.95, "learning_rate": 4.738033439971027e-05, "loss": 1.0644, "step": 15708 }, { "epoch": 0.95, "learning_rate": 4.738637049556347e-05, "loss": 1.257, "step": 15710 }, { "epoch": 0.95, "learning_rate": 4.739240659141667e-05, "loss": 1.1674, "step": 15712 }, { "epoch": 0.95, "learning_rate": 4.739844268726988e-05, "loss": 1.3385, "step": 15714 }, { "epoch": 0.95, "learning_rate": 4.740447878312308e-05, "loss": 1.0178, "step": 15716 }, { "epoch": 0.95, "learning_rate": 4.741051487897628e-05, "loss": 1.8575, "step": 15718 }, { "epoch": 0.95, "learning_rate": 4.741655097482948e-05, "loss": 1.394, "step": 15720 }, { "epoch": 0.95, "learning_rate": 4.7422587070682684e-05, "loss": 0.9952, "step": 15722 }, { "epoch": 0.95, "learning_rate": 4.7428623166535884e-05, "loss": 1.1241, "step": 15724 }, { "epoch": 0.95, "learning_rate": 4.743465926238909e-05, "loss": 1.3636, "step": 15726 }, { "epoch": 0.95, "learning_rate": 4.744069535824229e-05, "loss": 1.1039, "step": 15728 }, { "epoch": 0.95, "learning_rate": 4.7446731454095494e-05, "loss": 1.1011, "step": 15730 }, { "epoch": 0.95, "learning_rate": 4.7452767549948695e-05, "loss": 1.333, "step": 15732 }, { "epoch": 0.95, "learning_rate": 4.7458803645801895e-05, "loss": 1.2381, "step": 15734 }, { "epoch": 0.95, "learning_rate": 4.7464839741655096e-05, "loss": 1.088, "step": 15736 }, { "epoch": 0.95, "learning_rate": 4.74708758375083e-05, "loss": 1.2702, "step": 15738 }, { "epoch": 0.95, "learning_rate": 4.7476911933361505e-05, "loss": 1.5459, "step": 15740 }, { "epoch": 0.95, "learning_rate": 4.7482948029214705e-05, "loss": 1.2538, "step": 15742 }, { "epoch": 0.95, "learning_rate": 4.7488984125067906e-05, "loss": 1.3676, "step": 15744 }, { "epoch": 0.95, "learning_rate": 4.749502022092111e-05, "loss": 1.416, "step": 15746 }, { "epoch": 0.95, "learning_rate": 4.7501056316774315e-05, "loss": 1.1832, "step": 15748 }, { "epoch": 0.95, "learning_rate": 4.7507092412627515e-05, "loss": 0.9293, "step": 15750 }, { "epoch": 0.95, "learning_rate": 4.7513128508480716e-05, "loss": 1.0006, "step": 15752 }, { "epoch": 0.95, "learning_rate": 4.7519164604333924e-05, "loss": 0.8539, "step": 15754 }, { "epoch": 0.95, "learning_rate": 4.7525200700187125e-05, "loss": 1.1583, "step": 15756 }, { "epoch": 0.95, "learning_rate": 4.7531236796040325e-05, "loss": 1.2959, "step": 15758 }, { "epoch": 0.95, "learning_rate": 4.7537272891893526e-05, "loss": 1.1425, "step": 15760 }, { "epoch": 0.95, "learning_rate": 4.754330898774673e-05, "loss": 1.2399, "step": 15762 }, { "epoch": 0.95, "learning_rate": 4.754934508359993e-05, "loss": 1.0655, "step": 15764 }, { "epoch": 0.95, "learning_rate": 4.7555381179453136e-05, "loss": 1.2351, "step": 15766 }, { "epoch": 0.95, "learning_rate": 4.7561417275306336e-05, "loss": 1.2302, "step": 15768 }, { "epoch": 0.95, "learning_rate": 4.756745337115954e-05, "loss": 1.5414, "step": 15770 }, { "epoch": 0.95, "learning_rate": 4.757348946701274e-05, "loss": 1.7299, "step": 15772 }, { "epoch": 0.95, "learning_rate": 4.757952556286594e-05, "loss": 1.3826, "step": 15774 }, { "epoch": 0.95, "learning_rate": 4.758556165871914e-05, "loss": 1.2143, "step": 15776 }, { "epoch": 0.95, "learning_rate": 4.759159775457234e-05, "loss": 1.4699, "step": 15778 }, { "epoch": 0.95, "learning_rate": 4.759763385042555e-05, "loss": 1.6117, "step": 15780 }, { "epoch": 0.95, "learning_rate": 4.760366994627875e-05, "loss": 1.1179, "step": 15782 }, { "epoch": 0.95, "learning_rate": 4.760970604213195e-05, "loss": 1.2735, "step": 15784 }, { "epoch": 0.95, "learning_rate": 4.761574213798515e-05, "loss": 1.6215, "step": 15786 }, { "epoch": 0.95, "learning_rate": 4.762177823383835e-05, "loss": 1.1564, "step": 15788 }, { "epoch": 0.95, "learning_rate": 4.762781432969156e-05, "loss": 1.0536, "step": 15790 }, { "epoch": 0.95, "learning_rate": 4.763385042554476e-05, "loss": 1.243, "step": 15792 }, { "epoch": 0.95, "learning_rate": 4.763988652139796e-05, "loss": 1.3088, "step": 15794 }, { "epoch": 0.95, "learning_rate": 4.764592261725117e-05, "loss": 1.1265, "step": 15796 }, { "epoch": 0.95, "learning_rate": 4.765195871310437e-05, "loss": 1.255, "step": 15798 }, { "epoch": 0.95, "learning_rate": 4.765799480895757e-05, "loss": 1.1785, "step": 15800 }, { "epoch": 0.95, "learning_rate": 4.766403090481077e-05, "loss": 1.0867, "step": 15802 }, { "epoch": 0.95, "learning_rate": 4.767006700066397e-05, "loss": 1.0644, "step": 15804 }, { "epoch": 0.95, "learning_rate": 4.767610309651718e-05, "loss": 1.5844, "step": 15806 }, { "epoch": 0.95, "learning_rate": 4.768213919237038e-05, "loss": 1.4991, "step": 15808 }, { "epoch": 0.95, "learning_rate": 4.768817528822358e-05, "loss": 1.6097, "step": 15810 }, { "epoch": 0.95, "learning_rate": 4.769421138407678e-05, "loss": 1.5733, "step": 15812 }, { "epoch": 0.95, "learning_rate": 4.770024747992998e-05, "loss": 1.4944, "step": 15814 }, { "epoch": 0.95, "learning_rate": 4.770628357578318e-05, "loss": 1.4825, "step": 15816 }, { "epoch": 0.95, "learning_rate": 4.7712319671636384e-05, "loss": 1.0203, "step": 15818 }, { "epoch": 0.95, "learning_rate": 4.771835576748959e-05, "loss": 1.1287, "step": 15820 }, { "epoch": 0.95, "learning_rate": 4.772439186334279e-05, "loss": 1.2133, "step": 15822 }, { "epoch": 0.96, "learning_rate": 4.773042795919599e-05, "loss": 1.1366, "step": 15824 }, { "epoch": 0.96, "learning_rate": 4.7736464055049194e-05, "loss": 1.3393, "step": 15826 }, { "epoch": 0.96, "learning_rate": 4.7742500150902395e-05, "loss": 1.0916, "step": 15828 }, { "epoch": 0.96, "learning_rate": 4.7748536246755596e-05, "loss": 1.0151, "step": 15830 }, { "epoch": 0.96, "learning_rate": 4.77545723426088e-05, "loss": 1.4056, "step": 15832 }, { "epoch": 0.96, "learning_rate": 4.7760608438462004e-05, "loss": 0.7962, "step": 15834 }, { "epoch": 0.96, "learning_rate": 4.7766644534315205e-05, "loss": 1.4874, "step": 15836 }, { "epoch": 0.96, "learning_rate": 4.777268063016841e-05, "loss": 1.0586, "step": 15838 }, { "epoch": 0.96, "learning_rate": 4.777871672602161e-05, "loss": 1.1534, "step": 15840 }, { "epoch": 0.96, "learning_rate": 4.7784752821874814e-05, "loss": 1.2959, "step": 15842 }, { "epoch": 0.96, "learning_rate": 4.7790788917728015e-05, "loss": 1.1235, "step": 15844 }, { "epoch": 0.96, "learning_rate": 4.779682501358122e-05, "loss": 1.3679, "step": 15846 }, { "epoch": 0.96, "learning_rate": 4.780286110943442e-05, "loss": 1.213, "step": 15848 }, { "epoch": 0.96, "learning_rate": 4.7808897205287624e-05, "loss": 1.3029, "step": 15850 }, { "epoch": 0.96, "learning_rate": 4.7814933301140825e-05, "loss": 2.0082, "step": 15852 }, { "epoch": 0.96, "learning_rate": 4.7820969396994026e-05, "loss": 1.5629, "step": 15854 }, { "epoch": 0.96, "learning_rate": 4.7827005492847227e-05, "loss": 0.9463, "step": 15856 }, { "epoch": 0.96, "learning_rate": 4.7833041588700434e-05, "loss": 1.0319, "step": 15858 }, { "epoch": 0.96, "learning_rate": 4.7839077684553635e-05, "loss": 1.1102, "step": 15860 }, { "epoch": 0.96, "learning_rate": 4.7845113780406836e-05, "loss": 1.5358, "step": 15862 }, { "epoch": 0.96, "learning_rate": 4.785114987626004e-05, "loss": 1.5061, "step": 15864 }, { "epoch": 0.96, "learning_rate": 4.785718597211324e-05, "loss": 1.0195, "step": 15866 }, { "epoch": 0.96, "learning_rate": 4.786322206796644e-05, "loss": 0.9817, "step": 15868 }, { "epoch": 0.96, "learning_rate": 4.786925816381964e-05, "loss": 1.1985, "step": 15870 }, { "epoch": 0.96, "learning_rate": 4.787529425967285e-05, "loss": 1.0605, "step": 15872 }, { "epoch": 0.96, "learning_rate": 4.788133035552605e-05, "loss": 0.9963, "step": 15874 }, { "epoch": 0.96, "learning_rate": 4.788736645137925e-05, "loss": 1.417, "step": 15876 }, { "epoch": 0.96, "learning_rate": 4.789340254723245e-05, "loss": 1.6627, "step": 15878 }, { "epoch": 0.96, "learning_rate": 4.789943864308566e-05, "loss": 1.346, "step": 15880 }, { "epoch": 0.96, "learning_rate": 4.790547473893886e-05, "loss": 1.0644, "step": 15882 }, { "epoch": 0.96, "learning_rate": 4.791151083479206e-05, "loss": 1.3031, "step": 15884 }, { "epoch": 0.96, "learning_rate": 4.7917546930645266e-05, "loss": 1.2753, "step": 15886 }, { "epoch": 0.96, "learning_rate": 4.792358302649847e-05, "loss": 1.1673, "step": 15888 }, { "epoch": 0.96, "learning_rate": 4.792961912235167e-05, "loss": 1.3026, "step": 15890 }, { "epoch": 0.96, "learning_rate": 4.793565521820487e-05, "loss": 1.6552, "step": 15892 }, { "epoch": 0.96, "learning_rate": 4.794169131405807e-05, "loss": 0.8734, "step": 15894 }, { "epoch": 0.96, "learning_rate": 4.794772740991127e-05, "loss": 1.4954, "step": 15896 }, { "epoch": 0.96, "learning_rate": 4.795376350576448e-05, "loss": 1.3206, "step": 15898 }, { "epoch": 0.96, "learning_rate": 4.795979960161768e-05, "loss": 1.8556, "step": 15900 }, { "epoch": 0.96, "learning_rate": 4.796583569747088e-05, "loss": 1.5292, "step": 15902 }, { "epoch": 0.96, "learning_rate": 4.797187179332408e-05, "loss": 1.7428, "step": 15904 }, { "epoch": 0.96, "learning_rate": 4.797790788917728e-05, "loss": 1.1014, "step": 15906 }, { "epoch": 0.96, "learning_rate": 4.798394398503048e-05, "loss": 1.0586, "step": 15908 }, { "epoch": 0.96, "learning_rate": 4.798998008088368e-05, "loss": 0.9558, "step": 15910 }, { "epoch": 0.96, "learning_rate": 4.799601617673689e-05, "loss": 1.3121, "step": 15912 }, { "epoch": 0.96, "learning_rate": 4.800205227259009e-05, "loss": 1.4439, "step": 15914 }, { "epoch": 0.96, "learning_rate": 4.800808836844329e-05, "loss": 1.1314, "step": 15916 }, { "epoch": 0.96, "learning_rate": 4.801412446429649e-05, "loss": 1.2899, "step": 15918 }, { "epoch": 0.96, "learning_rate": 4.8020160560149693e-05, "loss": 1.4334, "step": 15920 }, { "epoch": 0.96, "learning_rate": 4.8026196656002894e-05, "loss": 1.234, "step": 15922 }, { "epoch": 0.96, "learning_rate": 4.80322327518561e-05, "loss": 1.5646, "step": 15924 }, { "epoch": 0.96, "learning_rate": 4.80382688477093e-05, "loss": 1.4693, "step": 15926 }, { "epoch": 0.96, "learning_rate": 4.804430494356251e-05, "loss": 1.7819, "step": 15928 }, { "epoch": 0.96, "learning_rate": 4.805034103941571e-05, "loss": 1.1923, "step": 15930 }, { "epoch": 0.96, "learning_rate": 4.805637713526891e-05, "loss": 1.4381, "step": 15932 }, { "epoch": 0.96, "learning_rate": 4.806241323112211e-05, "loss": 1.4167, "step": 15934 }, { "epoch": 0.96, "learning_rate": 4.8068449326975314e-05, "loss": 1.15, "step": 15936 }, { "epoch": 0.96, "learning_rate": 4.807448542282852e-05, "loss": 1.4332, "step": 15938 }, { "epoch": 0.96, "learning_rate": 4.808052151868172e-05, "loss": 1.5034, "step": 15940 }, { "epoch": 0.96, "learning_rate": 4.808655761453492e-05, "loss": 1.1911, "step": 15942 }, { "epoch": 0.96, "learning_rate": 4.8092593710388124e-05, "loss": 1.2364, "step": 15944 }, { "epoch": 0.96, "learning_rate": 4.8098629806241324e-05, "loss": 1.3924, "step": 15946 }, { "epoch": 0.96, "learning_rate": 4.8104665902094525e-05, "loss": 1.4596, "step": 15948 }, { "epoch": 0.96, "learning_rate": 4.811070199794773e-05, "loss": 1.4786, "step": 15950 }, { "epoch": 0.96, "learning_rate": 4.8116738093800934e-05, "loss": 1.2141, "step": 15952 }, { "epoch": 0.96, "learning_rate": 4.8122774189654134e-05, "loss": 1.1272, "step": 15954 }, { "epoch": 0.96, "learning_rate": 4.8128810285507335e-05, "loss": 1.5695, "step": 15956 }, { "epoch": 0.96, "learning_rate": 4.8134846381360536e-05, "loss": 1.2141, "step": 15958 }, { "epoch": 0.96, "learning_rate": 4.814088247721374e-05, "loss": 1.2234, "step": 15960 }, { "epoch": 0.96, "learning_rate": 4.814691857306694e-05, "loss": 1.0086, "step": 15962 }, { "epoch": 0.96, "learning_rate": 4.8152954668920145e-05, "loss": 1.0613, "step": 15964 }, { "epoch": 0.96, "learning_rate": 4.8158990764773346e-05, "loss": 1.2981, "step": 15966 }, { "epoch": 0.96, "learning_rate": 4.816502686062655e-05, "loss": 1.3505, "step": 15968 }, { "epoch": 0.96, "learning_rate": 4.817106295647975e-05, "loss": 1.1759, "step": 15970 }, { "epoch": 0.96, "learning_rate": 4.817408100440635e-05, "loss": 1.5481, "step": 15972 }, { "epoch": 0.96, "learning_rate": 4.818011710025955e-05, "loss": 1.4011, "step": 15974 }, { "epoch": 0.96, "learning_rate": 4.818615319611275e-05, "loss": 1.4352, "step": 15976 }, { "epoch": 0.96, "learning_rate": 4.819218929196596e-05, "loss": 1.559, "step": 15978 }, { "epoch": 0.96, "learning_rate": 4.819822538781916e-05, "loss": 1.1148, "step": 15980 }, { "epoch": 0.96, "learning_rate": 4.820426148367236e-05, "loss": 1.1741, "step": 15982 }, { "epoch": 0.96, "learning_rate": 4.821029757952556e-05, "loss": 1.5579, "step": 15984 }, { "epoch": 0.96, "learning_rate": 4.8216333675378764e-05, "loss": 1.3624, "step": 15986 }, { "epoch": 0.96, "learning_rate": 4.8222369771231965e-05, "loss": 1.1168, "step": 15988 }, { "epoch": 0.97, "learning_rate": 4.8228405867085166e-05, "loss": 1.0098, "step": 15990 }, { "epoch": 0.97, "learning_rate": 4.823444196293837e-05, "loss": 1.4758, "step": 15992 }, { "epoch": 0.97, "learning_rate": 4.8240478058791574e-05, "loss": 1.383, "step": 15994 }, { "epoch": 0.97, "learning_rate": 4.824651415464478e-05, "loss": 1.6595, "step": 15996 }, { "epoch": 0.97, "learning_rate": 4.825255025049798e-05, "loss": 1.6573, "step": 15998 }, { "epoch": 0.97, "learning_rate": 4.825858634635118e-05, "loss": 1.2053, "step": 16000 }, { "epoch": 0.97, "learning_rate": 4.8264622442204384e-05, "loss": 1.7771, "step": 16002 }, { "epoch": 0.97, "learning_rate": 4.827065853805759e-05, "loss": 1.0502, "step": 16004 }, { "epoch": 0.97, "learning_rate": 4.827669463391079e-05, "loss": 1.3027, "step": 16006 }, { "epoch": 0.97, "learning_rate": 4.8282730729763993e-05, "loss": 1.2348, "step": 16008 }, { "epoch": 0.97, "learning_rate": 4.8288766825617194e-05, "loss": 1.4828, "step": 16010 }, { "epoch": 0.97, "learning_rate": 4.8294802921470395e-05, "loss": 1.3718, "step": 16012 }, { "epoch": 0.97, "learning_rate": 4.8300839017323596e-05, "loss": 1.206, "step": 16014 }, { "epoch": 0.97, "learning_rate": 4.83068751131768e-05, "loss": 1.292, "step": 16016 }, { "epoch": 0.97, "learning_rate": 4.8312911209030004e-05, "loss": 1.3138, "step": 16018 }, { "epoch": 0.97, "learning_rate": 4.8318947304883205e-05, "loss": 1.182, "step": 16020 }, { "epoch": 0.97, "learning_rate": 4.8324983400736406e-05, "loss": 1.2259, "step": 16022 }, { "epoch": 0.97, "learning_rate": 4.833101949658961e-05, "loss": 1.4842, "step": 16024 }, { "epoch": 0.97, "learning_rate": 4.833705559244281e-05, "loss": 1.1702, "step": 16026 }, { "epoch": 0.97, "learning_rate": 4.834309168829601e-05, "loss": 1.4178, "step": 16028 }, { "epoch": 0.97, "learning_rate": 4.834912778414921e-05, "loss": 1.4119, "step": 16030 }, { "epoch": 0.97, "learning_rate": 4.835516388000242e-05, "loss": 1.1794, "step": 16032 }, { "epoch": 0.97, "learning_rate": 4.836119997585562e-05, "loss": 1.5987, "step": 16034 }, { "epoch": 0.97, "learning_rate": 4.836723607170882e-05, "loss": 1.9002, "step": 16036 }, { "epoch": 0.97, "learning_rate": 4.837327216756202e-05, "loss": 1.1487, "step": 16038 }, { "epoch": 0.97, "learning_rate": 4.837930826341523e-05, "loss": 1.5289, "step": 16040 }, { "epoch": 0.97, "learning_rate": 4.838534435926843e-05, "loss": 1.1668, "step": 16042 }, { "epoch": 0.97, "learning_rate": 4.8391380455121635e-05, "loss": 1.6907, "step": 16044 }, { "epoch": 0.97, "learning_rate": 4.8397416550974836e-05, "loss": 1.3301, "step": 16046 }, { "epoch": 0.97, "learning_rate": 4.840345264682804e-05, "loss": 1.3794, "step": 16048 }, { "epoch": 0.97, "learning_rate": 4.840948874268124e-05, "loss": 1.6492, "step": 16050 }, { "epoch": 0.97, "learning_rate": 4.841552483853444e-05, "loss": 1.0051, "step": 16052 }, { "epoch": 0.97, "learning_rate": 4.842156093438764e-05, "loss": 1.4104, "step": 16054 }, { "epoch": 0.97, "learning_rate": 4.842759703024084e-05, "loss": 1.579, "step": 16056 }, { "epoch": 0.97, "learning_rate": 4.843363312609405e-05, "loss": 1.2064, "step": 16058 }, { "epoch": 0.97, "learning_rate": 4.843966922194725e-05, "loss": 0.9141, "step": 16060 }, { "epoch": 0.97, "learning_rate": 4.844570531780045e-05, "loss": 1.1785, "step": 16062 }, { "epoch": 0.97, "learning_rate": 4.845174141365365e-05, "loss": 1.0583, "step": 16064 }, { "epoch": 0.97, "learning_rate": 4.845777750950685e-05, "loss": 1.4428, "step": 16066 }, { "epoch": 0.97, "learning_rate": 4.846381360536005e-05, "loss": 1.0731, "step": 16068 }, { "epoch": 0.97, "learning_rate": 4.846984970121326e-05, "loss": 1.3199, "step": 16070 }, { "epoch": 0.97, "learning_rate": 4.847588579706646e-05, "loss": 1.3558, "step": 16072 }, { "epoch": 0.97, "learning_rate": 4.848192189291966e-05, "loss": 1.2434, "step": 16074 }, { "epoch": 0.97, "learning_rate": 4.848795798877286e-05, "loss": 1.0333, "step": 16076 }, { "epoch": 0.97, "learning_rate": 4.849399408462606e-05, "loss": 1.1183, "step": 16078 }, { "epoch": 0.97, "learning_rate": 4.8500030180479264e-05, "loss": 0.9513, "step": 16080 }, { "epoch": 0.97, "learning_rate": 4.850606627633247e-05, "loss": 1.1575, "step": 16082 }, { "epoch": 0.97, "learning_rate": 4.851210237218567e-05, "loss": 1.3332, "step": 16084 }, { "epoch": 0.97, "learning_rate": 4.851813846803888e-05, "loss": 1.0493, "step": 16086 }, { "epoch": 0.97, "learning_rate": 4.852417456389208e-05, "loss": 1.0216, "step": 16088 }, { "epoch": 0.97, "learning_rate": 4.853021065974528e-05, "loss": 1.4126, "step": 16090 }, { "epoch": 0.97, "learning_rate": 4.853624675559848e-05, "loss": 1.4645, "step": 16092 }, { "epoch": 0.97, "learning_rate": 4.854228285145168e-05, "loss": 1.7531, "step": 16094 }, { "epoch": 0.97, "learning_rate": 4.854831894730489e-05, "loss": 1.4929, "step": 16096 }, { "epoch": 0.97, "learning_rate": 4.855435504315809e-05, "loss": 1.0877, "step": 16098 }, { "epoch": 0.97, "learning_rate": 4.856039113901129e-05, "loss": 1.8104, "step": 16100 }, { "epoch": 0.97, "learning_rate": 4.856642723486449e-05, "loss": 1.4454, "step": 16102 }, { "epoch": 0.97, "learning_rate": 4.8572463330717694e-05, "loss": 0.8703, "step": 16104 }, { "epoch": 0.97, "learning_rate": 4.8578499426570895e-05, "loss": 1.25, "step": 16106 }, { "epoch": 0.97, "learning_rate": 4.8584535522424095e-05, "loss": 1.2376, "step": 16108 }, { "epoch": 0.97, "learning_rate": 4.85905716182773e-05, "loss": 1.2221, "step": 16110 }, { "epoch": 0.97, "learning_rate": 4.8596607714130504e-05, "loss": 1.3342, "step": 16112 }, { "epoch": 0.97, "learning_rate": 4.8602643809983705e-05, "loss": 1.2125, "step": 16114 }, { "epoch": 0.97, "learning_rate": 4.8608679905836905e-05, "loss": 1.252, "step": 16116 }, { "epoch": 0.97, "learning_rate": 4.8614716001690106e-05, "loss": 0.8255, "step": 16118 }, { "epoch": 0.97, "learning_rate": 4.862075209754331e-05, "loss": 1.1899, "step": 16120 }, { "epoch": 0.97, "learning_rate": 4.862678819339651e-05, "loss": 1.1436, "step": 16122 }, { "epoch": 0.97, "learning_rate": 4.8632824289249715e-05, "loss": 1.4996, "step": 16124 }, { "epoch": 0.97, "learning_rate": 4.8638860385102916e-05, "loss": 1.3495, "step": 16126 }, { "epoch": 0.97, "learning_rate": 4.864489648095612e-05, "loss": 1.2327, "step": 16128 }, { "epoch": 0.97, "learning_rate": 4.8650932576809325e-05, "loss": 0.9934, "step": 16130 }, { "epoch": 0.97, "learning_rate": 4.8656968672662525e-05, "loss": 1.1037, "step": 16132 }, { "epoch": 0.97, "learning_rate": 4.8663004768515726e-05, "loss": 0.8717, "step": 16134 }, { "epoch": 0.97, "learning_rate": 4.8669040864368934e-05, "loss": 1.0006, "step": 16136 }, { "epoch": 0.97, "learning_rate": 4.8675076960222135e-05, "loss": 0.9855, "step": 16138 }, { "epoch": 0.97, "learning_rate": 4.8681113056075336e-05, "loss": 1.6941, "step": 16140 }, { "epoch": 0.97, "learning_rate": 4.8687149151928536e-05, "loss": 1.4184, "step": 16142 }, { "epoch": 0.97, "learning_rate": 4.869318524778174e-05, "loss": 1.3581, "step": 16144 }, { "epoch": 0.97, "learning_rate": 4.869922134363494e-05, "loss": 1.2113, "step": 16146 }, { "epoch": 0.97, "learning_rate": 4.870525743948814e-05, "loss": 1.3222, "step": 16148 }, { "epoch": 0.97, "learning_rate": 4.8711293535341346e-05, "loss": 1.2217, "step": 16150 }, { "epoch": 0.97, "learning_rate": 4.871732963119455e-05, "loss": 1.1817, "step": 16152 }, { "epoch": 0.98, "learning_rate": 4.872336572704775e-05, "loss": 1.3289, "step": 16154 }, { "epoch": 0.98, "learning_rate": 4.872940182290095e-05, "loss": 1.0692, "step": 16156 }, { "epoch": 0.98, "learning_rate": 4.873543791875415e-05, "loss": 1.2865, "step": 16158 }, { "epoch": 0.98, "learning_rate": 4.874147401460735e-05, "loss": 1.1486, "step": 16160 }, { "epoch": 0.98, "learning_rate": 4.874751011046056e-05, "loss": 1.1647, "step": 16162 }, { "epoch": 0.98, "learning_rate": 4.875354620631376e-05, "loss": 1.1763, "step": 16164 }, { "epoch": 0.98, "learning_rate": 4.875958230216696e-05, "loss": 1.235, "step": 16166 }, { "epoch": 0.98, "learning_rate": 4.876561839802016e-05, "loss": 1.5131, "step": 16168 }, { "epoch": 0.98, "learning_rate": 4.877165449387336e-05, "loss": 1.1599, "step": 16170 }, { "epoch": 0.98, "learning_rate": 4.877769058972657e-05, "loss": 1.3529, "step": 16172 }, { "epoch": 0.98, "learning_rate": 4.878372668557977e-05, "loss": 1.074, "step": 16174 }, { "epoch": 0.98, "learning_rate": 4.878976278143297e-05, "loss": 1.4252, "step": 16176 }, { "epoch": 0.98, "learning_rate": 4.879579887728618e-05, "loss": 1.578, "step": 16178 }, { "epoch": 0.98, "learning_rate": 4.880183497313938e-05, "loss": 1.2245, "step": 16180 }, { "epoch": 0.98, "learning_rate": 4.880787106899258e-05, "loss": 0.9228, "step": 16182 }, { "epoch": 0.98, "learning_rate": 4.881390716484578e-05, "loss": 0.8232, "step": 16184 }, { "epoch": 0.98, "learning_rate": 4.881994326069898e-05, "loss": 1.0144, "step": 16186 }, { "epoch": 0.98, "learning_rate": 4.882597935655218e-05, "loss": 1.248, "step": 16188 }, { "epoch": 0.98, "learning_rate": 4.883201545240539e-05, "loss": 1.1443, "step": 16190 }, { "epoch": 0.98, "learning_rate": 4.883805154825859e-05, "loss": 1.3399, "step": 16192 }, { "epoch": 0.98, "learning_rate": 4.884408764411179e-05, "loss": 1.2771, "step": 16194 }, { "epoch": 0.98, "learning_rate": 4.885012373996499e-05, "loss": 1.6566, "step": 16196 }, { "epoch": 0.98, "learning_rate": 4.885615983581819e-05, "loss": 1.2414, "step": 16198 }, { "epoch": 0.98, "learning_rate": 4.8862195931671394e-05, "loss": 1.8976, "step": 16200 }, { "epoch": 0.98, "learning_rate": 4.88682320275246e-05, "loss": 1.3606, "step": 16202 }, { "epoch": 0.98, "learning_rate": 4.88742681233778e-05, "loss": 1.3483, "step": 16204 }, { "epoch": 0.98, "learning_rate": 4.8880304219231e-05, "loss": 1.4131, "step": 16206 }, { "epoch": 0.98, "learning_rate": 4.8886340315084204e-05, "loss": 1.4812, "step": 16208 }, { "epoch": 0.98, "learning_rate": 4.8892376410937405e-05, "loss": 1.2451, "step": 16210 }, { "epoch": 0.98, "learning_rate": 4.8898412506790606e-05, "loss": 1.5167, "step": 16212 }, { "epoch": 0.98, "learning_rate": 4.8904448602643806e-05, "loss": 1.4431, "step": 16214 }, { "epoch": 0.98, "learning_rate": 4.8910484698497014e-05, "loss": 1.8727, "step": 16216 }, { "epoch": 0.98, "learning_rate": 4.8916520794350215e-05, "loss": 1.8572, "step": 16218 }, { "epoch": 0.98, "learning_rate": 4.892255689020342e-05, "loss": 1.1871, "step": 16220 }, { "epoch": 0.98, "learning_rate": 4.892859298605662e-05, "loss": 1.1468, "step": 16222 }, { "epoch": 0.98, "learning_rate": 4.8934629081909824e-05, "loss": 1.4149, "step": 16224 }, { "epoch": 0.98, "learning_rate": 4.8940665177763025e-05, "loss": 1.4332, "step": 16226 }, { "epoch": 0.98, "learning_rate": 4.894670127361623e-05, "loss": 1.522, "step": 16228 }, { "epoch": 0.98, "learning_rate": 4.895273736946943e-05, "loss": 1.1377, "step": 16230 }, { "epoch": 0.98, "learning_rate": 4.8958773465322634e-05, "loss": 1.5641, "step": 16232 }, { "epoch": 0.98, "learning_rate": 4.8964809561175835e-05, "loss": 1.8444, "step": 16234 }, { "epoch": 0.98, "learning_rate": 4.8970845657029036e-05, "loss": 0.9997, "step": 16236 }, { "epoch": 0.98, "learning_rate": 4.897688175288224e-05, "loss": 0.9225, "step": 16238 }, { "epoch": 0.98, "learning_rate": 4.898291784873544e-05, "loss": 1.1854, "step": 16240 }, { "epoch": 0.98, "learning_rate": 4.8988953944588645e-05, "loss": 1.5652, "step": 16242 }, { "epoch": 0.98, "learning_rate": 4.8994990040441846e-05, "loss": 0.8471, "step": 16244 }, { "epoch": 0.98, "learning_rate": 4.900102613629505e-05, "loss": 1.1119, "step": 16246 }, { "epoch": 0.98, "learning_rate": 4.900706223214825e-05, "loss": 1.2157, "step": 16248 }, { "epoch": 0.98, "learning_rate": 4.901309832800145e-05, "loss": 1.3241, "step": 16250 }, { "epoch": 0.98, "learning_rate": 4.901913442385465e-05, "loss": 1.1547, "step": 16252 }, { "epoch": 0.98, "learning_rate": 4.902517051970785e-05, "loss": 1.4727, "step": 16254 }, { "epoch": 0.98, "learning_rate": 4.903120661556106e-05, "loss": 1.1637, "step": 16256 }, { "epoch": 0.98, "learning_rate": 4.903724271141426e-05, "loss": 1.4395, "step": 16258 }, { "epoch": 0.98, "learning_rate": 4.904327880726746e-05, "loss": 1.2949, "step": 16260 }, { "epoch": 0.98, "learning_rate": 4.904931490312066e-05, "loss": 1.2758, "step": 16262 }, { "epoch": 0.98, "learning_rate": 4.905535099897387e-05, "loss": 1.6069, "step": 16264 }, { "epoch": 0.98, "learning_rate": 4.906138709482707e-05, "loss": 1.2076, "step": 16266 }, { "epoch": 0.98, "learning_rate": 4.9067423190680276e-05, "loss": 1.6939, "step": 16268 }, { "epoch": 0.98, "learning_rate": 4.907345928653348e-05, "loss": 1.5879, "step": 16270 }, { "epoch": 0.98, "learning_rate": 4.907949538238668e-05, "loss": 1.5316, "step": 16272 }, { "epoch": 0.98, "learning_rate": 4.908553147823988e-05, "loss": 1.1917, "step": 16274 }, { "epoch": 0.98, "learning_rate": 4.909156757409308e-05, "loss": 1.3028, "step": 16276 }, { "epoch": 0.98, "learning_rate": 4.909760366994628e-05, "loss": 0.8277, "step": 16278 }, { "epoch": 0.98, "learning_rate": 4.910363976579948e-05, "loss": 1.6454, "step": 16280 }, { "epoch": 0.98, "learning_rate": 4.910967586165269e-05, "loss": 1.6714, "step": 16282 }, { "epoch": 0.98, "learning_rate": 4.911571195750589e-05, "loss": 1.273, "step": 16284 }, { "epoch": 0.98, "learning_rate": 4.912174805335909e-05, "loss": 1.1263, "step": 16286 }, { "epoch": 0.98, "learning_rate": 4.912778414921229e-05, "loss": 1.6091, "step": 16288 }, { "epoch": 0.98, "learning_rate": 4.913382024506549e-05, "loss": 0.9187, "step": 16290 }, { "epoch": 0.98, "learning_rate": 4.913985634091869e-05, "loss": 1.5503, "step": 16292 }, { "epoch": 0.98, "learning_rate": 4.91458924367719e-05, "loss": 1.1456, "step": 16294 }, { "epoch": 0.98, "learning_rate": 4.91519285326251e-05, "loss": 0.9421, "step": 16296 }, { "epoch": 0.98, "learning_rate": 4.91579646284783e-05, "loss": 1.3878, "step": 16298 }, { "epoch": 0.98, "learning_rate": 4.91640007243315e-05, "loss": 0.8352, "step": 16300 }, { "epoch": 0.98, "learning_rate": 4.9170036820184703e-05, "loss": 0.9925, "step": 16302 }, { "epoch": 0.98, "learning_rate": 4.9176072916037904e-05, "loss": 1.5367, "step": 16304 }, { "epoch": 0.98, "learning_rate": 4.918210901189111e-05, "loss": 1.3186, "step": 16306 }, { "epoch": 0.98, "learning_rate": 4.918814510774431e-05, "loss": 1.5969, "step": 16308 }, { "epoch": 0.98, "learning_rate": 4.9194181203597514e-05, "loss": 1.393, "step": 16310 }, { "epoch": 0.98, "learning_rate": 4.920021729945072e-05, "loss": 1.0964, "step": 16312 }, { "epoch": 0.98, "learning_rate": 4.920625339530392e-05, "loss": 1.414, "step": 16314 }, { "epoch": 0.98, "learning_rate": 4.921228949115712e-05, "loss": 1.2664, "step": 16316 }, { "epoch": 0.98, "learning_rate": 4.9218325587010324e-05, "loss": 1.2347, "step": 16318 }, { "epoch": 0.99, "learning_rate": 4.922436168286353e-05, "loss": 1.0984, "step": 16320 }, { "epoch": 0.99, "learning_rate": 4.923039777871673e-05, "loss": 0.9802, "step": 16322 }, { "epoch": 0.99, "learning_rate": 4.923643387456993e-05, "loss": 1.2193, "step": 16324 }, { "epoch": 0.99, "learning_rate": 4.9242469970423134e-05, "loss": 1.3318, "step": 16326 }, { "epoch": 0.99, "learning_rate": 4.9248506066276334e-05, "loss": 1.1988, "step": 16328 }, { "epoch": 0.99, "learning_rate": 4.9254542162129535e-05, "loss": 1.0383, "step": 16330 }, { "epoch": 0.99, "learning_rate": 4.9260578257982736e-05, "loss": 1.345, "step": 16332 }, { "epoch": 0.99, "learning_rate": 4.9266614353835944e-05, "loss": 1.1687, "step": 16334 }, { "epoch": 0.99, "learning_rate": 4.9272650449689145e-05, "loss": 1.329, "step": 16336 }, { "epoch": 0.99, "learning_rate": 4.9278686545542345e-05, "loss": 1.5706, "step": 16338 }, { "epoch": 0.99, "learning_rate": 4.9284722641395546e-05, "loss": 1.6316, "step": 16340 }, { "epoch": 0.99, "learning_rate": 4.929075873724875e-05, "loss": 1.2921, "step": 16342 }, { "epoch": 0.99, "learning_rate": 4.929679483310195e-05, "loss": 1.2067, "step": 16344 }, { "epoch": 0.99, "learning_rate": 4.930283092895515e-05, "loss": 1.3365, "step": 16346 }, { "epoch": 0.99, "learning_rate": 4.9308867024808356e-05, "loss": 1.2615, "step": 16348 }, { "epoch": 0.99, "learning_rate": 4.931490312066156e-05, "loss": 1.3229, "step": 16350 }, { "epoch": 0.99, "learning_rate": 4.932093921651476e-05, "loss": 0.9367, "step": 16352 }, { "epoch": 0.99, "learning_rate": 4.9326975312367965e-05, "loss": 1.364, "step": 16354 }, { "epoch": 0.99, "learning_rate": 4.9333011408221166e-05, "loss": 1.6867, "step": 16356 }, { "epoch": 0.99, "learning_rate": 4.933904750407437e-05, "loss": 1.3618, "step": 16358 }, { "epoch": 0.99, "learning_rate": 4.9345083599927575e-05, "loss": 1.5031, "step": 16360 }, { "epoch": 0.99, "learning_rate": 4.9351119695780775e-05, "loss": 1.1591, "step": 16362 }, { "epoch": 0.99, "learning_rate": 4.9357155791633976e-05, "loss": 1.0697, "step": 16364 }, { "epoch": 0.99, "learning_rate": 4.936319188748718e-05, "loss": 1.3438, "step": 16366 }, { "epoch": 0.99, "learning_rate": 4.936922798334038e-05, "loss": 1.2465, "step": 16368 }, { "epoch": 0.99, "learning_rate": 4.937526407919358e-05, "loss": 1.5608, "step": 16370 }, { "epoch": 0.99, "learning_rate": 4.938130017504678e-05, "loss": 1.4197, "step": 16372 }, { "epoch": 0.99, "learning_rate": 4.938733627089999e-05, "loss": 1.4824, "step": 16374 }, { "epoch": 0.99, "learning_rate": 4.939337236675319e-05, "loss": 1.4481, "step": 16376 }, { "epoch": 0.99, "learning_rate": 4.939940846260639e-05, "loss": 1.1396, "step": 16378 }, { "epoch": 0.99, "learning_rate": 4.940544455845959e-05, "loss": 1.1476, "step": 16380 }, { "epoch": 0.99, "learning_rate": 4.941148065431279e-05, "loss": 1.2566, "step": 16382 }, { "epoch": 0.99, "learning_rate": 4.941751675016599e-05, "loss": 1.0335, "step": 16384 }, { "epoch": 0.99, "learning_rate": 4.94235528460192e-05, "loss": 1.2755, "step": 16386 }, { "epoch": 0.99, "learning_rate": 4.94295889418724e-05, "loss": 1.124, "step": 16388 }, { "epoch": 0.99, "learning_rate": 4.94356250377256e-05, "loss": 1.4465, "step": 16390 }, { "epoch": 0.99, "learning_rate": 4.94416611335788e-05, "loss": 1.5169, "step": 16392 }, { "epoch": 0.99, "learning_rate": 4.9447697229432e-05, "loss": 1.062, "step": 16394 }, { "epoch": 0.99, "learning_rate": 4.945373332528521e-05, "loss": 1.1948, "step": 16396 }, { "epoch": 0.99, "learning_rate": 4.945976942113841e-05, "loss": 1.4669, "step": 16398 }, { "epoch": 0.99, "learning_rate": 4.946580551699161e-05, "loss": 1.35, "step": 16400 }, { "epoch": 0.99, "learning_rate": 4.947184161284482e-05, "loss": 1.2775, "step": 16402 }, { "epoch": 0.99, "learning_rate": 4.947787770869802e-05, "loss": 1.5622, "step": 16404 }, { "epoch": 0.99, "learning_rate": 4.948391380455122e-05, "loss": 1.5182, "step": 16406 }, { "epoch": 0.99, "learning_rate": 4.948994990040442e-05, "loss": 1.5776, "step": 16408 }, { "epoch": 0.99, "learning_rate": 4.949598599625762e-05, "loss": 1.1741, "step": 16410 }, { "epoch": 0.99, "learning_rate": 4.950202209211082e-05, "loss": 1.1568, "step": 16412 }, { "epoch": 0.99, "learning_rate": 4.950805818796403e-05, "loss": 1.1329, "step": 16414 }, { "epoch": 0.99, "learning_rate": 4.951409428381723e-05, "loss": 1.2963, "step": 16416 }, { "epoch": 0.99, "learning_rate": 4.952013037967043e-05, "loss": 1.221, "step": 16418 }, { "epoch": 0.99, "learning_rate": 4.952616647552363e-05, "loss": 1.1975, "step": 16420 }, { "epoch": 0.99, "learning_rate": 4.9532202571376834e-05, "loss": 1.2286, "step": 16422 }, { "epoch": 0.99, "learning_rate": 4.9538238667230035e-05, "loss": 1.3984, "step": 16424 }, { "epoch": 0.99, "learning_rate": 4.954427476308324e-05, "loss": 1.0204, "step": 16426 }, { "epoch": 0.99, "learning_rate": 4.955031085893644e-05, "loss": 1.4422, "step": 16428 }, { "epoch": 0.99, "learning_rate": 4.9556346954789644e-05, "loss": 1.3534, "step": 16430 }, { "epoch": 0.99, "learning_rate": 4.9562383050642845e-05, "loss": 0.9906, "step": 16432 }, { "epoch": 0.99, "learning_rate": 4.9568419146496046e-05, "loss": 0.9412, "step": 16434 }, { "epoch": 0.99, "learning_rate": 4.9574455242349246e-05, "loss": 1.3987, "step": 16436 }, { "epoch": 0.99, "learning_rate": 4.958049133820245e-05, "loss": 1.3909, "step": 16438 }, { "epoch": 0.99, "learning_rate": 4.9586527434055655e-05, "loss": 1.2509, "step": 16440 }, { "epoch": 0.99, "learning_rate": 4.9592563529908856e-05, "loss": 1.3986, "step": 16442 }, { "epoch": 0.99, "learning_rate": 4.959859962576206e-05, "loss": 1.4895, "step": 16444 }, { "epoch": 0.99, "learning_rate": 4.9604635721615264e-05, "loss": 1.1964, "step": 16446 }, { "epoch": 0.99, "learning_rate": 4.9610671817468465e-05, "loss": 1.1982, "step": 16448 }, { "epoch": 0.99, "learning_rate": 4.9616707913321666e-05, "loss": 1.0323, "step": 16450 }, { "epoch": 0.99, "learning_rate": 4.962274400917487e-05, "loss": 0.99, "step": 16452 }, { "epoch": 0.99, "learning_rate": 4.9628780105028074e-05, "loss": 1.11, "step": 16454 }, { "epoch": 0.99, "learning_rate": 4.9634816200881275e-05, "loss": 1.1985, "step": 16456 }, { "epoch": 0.99, "learning_rate": 4.9640852296734476e-05, "loss": 1.3911, "step": 16458 }, { "epoch": 0.99, "learning_rate": 4.9646888392587677e-05, "loss": 1.2112, "step": 16460 }, { "epoch": 0.99, "learning_rate": 4.965292448844088e-05, "loss": 1.3959, "step": 16462 }, { "epoch": 0.99, "learning_rate": 4.965896058429408e-05, "loss": 1.2079, "step": 16464 }, { "epoch": 0.99, "learning_rate": 4.9664996680147286e-05, "loss": 1.212, "step": 16466 }, { "epoch": 0.99, "learning_rate": 4.967103277600049e-05, "loss": 1.528, "step": 16468 }, { "epoch": 0.99, "learning_rate": 4.967706887185369e-05, "loss": 1.6106, "step": 16470 }, { "epoch": 0.99, "learning_rate": 4.968310496770689e-05, "loss": 1.2121, "step": 16472 }, { "epoch": 0.99, "learning_rate": 4.968914106356009e-05, "loss": 1.078, "step": 16474 }, { "epoch": 0.99, "learning_rate": 4.969517715941329e-05, "loss": 1.2896, "step": 16476 }, { "epoch": 0.99, "learning_rate": 4.970121325526649e-05, "loss": 1.0488, "step": 16478 }, { "epoch": 0.99, "learning_rate": 4.97072493511197e-05, "loss": 1.1942, "step": 16480 }, { "epoch": 0.99, "learning_rate": 4.97132854469729e-05, "loss": 1.1136, "step": 16482 }, { "epoch": 0.99, "learning_rate": 4.97193215428261e-05, "loss": 1.0635, "step": 16484 }, { "epoch": 1.0, "learning_rate": 4.97253576386793e-05, "loss": 1.2554, "step": 16486 }, { "epoch": 1.0, "learning_rate": 4.973139373453251e-05, "loss": 1.0789, "step": 16488 }, { "epoch": 1.0, "learning_rate": 4.973742983038571e-05, "loss": 1.2043, "step": 16490 }, { "epoch": 1.0, "learning_rate": 4.974346592623892e-05, "loss": 1.7456, "step": 16492 }, { "epoch": 1.0, "learning_rate": 4.974950202209212e-05, "loss": 1.3377, "step": 16494 }, { "epoch": 1.0, "learning_rate": 4.975553811794532e-05, "loss": 1.1365, "step": 16496 }, { "epoch": 1.0, "learning_rate": 4.976157421379852e-05, "loss": 1.0463, "step": 16498 }, { "epoch": 1.0, "learning_rate": 4.976761030965172e-05, "loss": 1.4268, "step": 16500 }, { "epoch": 1.0, "learning_rate": 4.977364640550492e-05, "loss": 1.1794, "step": 16502 }, { "epoch": 1.0, "learning_rate": 4.977968250135812e-05, "loss": 1.3917, "step": 16504 }, { "epoch": 1.0, "learning_rate": 4.978571859721133e-05, "loss": 1.1619, "step": 16506 }, { "epoch": 1.0, "learning_rate": 4.979175469306453e-05, "loss": 1.0923, "step": 16508 }, { "epoch": 1.0, "learning_rate": 4.979779078891773e-05, "loss": 1.1467, "step": 16510 }, { "epoch": 1.0, "learning_rate": 4.980382688477093e-05, "loss": 0.8944, "step": 16512 }, { "epoch": 1.0, "learning_rate": 4.980986298062413e-05, "loss": 0.9574, "step": 16514 }, { "epoch": 1.0, "learning_rate": 4.981589907647733e-05, "loss": 1.0752, "step": 16516 }, { "epoch": 1.0, "learning_rate": 4.982193517233054e-05, "loss": 1.3599, "step": 16518 }, { "epoch": 1.0, "learning_rate": 4.982797126818374e-05, "loss": 1.2962, "step": 16520 }, { "epoch": 1.0, "learning_rate": 4.983400736403694e-05, "loss": 1.3076, "step": 16522 }, { "epoch": 1.0, "learning_rate": 4.9840043459890143e-05, "loss": 1.2433, "step": 16524 }, { "epoch": 1.0, "learning_rate": 4.9846079555743344e-05, "loss": 1.2022, "step": 16526 }, { "epoch": 1.0, "learning_rate": 4.9852115651596545e-05, "loss": 1.2118, "step": 16528 }, { "epoch": 1.0, "learning_rate": 4.985815174744975e-05, "loss": 1.7731, "step": 16530 }, { "epoch": 1.0, "learning_rate": 4.9864187843302953e-05, "loss": 1.1001, "step": 16532 }, { "epoch": 1.0, "learning_rate": 4.9870223939156154e-05, "loss": 1.3323, "step": 16534 }, { "epoch": 1.0, "learning_rate": 4.987626003500936e-05, "loss": 1.4287, "step": 16536 }, { "epoch": 1.0, "learning_rate": 4.988229613086256e-05, "loss": 1.0658, "step": 16538 }, { "epoch": 1.0, "learning_rate": 4.9888332226715764e-05, "loss": 1.018, "step": 16540 }, { "epoch": 1.0, "learning_rate": 4.9894368322568964e-05, "loss": 1.3664, "step": 16542 }, { "epoch": 1.0, "learning_rate": 4.9900404418422165e-05, "loss": 1.0516, "step": 16544 }, { "epoch": 1.0, "learning_rate": 4.990644051427537e-05, "loss": 1.5642, "step": 16546 }, { "epoch": 1.0, "learning_rate": 4.9912476610128574e-05, "loss": 1.2376, "step": 16548 }, { "epoch": 1.0, "learning_rate": 4.9918512705981774e-05, "loss": 1.1279, "step": 16550 }, { "epoch": 1.0, "learning_rate": 4.9924548801834975e-05, "loss": 1.5116, "step": 16552 }, { "epoch": 1.0, "learning_rate": 4.9930584897688176e-05, "loss": 1.2091, "step": 16554 }, { "epoch": 1.0, "learning_rate": 4.993662099354138e-05, "loss": 1.6877, "step": 16556 }, { "epoch": 1.0, "learning_rate": 4.9942657089394584e-05, "loss": 1.1518, "step": 16558 }, { "epoch": 1.0, "learning_rate": 4.9948693185247785e-05, "loss": 1.0131, "step": 16560 }, { "epoch": 1.0, "learning_rate": 4.9954729281100986e-05, "loss": 1.1463, "step": 16562 }, { "epoch": 1.0, "learning_rate": 4.996076537695419e-05, "loss": 1.0967, "step": 16564 }, { "epoch": 1.0, "learning_rate": 4.996680147280739e-05, "loss": 1.1777, "step": 16566 }, { "epoch": 1.0, "eval_cer": 0.185656877547082, "eval_loss": 1.197560429573059, "eval_runtime": 9462.7962, "eval_samples_per_second": 1.751, "eval_steps_per_second": 0.219, "step": 16567 }, { "epoch": 1.0, "learning_rate": 4.997283756866059e-05, "loss": 1.2985, "step": 16568 }, { "epoch": 1.0, "learning_rate": 4.997887366451379e-05, "loss": 1.1668, "step": 16570 }, { "epoch": 1.0, "learning_rate": 4.9984909760367e-05, "loss": 0.9674, "step": 16572 }, { "epoch": 1.0, "learning_rate": 4.99909458562202e-05, "loss": 0.8822, "step": 16574 }, { "epoch": 1.0, "learning_rate": 4.99969819520734e-05, "loss": 1.1729, "step": 16576 }, { "epoch": 1.0, "learning_rate": 4.999924554494289e-05, "loss": 1.1556, "step": 16578 }, { "epoch": 1.0, "learning_rate": 4.9997736634828665e-05, "loss": 1.4289, "step": 16580 }, { "epoch": 1.0, "learning_rate": 4.999622772471444e-05, "loss": 1.3299, "step": 16582 }, { "epoch": 1.0, "learning_rate": 4.999471881460022e-05, "loss": 0.9887, "step": 16584 }, { "epoch": 1.0, "learning_rate": 4.999320990448599e-05, "loss": 1.0923, "step": 16586 }, { "epoch": 1.0, "learning_rate": 4.9991700994371765e-05, "loss": 1.2962, "step": 16588 }, { "epoch": 1.0, "learning_rate": 4.9990192084257545e-05, "loss": 1.0818, "step": 16590 }, { "epoch": 1.0, "learning_rate": 4.998868317414332e-05, "loss": 1.3269, "step": 16592 }, { "epoch": 1.0, "learning_rate": 4.99871742640291e-05, "loss": 0.9592, "step": 16594 }, { "epoch": 1.0, "learning_rate": 4.998566535391487e-05, "loss": 1.3013, "step": 16596 }, { "epoch": 1.0, "learning_rate": 4.9984156443800644e-05, "loss": 1.2538, "step": 16598 }, { "epoch": 1.0, "learning_rate": 4.9982647533686424e-05, "loss": 1.0232, "step": 16600 }, { "epoch": 1.0, "learning_rate": 4.998113862357219e-05, "loss": 1.4562, "step": 16602 }, { "epoch": 1.0, "learning_rate": 4.997962971345797e-05, "loss": 1.264, "step": 16604 }, { "epoch": 1.0, "learning_rate": 4.997812080334375e-05, "loss": 0.887, "step": 16606 }, { "epoch": 1.0, "learning_rate": 4.997661189322952e-05, "loss": 1.1165, "step": 16608 }, { "epoch": 1.0, "learning_rate": 4.9975102983115296e-05, "loss": 0.8504, "step": 16610 }, { "epoch": 1.0, "learning_rate": 4.9973594073001076e-05, "loss": 0.8747, "step": 16612 }, { "epoch": 1.0, "learning_rate": 4.997208516288685e-05, "loss": 1.6842, "step": 16614 }, { "epoch": 1.0, "learning_rate": 4.997057625277263e-05, "loss": 0.9771, "step": 16616 }, { "epoch": 1.0, "learning_rate": 4.9969067342658396e-05, "loss": 1.4765, "step": 16618 }, { "epoch": 1.0, "learning_rate": 4.9967558432544175e-05, "loss": 1.0959, "step": 16620 }, { "epoch": 1.0, "learning_rate": 4.9966049522429955e-05, "loss": 1.3651, "step": 16622 }, { "epoch": 1.0, "learning_rate": 4.996454061231572e-05, "loss": 1.1024, "step": 16624 }, { "epoch": 1.0, "learning_rate": 4.99630317022015e-05, "loss": 1.2024, "step": 16626 }, { "epoch": 1.0, "learning_rate": 4.996152279208728e-05, "loss": 1.1086, "step": 16628 }, { "epoch": 1.0, "learning_rate": 4.9960013881973055e-05, "loss": 1.3107, "step": 16630 }, { "epoch": 1.0, "learning_rate": 4.995850497185883e-05, "loss": 1.2999, "step": 16632 }, { "epoch": 1.0, "learning_rate": 4.99569960617446e-05, "loss": 1.5624, "step": 16634 }, { "epoch": 1.0, "learning_rate": 4.995548715163038e-05, "loss": 0.8672, "step": 16636 }, { "epoch": 1.0, "learning_rate": 4.9953978241516154e-05, "loss": 1.2176, "step": 16638 }, { "epoch": 1.0, "learning_rate": 4.995246933140193e-05, "loss": 1.2737, "step": 16640 }, { "epoch": 1.0, "learning_rate": 4.995096042128771e-05, "loss": 1.3446, "step": 16642 }, { "epoch": 1.0, "learning_rate": 4.994945151117349e-05, "loss": 1.0192, "step": 16644 }, { "epoch": 1.0, "learning_rate": 4.994794260105925e-05, "loss": 1.0468, "step": 16646 }, { "epoch": 1.0, "learning_rate": 4.994643369094503e-05, "loss": 1.0934, "step": 16648 }, { "epoch": 1.0, "learning_rate": 4.994492478083081e-05, "loss": 1.2286, "step": 16650 }, { "epoch": 1.01, "learning_rate": 4.994341587071658e-05, "loss": 1.2858, "step": 16652 }, { "epoch": 1.01, "learning_rate": 4.994190696060236e-05, "loss": 1.3203, "step": 16654 }, { "epoch": 1.01, "learning_rate": 4.994039805048813e-05, "loss": 1.0123, "step": 16656 }, { "epoch": 1.01, "learning_rate": 4.993888914037391e-05, "loss": 1.341, "step": 16658 }, { "epoch": 1.01, "learning_rate": 4.9937380230259685e-05, "loss": 1.2808, "step": 16660 }, { "epoch": 1.01, "learning_rate": 4.993587132014546e-05, "loss": 1.1716, "step": 16662 }, { "epoch": 1.01, "learning_rate": 4.993436241003124e-05, "loss": 0.992, "step": 16664 }, { "epoch": 1.01, "learning_rate": 4.993285349991702e-05, "loss": 1.8329, "step": 16666 }, { "epoch": 1.01, "learning_rate": 4.9931344589802785e-05, "loss": 1.339, "step": 16668 }, { "epoch": 1.01, "learning_rate": 4.9929835679688565e-05, "loss": 1.3415, "step": 16670 }, { "epoch": 1.01, "learning_rate": 4.992832676957434e-05, "loss": 1.2622, "step": 16672 }, { "epoch": 1.01, "learning_rate": 4.992681785946011e-05, "loss": 1.0766, "step": 16674 }, { "epoch": 1.01, "learning_rate": 4.992530894934589e-05, "loss": 1.2823, "step": 16676 }, { "epoch": 1.01, "learning_rate": 4.9923800039231664e-05, "loss": 1.3473, "step": 16678 }, { "epoch": 1.01, "learning_rate": 4.9922291129117444e-05, "loss": 1.1455, "step": 16680 }, { "epoch": 1.01, "learning_rate": 4.992078221900322e-05, "loss": 1.0429, "step": 16682 }, { "epoch": 1.01, "learning_rate": 4.991927330888899e-05, "loss": 1.1406, "step": 16684 }, { "epoch": 1.01, "learning_rate": 4.991776439877477e-05, "loss": 0.9398, "step": 16686 }, { "epoch": 1.01, "learning_rate": 4.991625548866054e-05, "loss": 1.551, "step": 16688 }, { "epoch": 1.01, "learning_rate": 4.9914746578546316e-05, "loss": 1.0986, "step": 16690 }, { "epoch": 1.01, "learning_rate": 4.9913237668432096e-05, "loss": 1.6212, "step": 16692 }, { "epoch": 1.01, "learning_rate": 4.991172875831787e-05, "loss": 1.5387, "step": 16694 }, { "epoch": 1.01, "learning_rate": 4.991021984820364e-05, "loss": 1.5181, "step": 16696 }, { "epoch": 1.01, "learning_rate": 4.990871093808942e-05, "loss": 1.5057, "step": 16698 }, { "epoch": 1.01, "learning_rate": 4.9907202027975195e-05, "loss": 1.2122, "step": 16700 }, { "epoch": 1.01, "learning_rate": 4.990569311786097e-05, "loss": 1.322, "step": 16702 }, { "epoch": 1.01, "learning_rate": 4.990418420774675e-05, "loss": 1.0952, "step": 16704 }, { "epoch": 1.01, "learning_rate": 4.990267529763252e-05, "loss": 1.0754, "step": 16706 }, { "epoch": 1.01, "learning_rate": 4.99011663875183e-05, "loss": 1.3393, "step": 16708 }, { "epoch": 1.01, "learning_rate": 4.9899657477404074e-05, "loss": 1.6122, "step": 16710 }, { "epoch": 1.01, "learning_rate": 4.989814856728985e-05, "loss": 1.2917, "step": 16712 }, { "epoch": 1.01, "learning_rate": 4.989663965717563e-05, "loss": 1.4209, "step": 16714 }, { "epoch": 1.01, "learning_rate": 4.98951307470614e-05, "loss": 1.3574, "step": 16716 }, { "epoch": 1.01, "learning_rate": 4.9893621836947174e-05, "loss": 1.3488, "step": 16718 }, { "epoch": 1.01, "learning_rate": 4.9892112926832954e-05, "loss": 1.6003, "step": 16720 }, { "epoch": 1.01, "learning_rate": 4.989060401671873e-05, "loss": 1.1707, "step": 16722 }, { "epoch": 1.01, "learning_rate": 4.98890951066045e-05, "loss": 1.2296, "step": 16724 }, { "epoch": 1.01, "learning_rate": 4.988758619649028e-05, "loss": 1.385, "step": 16726 }, { "epoch": 1.01, "learning_rate": 4.988607728637605e-05, "loss": 1.7183, "step": 16728 }, { "epoch": 1.01, "learning_rate": 4.988456837626183e-05, "loss": 1.5289, "step": 16730 }, { "epoch": 1.01, "learning_rate": 4.98830594661476e-05, "loss": 0.9807, "step": 16732 }, { "epoch": 1.01, "learning_rate": 4.988155055603338e-05, "loss": 1.2583, "step": 16734 }, { "epoch": 1.01, "learning_rate": 4.988004164591916e-05, "loss": 1.1156, "step": 16736 }, { "epoch": 1.01, "learning_rate": 4.9878532735804925e-05, "loss": 1.1273, "step": 16738 }, { "epoch": 1.01, "learning_rate": 4.9877023825690705e-05, "loss": 0.9928, "step": 16740 }, { "epoch": 1.01, "learning_rate": 4.9875514915576485e-05, "loss": 1.4141, "step": 16742 }, { "epoch": 1.01, "learning_rate": 4.987400600546226e-05, "loss": 1.0172, "step": 16744 }, { "epoch": 1.01, "learning_rate": 4.987249709534803e-05, "loss": 1.5524, "step": 16746 }, { "epoch": 1.01, "learning_rate": 4.9870988185233805e-05, "loss": 1.104, "step": 16748 }, { "epoch": 1.01, "learning_rate": 4.9869479275119584e-05, "loss": 1.1061, "step": 16750 }, { "epoch": 1.01, "learning_rate": 4.986797036500536e-05, "loss": 1.4766, "step": 16752 }, { "epoch": 1.01, "learning_rate": 4.986646145489113e-05, "loss": 1.1088, "step": 16754 }, { "epoch": 1.01, "learning_rate": 4.986495254477691e-05, "loss": 1.2025, "step": 16756 }, { "epoch": 1.01, "learning_rate": 4.986344363466269e-05, "loss": 1.3708, "step": 16758 }, { "epoch": 1.01, "learning_rate": 4.986193472454846e-05, "loss": 0.8898, "step": 16760 }, { "epoch": 1.01, "learning_rate": 4.986042581443424e-05, "loss": 1.0264, "step": 16762 }, { "epoch": 1.01, "learning_rate": 4.985891690432002e-05, "loss": 1.1574, "step": 16764 }, { "epoch": 1.01, "learning_rate": 4.985740799420578e-05, "loss": 1.1169, "step": 16766 }, { "epoch": 1.01, "learning_rate": 4.985589908409156e-05, "loss": 1.1066, "step": 16768 }, { "epoch": 1.01, "learning_rate": 4.9854390173977336e-05, "loss": 1.2258, "step": 16770 }, { "epoch": 1.01, "learning_rate": 4.9852881263863116e-05, "loss": 1.2223, "step": 16772 }, { "epoch": 1.01, "learning_rate": 4.985137235374889e-05, "loss": 0.9089, "step": 16774 }, { "epoch": 1.01, "learning_rate": 4.984986344363466e-05, "loss": 1.1595, "step": 16776 }, { "epoch": 1.01, "learning_rate": 4.984835453352044e-05, "loss": 1.3478, "step": 16778 }, { "epoch": 1.01, "learning_rate": 4.984684562340622e-05, "loss": 1.0614, "step": 16780 }, { "epoch": 1.01, "learning_rate": 4.984533671329199e-05, "loss": 1.3272, "step": 16782 }, { "epoch": 1.01, "learning_rate": 4.984382780317777e-05, "loss": 0.9912, "step": 16784 }, { "epoch": 1.01, "learning_rate": 4.984231889306354e-05, "loss": 1.0689, "step": 16786 }, { "epoch": 1.01, "learning_rate": 4.9840809982949314e-05, "loss": 1.3215, "step": 16788 }, { "epoch": 1.01, "learning_rate": 4.9839301072835094e-05, "loss": 1.1857, "step": 16790 }, { "epoch": 1.01, "learning_rate": 4.983779216272087e-05, "loss": 0.8754, "step": 16792 }, { "epoch": 1.01, "learning_rate": 4.983628325260665e-05, "loss": 0.9522, "step": 16794 }, { "epoch": 1.01, "learning_rate": 4.983477434249242e-05, "loss": 1.0576, "step": 16796 }, { "epoch": 1.01, "learning_rate": 4.9833265432378194e-05, "loss": 0.8009, "step": 16798 }, { "epoch": 1.01, "learning_rate": 4.9831756522263974e-05, "loss": 1.4649, "step": 16800 }, { "epoch": 1.01, "learning_rate": 4.983024761214975e-05, "loss": 1.0231, "step": 16802 }, { "epoch": 1.01, "learning_rate": 4.982873870203552e-05, "loss": 1.0991, "step": 16804 }, { "epoch": 1.01, "learning_rate": 4.98272297919213e-05, "loss": 0.9413, "step": 16806 }, { "epoch": 1.01, "learning_rate": 4.982572088180707e-05, "loss": 1.1486, "step": 16808 }, { "epoch": 1.01, "learning_rate": 4.9824211971692846e-05, "loss": 1.4191, "step": 16810 }, { "epoch": 1.01, "learning_rate": 4.9822703061578626e-05, "loss": 1.1218, "step": 16812 }, { "epoch": 1.01, "learning_rate": 4.98211941514644e-05, "loss": 1.0683, "step": 16814 }, { "epoch": 1.01, "learning_rate": 4.981968524135017e-05, "loss": 0.944, "step": 16816 }, { "epoch": 1.02, "learning_rate": 4.981817633123595e-05, "loss": 1.4263, "step": 16818 }, { "epoch": 1.02, "learning_rate": 4.9816667421121725e-05, "loss": 1.0614, "step": 16820 }, { "epoch": 1.02, "learning_rate": 4.9815158511007505e-05, "loss": 1.2175, "step": 16822 }, { "epoch": 1.02, "learning_rate": 4.981364960089328e-05, "loss": 1.1649, "step": 16824 }, { "epoch": 1.02, "learning_rate": 4.981214069077905e-05, "loss": 1.1209, "step": 16826 }, { "epoch": 1.02, "learning_rate": 4.981063178066483e-05, "loss": 1.1896, "step": 16828 }, { "epoch": 1.02, "learning_rate": 4.9809122870550604e-05, "loss": 1.2261, "step": 16830 }, { "epoch": 1.02, "learning_rate": 4.980761396043638e-05, "loss": 1.2564, "step": 16832 }, { "epoch": 1.02, "learning_rate": 4.980610505032216e-05, "loss": 1.2929, "step": 16834 }, { "epoch": 1.02, "learning_rate": 4.980459614020793e-05, "loss": 1.1945, "step": 16836 }, { "epoch": 1.02, "learning_rate": 4.9803087230093704e-05, "loss": 1.2387, "step": 16838 }, { "epoch": 1.02, "learning_rate": 4.9801578319979483e-05, "loss": 1.6714, "step": 16840 }, { "epoch": 1.02, "learning_rate": 4.980006940986526e-05, "loss": 1.3993, "step": 16842 }, { "epoch": 1.02, "learning_rate": 4.9798560499751037e-05, "loss": 1.0462, "step": 16844 }, { "epoch": 1.02, "learning_rate": 4.97970515896368e-05, "loss": 1.0917, "step": 16846 }, { "epoch": 1.02, "learning_rate": 4.979554267952258e-05, "loss": 0.7029, "step": 16848 }, { "epoch": 1.02, "learning_rate": 4.979403376940836e-05, "loss": 1.2014, "step": 16850 }, { "epoch": 1.02, "learning_rate": 4.979252485929413e-05, "loss": 0.8478, "step": 16852 }, { "epoch": 1.02, "learning_rate": 4.979101594917991e-05, "loss": 1.0943, "step": 16854 }, { "epoch": 1.02, "learning_rate": 4.978950703906569e-05, "loss": 1.3746, "step": 16856 }, { "epoch": 1.02, "learning_rate": 4.978799812895146e-05, "loss": 1.4389, "step": 16858 }, { "epoch": 1.02, "learning_rate": 4.9786489218837235e-05, "loss": 1.1937, "step": 16860 }, { "epoch": 1.02, "learning_rate": 4.978498030872301e-05, "loss": 1.0116, "step": 16862 }, { "epoch": 1.02, "learning_rate": 4.978347139860879e-05, "loss": 1.0319, "step": 16864 }, { "epoch": 1.02, "learning_rate": 4.978196248849456e-05, "loss": 0.9057, "step": 16866 }, { "epoch": 1.02, "learning_rate": 4.9780453578380334e-05, "loss": 1.1474, "step": 16868 }, { "epoch": 1.02, "learning_rate": 4.9778944668266114e-05, "loss": 0.9803, "step": 16870 }, { "epoch": 1.02, "learning_rate": 4.9777435758151894e-05, "loss": 1.1478, "step": 16872 }, { "epoch": 1.02, "learning_rate": 4.977592684803766e-05, "loss": 1.1124, "step": 16874 }, { "epoch": 1.02, "learning_rate": 4.977441793792344e-05, "loss": 1.0344, "step": 16876 }, { "epoch": 1.02, "learning_rate": 4.977290902780922e-05, "loss": 1.3799, "step": 16878 }, { "epoch": 1.02, "learning_rate": 4.9771400117694993e-05, "loss": 1.1871, "step": 16880 }, { "epoch": 1.02, "learning_rate": 4.9769891207580767e-05, "loss": 1.2555, "step": 16882 }, { "epoch": 1.02, "learning_rate": 4.976838229746654e-05, "loss": 1.2562, "step": 16884 }, { "epoch": 1.02, "learning_rate": 4.976687338735232e-05, "loss": 1.605, "step": 16886 }, { "epoch": 1.02, "learning_rate": 4.976536447723809e-05, "loss": 1.1694, "step": 16888 }, { "epoch": 1.02, "learning_rate": 4.9763855567123866e-05, "loss": 1.1425, "step": 16890 }, { "epoch": 1.02, "learning_rate": 4.9762346657009646e-05, "loss": 1.1137, "step": 16892 }, { "epoch": 1.02, "learning_rate": 4.9760837746895426e-05, "loss": 1.4219, "step": 16894 }, { "epoch": 1.02, "learning_rate": 4.975932883678119e-05, "loss": 1.2661, "step": 16896 }, { "epoch": 1.02, "learning_rate": 4.975781992666697e-05, "loss": 0.9536, "step": 16898 }, { "epoch": 1.02, "learning_rate": 4.9756311016552745e-05, "loss": 1.2052, "step": 16900 }, { "epoch": 1.02, "learning_rate": 4.975480210643852e-05, "loss": 1.1344, "step": 16902 }, { "epoch": 1.02, "learning_rate": 4.97532931963243e-05, "loss": 0.997, "step": 16904 }, { "epoch": 1.02, "learning_rate": 4.975178428621007e-05, "loss": 1.0564, "step": 16906 }, { "epoch": 1.02, "learning_rate": 4.975027537609585e-05, "loss": 1.4251, "step": 16908 }, { "epoch": 1.02, "learning_rate": 4.9748766465981624e-05, "loss": 1.2163, "step": 16910 }, { "epoch": 1.02, "learning_rate": 4.97472575558674e-05, "loss": 1.0962, "step": 16912 }, { "epoch": 1.02, "learning_rate": 4.974574864575318e-05, "loss": 1.1945, "step": 16914 }, { "epoch": 1.02, "learning_rate": 4.974423973563895e-05, "loss": 1.2145, "step": 16916 }, { "epoch": 1.02, "learning_rate": 4.9742730825524723e-05, "loss": 1.5357, "step": 16918 }, { "epoch": 1.02, "learning_rate": 4.97412219154105e-05, "loss": 0.9435, "step": 16920 }, { "epoch": 1.02, "learning_rate": 4.9739713005296277e-05, "loss": 1.0847, "step": 16922 }, { "epoch": 1.02, "learning_rate": 4.973820409518205e-05, "loss": 1.5163, "step": 16924 }, { "epoch": 1.02, "learning_rate": 4.973669518506783e-05, "loss": 1.2114, "step": 16926 }, { "epoch": 1.02, "learning_rate": 4.97351862749536e-05, "loss": 0.7519, "step": 16928 }, { "epoch": 1.02, "learning_rate": 4.973367736483938e-05, "loss": 1.1669, "step": 16930 }, { "epoch": 1.02, "learning_rate": 4.9732168454725156e-05, "loss": 1.1981, "step": 16932 }, { "epoch": 1.02, "learning_rate": 4.973065954461093e-05, "loss": 1.3823, "step": 16934 }, { "epoch": 1.02, "learning_rate": 4.972915063449671e-05, "loss": 1.3612, "step": 16936 }, { "epoch": 1.02, "learning_rate": 4.972764172438248e-05, "loss": 0.7046, "step": 16938 }, { "epoch": 1.02, "learning_rate": 4.9726132814268255e-05, "loss": 0.9535, "step": 16940 }, { "epoch": 1.02, "learning_rate": 4.9724623904154035e-05, "loss": 1.3853, "step": 16942 }, { "epoch": 1.02, "learning_rate": 4.972311499403981e-05, "loss": 1.2524, "step": 16944 }, { "epoch": 1.02, "learning_rate": 4.972160608392558e-05, "loss": 1.0363, "step": 16946 }, { "epoch": 1.02, "learning_rate": 4.972009717381136e-05, "loss": 1.1042, "step": 16948 }, { "epoch": 1.02, "learning_rate": 4.9718588263697134e-05, "loss": 1.3305, "step": 16950 }, { "epoch": 1.02, "learning_rate": 4.971707935358291e-05, "loss": 1.1531, "step": 16952 }, { "epoch": 1.02, "learning_rate": 4.971557044346869e-05, "loss": 1.4472, "step": 16954 }, { "epoch": 1.02, "learning_rate": 4.971406153335446e-05, "loss": 1.3937, "step": 16956 }, { "epoch": 1.02, "learning_rate": 4.971255262324024e-05, "loss": 1.1277, "step": 16958 }, { "epoch": 1.02, "learning_rate": 4.9711043713126007e-05, "loss": 1.1885, "step": 16960 }, { "epoch": 1.02, "learning_rate": 4.9709534803011786e-05, "loss": 1.2998, "step": 16962 }, { "epoch": 1.02, "learning_rate": 4.9708025892897566e-05, "loss": 1.1439, "step": 16964 }, { "epoch": 1.02, "learning_rate": 4.970651698278333e-05, "loss": 1.3245, "step": 16966 }, { "epoch": 1.02, "learning_rate": 4.970500807266911e-05, "loss": 1.166, "step": 16968 }, { "epoch": 1.02, "learning_rate": 4.970349916255489e-05, "loss": 1.133, "step": 16970 }, { "epoch": 1.02, "learning_rate": 4.9701990252440666e-05, "loss": 0.9828, "step": 16972 }, { "epoch": 1.02, "learning_rate": 4.970048134232644e-05, "loss": 1.2876, "step": 16974 }, { "epoch": 1.02, "learning_rate": 4.969897243221221e-05, "loss": 0.8537, "step": 16976 }, { "epoch": 1.02, "learning_rate": 4.969746352209799e-05, "loss": 1.2713, "step": 16978 }, { "epoch": 1.02, "learning_rate": 4.9695954611983765e-05, "loss": 1.3791, "step": 16980 }, { "epoch": 1.02, "learning_rate": 4.969444570186954e-05, "loss": 0.8359, "step": 16982 }, { "epoch": 1.03, "learning_rate": 4.969293679175532e-05, "loss": 1.2684, "step": 16984 }, { "epoch": 1.03, "learning_rate": 4.96914278816411e-05, "loss": 1.3677, "step": 16986 }, { "epoch": 1.03, "learning_rate": 4.9689918971526864e-05, "loss": 1.3275, "step": 16988 }, { "epoch": 1.03, "learning_rate": 4.9688410061412644e-05, "loss": 1.1507, "step": 16990 }, { "epoch": 1.03, "learning_rate": 4.9686901151298424e-05, "loss": 1.098, "step": 16992 }, { "epoch": 1.03, "learning_rate": 4.96853922411842e-05, "loss": 0.9192, "step": 16994 }, { "epoch": 1.03, "learning_rate": 4.968388333106997e-05, "loss": 1.3074, "step": 16996 }, { "epoch": 1.03, "learning_rate": 4.968237442095574e-05, "loss": 0.9791, "step": 16998 }, { "epoch": 1.03, "learning_rate": 4.968086551084152e-05, "loss": 1.1175, "step": 17000 }, { "epoch": 1.03, "learning_rate": 4.9679356600727296e-05, "loss": 1.3441, "step": 17002 }, { "epoch": 1.03, "learning_rate": 4.967784769061307e-05, "loss": 1.3301, "step": 17004 }, { "epoch": 1.03, "learning_rate": 4.967633878049885e-05, "loss": 1.1371, "step": 17006 }, { "epoch": 1.03, "learning_rate": 4.967482987038463e-05, "loss": 1.106, "step": 17008 }, { "epoch": 1.03, "learning_rate": 4.9673320960270396e-05, "loss": 1.3828, "step": 17010 }, { "epoch": 1.03, "learning_rate": 4.9671812050156176e-05, "loss": 1.456, "step": 17012 }, { "epoch": 1.03, "learning_rate": 4.967030314004195e-05, "loss": 1.4455, "step": 17014 }, { "epoch": 1.03, "learning_rate": 4.966879422992772e-05, "loss": 1.5135, "step": 17016 }, { "epoch": 1.03, "learning_rate": 4.96672853198135e-05, "loss": 1.1488, "step": 17018 }, { "epoch": 1.03, "learning_rate": 4.9665776409699275e-05, "loss": 1.3814, "step": 17020 }, { "epoch": 1.03, "learning_rate": 4.9664267499585055e-05, "loss": 1.0172, "step": 17022 }, { "epoch": 1.03, "learning_rate": 4.966275858947083e-05, "loss": 1.0766, "step": 17024 }, { "epoch": 1.03, "learning_rate": 4.96612496793566e-05, "loss": 1.2773, "step": 17026 }, { "epoch": 1.03, "learning_rate": 4.965974076924238e-05, "loss": 1.0588, "step": 17028 }, { "epoch": 1.03, "learning_rate": 4.9658231859128154e-05, "loss": 1.0397, "step": 17030 }, { "epoch": 1.03, "learning_rate": 4.965672294901393e-05, "loss": 0.8547, "step": 17032 }, { "epoch": 1.03, "learning_rate": 4.965521403889971e-05, "loss": 0.905, "step": 17034 }, { "epoch": 1.03, "learning_rate": 4.965370512878548e-05, "loss": 1.4768, "step": 17036 }, { "epoch": 1.03, "learning_rate": 4.965219621867125e-05, "loss": 1.5531, "step": 17038 }, { "epoch": 1.03, "learning_rate": 4.965068730855703e-05, "loss": 1.5489, "step": 17040 }, { "epoch": 1.03, "learning_rate": 4.9649178398442806e-05, "loss": 1.0099, "step": 17042 }, { "epoch": 1.03, "learning_rate": 4.9647669488328586e-05, "loss": 1.1641, "step": 17044 }, { "epoch": 1.03, "learning_rate": 4.964616057821436e-05, "loss": 1.4399, "step": 17046 }, { "epoch": 1.03, "learning_rate": 4.964465166810013e-05, "loss": 1.1352, "step": 17048 }, { "epoch": 1.03, "learning_rate": 4.964314275798591e-05, "loss": 1.1884, "step": 17050 }, { "epoch": 1.03, "learning_rate": 4.9641633847871686e-05, "loss": 1.0313, "step": 17052 }, { "epoch": 1.03, "learning_rate": 4.964012493775746e-05, "loss": 1.2858, "step": 17054 }, { "epoch": 1.03, "learning_rate": 4.963861602764324e-05, "loss": 0.7806, "step": 17056 }, { "epoch": 1.03, "learning_rate": 4.963710711752901e-05, "loss": 0.7776, "step": 17058 }, { "epoch": 1.03, "learning_rate": 4.9635598207414785e-05, "loss": 1.2776, "step": 17060 }, { "epoch": 1.03, "learning_rate": 4.9634089297300565e-05, "loss": 1.1037, "step": 17062 }, { "epoch": 1.03, "learning_rate": 4.963258038718634e-05, "loss": 1.6812, "step": 17064 }, { "epoch": 1.03, "learning_rate": 4.963107147707211e-05, "loss": 0.9632, "step": 17066 }, { "epoch": 1.03, "learning_rate": 4.962956256695789e-05, "loss": 1.2283, "step": 17068 }, { "epoch": 1.03, "learning_rate": 4.9628053656843664e-05, "loss": 1.0495, "step": 17070 }, { "epoch": 1.03, "learning_rate": 4.9626544746729444e-05, "loss": 1.3853, "step": 17072 }, { "epoch": 1.03, "learning_rate": 4.962503583661521e-05, "loss": 1.2091, "step": 17074 }, { "epoch": 1.03, "learning_rate": 4.962352692650099e-05, "loss": 1.2142, "step": 17076 }, { "epoch": 1.03, "learning_rate": 4.962201801638677e-05, "loss": 1.4155, "step": 17078 }, { "epoch": 1.03, "learning_rate": 4.9620509106272536e-05, "loss": 1.3359, "step": 17080 }, { "epoch": 1.03, "learning_rate": 4.9619000196158316e-05, "loss": 0.8716, "step": 17082 }, { "epoch": 1.03, "learning_rate": 4.9617491286044096e-05, "loss": 1.3569, "step": 17084 }, { "epoch": 1.03, "learning_rate": 4.961598237592987e-05, "loss": 1.1808, "step": 17086 }, { "epoch": 1.03, "learning_rate": 4.961447346581564e-05, "loss": 1.0187, "step": 17088 }, { "epoch": 1.03, "learning_rate": 4.9612964555701416e-05, "loss": 1.1137, "step": 17090 }, { "epoch": 1.03, "learning_rate": 4.9611455645587195e-05, "loss": 0.8985, "step": 17092 }, { "epoch": 1.03, "learning_rate": 4.9609946735472975e-05, "loss": 1.1962, "step": 17094 }, { "epoch": 1.03, "learning_rate": 4.960843782535874e-05, "loss": 1.3185, "step": 17096 }, { "epoch": 1.03, "learning_rate": 4.960692891524452e-05, "loss": 0.887, "step": 17098 }, { "epoch": 1.03, "learning_rate": 4.96054200051303e-05, "loss": 1.3374, "step": 17100 }, { "epoch": 1.03, "learning_rate": 4.960391109501607e-05, "loss": 1.3508, "step": 17102 }, { "epoch": 1.03, "learning_rate": 4.960240218490185e-05, "loss": 1.4815, "step": 17104 }, { "epoch": 1.03, "learning_rate": 4.960089327478762e-05, "loss": 0.7909, "step": 17106 }, { "epoch": 1.03, "learning_rate": 4.95993843646734e-05, "loss": 1.3239, "step": 17108 }, { "epoch": 1.03, "learning_rate": 4.9597875454559174e-05, "loss": 1.2597, "step": 17110 }, { "epoch": 1.03, "learning_rate": 4.959636654444495e-05, "loss": 1.082, "step": 17112 }, { "epoch": 1.03, "learning_rate": 4.959485763433073e-05, "loss": 1.2759, "step": 17114 }, { "epoch": 1.03, "learning_rate": 4.95933487242165e-05, "loss": 1.221, "step": 17116 }, { "epoch": 1.03, "learning_rate": 4.959183981410227e-05, "loss": 1.1233, "step": 17118 }, { "epoch": 1.03, "learning_rate": 4.959033090398805e-05, "loss": 0.8498, "step": 17120 }, { "epoch": 1.03, "learning_rate": 4.958882199387383e-05, "loss": 1.2564, "step": 17122 }, { "epoch": 1.03, "learning_rate": 4.95873130837596e-05, "loss": 1.1547, "step": 17124 }, { "epoch": 1.03, "learning_rate": 4.958580417364538e-05, "loss": 1.2397, "step": 17126 }, { "epoch": 1.03, "learning_rate": 4.958429526353115e-05, "loss": 1.6178, "step": 17128 }, { "epoch": 1.03, "learning_rate": 4.9582786353416926e-05, "loss": 0.9612, "step": 17130 }, { "epoch": 1.03, "learning_rate": 4.9581277443302705e-05, "loss": 1.2091, "step": 17132 }, { "epoch": 1.03, "learning_rate": 4.957976853318848e-05, "loss": 1.2539, "step": 17134 }, { "epoch": 1.03, "learning_rate": 4.957825962307426e-05, "loss": 1.1269, "step": 17136 }, { "epoch": 1.03, "learning_rate": 4.957675071296003e-05, "loss": 1.2783, "step": 17138 }, { "epoch": 1.03, "learning_rate": 4.9575241802845805e-05, "loss": 1.0519, "step": 17140 }, { "epoch": 1.03, "learning_rate": 4.9573732892731585e-05, "loss": 1.276, "step": 17142 }, { "epoch": 1.03, "learning_rate": 4.957222398261736e-05, "loss": 0.9582, "step": 17144 }, { "epoch": 1.03, "learning_rate": 4.957071507250313e-05, "loss": 0.8341, "step": 17146 }, { "epoch": 1.04, "learning_rate": 4.956920616238891e-05, "loss": 1.1734, "step": 17148 }, { "epoch": 1.04, "learning_rate": 4.9567697252274684e-05, "loss": 1.4615, "step": 17150 }, { "epoch": 1.04, "learning_rate": 4.956618834216046e-05, "loss": 1.2365, "step": 17152 }, { "epoch": 1.04, "learning_rate": 4.956467943204624e-05, "loss": 1.2017, "step": 17154 }, { "epoch": 1.04, "learning_rate": 4.956317052193201e-05, "loss": 1.1234, "step": 17156 }, { "epoch": 1.04, "learning_rate": 4.956166161181779e-05, "loss": 1.1263, "step": 17158 }, { "epoch": 1.04, "learning_rate": 4.956015270170356e-05, "loss": 0.9499, "step": 17160 }, { "epoch": 1.04, "learning_rate": 4.9558643791589336e-05, "loss": 1.0918, "step": 17162 }, { "epoch": 1.04, "learning_rate": 4.9557134881475116e-05, "loss": 1.109, "step": 17164 }, { "epoch": 1.04, "learning_rate": 4.955562597136089e-05, "loss": 0.982, "step": 17166 }, { "epoch": 1.04, "learning_rate": 4.955411706124666e-05, "loss": 1.0679, "step": 17168 }, { "epoch": 1.04, "learning_rate": 4.955260815113244e-05, "loss": 1.2536, "step": 17170 }, { "epoch": 1.04, "learning_rate": 4.9551099241018215e-05, "loss": 1.0624, "step": 17172 }, { "epoch": 1.04, "learning_rate": 4.954959033090399e-05, "loss": 0.9707, "step": 17174 }, { "epoch": 1.04, "learning_rate": 4.954808142078977e-05, "loss": 1.0008, "step": 17176 }, { "epoch": 1.04, "learning_rate": 4.954657251067554e-05, "loss": 0.8519, "step": 17178 }, { "epoch": 1.04, "learning_rate": 4.9545063600561315e-05, "loss": 1.5721, "step": 17180 }, { "epoch": 1.04, "learning_rate": 4.9543554690447095e-05, "loss": 1.2257, "step": 17182 }, { "epoch": 1.04, "learning_rate": 4.954204578033287e-05, "loss": 1.2165, "step": 17184 }, { "epoch": 1.04, "learning_rate": 4.954053687021865e-05, "loss": 0.8144, "step": 17186 }, { "epoch": 1.04, "learning_rate": 4.9539027960104414e-05, "loss": 1.0821, "step": 17188 }, { "epoch": 1.04, "learning_rate": 4.9537519049990194e-05, "loss": 1.0384, "step": 17190 }, { "epoch": 1.04, "learning_rate": 4.9536010139875974e-05, "loss": 1.3507, "step": 17192 }, { "epoch": 1.04, "learning_rate": 4.953450122976174e-05, "loss": 1.0558, "step": 17194 }, { "epoch": 1.04, "learning_rate": 4.953299231964752e-05, "loss": 0.9476, "step": 17196 }, { "epoch": 1.04, "learning_rate": 4.95314834095333e-05, "loss": 0.8431, "step": 17198 }, { "epoch": 1.04, "learning_rate": 4.952997449941907e-05, "loss": 1.2134, "step": 17200 }, { "epoch": 1.04, "learning_rate": 4.9528465589304846e-05, "loss": 1.0959, "step": 17202 }, { "epoch": 1.04, "learning_rate": 4.952695667919062e-05, "loss": 0.9496, "step": 17204 }, { "epoch": 1.04, "learning_rate": 4.95254477690764e-05, "loss": 1.0773, "step": 17206 }, { "epoch": 1.04, "learning_rate": 4.952393885896218e-05, "loss": 1.2739, "step": 17208 }, { "epoch": 1.04, "learning_rate": 4.9522429948847945e-05, "loss": 1.3203, "step": 17210 }, { "epoch": 1.04, "learning_rate": 4.9520921038733725e-05, "loss": 1.0498, "step": 17212 }, { "epoch": 1.04, "learning_rate": 4.9519412128619505e-05, "loss": 1.447, "step": 17214 }, { "epoch": 1.04, "learning_rate": 4.951790321850527e-05, "loss": 0.9691, "step": 17216 }, { "epoch": 1.04, "learning_rate": 4.951639430839105e-05, "loss": 1.1169, "step": 17218 }, { "epoch": 1.04, "learning_rate": 4.9514885398276825e-05, "loss": 1.6439, "step": 17220 }, { "epoch": 1.04, "learning_rate": 4.9513376488162604e-05, "loss": 1.0407, "step": 17222 }, { "epoch": 1.04, "learning_rate": 4.951186757804838e-05, "loss": 1.1322, "step": 17224 }, { "epoch": 1.04, "learning_rate": 4.951035866793415e-05, "loss": 1.2954, "step": 17226 }, { "epoch": 1.04, "learning_rate": 4.950884975781993e-05, "loss": 1.5119, "step": 17228 }, { "epoch": 1.04, "learning_rate": 4.9507340847705704e-05, "loss": 1.566, "step": 17230 }, { "epoch": 1.04, "learning_rate": 4.950583193759148e-05, "loss": 1.3081, "step": 17232 }, { "epoch": 1.04, "learning_rate": 4.950432302747726e-05, "loss": 1.1458, "step": 17234 }, { "epoch": 1.04, "learning_rate": 4.950281411736304e-05, "loss": 1.4889, "step": 17236 }, { "epoch": 1.04, "learning_rate": 4.95013052072488e-05, "loss": 1.5322, "step": 17238 }, { "epoch": 1.04, "learning_rate": 4.949979629713458e-05, "loss": 0.909, "step": 17240 }, { "epoch": 1.04, "learning_rate": 4.9498287387020356e-05, "loss": 1.129, "step": 17242 }, { "epoch": 1.04, "learning_rate": 4.949677847690613e-05, "loss": 1.2234, "step": 17244 }, { "epoch": 1.04, "learning_rate": 4.949526956679191e-05, "loss": 1.0021, "step": 17246 }, { "epoch": 1.04, "learning_rate": 4.949376065667768e-05, "loss": 1.0355, "step": 17248 }, { "epoch": 1.04, "learning_rate": 4.949225174656346e-05, "loss": 1.6651, "step": 17250 }, { "epoch": 1.04, "learning_rate": 4.9490742836449235e-05, "loss": 1.1037, "step": 17252 }, { "epoch": 1.04, "learning_rate": 4.948923392633501e-05, "loss": 0.9761, "step": 17254 }, { "epoch": 1.04, "learning_rate": 4.948772501622079e-05, "loss": 1.3179, "step": 17256 }, { "epoch": 1.04, "learning_rate": 4.948621610610656e-05, "loss": 1.3869, "step": 17258 }, { "epoch": 1.04, "learning_rate": 4.9484707195992335e-05, "loss": 1.3, "step": 17260 }, { "epoch": 1.04, "learning_rate": 4.9483198285878114e-05, "loss": 1.1257, "step": 17262 }, { "epoch": 1.04, "learning_rate": 4.948168937576389e-05, "loss": 1.1622, "step": 17264 }, { "epoch": 1.04, "learning_rate": 4.948018046564966e-05, "loss": 1.2587, "step": 17266 }, { "epoch": 1.04, "learning_rate": 4.947867155553544e-05, "loss": 1.1348, "step": 17268 }, { "epoch": 1.04, "learning_rate": 4.9477162645421214e-05, "loss": 0.8875, "step": 17270 }, { "epoch": 1.04, "learning_rate": 4.9475653735306994e-05, "loss": 0.8427, "step": 17272 }, { "epoch": 1.04, "learning_rate": 4.947414482519277e-05, "loss": 1.0356, "step": 17274 }, { "epoch": 1.04, "learning_rate": 4.947263591507854e-05, "loss": 0.9751, "step": 17276 }, { "epoch": 1.04, "learning_rate": 4.947112700496432e-05, "loss": 1.1891, "step": 17278 }, { "epoch": 1.04, "learning_rate": 4.946961809485009e-05, "loss": 1.6413, "step": 17280 }, { "epoch": 1.04, "learning_rate": 4.9468109184735866e-05, "loss": 1.2894, "step": 17282 }, { "epoch": 1.04, "learning_rate": 4.9466600274621646e-05, "loss": 1.6092, "step": 17284 }, { "epoch": 1.04, "learning_rate": 4.946509136450742e-05, "loss": 1.398, "step": 17286 }, { "epoch": 1.04, "learning_rate": 4.946358245439319e-05, "loss": 1.1932, "step": 17288 }, { "epoch": 1.04, "learning_rate": 4.946207354427897e-05, "loss": 1.1361, "step": 17290 }, { "epoch": 1.04, "learning_rate": 4.9460564634164745e-05, "loss": 1.0523, "step": 17292 }, { "epoch": 1.04, "learning_rate": 4.945905572405052e-05, "loss": 1.2036, "step": 17294 }, { "epoch": 1.04, "learning_rate": 4.94575468139363e-05, "loss": 0.8791, "step": 17296 }, { "epoch": 1.04, "learning_rate": 4.945603790382207e-05, "loss": 1.1997, "step": 17298 }, { "epoch": 1.04, "learning_rate": 4.945452899370785e-05, "loss": 0.9307, "step": 17300 }, { "epoch": 1.04, "learning_rate": 4.945302008359362e-05, "loss": 1.0015, "step": 17302 }, { "epoch": 1.04, "learning_rate": 4.94515111734794e-05, "loss": 1.1304, "step": 17304 }, { "epoch": 1.04, "learning_rate": 4.945000226336518e-05, "loss": 1.1112, "step": 17306 }, { "epoch": 1.04, "learning_rate": 4.9448493353250944e-05, "loss": 1.252, "step": 17308 }, { "epoch": 1.04, "learning_rate": 4.9446984443136724e-05, "loss": 1.1862, "step": 17310 }, { "epoch": 1.04, "learning_rate": 4.9445475533022504e-05, "loss": 1.4085, "step": 17312 }, { "epoch": 1.05, "learning_rate": 4.944396662290828e-05, "loss": 1.1328, "step": 17314 }, { "epoch": 1.05, "learning_rate": 4.944245771279405e-05, "loss": 0.956, "step": 17316 }, { "epoch": 1.05, "learning_rate": 4.944094880267982e-05, "loss": 1.0147, "step": 17318 }, { "epoch": 1.05, "learning_rate": 4.94394398925656e-05, "loss": 0.9656, "step": 17320 }, { "epoch": 1.05, "learning_rate": 4.943793098245138e-05, "loss": 1.3361, "step": 17322 }, { "epoch": 1.05, "learning_rate": 4.943642207233715e-05, "loss": 0.9834, "step": 17324 }, { "epoch": 1.05, "learning_rate": 4.943491316222293e-05, "loss": 0.7005, "step": 17326 }, { "epoch": 1.05, "learning_rate": 4.943340425210871e-05, "loss": 0.9225, "step": 17328 }, { "epoch": 1.05, "learning_rate": 4.9431895341994475e-05, "loss": 1.3931, "step": 17330 }, { "epoch": 1.05, "learning_rate": 4.9430386431880255e-05, "loss": 1.2067, "step": 17332 }, { "epoch": 1.05, "learning_rate": 4.942887752176603e-05, "loss": 0.9023, "step": 17334 }, { "epoch": 1.05, "learning_rate": 4.942736861165181e-05, "loss": 1.266, "step": 17336 }, { "epoch": 1.05, "learning_rate": 4.942585970153758e-05, "loss": 1.4055, "step": 17338 }, { "epoch": 1.05, "learning_rate": 4.9424350791423354e-05, "loss": 1.4263, "step": 17340 }, { "epoch": 1.05, "learning_rate": 4.9422841881309134e-05, "loss": 1.3815, "step": 17342 }, { "epoch": 1.05, "learning_rate": 4.942133297119491e-05, "loss": 1.5236, "step": 17344 }, { "epoch": 1.05, "learning_rate": 4.941982406108068e-05, "loss": 1.3699, "step": 17346 }, { "epoch": 1.05, "learning_rate": 4.941831515096646e-05, "loss": 1.4276, "step": 17348 }, { "epoch": 1.05, "learning_rate": 4.941680624085224e-05, "loss": 1.3617, "step": 17350 }, { "epoch": 1.05, "learning_rate": 4.941529733073801e-05, "loss": 1.0795, "step": 17352 }, { "epoch": 1.05, "learning_rate": 4.9413788420623787e-05, "loss": 0.8852, "step": 17354 }, { "epoch": 1.05, "learning_rate": 4.941227951050956e-05, "loss": 1.3066, "step": 17356 }, { "epoch": 1.05, "learning_rate": 4.941077060039533e-05, "loss": 1.1448, "step": 17358 }, { "epoch": 1.05, "learning_rate": 4.940926169028111e-05, "loss": 1.1565, "step": 17360 }, { "epoch": 1.05, "learning_rate": 4.9407752780166886e-05, "loss": 1.1134, "step": 17362 }, { "epoch": 1.05, "learning_rate": 4.9406243870052666e-05, "loss": 0.8899, "step": 17364 }, { "epoch": 1.05, "learning_rate": 4.940473495993844e-05, "loss": 1.3714, "step": 17366 }, { "epoch": 1.05, "learning_rate": 4.940322604982421e-05, "loss": 0.9798, "step": 17368 }, { "epoch": 1.05, "learning_rate": 4.940171713970999e-05, "loss": 1.1523, "step": 17370 }, { "epoch": 1.05, "learning_rate": 4.9400208229595765e-05, "loss": 0.8234, "step": 17372 }, { "epoch": 1.05, "learning_rate": 4.939869931948154e-05, "loss": 1.5664, "step": 17374 }, { "epoch": 1.05, "learning_rate": 4.939719040936732e-05, "loss": 0.9355, "step": 17376 }, { "epoch": 1.05, "learning_rate": 4.939568149925309e-05, "loss": 1.3172, "step": 17378 }, { "epoch": 1.05, "learning_rate": 4.9394172589138864e-05, "loss": 0.8258, "step": 17380 }, { "epoch": 1.05, "learning_rate": 4.9392663679024644e-05, "loss": 1.6005, "step": 17382 }, { "epoch": 1.05, "learning_rate": 4.939115476891042e-05, "loss": 1.0801, "step": 17384 }, { "epoch": 1.05, "learning_rate": 4.93896458587962e-05, "loss": 1.0917, "step": 17386 }, { "epoch": 1.05, "learning_rate": 4.938813694868197e-05, "loss": 1.2603, "step": 17388 }, { "epoch": 1.05, "learning_rate": 4.9386628038567744e-05, "loss": 1.141, "step": 17390 }, { "epoch": 1.05, "learning_rate": 4.9385119128453523e-05, "loss": 1.0908, "step": 17392 }, { "epoch": 1.05, "learning_rate": 4.9383610218339297e-05, "loss": 0.9803, "step": 17394 }, { "epoch": 1.05, "learning_rate": 4.938210130822507e-05, "loss": 1.1187, "step": 17396 }, { "epoch": 1.05, "learning_rate": 4.938059239811085e-05, "loss": 0.7855, "step": 17398 }, { "epoch": 1.05, "learning_rate": 4.937908348799662e-05, "loss": 1.5001, "step": 17400 }, { "epoch": 1.05, "learning_rate": 4.9377574577882396e-05, "loss": 1.4064, "step": 17402 }, { "epoch": 1.05, "learning_rate": 4.9376065667768176e-05, "loss": 0.8251, "step": 17404 }, { "epoch": 1.05, "learning_rate": 4.937455675765395e-05, "loss": 1.0936, "step": 17406 }, { "epoch": 1.05, "learning_rate": 4.937304784753972e-05, "loss": 0.993, "step": 17408 }, { "epoch": 1.05, "learning_rate": 4.93715389374255e-05, "loss": 1.3431, "step": 17410 }, { "epoch": 1.05, "learning_rate": 4.9370030027311275e-05, "loss": 1.0242, "step": 17412 }, { "epoch": 1.05, "learning_rate": 4.9368521117197055e-05, "loss": 1.0654, "step": 17414 }, { "epoch": 1.05, "learning_rate": 4.936701220708282e-05, "loss": 1.3107, "step": 17416 }, { "epoch": 1.05, "learning_rate": 4.93655032969686e-05, "loss": 1.372, "step": 17418 }, { "epoch": 1.05, "learning_rate": 4.936399438685438e-05, "loss": 1.0661, "step": 17420 }, { "epoch": 1.05, "learning_rate": 4.9362485476740154e-05, "loss": 1.0858, "step": 17422 }, { "epoch": 1.05, "learning_rate": 4.936097656662593e-05, "loss": 1.2035, "step": 17424 }, { "epoch": 1.05, "learning_rate": 4.935946765651171e-05, "loss": 0.974, "step": 17426 }, { "epoch": 1.05, "learning_rate": 4.935795874639748e-05, "loss": 1.3556, "step": 17428 }, { "epoch": 1.05, "learning_rate": 4.9356449836283253e-05, "loss": 0.9767, "step": 17430 }, { "epoch": 1.05, "learning_rate": 4.9354940926169027e-05, "loss": 1.1843, "step": 17432 }, { "epoch": 1.05, "learning_rate": 4.9353432016054806e-05, "loss": 1.3588, "step": 17434 }, { "epoch": 1.05, "learning_rate": 4.9351923105940586e-05, "loss": 0.9723, "step": 17436 }, { "epoch": 1.05, "learning_rate": 4.935041419582635e-05, "loss": 0.8336, "step": 17438 }, { "epoch": 1.05, "learning_rate": 4.934890528571213e-05, "loss": 1.0022, "step": 17440 }, { "epoch": 1.05, "learning_rate": 4.934739637559791e-05, "loss": 1.0731, "step": 17442 }, { "epoch": 1.05, "learning_rate": 4.934588746548368e-05, "loss": 1.3587, "step": 17444 }, { "epoch": 1.05, "learning_rate": 4.934437855536946e-05, "loss": 0.9432, "step": 17446 }, { "epoch": 1.05, "learning_rate": 4.934286964525523e-05, "loss": 1.0662, "step": 17448 }, { "epoch": 1.05, "learning_rate": 4.934136073514101e-05, "loss": 1.138, "step": 17450 }, { "epoch": 1.05, "learning_rate": 4.9339851825026785e-05, "loss": 0.814, "step": 17452 }, { "epoch": 1.05, "learning_rate": 4.933834291491256e-05, "loss": 1.256, "step": 17454 }, { "epoch": 1.05, "learning_rate": 4.933683400479834e-05, "loss": 1.1976, "step": 17456 }, { "epoch": 1.05, "learning_rate": 4.933532509468411e-05, "loss": 1.3793, "step": 17458 }, { "epoch": 1.05, "learning_rate": 4.9333816184569884e-05, "loss": 1.6597, "step": 17460 }, { "epoch": 1.05, "learning_rate": 4.9332307274455664e-05, "loss": 1.1946, "step": 17462 }, { "epoch": 1.05, "learning_rate": 4.933079836434144e-05, "loss": 1.1051, "step": 17464 }, { "epoch": 1.05, "learning_rate": 4.932928945422721e-05, "loss": 1.5068, "step": 17466 }, { "epoch": 1.05, "learning_rate": 4.932778054411299e-05, "loss": 1.0012, "step": 17468 }, { "epoch": 1.05, "learning_rate": 4.9326271633998763e-05, "loss": 1.0227, "step": 17470 }, { "epoch": 1.05, "learning_rate": 4.932476272388454e-05, "loss": 0.8389, "step": 17472 }, { "epoch": 1.05, "learning_rate": 4.9323253813770316e-05, "loss": 0.7919, "step": 17474 }, { "epoch": 1.05, "learning_rate": 4.932174490365609e-05, "loss": 1.1187, "step": 17476 }, { "epoch": 1.05, "learning_rate": 4.932023599354187e-05, "loss": 1.064, "step": 17478 }, { "epoch": 1.06, "learning_rate": 4.931872708342764e-05, "loss": 1.2024, "step": 17480 }, { "epoch": 1.06, "learning_rate": 4.9317218173313416e-05, "loss": 1.1199, "step": 17482 }, { "epoch": 1.06, "learning_rate": 4.9315709263199196e-05, "loss": 1.2347, "step": 17484 }, { "epoch": 1.06, "learning_rate": 4.931420035308497e-05, "loss": 1.0005, "step": 17486 }, { "epoch": 1.06, "learning_rate": 4.931269144297074e-05, "loss": 1.3462, "step": 17488 }, { "epoch": 1.06, "learning_rate": 4.931118253285652e-05, "loss": 0.9961, "step": 17490 }, { "epoch": 1.06, "learning_rate": 4.9309673622742295e-05, "loss": 1.0143, "step": 17492 }, { "epoch": 1.06, "learning_rate": 4.930816471262807e-05, "loss": 1.0019, "step": 17494 }, { "epoch": 1.06, "learning_rate": 4.930665580251385e-05, "loss": 1.4148, "step": 17496 }, { "epoch": 1.06, "learning_rate": 4.930514689239962e-05, "loss": 1.0236, "step": 17498 }, { "epoch": 1.06, "learning_rate": 4.93036379822854e-05, "loss": 1.1872, "step": 17500 }, { "epoch": 1.06, "learning_rate": 4.9302129072171174e-05, "loss": 1.126, "step": 17502 }, { "epoch": 1.06, "learning_rate": 4.930062016205695e-05, "loss": 1.0865, "step": 17504 }, { "epoch": 1.06, "learning_rate": 4.929911125194273e-05, "loss": 1.3545, "step": 17506 }, { "epoch": 1.06, "learning_rate": 4.92976023418285e-05, "loss": 1.034, "step": 17508 }, { "epoch": 1.06, "learning_rate": 4.929609343171427e-05, "loss": 0.9739, "step": 17510 }, { "epoch": 1.06, "learning_rate": 4.929458452160005e-05, "loss": 1.174, "step": 17512 }, { "epoch": 1.06, "learning_rate": 4.9293075611485826e-05, "loss": 1.0133, "step": 17514 }, { "epoch": 1.06, "learning_rate": 4.92915667013716e-05, "loss": 1.0844, "step": 17516 }, { "epoch": 1.06, "learning_rate": 4.929005779125738e-05, "loss": 1.0858, "step": 17518 }, { "epoch": 1.06, "learning_rate": 4.928854888114315e-05, "loss": 1.1897, "step": 17520 }, { "epoch": 1.06, "learning_rate": 4.9287039971028926e-05, "loss": 1.1931, "step": 17522 }, { "epoch": 1.06, "learning_rate": 4.9285531060914706e-05, "loss": 1.412, "step": 17524 }, { "epoch": 1.06, "learning_rate": 4.928402215080048e-05, "loss": 0.9346, "step": 17526 }, { "epoch": 1.06, "learning_rate": 4.928251324068626e-05, "loss": 1.2341, "step": 17528 }, { "epoch": 1.06, "learning_rate": 4.9281004330572025e-05, "loss": 1.2025, "step": 17530 }, { "epoch": 1.06, "learning_rate": 4.9279495420457805e-05, "loss": 0.8836, "step": 17532 }, { "epoch": 1.06, "learning_rate": 4.9277986510343585e-05, "loss": 1.22, "step": 17534 }, { "epoch": 1.06, "learning_rate": 4.927647760022936e-05, "loss": 0.9305, "step": 17536 }, { "epoch": 1.06, "learning_rate": 4.927496869011513e-05, "loss": 1.3796, "step": 17538 }, { "epoch": 1.06, "learning_rate": 4.927345978000091e-05, "loss": 2.026, "step": 17540 }, { "epoch": 1.06, "learning_rate": 4.9271950869886684e-05, "loss": 1.1347, "step": 17542 }, { "epoch": 1.06, "learning_rate": 4.927044195977246e-05, "loss": 1.2715, "step": 17544 }, { "epoch": 1.06, "learning_rate": 4.926893304965823e-05, "loss": 1.4464, "step": 17546 }, { "epoch": 1.06, "learning_rate": 4.926742413954401e-05, "loss": 1.0207, "step": 17548 }, { "epoch": 1.06, "learning_rate": 4.926591522942979e-05, "loss": 1.5385, "step": 17550 }, { "epoch": 1.06, "learning_rate": 4.9264406319315556e-05, "loss": 1.2236, "step": 17552 }, { "epoch": 1.06, "learning_rate": 4.9262897409201336e-05, "loss": 1.3528, "step": 17554 }, { "epoch": 1.06, "learning_rate": 4.9261388499087116e-05, "loss": 1.1738, "step": 17556 }, { "epoch": 1.06, "learning_rate": 4.925987958897288e-05, "loss": 1.2884, "step": 17558 }, { "epoch": 1.06, "learning_rate": 4.925837067885866e-05, "loss": 1.2204, "step": 17560 }, { "epoch": 1.06, "learning_rate": 4.9256861768744436e-05, "loss": 1.3914, "step": 17562 }, { "epoch": 1.06, "learning_rate": 4.9255352858630216e-05, "loss": 1.3008, "step": 17564 }, { "epoch": 1.06, "learning_rate": 4.925384394851599e-05, "loss": 0.9648, "step": 17566 }, { "epoch": 1.06, "learning_rate": 4.925233503840176e-05, "loss": 0.9502, "step": 17568 }, { "epoch": 1.06, "learning_rate": 4.925082612828754e-05, "loss": 1.2239, "step": 17570 }, { "epoch": 1.06, "learning_rate": 4.9249317218173315e-05, "loss": 1.1264, "step": 17572 }, { "epoch": 1.06, "learning_rate": 4.924780830805909e-05, "loss": 0.9302, "step": 17574 }, { "epoch": 1.06, "learning_rate": 4.924629939794487e-05, "loss": 0.8067, "step": 17576 }, { "epoch": 1.06, "learning_rate": 4.924479048783064e-05, "loss": 1.5434, "step": 17578 }, { "epoch": 1.06, "learning_rate": 4.9243281577716414e-05, "loss": 1.2789, "step": 17580 }, { "epoch": 1.06, "learning_rate": 4.9241772667602194e-05, "loss": 1.2425, "step": 17582 }, { "epoch": 1.06, "learning_rate": 4.924026375748797e-05, "loss": 1.24, "step": 17584 }, { "epoch": 1.06, "learning_rate": 4.923875484737375e-05, "loss": 0.8877, "step": 17586 }, { "epoch": 1.06, "learning_rate": 4.923724593725952e-05, "loss": 1.2842, "step": 17588 }, { "epoch": 1.06, "learning_rate": 4.923573702714529e-05, "loss": 1.3125, "step": 17590 }, { "epoch": 1.06, "learning_rate": 4.923422811703107e-05, "loss": 1.3075, "step": 17592 }, { "epoch": 1.06, "learning_rate": 4.9232719206916846e-05, "loss": 1.1598, "step": 17594 }, { "epoch": 1.06, "learning_rate": 4.923121029680262e-05, "loss": 1.3427, "step": 17596 }, { "epoch": 1.06, "learning_rate": 4.92297013866884e-05, "loss": 1.0044, "step": 17598 }, { "epoch": 1.06, "learning_rate": 4.922819247657417e-05, "loss": 1.2936, "step": 17600 }, { "epoch": 1.06, "learning_rate": 4.9226683566459946e-05, "loss": 1.4592, "step": 17602 }, { "epoch": 1.06, "learning_rate": 4.9225174656345725e-05, "loss": 1.0065, "step": 17604 }, { "epoch": 1.06, "learning_rate": 4.92236657462315e-05, "loss": 1.0176, "step": 17606 }, { "epoch": 1.06, "learning_rate": 4.922215683611727e-05, "loss": 1.4109, "step": 17608 }, { "epoch": 1.06, "learning_rate": 4.922064792600305e-05, "loss": 1.3608, "step": 17610 }, { "epoch": 1.06, "learning_rate": 4.9219139015888825e-05, "loss": 1.0258, "step": 17612 }, { "epoch": 1.06, "learning_rate": 4.9217630105774605e-05, "loss": 1.009, "step": 17614 }, { "epoch": 1.06, "learning_rate": 4.921612119566038e-05, "loss": 1.3052, "step": 17616 }, { "epoch": 1.06, "learning_rate": 4.921461228554615e-05, "loss": 1.5622, "step": 17618 }, { "epoch": 1.06, "learning_rate": 4.921310337543193e-05, "loss": 2.1623, "step": 17620 }, { "epoch": 1.06, "learning_rate": 4.9211594465317704e-05, "loss": 0.8909, "step": 17622 }, { "epoch": 1.06, "learning_rate": 4.921008555520348e-05, "loss": 1.2357, "step": 17624 }, { "epoch": 1.06, "learning_rate": 4.920857664508926e-05, "loss": 0.9976, "step": 17626 }, { "epoch": 1.06, "learning_rate": 4.920706773497503e-05, "loss": 0.9436, "step": 17628 }, { "epoch": 1.06, "learning_rate": 4.92055588248608e-05, "loss": 0.7509, "step": 17630 }, { "epoch": 1.06, "learning_rate": 4.920404991474658e-05, "loss": 1.0568, "step": 17632 }, { "epoch": 1.06, "learning_rate": 4.9202541004632356e-05, "loss": 1.1981, "step": 17634 }, { "epoch": 1.06, "learning_rate": 4.9201032094518136e-05, "loss": 1.5272, "step": 17636 }, { "epoch": 1.06, "learning_rate": 4.919952318440391e-05, "loss": 1.1545, "step": 17638 }, { "epoch": 1.06, "learning_rate": 4.919801427428968e-05, "loss": 1.4187, "step": 17640 }, { "epoch": 1.06, "learning_rate": 4.919650536417546e-05, "loss": 1.0982, "step": 17642 }, { "epoch": 1.06, "learning_rate": 4.919499645406123e-05, "loss": 0.9048, "step": 17644 }, { "epoch": 1.07, "learning_rate": 4.919348754394701e-05, "loss": 0.9654, "step": 17646 }, { "epoch": 1.07, "learning_rate": 4.919197863383279e-05, "loss": 1.0476, "step": 17648 }, { "epoch": 1.07, "learning_rate": 4.919046972371856e-05, "loss": 1.0513, "step": 17650 }, { "epoch": 1.07, "learning_rate": 4.9188960813604335e-05, "loss": 1.1144, "step": 17652 }, { "epoch": 1.07, "learning_rate": 4.9187451903490115e-05, "loss": 1.1328, "step": 17654 }, { "epoch": 1.07, "learning_rate": 4.918594299337589e-05, "loss": 1.6623, "step": 17656 }, { "epoch": 1.07, "learning_rate": 4.918443408326166e-05, "loss": 1.0403, "step": 17658 }, { "epoch": 1.07, "learning_rate": 4.9182925173147434e-05, "loss": 1.1524, "step": 17660 }, { "epoch": 1.07, "learning_rate": 4.9181416263033214e-05, "loss": 1.2615, "step": 17662 }, { "epoch": 1.07, "learning_rate": 4.9179907352918994e-05, "loss": 1.1161, "step": 17664 }, { "epoch": 1.07, "learning_rate": 4.917839844280476e-05, "loss": 1.1541, "step": 17666 }, { "epoch": 1.07, "learning_rate": 4.917688953269054e-05, "loss": 1.1873, "step": 17668 }, { "epoch": 1.07, "learning_rate": 4.917538062257632e-05, "loss": 0.8892, "step": 17670 }, { "epoch": 1.07, "learning_rate": 4.9173871712462086e-05, "loss": 1.3046, "step": 17672 }, { "epoch": 1.07, "learning_rate": 4.9172362802347866e-05, "loss": 1.3753, "step": 17674 }, { "epoch": 1.07, "learning_rate": 4.917085389223364e-05, "loss": 1.1274, "step": 17676 }, { "epoch": 1.07, "learning_rate": 4.916934498211942e-05, "loss": 1.1188, "step": 17678 }, { "epoch": 1.07, "learning_rate": 4.916783607200519e-05, "loss": 1.2814, "step": 17680 }, { "epoch": 1.07, "learning_rate": 4.9166327161890965e-05, "loss": 1.0009, "step": 17682 }, { "epoch": 1.07, "learning_rate": 4.9164818251776745e-05, "loss": 0.8704, "step": 17684 }, { "epoch": 1.07, "learning_rate": 4.916330934166252e-05, "loss": 0.9241, "step": 17686 }, { "epoch": 1.07, "learning_rate": 4.916180043154829e-05, "loss": 1.276, "step": 17688 }, { "epoch": 1.07, "learning_rate": 4.916029152143407e-05, "loss": 0.9548, "step": 17690 }, { "epoch": 1.07, "learning_rate": 4.9158782611319845e-05, "loss": 1.0086, "step": 17692 }, { "epoch": 1.07, "learning_rate": 4.915727370120562e-05, "loss": 1.2894, "step": 17694 }, { "epoch": 1.07, "learning_rate": 4.91557647910914e-05, "loss": 1.2862, "step": 17696 }, { "epoch": 1.07, "learning_rate": 4.915425588097717e-05, "loss": 1.4538, "step": 17698 }, { "epoch": 1.07, "learning_rate": 4.915274697086295e-05, "loss": 1.0085, "step": 17700 }, { "epoch": 1.07, "learning_rate": 4.9151238060748724e-05, "loss": 1.3073, "step": 17702 }, { "epoch": 1.07, "learning_rate": 4.91497291506345e-05, "loss": 1.033, "step": 17704 }, { "epoch": 1.07, "learning_rate": 4.914822024052028e-05, "loss": 1.5528, "step": 17706 }, { "epoch": 1.07, "learning_rate": 4.914671133040605e-05, "loss": 1.2155, "step": 17708 }, { "epoch": 1.07, "learning_rate": 4.914520242029182e-05, "loss": 0.9676, "step": 17710 }, { "epoch": 1.07, "learning_rate": 4.91436935101776e-05, "loss": 1.3619, "step": 17712 }, { "epoch": 1.07, "learning_rate": 4.9142184600063376e-05, "loss": 1.0773, "step": 17714 }, { "epoch": 1.07, "learning_rate": 4.914067568994915e-05, "loss": 1.077, "step": 17716 }, { "epoch": 1.07, "learning_rate": 4.913916677983493e-05, "loss": 1.0427, "step": 17718 }, { "epoch": 1.07, "learning_rate": 4.91376578697207e-05, "loss": 1.436, "step": 17720 }, { "epoch": 1.07, "learning_rate": 4.9136148959606475e-05, "loss": 1.0651, "step": 17722 }, { "epoch": 1.07, "learning_rate": 4.9134640049492255e-05, "loss": 0.987, "step": 17724 }, { "epoch": 1.07, "learning_rate": 4.913313113937803e-05, "loss": 1.2688, "step": 17726 }, { "epoch": 1.07, "learning_rate": 4.913162222926381e-05, "loss": 1.2911, "step": 17728 }, { "epoch": 1.07, "learning_rate": 4.913011331914958e-05, "loss": 1.1475, "step": 17730 }, { "epoch": 1.07, "learning_rate": 4.9128604409035355e-05, "loss": 1.1599, "step": 17732 }, { "epoch": 1.07, "learning_rate": 4.9127095498921134e-05, "loss": 1.1004, "step": 17734 }, { "epoch": 1.07, "learning_rate": 4.912558658880691e-05, "loss": 1.1622, "step": 17736 }, { "epoch": 1.07, "learning_rate": 4.912407767869268e-05, "loss": 1.3763, "step": 17738 }, { "epoch": 1.07, "learning_rate": 4.912256876857846e-05, "loss": 1.1232, "step": 17740 }, { "epoch": 1.07, "learning_rate": 4.9121059858464234e-05, "loss": 1.7681, "step": 17742 }, { "epoch": 1.07, "learning_rate": 4.911955094835001e-05, "loss": 0.9733, "step": 17744 }, { "epoch": 1.07, "learning_rate": 4.911804203823579e-05, "loss": 1.2486, "step": 17746 }, { "epoch": 1.07, "learning_rate": 4.911653312812156e-05, "loss": 1.1834, "step": 17748 }, { "epoch": 1.07, "learning_rate": 4.911502421800734e-05, "loss": 1.0923, "step": 17750 }, { "epoch": 1.07, "learning_rate": 4.911351530789311e-05, "loss": 1.0718, "step": 17752 }, { "epoch": 1.07, "learning_rate": 4.9112006397778886e-05, "loss": 1.0257, "step": 17754 }, { "epoch": 1.07, "learning_rate": 4.9110497487664666e-05, "loss": 1.3824, "step": 17756 }, { "epoch": 1.07, "learning_rate": 4.910898857755043e-05, "loss": 1.2973, "step": 17758 }, { "epoch": 1.07, "learning_rate": 4.910747966743621e-05, "loss": 1.4762, "step": 17760 }, { "epoch": 1.07, "learning_rate": 4.910597075732199e-05, "loss": 1.1977, "step": 17762 }, { "epoch": 1.07, "learning_rate": 4.9104461847207765e-05, "loss": 1.2137, "step": 17764 }, { "epoch": 1.07, "learning_rate": 4.910295293709354e-05, "loss": 1.46, "step": 17766 }, { "epoch": 1.07, "learning_rate": 4.910144402697932e-05, "loss": 1.2553, "step": 17768 }, { "epoch": 1.07, "learning_rate": 4.909993511686509e-05, "loss": 1.3145, "step": 17770 }, { "epoch": 1.07, "learning_rate": 4.9098426206750865e-05, "loss": 1.1047, "step": 17772 }, { "epoch": 1.07, "learning_rate": 4.909691729663664e-05, "loss": 1.0603, "step": 17774 }, { "epoch": 1.07, "learning_rate": 4.909540838652242e-05, "loss": 1.1378, "step": 17776 }, { "epoch": 1.07, "learning_rate": 4.90938994764082e-05, "loss": 1.1372, "step": 17778 }, { "epoch": 1.07, "learning_rate": 4.9092390566293964e-05, "loss": 1.2497, "step": 17780 }, { "epoch": 1.07, "learning_rate": 4.9090881656179744e-05, "loss": 1.1421, "step": 17782 }, { "epoch": 1.07, "learning_rate": 4.9089372746065524e-05, "loss": 1.0089, "step": 17784 }, { "epoch": 1.07, "learning_rate": 4.908786383595129e-05, "loss": 1.068, "step": 17786 }, { "epoch": 1.07, "learning_rate": 4.908635492583707e-05, "loss": 1.065, "step": 17788 }, { "epoch": 1.07, "learning_rate": 4.908484601572284e-05, "loss": 1.1471, "step": 17790 }, { "epoch": 1.07, "learning_rate": 4.908333710560862e-05, "loss": 1.1995, "step": 17792 }, { "epoch": 1.07, "learning_rate": 4.9081828195494396e-05, "loss": 0.9754, "step": 17794 }, { "epoch": 1.07, "learning_rate": 4.908031928538017e-05, "loss": 1.3069, "step": 17796 }, { "epoch": 1.07, "learning_rate": 4.907881037526595e-05, "loss": 1.5173, "step": 17798 }, { "epoch": 1.07, "learning_rate": 4.907730146515173e-05, "loss": 1.011, "step": 17800 }, { "epoch": 1.07, "learning_rate": 4.9075792555037495e-05, "loss": 1.097, "step": 17802 }, { "epoch": 1.07, "learning_rate": 4.9074283644923275e-05, "loss": 1.2374, "step": 17804 }, { "epoch": 1.07, "learning_rate": 4.907277473480905e-05, "loss": 1.0391, "step": 17806 }, { "epoch": 1.07, "learning_rate": 4.907126582469482e-05, "loss": 1.0324, "step": 17808 }, { "epoch": 1.07, "learning_rate": 4.90697569145806e-05, "loss": 1.1863, "step": 17810 }, { "epoch": 1.08, "learning_rate": 4.9068248004466374e-05, "loss": 1.0759, "step": 17812 }, { "epoch": 1.08, "learning_rate": 4.9066739094352154e-05, "loss": 1.4118, "step": 17814 }, { "epoch": 1.08, "learning_rate": 4.906523018423793e-05, "loss": 0.9594, "step": 17816 }, { "epoch": 1.08, "learning_rate": 4.90637212741237e-05, "loss": 1.4005, "step": 17818 }, { "epoch": 1.08, "learning_rate": 4.906221236400948e-05, "loss": 1.0635, "step": 17820 }, { "epoch": 1.08, "learning_rate": 4.9060703453895254e-05, "loss": 0.8268, "step": 17822 }, { "epoch": 1.08, "learning_rate": 4.905919454378103e-05, "loss": 1.0307, "step": 17824 }, { "epoch": 1.08, "learning_rate": 4.905768563366681e-05, "loss": 0.8553, "step": 17826 }, { "epoch": 1.08, "learning_rate": 4.905617672355258e-05, "loss": 1.1275, "step": 17828 }, { "epoch": 1.08, "learning_rate": 4.905466781343835e-05, "loss": 1.2628, "step": 17830 }, { "epoch": 1.08, "learning_rate": 4.905315890332413e-05, "loss": 1.2609, "step": 17832 }, { "epoch": 1.08, "learning_rate": 4.9051649993209906e-05, "loss": 1.0855, "step": 17834 }, { "epoch": 1.08, "learning_rate": 4.905014108309568e-05, "loss": 1.0799, "step": 17836 }, { "epoch": 1.08, "learning_rate": 4.904863217298146e-05, "loss": 0.764, "step": 17838 }, { "epoch": 1.08, "learning_rate": 4.904712326286723e-05, "loss": 1.3039, "step": 17840 }, { "epoch": 1.08, "learning_rate": 4.904561435275301e-05, "loss": 1.1099, "step": 17842 }, { "epoch": 1.08, "learning_rate": 4.9044105442638785e-05, "loss": 1.0059, "step": 17844 }, { "epoch": 1.08, "learning_rate": 4.904259653252456e-05, "loss": 1.1107, "step": 17846 }, { "epoch": 1.08, "learning_rate": 4.904108762241034e-05, "loss": 1.1609, "step": 17848 }, { "epoch": 1.08, "learning_rate": 4.903957871229611e-05, "loss": 0.8438, "step": 17850 }, { "epoch": 1.08, "learning_rate": 4.9038069802181884e-05, "loss": 1.1709, "step": 17852 }, { "epoch": 1.08, "learning_rate": 4.9036560892067664e-05, "loss": 0.8439, "step": 17854 }, { "epoch": 1.08, "learning_rate": 4.903505198195344e-05, "loss": 1.0172, "step": 17856 }, { "epoch": 1.08, "learning_rate": 4.903354307183921e-05, "loss": 0.9453, "step": 17858 }, { "epoch": 1.08, "learning_rate": 4.903203416172499e-05, "loss": 1.2443, "step": 17860 }, { "epoch": 1.08, "learning_rate": 4.9030525251610764e-05, "loss": 1.2341, "step": 17862 }, { "epoch": 1.08, "learning_rate": 4.9029016341496543e-05, "loss": 0.7718, "step": 17864 }, { "epoch": 1.08, "learning_rate": 4.9027507431382317e-05, "loss": 0.9757, "step": 17866 }, { "epoch": 1.08, "learning_rate": 4.902599852126809e-05, "loss": 1.0455, "step": 17868 }, { "epoch": 1.08, "learning_rate": 4.902448961115387e-05, "loss": 1.2589, "step": 17870 }, { "epoch": 1.08, "learning_rate": 4.902373515609675e-05, "loss": 1.4118, "step": 17872 }, { "epoch": 1.08, "learning_rate": 4.9022226245982526e-05, "loss": 1.0021, "step": 17874 }, { "epoch": 1.08, "learning_rate": 4.9020717335868306e-05, "loss": 1.2386, "step": 17876 }, { "epoch": 1.08, "learning_rate": 4.901920842575408e-05, "loss": 1.1943, "step": 17878 }, { "epoch": 1.08, "learning_rate": 4.901769951563985e-05, "loss": 1.0872, "step": 17880 }, { "epoch": 1.08, "learning_rate": 4.901619060552563e-05, "loss": 1.3789, "step": 17882 }, { "epoch": 1.08, "learning_rate": 4.9014681695411405e-05, "loss": 1.2558, "step": 17884 }, { "epoch": 1.08, "learning_rate": 4.9013172785297185e-05, "loss": 0.9561, "step": 17886 }, { "epoch": 1.08, "learning_rate": 4.901166387518296e-05, "loss": 1.0966, "step": 17888 }, { "epoch": 1.08, "learning_rate": 4.901015496506873e-05, "loss": 1.1254, "step": 17890 }, { "epoch": 1.08, "learning_rate": 4.900864605495451e-05, "loss": 1.0724, "step": 17892 }, { "epoch": 1.08, "learning_rate": 4.9007137144840284e-05, "loss": 1.0976, "step": 17894 }, { "epoch": 1.08, "learning_rate": 4.900562823472606e-05, "loss": 1.0306, "step": 17896 }, { "epoch": 1.08, "learning_rate": 4.900411932461184e-05, "loss": 1.1126, "step": 17898 }, { "epoch": 1.08, "learning_rate": 4.900261041449761e-05, "loss": 1.1506, "step": 17900 }, { "epoch": 1.08, "learning_rate": 4.9001101504383384e-05, "loss": 1.2567, "step": 17902 }, { "epoch": 1.08, "learning_rate": 4.8999592594269163e-05, "loss": 1.428, "step": 17904 }, { "epoch": 1.08, "learning_rate": 4.8998083684154937e-05, "loss": 0.8659, "step": 17906 }, { "epoch": 1.08, "learning_rate": 4.8996574774040716e-05, "loss": 0.9573, "step": 17908 }, { "epoch": 1.08, "learning_rate": 4.899506586392649e-05, "loss": 1.1702, "step": 17910 }, { "epoch": 1.08, "learning_rate": 4.899355695381226e-05, "loss": 1.0897, "step": 17912 }, { "epoch": 1.08, "learning_rate": 4.899204804369804e-05, "loss": 0.8729, "step": 17914 }, { "epoch": 1.08, "learning_rate": 4.8990539133583816e-05, "loss": 1.4654, "step": 17916 }, { "epoch": 1.08, "learning_rate": 4.898903022346959e-05, "loss": 0.693, "step": 17918 }, { "epoch": 1.08, "learning_rate": 4.898752131335537e-05, "loss": 1.0429, "step": 17920 }, { "epoch": 1.08, "learning_rate": 4.898601240324114e-05, "loss": 1.2586, "step": 17922 }, { "epoch": 1.08, "learning_rate": 4.8984503493126915e-05, "loss": 1.3219, "step": 17924 }, { "epoch": 1.08, "learning_rate": 4.8982994583012695e-05, "loss": 1.2548, "step": 17926 }, { "epoch": 1.08, "learning_rate": 4.898148567289847e-05, "loss": 0.9733, "step": 17928 }, { "epoch": 1.08, "learning_rate": 4.897997676278424e-05, "loss": 0.9762, "step": 17930 }, { "epoch": 1.08, "learning_rate": 4.897846785267002e-05, "loss": 1.1779, "step": 17932 }, { "epoch": 1.08, "learning_rate": 4.8976958942555794e-05, "loss": 1.3696, "step": 17934 }, { "epoch": 1.08, "learning_rate": 4.8975450032441574e-05, "loss": 1.0356, "step": 17936 }, { "epoch": 1.08, "learning_rate": 4.897394112232734e-05, "loss": 0.9654, "step": 17938 }, { "epoch": 1.08, "learning_rate": 4.897243221221312e-05, "loss": 1.3697, "step": 17940 }, { "epoch": 1.08, "learning_rate": 4.89709233020989e-05, "loss": 0.8805, "step": 17942 }, { "epoch": 1.08, "learning_rate": 4.896941439198467e-05, "loss": 0.9694, "step": 17944 }, { "epoch": 1.08, "learning_rate": 4.8967905481870447e-05, "loss": 0.7978, "step": 17946 }, { "epoch": 1.08, "learning_rate": 4.8966396571756226e-05, "loss": 1.1901, "step": 17948 }, { "epoch": 1.08, "learning_rate": 4.8964887661642e-05, "loss": 1.2487, "step": 17950 }, { "epoch": 1.08, "learning_rate": 4.896337875152777e-05, "loss": 1.2699, "step": 17952 }, { "epoch": 1.08, "learning_rate": 4.8961869841413546e-05, "loss": 1.2376, "step": 17954 }, { "epoch": 1.08, "learning_rate": 4.8960360931299326e-05, "loss": 1.2662, "step": 17956 }, { "epoch": 1.08, "learning_rate": 4.89588520211851e-05, "loss": 1.3784, "step": 17958 }, { "epoch": 1.08, "learning_rate": 4.895734311107087e-05, "loss": 1.2555, "step": 17960 }, { "epoch": 1.08, "learning_rate": 4.895583420095665e-05, "loss": 1.1578, "step": 17962 }, { "epoch": 1.08, "learning_rate": 4.895432529084243e-05, "loss": 1.0997, "step": 17964 }, { "epoch": 1.08, "learning_rate": 4.89528163807282e-05, "loss": 1.3313, "step": 17966 }, { "epoch": 1.08, "learning_rate": 4.895130747061398e-05, "loss": 1.2143, "step": 17968 }, { "epoch": 1.08, "learning_rate": 4.894979856049975e-05, "loss": 1.2104, "step": 17970 }, { "epoch": 1.08, "learning_rate": 4.894828965038553e-05, "loss": 0.9589, "step": 17972 }, { "epoch": 1.08, "learning_rate": 4.8946780740271304e-05, "loss": 0.9839, "step": 17974 }, { "epoch": 1.08, "learning_rate": 4.894527183015708e-05, "loss": 1.2741, "step": 17976 }, { "epoch": 1.09, "learning_rate": 4.894376292004286e-05, "loss": 0.9525, "step": 17978 }, { "epoch": 1.09, "learning_rate": 4.894225400992863e-05, "loss": 1.2273, "step": 17980 }, { "epoch": 1.09, "learning_rate": 4.8940745099814403e-05, "loss": 1.3163, "step": 17982 }, { "epoch": 1.09, "learning_rate": 4.893923618970018e-05, "loss": 1.1036, "step": 17984 }, { "epoch": 1.09, "learning_rate": 4.8937727279585956e-05, "loss": 0.9837, "step": 17986 }, { "epoch": 1.09, "learning_rate": 4.893621836947173e-05, "loss": 1.238, "step": 17988 }, { "epoch": 1.09, "learning_rate": 4.893470945935751e-05, "loss": 1.1902, "step": 17990 }, { "epoch": 1.09, "learning_rate": 4.893320054924328e-05, "loss": 1.3324, "step": 17992 }, { "epoch": 1.09, "learning_rate": 4.8931691639129056e-05, "loss": 1.0147, "step": 17994 }, { "epoch": 1.09, "learning_rate": 4.8930182729014836e-05, "loss": 1.236, "step": 17996 }, { "epoch": 1.09, "learning_rate": 4.892867381890061e-05, "loss": 1.0106, "step": 17998 }, { "epoch": 1.09, "learning_rate": 4.892716490878639e-05, "loss": 1.1026, "step": 18000 }, { "epoch": 1.09, "learning_rate": 4.892565599867216e-05, "loss": 0.8904, "step": 18002 }, { "epoch": 1.09, "learning_rate": 4.8924147088557935e-05, "loss": 0.9962, "step": 18004 }, { "epoch": 1.09, "learning_rate": 4.8922638178443715e-05, "loss": 1.2594, "step": 18006 }, { "epoch": 1.09, "learning_rate": 4.892112926832949e-05, "loss": 1.0706, "step": 18008 }, { "epoch": 1.09, "learning_rate": 4.891962035821526e-05, "loss": 1.0427, "step": 18010 }, { "epoch": 1.09, "learning_rate": 4.891811144810104e-05, "loss": 0.8596, "step": 18012 }, { "epoch": 1.09, "learning_rate": 4.8916602537986814e-05, "loss": 1.0573, "step": 18014 }, { "epoch": 1.09, "learning_rate": 4.891509362787259e-05, "loss": 1.1535, "step": 18016 }, { "epoch": 1.09, "learning_rate": 4.891358471775837e-05, "loss": 1.2749, "step": 18018 }, { "epoch": 1.09, "learning_rate": 4.891207580764414e-05, "loss": 1.4113, "step": 18020 }, { "epoch": 1.09, "learning_rate": 4.891056689752992e-05, "loss": 1.0434, "step": 18022 }, { "epoch": 1.09, "learning_rate": 4.890905798741569e-05, "loss": 1.2688, "step": 18024 }, { "epoch": 1.09, "learning_rate": 4.8907549077301466e-05, "loss": 1.0663, "step": 18026 }, { "epoch": 1.09, "learning_rate": 4.8906040167187246e-05, "loss": 1.0429, "step": 18028 }, { "epoch": 1.09, "learning_rate": 4.890453125707302e-05, "loss": 1.0658, "step": 18030 }, { "epoch": 1.09, "learning_rate": 4.890302234695879e-05, "loss": 1.1588, "step": 18032 }, { "epoch": 1.09, "learning_rate": 4.890151343684457e-05, "loss": 0.8571, "step": 18034 }, { "epoch": 1.09, "learning_rate": 4.8900004526730346e-05, "loss": 1.1511, "step": 18036 }, { "epoch": 1.09, "learning_rate": 4.889849561661612e-05, "loss": 1.1735, "step": 18038 }, { "epoch": 1.09, "learning_rate": 4.88969867065019e-05, "loss": 1.0359, "step": 18040 }, { "epoch": 1.09, "learning_rate": 4.889547779638767e-05, "loss": 1.2009, "step": 18042 }, { "epoch": 1.09, "learning_rate": 4.8893968886273445e-05, "loss": 1.0326, "step": 18044 }, { "epoch": 1.09, "learning_rate": 4.8892459976159225e-05, "loss": 1.1095, "step": 18046 }, { "epoch": 1.09, "learning_rate": 4.8890951066045e-05, "loss": 1.3209, "step": 18048 }, { "epoch": 1.09, "learning_rate": 4.888944215593078e-05, "loss": 1.1855, "step": 18050 }, { "epoch": 1.09, "learning_rate": 4.8887933245816544e-05, "loss": 1.2335, "step": 18052 }, { "epoch": 1.09, "learning_rate": 4.8886424335702324e-05, "loss": 1.1707, "step": 18054 }, { "epoch": 1.09, "learning_rate": 4.8884915425588104e-05, "loss": 1.0615, "step": 18056 }, { "epoch": 1.09, "learning_rate": 4.888340651547387e-05, "loss": 1.1967, "step": 18058 }, { "epoch": 1.09, "learning_rate": 4.888189760535965e-05, "loss": 0.9377, "step": 18060 }, { "epoch": 1.09, "learning_rate": 4.888038869524543e-05, "loss": 0.9806, "step": 18062 }, { "epoch": 1.09, "learning_rate": 4.88788797851312e-05, "loss": 1.1243, "step": 18064 }, { "epoch": 1.09, "learning_rate": 4.8877370875016976e-05, "loss": 1.1175, "step": 18066 }, { "epoch": 1.09, "learning_rate": 4.887586196490275e-05, "loss": 1.2074, "step": 18068 }, { "epoch": 1.09, "learning_rate": 4.887435305478853e-05, "loss": 1.28, "step": 18070 }, { "epoch": 1.09, "learning_rate": 4.887284414467431e-05, "loss": 1.1093, "step": 18072 }, { "epoch": 1.09, "learning_rate": 4.8871335234560076e-05, "loss": 1.4407, "step": 18074 }, { "epoch": 1.09, "learning_rate": 4.8869826324445856e-05, "loss": 0.7772, "step": 18076 }, { "epoch": 1.09, "learning_rate": 4.8868317414331635e-05, "loss": 0.9636, "step": 18078 }, { "epoch": 1.09, "learning_rate": 4.88668085042174e-05, "loss": 1.1706, "step": 18080 }, { "epoch": 1.09, "learning_rate": 4.886529959410318e-05, "loss": 1.1353, "step": 18082 }, { "epoch": 1.09, "learning_rate": 4.8863790683988955e-05, "loss": 1.1014, "step": 18084 }, { "epoch": 1.09, "learning_rate": 4.8862281773874735e-05, "loss": 1.5226, "step": 18086 }, { "epoch": 1.09, "learning_rate": 4.886077286376051e-05, "loss": 0.9334, "step": 18088 }, { "epoch": 1.09, "learning_rate": 4.885926395364628e-05, "loss": 1.0591, "step": 18090 }, { "epoch": 1.09, "learning_rate": 4.885775504353206e-05, "loss": 1.5572, "step": 18092 }, { "epoch": 1.09, "learning_rate": 4.8856246133417834e-05, "loss": 0.9185, "step": 18094 }, { "epoch": 1.09, "learning_rate": 4.885473722330361e-05, "loss": 1.3369, "step": 18096 }, { "epoch": 1.09, "learning_rate": 4.885322831318939e-05, "loss": 1.293, "step": 18098 }, { "epoch": 1.09, "learning_rate": 4.885171940307516e-05, "loss": 0.8546, "step": 18100 }, { "epoch": 1.09, "learning_rate": 4.885021049296093e-05, "loss": 1.4778, "step": 18102 }, { "epoch": 1.09, "learning_rate": 4.884870158284671e-05, "loss": 1.0292, "step": 18104 }, { "epoch": 1.09, "learning_rate": 4.8847192672732486e-05, "loss": 1.4248, "step": 18106 }, { "epoch": 1.09, "learning_rate": 4.884568376261826e-05, "loss": 0.9365, "step": 18108 }, { "epoch": 1.09, "learning_rate": 4.884417485250404e-05, "loss": 0.6376, "step": 18110 }, { "epoch": 1.09, "learning_rate": 4.884266594238981e-05, "loss": 1.1006, "step": 18112 }, { "epoch": 1.09, "learning_rate": 4.884115703227559e-05, "loss": 1.3976, "step": 18114 }, { "epoch": 1.09, "learning_rate": 4.8839648122161365e-05, "loss": 1.2307, "step": 18116 }, { "epoch": 1.09, "learning_rate": 4.883813921204714e-05, "loss": 0.9114, "step": 18118 }, { "epoch": 1.09, "learning_rate": 4.883663030193292e-05, "loss": 1.1408, "step": 18120 }, { "epoch": 1.09, "learning_rate": 4.883512139181869e-05, "loss": 1.01, "step": 18122 }, { "epoch": 1.09, "learning_rate": 4.8833612481704465e-05, "loss": 1.3211, "step": 18124 }, { "epoch": 1.09, "learning_rate": 4.8832103571590245e-05, "loss": 1.1079, "step": 18126 }, { "epoch": 1.09, "learning_rate": 4.883059466147602e-05, "loss": 0.8548, "step": 18128 }, { "epoch": 1.09, "learning_rate": 4.882908575136179e-05, "loss": 1.2098, "step": 18130 }, { "epoch": 1.09, "learning_rate": 4.882757684124757e-05, "loss": 1.0909, "step": 18132 }, { "epoch": 1.09, "learning_rate": 4.8826067931133344e-05, "loss": 0.9365, "step": 18134 }, { "epoch": 1.09, "learning_rate": 4.8824559021019124e-05, "loss": 1.1657, "step": 18136 }, { "epoch": 1.09, "learning_rate": 4.88230501109049e-05, "loss": 0.9278, "step": 18138 }, { "epoch": 1.09, "learning_rate": 4.882154120079067e-05, "loss": 1.0943, "step": 18140 }, { "epoch": 1.1, "learning_rate": 4.882003229067645e-05, "loss": 0.9809, "step": 18142 }, { "epoch": 1.1, "learning_rate": 4.881852338056222e-05, "loss": 1.4709, "step": 18144 }, { "epoch": 1.1, "learning_rate": 4.8817014470447996e-05, "loss": 1.284, "step": 18146 }, { "epoch": 1.1, "learning_rate": 4.8815505560333776e-05, "loss": 0.9152, "step": 18148 }, { "epoch": 1.1, "learning_rate": 4.881399665021955e-05, "loss": 0.9301, "step": 18150 }, { "epoch": 1.1, "learning_rate": 4.881248774010532e-05, "loss": 1.0152, "step": 18152 }, { "epoch": 1.1, "learning_rate": 4.88109788299911e-05, "loss": 0.9343, "step": 18154 }, { "epoch": 1.1, "learning_rate": 4.8809469919876875e-05, "loss": 1.2082, "step": 18156 }, { "epoch": 1.1, "learning_rate": 4.8808715464819765e-05, "loss": 1.2719, "step": 18158 }, { "epoch": 1.1, "learning_rate": 4.880720655470554e-05, "loss": 0.8855, "step": 18160 }, { "epoch": 1.1, "learning_rate": 4.880569764459131e-05, "loss": 0.925, "step": 18162 }, { "epoch": 1.1, "learning_rate": 4.880418873447709e-05, "loss": 0.991, "step": 18164 }, { "epoch": 1.1, "learning_rate": 4.8802679824362865e-05, "loss": 1.2085, "step": 18166 }, { "epoch": 1.1, "learning_rate": 4.880117091424864e-05, "loss": 0.8654, "step": 18168 }, { "epoch": 1.1, "learning_rate": 4.879966200413442e-05, "loss": 0.8973, "step": 18170 }, { "epoch": 1.1, "learning_rate": 4.879815309402019e-05, "loss": 1.2649, "step": 18172 }, { "epoch": 1.1, "learning_rate": 4.8796644183905964e-05, "loss": 0.8938, "step": 18174 }, { "epoch": 1.1, "learning_rate": 4.8795135273791744e-05, "loss": 1.0537, "step": 18176 }, { "epoch": 1.1, "learning_rate": 4.879362636367752e-05, "loss": 1.4314, "step": 18178 }, { "epoch": 1.1, "learning_rate": 4.87921174535633e-05, "loss": 1.2057, "step": 18180 }, { "epoch": 1.1, "learning_rate": 4.879060854344907e-05, "loss": 1.1682, "step": 18182 }, { "epoch": 1.1, "learning_rate": 4.878909963333484e-05, "loss": 1.3064, "step": 18184 }, { "epoch": 1.1, "learning_rate": 4.878759072322062e-05, "loss": 0.9816, "step": 18186 }, { "epoch": 1.1, "learning_rate": 4.8786081813106396e-05, "loss": 1.4107, "step": 18188 }, { "epoch": 1.1, "learning_rate": 4.878457290299217e-05, "loss": 1.2913, "step": 18190 }, { "epoch": 1.1, "learning_rate": 4.878306399287795e-05, "loss": 1.1634, "step": 18192 }, { "epoch": 1.1, "learning_rate": 4.878155508276372e-05, "loss": 1.0898, "step": 18194 }, { "epoch": 1.1, "learning_rate": 4.8780046172649495e-05, "loss": 1.165, "step": 18196 }, { "epoch": 1.1, "learning_rate": 4.8778537262535275e-05, "loss": 0.8604, "step": 18198 }, { "epoch": 1.1, "learning_rate": 4.877702835242105e-05, "loss": 1.844, "step": 18200 }, { "epoch": 1.1, "learning_rate": 4.877551944230682e-05, "loss": 1.2838, "step": 18202 }, { "epoch": 1.1, "learning_rate": 4.87740105321926e-05, "loss": 1.3805, "step": 18204 }, { "epoch": 1.1, "learning_rate": 4.8772501622078375e-05, "loss": 1.0596, "step": 18206 }, { "epoch": 1.1, "learning_rate": 4.8770992711964155e-05, "loss": 0.7877, "step": 18208 }, { "epoch": 1.1, "learning_rate": 4.876948380184993e-05, "loss": 1.1898, "step": 18210 }, { "epoch": 1.1, "learning_rate": 4.87679748917357e-05, "loss": 1.2161, "step": 18212 }, { "epoch": 1.1, "learning_rate": 4.876646598162148e-05, "loss": 1.2248, "step": 18214 }, { "epoch": 1.1, "learning_rate": 4.876495707150725e-05, "loss": 1.517, "step": 18216 }, { "epoch": 1.1, "learning_rate": 4.876344816139303e-05, "loss": 1.4296, "step": 18218 }, { "epoch": 1.1, "learning_rate": 4.876193925127881e-05, "loss": 0.9816, "step": 18220 }, { "epoch": 1.1, "learning_rate": 4.876043034116458e-05, "loss": 1.2405, "step": 18222 }, { "epoch": 1.1, "learning_rate": 4.875892143105035e-05, "loss": 1.1228, "step": 18224 }, { "epoch": 1.1, "learning_rate": 4.875741252093613e-05, "loss": 0.9124, "step": 18226 }, { "epoch": 1.1, "learning_rate": 4.8755903610821906e-05, "loss": 1.0562, "step": 18228 }, { "epoch": 1.1, "learning_rate": 4.875439470070768e-05, "loss": 1.4552, "step": 18230 }, { "epoch": 1.1, "learning_rate": 4.875288579059345e-05, "loss": 0.7347, "step": 18232 }, { "epoch": 1.1, "learning_rate": 4.875137688047923e-05, "loss": 1.2887, "step": 18234 }, { "epoch": 1.1, "learning_rate": 4.874986797036501e-05, "loss": 1.352, "step": 18236 }, { "epoch": 1.1, "learning_rate": 4.874835906025078e-05, "loss": 0.9798, "step": 18238 }, { "epoch": 1.1, "learning_rate": 4.874685015013656e-05, "loss": 1.2645, "step": 18240 }, { "epoch": 1.1, "learning_rate": 4.874534124002234e-05, "loss": 0.9157, "step": 18242 }, { "epoch": 1.1, "learning_rate": 4.874383232990811e-05, "loss": 1.4031, "step": 18244 }, { "epoch": 1.1, "learning_rate": 4.8742323419793885e-05, "loss": 0.9932, "step": 18246 }, { "epoch": 1.1, "learning_rate": 4.874081450967966e-05, "loss": 1.3566, "step": 18248 }, { "epoch": 1.1, "learning_rate": 4.873930559956544e-05, "loss": 1.0481, "step": 18250 }, { "epoch": 1.1, "learning_rate": 4.873779668945121e-05, "loss": 1.1902, "step": 18252 }, { "epoch": 1.1, "learning_rate": 4.8736287779336984e-05, "loss": 1.3658, "step": 18254 }, { "epoch": 1.1, "learning_rate": 4.8734778869222764e-05, "loss": 1.231, "step": 18256 }, { "epoch": 1.1, "learning_rate": 4.8733269959108544e-05, "loss": 0.8906, "step": 18258 }, { "epoch": 1.1, "learning_rate": 4.873176104899431e-05, "loss": 1.0773, "step": 18260 }, { "epoch": 1.1, "learning_rate": 4.873025213888009e-05, "loss": 1.225, "step": 18262 }, { "epoch": 1.1, "learning_rate": 4.872874322876586e-05, "loss": 1.2421, "step": 18264 }, { "epoch": 1.1, "learning_rate": 4.8727234318651636e-05, "loss": 1.0363, "step": 18266 }, { "epoch": 1.1, "learning_rate": 4.8725725408537416e-05, "loss": 1.0388, "step": 18268 }, { "epoch": 1.1, "learning_rate": 4.872421649842319e-05, "loss": 0.9599, "step": 18270 }, { "epoch": 1.1, "learning_rate": 4.872270758830897e-05, "loss": 1.3011, "step": 18272 }, { "epoch": 1.1, "learning_rate": 4.872119867819474e-05, "loss": 0.7838, "step": 18274 }, { "epoch": 1.1, "learning_rate": 4.8719689768080515e-05, "loss": 1.2537, "step": 18276 }, { "epoch": 1.1, "learning_rate": 4.8718180857966295e-05, "loss": 1.3508, "step": 18278 }, { "epoch": 1.1, "learning_rate": 4.871667194785207e-05, "loss": 1.0997, "step": 18280 }, { "epoch": 1.1, "learning_rate": 4.871516303773784e-05, "loss": 1.3398, "step": 18282 }, { "epoch": 1.1, "learning_rate": 4.871365412762362e-05, "loss": 1.1936, "step": 18284 }, { "epoch": 1.1, "learning_rate": 4.8712145217509394e-05, "loss": 0.7444, "step": 18286 }, { "epoch": 1.1, "learning_rate": 4.871063630739517e-05, "loss": 1.2079, "step": 18288 }, { "epoch": 1.1, "learning_rate": 4.870912739728095e-05, "loss": 1.3043, "step": 18290 }, { "epoch": 1.1, "learning_rate": 4.870761848716672e-05, "loss": 1.0154, "step": 18292 }, { "epoch": 1.1, "learning_rate": 4.87061095770525e-05, "loss": 1.2895, "step": 18294 }, { "epoch": 1.1, "learning_rate": 4.8704600666938274e-05, "loss": 0.9468, "step": 18296 }, { "epoch": 1.1, "learning_rate": 4.870309175682405e-05, "loss": 0.9337, "step": 18298 }, { "epoch": 1.1, "learning_rate": 4.870158284670983e-05, "loss": 0.8317, "step": 18300 }, { "epoch": 1.1, "learning_rate": 4.87000739365956e-05, "loss": 0.9363, "step": 18302 }, { "epoch": 1.1, "learning_rate": 4.869856502648137e-05, "loss": 0.8483, "step": 18304 }, { "epoch": 1.1, "learning_rate": 4.869705611636715e-05, "loss": 0.9882, "step": 18306 }, { "epoch": 1.11, "learning_rate": 4.8695547206252926e-05, "loss": 1.4028, "step": 18308 }, { "epoch": 1.11, "learning_rate": 4.86940382961387e-05, "loss": 0.9819, "step": 18310 }, { "epoch": 1.11, "learning_rate": 4.869252938602448e-05, "loss": 0.9934, "step": 18312 }, { "epoch": 1.11, "learning_rate": 4.869102047591025e-05, "loss": 0.9437, "step": 18314 }, { "epoch": 1.11, "learning_rate": 4.8689511565796025e-05, "loss": 1.1604, "step": 18316 }, { "epoch": 1.11, "learning_rate": 4.8688002655681805e-05, "loss": 1.1718, "step": 18318 }, { "epoch": 1.11, "learning_rate": 4.868649374556758e-05, "loss": 0.8515, "step": 18320 }, { "epoch": 1.11, "learning_rate": 4.868498483545336e-05, "loss": 1.3886, "step": 18322 }, { "epoch": 1.11, "learning_rate": 4.868347592533913e-05, "loss": 1.2829, "step": 18324 }, { "epoch": 1.11, "learning_rate": 4.8681967015224904e-05, "loss": 1.2795, "step": 18326 }, { "epoch": 1.11, "learning_rate": 4.8680458105110684e-05, "loss": 1.1302, "step": 18328 }, { "epoch": 1.11, "learning_rate": 4.867894919499645e-05, "loss": 1.2544, "step": 18330 }, { "epoch": 1.11, "learning_rate": 4.867744028488223e-05, "loss": 1.4128, "step": 18332 }, { "epoch": 1.11, "learning_rate": 4.867593137476801e-05, "loss": 1.3235, "step": 18334 }, { "epoch": 1.11, "learning_rate": 4.8674422464653784e-05, "loss": 0.8146, "step": 18336 }, { "epoch": 1.11, "learning_rate": 4.867291355453956e-05, "loss": 0.8685, "step": 18338 }, { "epoch": 1.11, "learning_rate": 4.867140464442534e-05, "loss": 1.3052, "step": 18340 }, { "epoch": 1.11, "learning_rate": 4.866989573431111e-05, "loss": 1.2301, "step": 18342 }, { "epoch": 1.11, "learning_rate": 4.866838682419689e-05, "loss": 1.2177, "step": 18344 }, { "epoch": 1.11, "learning_rate": 4.8666877914082656e-05, "loss": 0.8789, "step": 18346 }, { "epoch": 1.11, "learning_rate": 4.8665369003968436e-05, "loss": 1.0797, "step": 18348 }, { "epoch": 1.11, "learning_rate": 4.8663860093854216e-05, "loss": 1.2395, "step": 18350 }, { "epoch": 1.11, "learning_rate": 4.866235118373998e-05, "loss": 1.23, "step": 18352 }, { "epoch": 1.11, "learning_rate": 4.866084227362576e-05, "loss": 1.0877, "step": 18354 }, { "epoch": 1.11, "learning_rate": 4.865933336351154e-05, "loss": 1.0771, "step": 18356 }, { "epoch": 1.11, "learning_rate": 4.8657824453397315e-05, "loss": 1.0812, "step": 18358 }, { "epoch": 1.11, "learning_rate": 4.865631554328309e-05, "loss": 0.8926, "step": 18360 }, { "epoch": 1.11, "learning_rate": 4.865480663316886e-05, "loss": 1.0089, "step": 18362 }, { "epoch": 1.11, "learning_rate": 4.865329772305464e-05, "loss": 0.99, "step": 18364 }, { "epoch": 1.11, "learning_rate": 4.8651788812940414e-05, "loss": 1.0188, "step": 18366 }, { "epoch": 1.11, "learning_rate": 4.865027990282619e-05, "loss": 1.4544, "step": 18368 }, { "epoch": 1.11, "learning_rate": 4.864877099271197e-05, "loss": 1.2773, "step": 18370 }, { "epoch": 1.11, "learning_rate": 4.864726208259775e-05, "loss": 1.116, "step": 18372 }, { "epoch": 1.11, "learning_rate": 4.8645753172483514e-05, "loss": 1.0809, "step": 18374 }, { "epoch": 1.11, "learning_rate": 4.8644244262369294e-05, "loss": 1.0776, "step": 18376 }, { "epoch": 1.11, "learning_rate": 4.864273535225507e-05, "loss": 1.4856, "step": 18378 }, { "epoch": 1.11, "learning_rate": 4.864122644214084e-05, "loss": 1.1014, "step": 18380 }, { "epoch": 1.11, "learning_rate": 4.863971753202662e-05, "loss": 1.2132, "step": 18382 }, { "epoch": 1.11, "learning_rate": 4.863820862191239e-05, "loss": 1.1053, "step": 18384 }, { "epoch": 1.11, "learning_rate": 4.863669971179817e-05, "loss": 0.9807, "step": 18386 }, { "epoch": 1.11, "learning_rate": 4.8635190801683946e-05, "loss": 0.8473, "step": 18388 }, { "epoch": 1.11, "learning_rate": 4.863368189156972e-05, "loss": 1.0798, "step": 18390 }, { "epoch": 1.11, "learning_rate": 4.86321729814555e-05, "loss": 1.2027, "step": 18392 }, { "epoch": 1.11, "learning_rate": 4.863066407134127e-05, "loss": 0.7393, "step": 18394 }, { "epoch": 1.11, "learning_rate": 4.8629155161227045e-05, "loss": 1.1174, "step": 18396 }, { "epoch": 1.11, "learning_rate": 4.8627646251112825e-05, "loss": 0.9248, "step": 18398 }, { "epoch": 1.11, "learning_rate": 4.86261373409986e-05, "loss": 1.3105, "step": 18400 }, { "epoch": 1.11, "learning_rate": 4.862462843088437e-05, "loss": 1.2062, "step": 18402 }, { "epoch": 1.11, "learning_rate": 4.862311952077015e-05, "loss": 1.2221, "step": 18404 }, { "epoch": 1.11, "learning_rate": 4.8621610610655924e-05, "loss": 1.1157, "step": 18406 }, { "epoch": 1.11, "learning_rate": 4.8620101700541704e-05, "loss": 1.2141, "step": 18408 }, { "epoch": 1.11, "learning_rate": 4.861859279042748e-05, "loss": 1.3467, "step": 18410 }, { "epoch": 1.11, "learning_rate": 4.861708388031325e-05, "loss": 1.1679, "step": 18412 }, { "epoch": 1.11, "learning_rate": 4.861557497019903e-05, "loss": 1.3339, "step": 18414 }, { "epoch": 1.11, "learning_rate": 4.8614066060084804e-05, "loss": 1.0274, "step": 18416 }, { "epoch": 1.11, "learning_rate": 4.861255714997058e-05, "loss": 1.1319, "step": 18418 }, { "epoch": 1.11, "learning_rate": 4.8611048239856357e-05, "loss": 1.0816, "step": 18420 }, { "epoch": 1.11, "learning_rate": 4.860953932974213e-05, "loss": 1.0209, "step": 18422 }, { "epoch": 1.11, "learning_rate": 4.86080304196279e-05, "loss": 0.8775, "step": 18424 }, { "epoch": 1.11, "learning_rate": 4.860652150951368e-05, "loss": 1.2306, "step": 18426 }, { "epoch": 1.11, "learning_rate": 4.8605012599399456e-05, "loss": 1.2395, "step": 18428 }, { "epoch": 1.11, "learning_rate": 4.860350368928523e-05, "loss": 1.4279, "step": 18430 }, { "epoch": 1.11, "learning_rate": 4.860199477917101e-05, "loss": 0.993, "step": 18432 }, { "epoch": 1.11, "learning_rate": 4.860048586905678e-05, "loss": 1.0398, "step": 18434 }, { "epoch": 1.11, "learning_rate": 4.859897695894256e-05, "loss": 1.2517, "step": 18436 }, { "epoch": 1.11, "learning_rate": 4.8597468048828335e-05, "loss": 0.8808, "step": 18438 }, { "epoch": 1.11, "learning_rate": 4.859595913871411e-05, "loss": 1.0764, "step": 18440 }, { "epoch": 1.11, "learning_rate": 4.859445022859989e-05, "loss": 0.8546, "step": 18442 }, { "epoch": 1.11, "learning_rate": 4.8592941318485654e-05, "loss": 1.206, "step": 18444 }, { "epoch": 1.11, "learning_rate": 4.8591432408371434e-05, "loss": 1.1126, "step": 18446 }, { "epoch": 1.11, "learning_rate": 4.8589923498257214e-05, "loss": 0.9102, "step": 18448 }, { "epoch": 1.11, "learning_rate": 4.858841458814299e-05, "loss": 1.1539, "step": 18450 }, { "epoch": 1.11, "learning_rate": 4.858690567802876e-05, "loss": 1.5605, "step": 18452 }, { "epoch": 1.11, "learning_rate": 4.858539676791454e-05, "loss": 1.1229, "step": 18454 }, { "epoch": 1.11, "learning_rate": 4.8583887857800313e-05, "loss": 1.1193, "step": 18456 }, { "epoch": 1.11, "learning_rate": 4.858237894768609e-05, "loss": 1.1907, "step": 18458 }, { "epoch": 1.11, "learning_rate": 4.858087003757186e-05, "loss": 0.6858, "step": 18460 }, { "epoch": 1.11, "learning_rate": 4.857936112745764e-05, "loss": 1.088, "step": 18462 }, { "epoch": 1.11, "learning_rate": 4.857785221734342e-05, "loss": 1.1838, "step": 18464 }, { "epoch": 1.11, "learning_rate": 4.8576343307229186e-05, "loss": 1.1098, "step": 18466 }, { "epoch": 1.11, "learning_rate": 4.8574834397114966e-05, "loss": 1.4511, "step": 18468 }, { "epoch": 1.11, "learning_rate": 4.8573325487000746e-05, "loss": 1.4974, "step": 18470 }, { "epoch": 1.11, "learning_rate": 4.857181657688652e-05, "loss": 0.8337, "step": 18472 }, { "epoch": 1.12, "learning_rate": 4.857030766677229e-05, "loss": 1.179, "step": 18474 }, { "epoch": 1.12, "learning_rate": 4.8568798756658065e-05, "loss": 1.1524, "step": 18476 }, { "epoch": 1.12, "learning_rate": 4.8567289846543845e-05, "loss": 1.3109, "step": 18478 }, { "epoch": 1.12, "learning_rate": 4.856578093642962e-05, "loss": 1.039, "step": 18480 }, { "epoch": 1.12, "learning_rate": 4.856427202631539e-05, "loss": 1.1499, "step": 18482 }, { "epoch": 1.12, "learning_rate": 4.856276311620117e-05, "loss": 1.077, "step": 18484 }, { "epoch": 1.12, "learning_rate": 4.856125420608695e-05, "loss": 1.1456, "step": 18486 }, { "epoch": 1.12, "learning_rate": 4.855974529597272e-05, "loss": 1.3282, "step": 18488 }, { "epoch": 1.12, "learning_rate": 4.85582363858585e-05, "loss": 1.6183, "step": 18490 }, { "epoch": 1.12, "learning_rate": 4.855672747574427e-05, "loss": 0.9002, "step": 18492 }, { "epoch": 1.12, "learning_rate": 4.8555218565630043e-05, "loss": 1.0394, "step": 18494 }, { "epoch": 1.12, "learning_rate": 4.8553709655515823e-05, "loss": 0.8004, "step": 18496 }, { "epoch": 1.12, "learning_rate": 4.8552200745401597e-05, "loss": 1.0156, "step": 18498 }, { "epoch": 1.12, "learning_rate": 4.8550691835287376e-05, "loss": 0.8295, "step": 18500 }, { "epoch": 1.12, "learning_rate": 4.854918292517315e-05, "loss": 0.8249, "step": 18502 }, { "epoch": 1.12, "learning_rate": 4.854767401505892e-05, "loss": 1.079, "step": 18504 }, { "epoch": 1.12, "learning_rate": 4.85461651049447e-05, "loss": 0.9716, "step": 18506 }, { "epoch": 1.12, "learning_rate": 4.8544656194830476e-05, "loss": 1.0583, "step": 18508 }, { "epoch": 1.12, "learning_rate": 4.854314728471625e-05, "loss": 1.2148, "step": 18510 }, { "epoch": 1.12, "learning_rate": 4.854163837460203e-05, "loss": 0.9047, "step": 18512 }, { "epoch": 1.12, "learning_rate": 4.85401294644878e-05, "loss": 0.998, "step": 18514 }, { "epoch": 1.12, "learning_rate": 4.8538620554373575e-05, "loss": 0.9743, "step": 18516 }, { "epoch": 1.12, "learning_rate": 4.8537111644259355e-05, "loss": 0.7968, "step": 18518 }, { "epoch": 1.12, "learning_rate": 4.853560273414513e-05, "loss": 1.2062, "step": 18520 }, { "epoch": 1.12, "learning_rate": 4.853409382403091e-05, "loss": 0.9486, "step": 18522 }, { "epoch": 1.12, "learning_rate": 4.853258491391668e-05, "loss": 0.8231, "step": 18524 }, { "epoch": 1.12, "learning_rate": 4.8531076003802454e-05, "loss": 1.1468, "step": 18526 }, { "epoch": 1.12, "learning_rate": 4.8529567093688234e-05, "loss": 1.3064, "step": 18528 }, { "epoch": 1.12, "learning_rate": 4.852805818357401e-05, "loss": 1.5148, "step": 18530 }, { "epoch": 1.12, "learning_rate": 4.852654927345978e-05, "loss": 0.9981, "step": 18532 }, { "epoch": 1.12, "learning_rate": 4.852504036334556e-05, "loss": 1.0416, "step": 18534 }, { "epoch": 1.12, "learning_rate": 4.852353145323133e-05, "loss": 0.8589, "step": 18536 }, { "epoch": 1.12, "learning_rate": 4.8522022543117106e-05, "loss": 1.2094, "step": 18538 }, { "epoch": 1.12, "learning_rate": 4.8520513633002886e-05, "loss": 0.8132, "step": 18540 }, { "epoch": 1.12, "learning_rate": 4.851900472288866e-05, "loss": 0.9848, "step": 18542 }, { "epoch": 1.12, "learning_rate": 4.851749581277443e-05, "loss": 0.9008, "step": 18544 }, { "epoch": 1.12, "learning_rate": 4.851598690266021e-05, "loss": 0.9419, "step": 18546 }, { "epoch": 1.12, "learning_rate": 4.8514477992545986e-05, "loss": 1.1852, "step": 18548 }, { "epoch": 1.12, "learning_rate": 4.8512969082431766e-05, "loss": 1.0939, "step": 18550 }, { "epoch": 1.12, "learning_rate": 4.851146017231754e-05, "loss": 0.9702, "step": 18552 }, { "epoch": 1.12, "learning_rate": 4.850995126220331e-05, "loss": 1.1364, "step": 18554 }, { "epoch": 1.12, "learning_rate": 4.850844235208909e-05, "loss": 1.1634, "step": 18556 }, { "epoch": 1.12, "learning_rate": 4.850693344197486e-05, "loss": 1.0107, "step": 18558 }, { "epoch": 1.12, "learning_rate": 4.850542453186064e-05, "loss": 1.1056, "step": 18560 }, { "epoch": 1.12, "learning_rate": 4.850391562174642e-05, "loss": 0.8987, "step": 18562 }, { "epoch": 1.12, "learning_rate": 4.850240671163219e-05, "loss": 0.8588, "step": 18564 }, { "epoch": 1.12, "learning_rate": 4.8500897801517964e-05, "loss": 1.4721, "step": 18566 }, { "epoch": 1.12, "learning_rate": 4.8499388891403744e-05, "loss": 1.1986, "step": 18568 }, { "epoch": 1.12, "learning_rate": 4.849787998128952e-05, "loss": 1.1074, "step": 18570 }, { "epoch": 1.12, "learning_rate": 4.84963710711753e-05, "loss": 1.41, "step": 18572 }, { "epoch": 1.12, "learning_rate": 4.849486216106106e-05, "loss": 1.3416, "step": 18574 }, { "epoch": 1.12, "learning_rate": 4.849335325094684e-05, "loss": 0.8953, "step": 18576 }, { "epoch": 1.12, "learning_rate": 4.849184434083262e-05, "loss": 1.237, "step": 18578 }, { "epoch": 1.12, "learning_rate": 4.849033543071839e-05, "loss": 0.9212, "step": 18580 }, { "epoch": 1.12, "learning_rate": 4.848882652060417e-05, "loss": 1.1632, "step": 18582 }, { "epoch": 1.12, "learning_rate": 4.848731761048995e-05, "loss": 1.5897, "step": 18584 }, { "epoch": 1.12, "learning_rate": 4.848580870037572e-05, "loss": 1.1847, "step": 18586 }, { "epoch": 1.12, "learning_rate": 4.8484299790261496e-05, "loss": 0.8633, "step": 18588 }, { "epoch": 1.12, "learning_rate": 4.848279088014727e-05, "loss": 1.2979, "step": 18590 }, { "epoch": 1.12, "learning_rate": 4.848128197003305e-05, "loss": 0.8881, "step": 18592 }, { "epoch": 1.12, "learning_rate": 4.847977305991882e-05, "loss": 1.0478, "step": 18594 }, { "epoch": 1.12, "learning_rate": 4.8478264149804595e-05, "loss": 1.0967, "step": 18596 }, { "epoch": 1.12, "learning_rate": 4.8476755239690375e-05, "loss": 1.3467, "step": 18598 }, { "epoch": 1.12, "learning_rate": 4.8475246329576155e-05, "loss": 1.173, "step": 18600 }, { "epoch": 1.12, "learning_rate": 4.847373741946192e-05, "loss": 1.1899, "step": 18602 }, { "epoch": 1.12, "learning_rate": 4.84722285093477e-05, "loss": 1.1691, "step": 18604 }, { "epoch": 1.12, "learning_rate": 4.8470719599233474e-05, "loss": 1.1126, "step": 18606 }, { "epoch": 1.12, "learning_rate": 4.846921068911925e-05, "loss": 0.9645, "step": 18608 }, { "epoch": 1.12, "learning_rate": 4.846770177900503e-05, "loss": 1.1131, "step": 18610 }, { "epoch": 1.12, "learning_rate": 4.84661928688908e-05, "loss": 1.158, "step": 18612 }, { "epoch": 1.12, "learning_rate": 4.846468395877658e-05, "loss": 1.181, "step": 18614 }, { "epoch": 1.12, "learning_rate": 4.846317504866235e-05, "loss": 1.0401, "step": 18616 }, { "epoch": 1.12, "learning_rate": 4.8461666138548126e-05, "loss": 1.3845, "step": 18618 }, { "epoch": 1.12, "learning_rate": 4.8460157228433906e-05, "loss": 1.1871, "step": 18620 }, { "epoch": 1.12, "learning_rate": 4.845864831831968e-05, "loss": 1.1008, "step": 18622 }, { "epoch": 1.12, "learning_rate": 4.845713940820545e-05, "loss": 1.0919, "step": 18624 }, { "epoch": 1.12, "learning_rate": 4.845563049809123e-05, "loss": 0.7814, "step": 18626 }, { "epoch": 1.12, "learning_rate": 4.8454121587977006e-05, "loss": 1.0868, "step": 18628 }, { "epoch": 1.12, "learning_rate": 4.845261267786278e-05, "loss": 1.2513, "step": 18630 }, { "epoch": 1.12, "learning_rate": 4.845110376774856e-05, "loss": 1.1529, "step": 18632 }, { "epoch": 1.12, "learning_rate": 4.844959485763433e-05, "loss": 1.2257, "step": 18634 }, { "epoch": 1.12, "learning_rate": 4.844808594752011e-05, "loss": 1.1689, "step": 18636 }, { "epoch": 1.12, "learning_rate": 4.8446577037405885e-05, "loss": 0.9518, "step": 18638 }, { "epoch": 1.13, "learning_rate": 4.844506812729166e-05, "loss": 1.3383, "step": 18640 }, { "epoch": 1.13, "learning_rate": 4.844355921717744e-05, "loss": 1.1471, "step": 18642 }, { "epoch": 1.13, "learning_rate": 4.844205030706321e-05, "loss": 1.2994, "step": 18644 }, { "epoch": 1.13, "learning_rate": 4.8440541396948984e-05, "loss": 0.9178, "step": 18646 }, { "epoch": 1.13, "learning_rate": 4.8439032486834764e-05, "loss": 1.2587, "step": 18648 }, { "epoch": 1.13, "learning_rate": 4.843752357672054e-05, "loss": 1.0766, "step": 18650 }, { "epoch": 1.13, "learning_rate": 4.843601466660631e-05, "loss": 0.7552, "step": 18652 }, { "epoch": 1.13, "learning_rate": 4.843450575649209e-05, "loss": 1.2721, "step": 18654 }, { "epoch": 1.13, "learning_rate": 4.843299684637786e-05, "loss": 1.3088, "step": 18656 }, { "epoch": 1.13, "learning_rate": 4.8431487936263636e-05, "loss": 1.2289, "step": 18658 }, { "epoch": 1.13, "learning_rate": 4.8429979026149416e-05, "loss": 1.4426, "step": 18660 }, { "epoch": 1.13, "learning_rate": 4.842847011603519e-05, "loss": 1.3859, "step": 18662 }, { "epoch": 1.13, "learning_rate": 4.842696120592097e-05, "loss": 0.8822, "step": 18664 }, { "epoch": 1.13, "learning_rate": 4.842545229580674e-05, "loss": 0.9819, "step": 18666 }, { "epoch": 1.13, "learning_rate": 4.8423943385692515e-05, "loss": 1.2378, "step": 18668 }, { "epoch": 1.13, "learning_rate": 4.8422434475578295e-05, "loss": 1.371, "step": 18670 }, { "epoch": 1.13, "learning_rate": 4.842092556546407e-05, "loss": 1.1052, "step": 18672 }, { "epoch": 1.13, "learning_rate": 4.841941665534984e-05, "loss": 1.3878, "step": 18674 }, { "epoch": 1.13, "learning_rate": 4.841790774523562e-05, "loss": 0.8724, "step": 18676 }, { "epoch": 1.13, "learning_rate": 4.8416398835121395e-05, "loss": 1.0525, "step": 18678 }, { "epoch": 1.13, "learning_rate": 4.841488992500717e-05, "loss": 1.1394, "step": 18680 }, { "epoch": 1.13, "learning_rate": 4.841338101489295e-05, "loss": 1.021, "step": 18682 }, { "epoch": 1.13, "learning_rate": 4.841187210477872e-05, "loss": 1.5198, "step": 18684 }, { "epoch": 1.13, "learning_rate": 4.84103631946645e-05, "loss": 0.8697, "step": 18686 }, { "epoch": 1.13, "learning_rate": 4.840885428455027e-05, "loss": 1.1859, "step": 18688 }, { "epoch": 1.13, "learning_rate": 4.840734537443605e-05, "loss": 1.152, "step": 18690 }, { "epoch": 1.13, "learning_rate": 4.840583646432183e-05, "loss": 1.0973, "step": 18692 }, { "epoch": 1.13, "learning_rate": 4.840432755420759e-05, "loss": 1.0918, "step": 18694 }, { "epoch": 1.13, "learning_rate": 4.840281864409337e-05, "loss": 0.9473, "step": 18696 }, { "epoch": 1.13, "learning_rate": 4.840130973397915e-05, "loss": 0.7914, "step": 18698 }, { "epoch": 1.13, "learning_rate": 4.8399800823864926e-05, "loss": 1.2231, "step": 18700 }, { "epoch": 1.13, "learning_rate": 4.83982919137507e-05, "loss": 0.9558, "step": 18702 }, { "epoch": 1.13, "learning_rate": 4.839678300363647e-05, "loss": 1.7909, "step": 18704 }, { "epoch": 1.13, "learning_rate": 4.839527409352225e-05, "loss": 1.1017, "step": 18706 }, { "epoch": 1.13, "learning_rate": 4.8393765183408025e-05, "loss": 1.0658, "step": 18708 }, { "epoch": 1.13, "learning_rate": 4.83922562732938e-05, "loss": 1.2922, "step": 18710 }, { "epoch": 1.13, "learning_rate": 4.839074736317958e-05, "loss": 1.3015, "step": 18712 }, { "epoch": 1.13, "learning_rate": 4.838923845306536e-05, "loss": 0.9754, "step": 18714 }, { "epoch": 1.13, "learning_rate": 4.8387729542951125e-05, "loss": 1.2757, "step": 18716 }, { "epoch": 1.13, "learning_rate": 4.8386220632836905e-05, "loss": 1.6536, "step": 18718 }, { "epoch": 1.13, "learning_rate": 4.838471172272268e-05, "loss": 1.1106, "step": 18720 }, { "epoch": 1.13, "learning_rate": 4.838320281260846e-05, "loss": 1.1694, "step": 18722 }, { "epoch": 1.13, "learning_rate": 4.838169390249423e-05, "loss": 1.2327, "step": 18724 }, { "epoch": 1.13, "learning_rate": 4.8380184992380004e-05, "loss": 1.1862, "step": 18726 }, { "epoch": 1.13, "learning_rate": 4.8378676082265784e-05, "loss": 1.3091, "step": 18728 }, { "epoch": 1.13, "learning_rate": 4.837716717215156e-05, "loss": 1.4255, "step": 18730 }, { "epoch": 1.13, "learning_rate": 4.837565826203733e-05, "loss": 0.7464, "step": 18732 }, { "epoch": 1.13, "learning_rate": 4.837414935192311e-05, "loss": 1.2771, "step": 18734 }, { "epoch": 1.13, "learning_rate": 4.837264044180888e-05, "loss": 1.4485, "step": 18736 }, { "epoch": 1.13, "learning_rate": 4.8371131531694656e-05, "loss": 0.9862, "step": 18738 }, { "epoch": 1.13, "learning_rate": 4.8369622621580436e-05, "loss": 0.9228, "step": 18740 }, { "epoch": 1.13, "learning_rate": 4.836811371146621e-05, "loss": 0.8755, "step": 18742 }, { "epoch": 1.13, "learning_rate": 4.836660480135198e-05, "loss": 0.9635, "step": 18744 }, { "epoch": 1.13, "learning_rate": 4.836509589123776e-05, "loss": 1.4237, "step": 18746 }, { "epoch": 1.13, "learning_rate": 4.8363586981123535e-05, "loss": 1.3242, "step": 18748 }, { "epoch": 1.13, "learning_rate": 4.8362078071009315e-05, "loss": 0.9546, "step": 18750 }, { "epoch": 1.13, "learning_rate": 4.836056916089509e-05, "loss": 0.8634, "step": 18752 }, { "epoch": 1.13, "learning_rate": 4.835906025078086e-05, "loss": 1.2961, "step": 18754 }, { "epoch": 1.13, "learning_rate": 4.835755134066664e-05, "loss": 1.3891, "step": 18756 }, { "epoch": 1.13, "learning_rate": 4.8356042430552415e-05, "loss": 1.4892, "step": 18758 }, { "epoch": 1.13, "learning_rate": 4.835453352043819e-05, "loss": 0.9401, "step": 18760 }, { "epoch": 1.13, "learning_rate": 4.835302461032397e-05, "loss": 0.9843, "step": 18762 }, { "epoch": 1.13, "learning_rate": 4.835151570020974e-05, "loss": 1.6098, "step": 18764 }, { "epoch": 1.13, "learning_rate": 4.8350006790095514e-05, "loss": 1.1594, "step": 18766 }, { "epoch": 1.13, "learning_rate": 4.8348497879981294e-05, "loss": 1.1535, "step": 18768 }, { "epoch": 1.13, "learning_rate": 4.834698896986707e-05, "loss": 0.9783, "step": 18770 }, { "epoch": 1.13, "learning_rate": 4.834548005975284e-05, "loss": 1.0795, "step": 18772 }, { "epoch": 1.13, "learning_rate": 4.834397114963862e-05, "loss": 1.2916, "step": 18774 }, { "epoch": 1.13, "learning_rate": 4.834246223952439e-05, "loss": 1.0325, "step": 18776 }, { "epoch": 1.13, "learning_rate": 4.834095332941017e-05, "loss": 1.4918, "step": 18778 }, { "epoch": 1.13, "learning_rate": 4.8339444419295946e-05, "loss": 1.0002, "step": 18780 }, { "epoch": 1.13, "learning_rate": 4.833793550918172e-05, "loss": 1.2111, "step": 18782 }, { "epoch": 1.13, "learning_rate": 4.83364265990675e-05, "loss": 1.0993, "step": 18784 }, { "epoch": 1.13, "learning_rate": 4.833491768895327e-05, "loss": 1.1103, "step": 18786 }, { "epoch": 1.13, "learning_rate": 4.8333408778839045e-05, "loss": 1.2634, "step": 18788 }, { "epoch": 1.13, "learning_rate": 4.8331899868724825e-05, "loss": 1.2265, "step": 18790 }, { "epoch": 1.13, "learning_rate": 4.83303909586106e-05, "loss": 1.0792, "step": 18792 }, { "epoch": 1.13, "learning_rate": 4.832888204849637e-05, "loss": 0.8247, "step": 18794 }, { "epoch": 1.13, "learning_rate": 4.832737313838215e-05, "loss": 1.3544, "step": 18796 }, { "epoch": 1.13, "learning_rate": 4.8325864228267924e-05, "loss": 1.4742, "step": 18798 }, { "epoch": 1.13, "learning_rate": 4.8324355318153704e-05, "loss": 1.0503, "step": 18800 }, { "epoch": 1.13, "learning_rate": 4.832284640803947e-05, "loss": 1.1974, "step": 18802 }, { "epoch": 1.13, "learning_rate": 4.832133749792525e-05, "loss": 0.9391, "step": 18804 }, { "epoch": 1.14, "learning_rate": 4.831982858781103e-05, "loss": 0.8764, "step": 18806 }, { "epoch": 1.14, "learning_rate": 4.83183196776968e-05, "loss": 1.1378, "step": 18808 }, { "epoch": 1.14, "learning_rate": 4.831681076758258e-05, "loss": 0.9259, "step": 18810 }, { "epoch": 1.14, "learning_rate": 4.831530185746836e-05, "loss": 1.1973, "step": 18812 }, { "epoch": 1.14, "learning_rate": 4.831379294735413e-05, "loss": 0.9188, "step": 18814 }, { "epoch": 1.14, "learning_rate": 4.83122840372399e-05, "loss": 1.1082, "step": 18816 }, { "epoch": 1.14, "learning_rate": 4.8310775127125676e-05, "loss": 1.5195, "step": 18818 }, { "epoch": 1.14, "learning_rate": 4.8309266217011456e-05, "loss": 1.0474, "step": 18820 }, { "epoch": 1.14, "learning_rate": 4.830775730689723e-05, "loss": 1.3486, "step": 18822 }, { "epoch": 1.14, "learning_rate": 4.8306248396783e-05, "loss": 1.3003, "step": 18824 }, { "epoch": 1.14, "learning_rate": 4.830473948666878e-05, "loss": 1.2951, "step": 18826 }, { "epoch": 1.14, "learning_rate": 4.830323057655456e-05, "loss": 0.9163, "step": 18828 }, { "epoch": 1.14, "learning_rate": 4.830172166644033e-05, "loss": 1.1417, "step": 18830 }, { "epoch": 1.14, "learning_rate": 4.830021275632611e-05, "loss": 1.0026, "step": 18832 }, { "epoch": 1.14, "learning_rate": 4.829870384621188e-05, "loss": 0.9547, "step": 18834 }, { "epoch": 1.14, "learning_rate": 4.829719493609766e-05, "loss": 0.866, "step": 18836 }, { "epoch": 1.14, "learning_rate": 4.8295686025983434e-05, "loss": 1.3532, "step": 18838 }, { "epoch": 1.14, "learning_rate": 4.829417711586921e-05, "loss": 1.3806, "step": 18840 }, { "epoch": 1.14, "learning_rate": 4.829266820575499e-05, "loss": 1.0366, "step": 18842 }, { "epoch": 1.14, "learning_rate": 4.829115929564076e-05, "loss": 1.3845, "step": 18844 }, { "epoch": 1.14, "learning_rate": 4.8289650385526534e-05, "loss": 1.1711, "step": 18846 }, { "epoch": 1.14, "learning_rate": 4.8288141475412314e-05, "loss": 1.0734, "step": 18848 }, { "epoch": 1.14, "learning_rate": 4.828663256529809e-05, "loss": 1.1959, "step": 18850 }, { "epoch": 1.14, "learning_rate": 4.828512365518386e-05, "loss": 1.167, "step": 18852 }, { "epoch": 1.14, "learning_rate": 4.828361474506964e-05, "loss": 1.4461, "step": 18854 }, { "epoch": 1.14, "learning_rate": 4.828210583495541e-05, "loss": 1.0662, "step": 18856 }, { "epoch": 1.14, "learning_rate": 4.8280596924841186e-05, "loss": 1.1001, "step": 18858 }, { "epoch": 1.14, "learning_rate": 4.8279088014726966e-05, "loss": 1.0983, "step": 18860 }, { "epoch": 1.14, "learning_rate": 4.827757910461274e-05, "loss": 0.9831, "step": 18862 }, { "epoch": 1.14, "learning_rate": 4.827607019449852e-05, "loss": 1.1893, "step": 18864 }, { "epoch": 1.14, "learning_rate": 4.827456128438429e-05, "loss": 1.2858, "step": 18866 }, { "epoch": 1.14, "learning_rate": 4.8273052374270065e-05, "loss": 1.3988, "step": 18868 }, { "epoch": 1.14, "learning_rate": 4.8271543464155845e-05, "loss": 1.3272, "step": 18870 }, { "epoch": 1.14, "learning_rate": 4.827003455404162e-05, "loss": 0.9457, "step": 18872 }, { "epoch": 1.14, "learning_rate": 4.826852564392739e-05, "loss": 1.171, "step": 18874 }, { "epoch": 1.14, "learning_rate": 4.826701673381317e-05, "loss": 1.092, "step": 18876 }, { "epoch": 1.14, "learning_rate": 4.8265507823698944e-05, "loss": 1.1001, "step": 18878 }, { "epoch": 1.14, "learning_rate": 4.826399891358472e-05, "loss": 1.0092, "step": 18880 }, { "epoch": 1.14, "learning_rate": 4.82624900034705e-05, "loss": 0.906, "step": 18882 }, { "epoch": 1.14, "learning_rate": 4.826098109335627e-05, "loss": 1.0548, "step": 18884 }, { "epoch": 1.14, "learning_rate": 4.825947218324205e-05, "loss": 1.0102, "step": 18886 }, { "epoch": 1.14, "learning_rate": 4.8257963273127824e-05, "loss": 1.1362, "step": 18888 }, { "epoch": 1.14, "learning_rate": 4.82564543630136e-05, "loss": 1.178, "step": 18890 }, { "epoch": 1.14, "learning_rate": 4.8254945452899377e-05, "loss": 0.7724, "step": 18892 }, { "epoch": 1.14, "learning_rate": 4.825343654278515e-05, "loss": 1.3353, "step": 18894 }, { "epoch": 1.14, "learning_rate": 4.825192763267092e-05, "loss": 0.9996, "step": 18896 }, { "epoch": 1.14, "learning_rate": 4.82504187225567e-05, "loss": 1.4228, "step": 18898 }, { "epoch": 1.14, "learning_rate": 4.8248909812442476e-05, "loss": 1.1972, "step": 18900 }, { "epoch": 1.14, "learning_rate": 4.824740090232825e-05, "loss": 1.1437, "step": 18902 }, { "epoch": 1.14, "learning_rate": 4.824589199221403e-05, "loss": 1.3321, "step": 18904 }, { "epoch": 1.14, "learning_rate": 4.82443830820998e-05, "loss": 1.24, "step": 18906 }, { "epoch": 1.14, "learning_rate": 4.8242874171985575e-05, "loss": 0.8011, "step": 18908 }, { "epoch": 1.14, "learning_rate": 4.8241365261871355e-05, "loss": 0.8527, "step": 18910 }, { "epoch": 1.14, "learning_rate": 4.823985635175713e-05, "loss": 1.5089, "step": 18912 }, { "epoch": 1.14, "learning_rate": 4.823834744164291e-05, "loss": 1.1987, "step": 18914 }, { "epoch": 1.14, "learning_rate": 4.8236838531528674e-05, "loss": 0.9227, "step": 18916 }, { "epoch": 1.14, "learning_rate": 4.8235329621414454e-05, "loss": 1.4738, "step": 18918 }, { "epoch": 1.14, "learning_rate": 4.8233820711300234e-05, "loss": 1.0116, "step": 18920 }, { "epoch": 1.14, "learning_rate": 4.8232311801186e-05, "loss": 1.0222, "step": 18922 }, { "epoch": 1.14, "learning_rate": 4.823080289107178e-05, "loss": 0.9998, "step": 18924 }, { "epoch": 1.14, "learning_rate": 4.822929398095756e-05, "loss": 1.0667, "step": 18926 }, { "epoch": 1.14, "learning_rate": 4.8227785070843333e-05, "loss": 1.531, "step": 18928 }, { "epoch": 1.14, "learning_rate": 4.822627616072911e-05, "loss": 1.2442, "step": 18930 }, { "epoch": 1.14, "learning_rate": 4.822476725061488e-05, "loss": 0.7503, "step": 18932 }, { "epoch": 1.14, "learning_rate": 4.822325834050066e-05, "loss": 1.195, "step": 18934 }, { "epoch": 1.14, "learning_rate": 4.822174943038643e-05, "loss": 1.3508, "step": 18936 }, { "epoch": 1.14, "learning_rate": 4.8220240520272206e-05, "loss": 1.1425, "step": 18938 }, { "epoch": 1.14, "learning_rate": 4.8218731610157986e-05, "loss": 1.569, "step": 18940 }, { "epoch": 1.14, "learning_rate": 4.8217222700043766e-05, "loss": 1.0495, "step": 18942 }, { "epoch": 1.14, "learning_rate": 4.821571378992953e-05, "loss": 1.0445, "step": 18944 }, { "epoch": 1.14, "learning_rate": 4.821420487981531e-05, "loss": 0.8663, "step": 18946 }, { "epoch": 1.14, "learning_rate": 4.8212695969701085e-05, "loss": 0.9931, "step": 18948 }, { "epoch": 1.14, "learning_rate": 4.8211187059586865e-05, "loss": 0.9258, "step": 18950 }, { "epoch": 1.14, "learning_rate": 4.820967814947264e-05, "loss": 1.1658, "step": 18952 }, { "epoch": 1.14, "learning_rate": 4.820816923935841e-05, "loss": 1.0437, "step": 18954 }, { "epoch": 1.14, "learning_rate": 4.820666032924419e-05, "loss": 1.0609, "step": 18956 }, { "epoch": 1.14, "learning_rate": 4.8205151419129964e-05, "loss": 1.2395, "step": 18958 }, { "epoch": 1.14, "learning_rate": 4.820364250901574e-05, "loss": 1.0521, "step": 18960 }, { "epoch": 1.14, "learning_rate": 4.820213359890152e-05, "loss": 1.1164, "step": 18962 }, { "epoch": 1.14, "learning_rate": 4.820062468878729e-05, "loss": 0.8584, "step": 18964 }, { "epoch": 1.14, "learning_rate": 4.8199115778673064e-05, "loss": 1.1794, "step": 18966 }, { "epoch": 1.14, "learning_rate": 4.8197606868558843e-05, "loss": 1.1376, "step": 18968 }, { "epoch": 1.14, "learning_rate": 4.8196097958444617e-05, "loss": 1.0702, "step": 18970 }, { "epoch": 1.15, "learning_rate": 4.819458904833039e-05, "loss": 1.317, "step": 18972 }, { "epoch": 1.15, "learning_rate": 4.819308013821617e-05, "loss": 0.9429, "step": 18974 }, { "epoch": 1.15, "learning_rate": 4.819157122810194e-05, "loss": 1.1196, "step": 18976 }, { "epoch": 1.15, "learning_rate": 4.819006231798772e-05, "loss": 0.9451, "step": 18978 }, { "epoch": 1.15, "learning_rate": 4.8188553407873496e-05, "loss": 1.423, "step": 18980 }, { "epoch": 1.15, "learning_rate": 4.818704449775927e-05, "loss": 1.2328, "step": 18982 }, { "epoch": 1.15, "learning_rate": 4.818553558764505e-05, "loss": 1.3064, "step": 18984 }, { "epoch": 1.15, "learning_rate": 4.818402667753082e-05, "loss": 1.5319, "step": 18986 }, { "epoch": 1.15, "learning_rate": 4.8182517767416595e-05, "loss": 1.002, "step": 18988 }, { "epoch": 1.15, "learning_rate": 4.8181008857302375e-05, "loss": 1.1723, "step": 18990 }, { "epoch": 1.15, "learning_rate": 4.817949994718815e-05, "loss": 1.0566, "step": 18992 }, { "epoch": 1.15, "learning_rate": 4.817799103707392e-05, "loss": 1.1698, "step": 18994 }, { "epoch": 1.15, "learning_rate": 4.81764821269597e-05, "loss": 0.9897, "step": 18996 }, { "epoch": 1.15, "learning_rate": 4.8174973216845474e-05, "loss": 0.88, "step": 18998 }, { "epoch": 1.15, "learning_rate": 4.8173464306731254e-05, "loss": 1.8444, "step": 19000 }, { "epoch": 1.15, "learning_rate": 4.817195539661703e-05, "loss": 1.3215, "step": 19002 }, { "epoch": 1.15, "learning_rate": 4.81704464865028e-05, "loss": 1.1968, "step": 19004 }, { "epoch": 1.15, "learning_rate": 4.816893757638858e-05, "loss": 0.8697, "step": 19006 }, { "epoch": 1.15, "learning_rate": 4.816742866627435e-05, "loss": 1.3968, "step": 19008 }, { "epoch": 1.15, "learning_rate": 4.8165919756160127e-05, "loss": 1.2573, "step": 19010 }, { "epoch": 1.15, "learning_rate": 4.8164410846045906e-05, "loss": 1.1409, "step": 19012 }, { "epoch": 1.15, "learning_rate": 4.816290193593168e-05, "loss": 0.9592, "step": 19014 }, { "epoch": 1.15, "learning_rate": 4.816139302581745e-05, "loss": 1.1919, "step": 19016 }, { "epoch": 1.15, "learning_rate": 4.815988411570323e-05, "loss": 1.0202, "step": 19018 }, { "epoch": 1.15, "learning_rate": 4.8158375205589006e-05, "loss": 1.1462, "step": 19020 }, { "epoch": 1.15, "learning_rate": 4.815686629547478e-05, "loss": 1.1123, "step": 19022 }, { "epoch": 1.15, "learning_rate": 4.815535738536056e-05, "loss": 0.9276, "step": 19024 }, { "epoch": 1.15, "learning_rate": 4.815384847524633e-05, "loss": 1.0105, "step": 19026 }, { "epoch": 1.15, "learning_rate": 4.815233956513211e-05, "loss": 1.2501, "step": 19028 }, { "epoch": 1.15, "learning_rate": 4.815083065501788e-05, "loss": 0.9251, "step": 19030 }, { "epoch": 1.15, "learning_rate": 4.814932174490366e-05, "loss": 1.0225, "step": 19032 }, { "epoch": 1.15, "learning_rate": 4.814781283478944e-05, "loss": 1.2817, "step": 19034 }, { "epoch": 1.15, "learning_rate": 4.8146303924675204e-05, "loss": 1.3039, "step": 19036 }, { "epoch": 1.15, "learning_rate": 4.8144795014560984e-05, "loss": 0.9648, "step": 19038 }, { "epoch": 1.15, "learning_rate": 4.8143286104446764e-05, "loss": 1.2261, "step": 19040 }, { "epoch": 1.15, "learning_rate": 4.814177719433254e-05, "loss": 1.0649, "step": 19042 }, { "epoch": 1.15, "learning_rate": 4.814026828421831e-05, "loss": 1.0836, "step": 19044 }, { "epoch": 1.15, "learning_rate": 4.8138759374104083e-05, "loss": 1.1092, "step": 19046 }, { "epoch": 1.15, "learning_rate": 4.813725046398986e-05, "loss": 1.3727, "step": 19048 }, { "epoch": 1.15, "learning_rate": 4.813574155387564e-05, "loss": 0.9649, "step": 19050 }, { "epoch": 1.15, "learning_rate": 4.813423264376141e-05, "loss": 1.016, "step": 19052 }, { "epoch": 1.15, "learning_rate": 4.813272373364719e-05, "loss": 1.0405, "step": 19054 }, { "epoch": 1.15, "learning_rate": 4.813121482353297e-05, "loss": 1.1834, "step": 19056 }, { "epoch": 1.15, "learning_rate": 4.8129705913418736e-05, "loss": 1.1291, "step": 19058 }, { "epoch": 1.15, "learning_rate": 4.8128197003304516e-05, "loss": 1.3553, "step": 19060 }, { "epoch": 1.15, "learning_rate": 4.812668809319029e-05, "loss": 1.1566, "step": 19062 }, { "epoch": 1.15, "learning_rate": 4.812517918307607e-05, "loss": 1.0879, "step": 19064 }, { "epoch": 1.15, "learning_rate": 4.812367027296184e-05, "loss": 0.7812, "step": 19066 }, { "epoch": 1.15, "learning_rate": 4.8122161362847615e-05, "loss": 1.3524, "step": 19068 }, { "epoch": 1.15, "learning_rate": 4.8120652452733395e-05, "loss": 0.9553, "step": 19070 }, { "epoch": 1.15, "learning_rate": 4.811914354261917e-05, "loss": 0.8441, "step": 19072 }, { "epoch": 1.15, "learning_rate": 4.811763463250494e-05, "loss": 1.0656, "step": 19074 }, { "epoch": 1.15, "learning_rate": 4.811612572239072e-05, "loss": 1.5649, "step": 19076 }, { "epoch": 1.15, "learning_rate": 4.8114616812276494e-05, "loss": 0.9587, "step": 19078 }, { "epoch": 1.15, "learning_rate": 4.811310790216227e-05, "loss": 1.2782, "step": 19080 }, { "epoch": 1.15, "learning_rate": 4.811159899204805e-05, "loss": 1.1001, "step": 19082 }, { "epoch": 1.15, "learning_rate": 4.811009008193382e-05, "loss": 1.1813, "step": 19084 }, { "epoch": 1.15, "learning_rate": 4.810858117181959e-05, "loss": 1.4501, "step": 19086 }, { "epoch": 1.15, "learning_rate": 4.810707226170537e-05, "loss": 1.2313, "step": 19088 }, { "epoch": 1.15, "learning_rate": 4.8105563351591146e-05, "loss": 0.7936, "step": 19090 }, { "epoch": 1.15, "learning_rate": 4.8104054441476926e-05, "loss": 1.0974, "step": 19092 }, { "epoch": 1.15, "learning_rate": 4.81025455313627e-05, "loss": 1.0322, "step": 19094 }, { "epoch": 1.15, "learning_rate": 4.810103662124847e-05, "loss": 0.9299, "step": 19096 }, { "epoch": 1.15, "learning_rate": 4.809952771113425e-05, "loss": 0.8985, "step": 19098 }, { "epoch": 1.15, "learning_rate": 4.8098018801020026e-05, "loss": 1.2932, "step": 19100 }, { "epoch": 1.15, "learning_rate": 4.80965098909058e-05, "loss": 1.0915, "step": 19102 }, { "epoch": 1.15, "learning_rate": 4.809500098079158e-05, "loss": 1.0264, "step": 19104 }, { "epoch": 1.15, "learning_rate": 4.809349207067735e-05, "loss": 1.0841, "step": 19106 }, { "epoch": 1.15, "learning_rate": 4.8091983160563125e-05, "loss": 0.8744, "step": 19108 }, { "epoch": 1.15, "learning_rate": 4.8090474250448905e-05, "loss": 1.0824, "step": 19110 }, { "epoch": 1.15, "learning_rate": 4.808896534033468e-05, "loss": 1.1913, "step": 19112 }, { "epoch": 1.15, "learning_rate": 4.808745643022046e-05, "loss": 1.1358, "step": 19114 }, { "epoch": 1.15, "learning_rate": 4.808594752010623e-05, "loss": 0.9545, "step": 19116 }, { "epoch": 1.15, "learning_rate": 4.8084438609992004e-05, "loss": 1.3849, "step": 19118 }, { "epoch": 1.15, "learning_rate": 4.8082929699877784e-05, "loss": 1.146, "step": 19120 }, { "epoch": 1.15, "learning_rate": 4.808142078976356e-05, "loss": 1.2123, "step": 19122 }, { "epoch": 1.15, "learning_rate": 4.807991187964933e-05, "loss": 0.9816, "step": 19124 }, { "epoch": 1.15, "learning_rate": 4.807840296953511e-05, "loss": 1.0303, "step": 19126 }, { "epoch": 1.15, "learning_rate": 4.807689405942088e-05, "loss": 1.0769, "step": 19128 }, { "epoch": 1.15, "learning_rate": 4.8075385149306656e-05, "loss": 1.1661, "step": 19130 }, { "epoch": 1.15, "learning_rate": 4.8073876239192436e-05, "loss": 1.0274, "step": 19132 }, { "epoch": 1.15, "learning_rate": 4.807236732907821e-05, "loss": 0.9757, "step": 19134 }, { "epoch": 1.15, "learning_rate": 4.807085841896398e-05, "loss": 1.1287, "step": 19136 }, { "epoch": 1.16, "learning_rate": 4.806934950884976e-05, "loss": 0.9808, "step": 19138 }, { "epoch": 1.16, "learning_rate": 4.8067840598735536e-05, "loss": 0.8342, "step": 19140 }, { "epoch": 1.16, "learning_rate": 4.8066331688621315e-05, "loss": 1.4058, "step": 19142 }, { "epoch": 1.16, "learning_rate": 4.806482277850708e-05, "loss": 0.8166, "step": 19144 }, { "epoch": 1.16, "learning_rate": 4.806331386839286e-05, "loss": 1.3344, "step": 19146 }, { "epoch": 1.16, "learning_rate": 4.806180495827864e-05, "loss": 1.0464, "step": 19148 }, { "epoch": 1.16, "learning_rate": 4.806029604816441e-05, "loss": 1.06, "step": 19150 }, { "epoch": 1.16, "learning_rate": 4.805878713805019e-05, "loss": 1.0165, "step": 19152 }, { "epoch": 1.16, "learning_rate": 4.805727822793597e-05, "loss": 1.3863, "step": 19154 }, { "epoch": 1.16, "learning_rate": 4.805576931782174e-05, "loss": 1.055, "step": 19156 }, { "epoch": 1.16, "learning_rate": 4.8054260407707514e-05, "loss": 1.0344, "step": 19158 }, { "epoch": 1.16, "learning_rate": 4.805275149759329e-05, "loss": 1.1374, "step": 19160 }, { "epoch": 1.16, "learning_rate": 4.805124258747907e-05, "loss": 1.0897, "step": 19162 }, { "epoch": 1.16, "learning_rate": 4.804973367736485e-05, "loss": 1.2114, "step": 19164 }, { "epoch": 1.16, "learning_rate": 4.804822476725061e-05, "loss": 1.0622, "step": 19166 }, { "epoch": 1.16, "learning_rate": 4.804671585713639e-05, "loss": 1.2674, "step": 19168 }, { "epoch": 1.16, "learning_rate": 4.804520694702217e-05, "loss": 1.1972, "step": 19170 }, { "epoch": 1.16, "learning_rate": 4.804369803690794e-05, "loss": 1.0637, "step": 19172 }, { "epoch": 1.16, "learning_rate": 4.804218912679372e-05, "loss": 1.0895, "step": 19174 }, { "epoch": 1.16, "learning_rate": 4.804068021667949e-05, "loss": 0.8775, "step": 19176 }, { "epoch": 1.16, "learning_rate": 4.803917130656527e-05, "loss": 1.1343, "step": 19178 }, { "epoch": 1.16, "learning_rate": 4.8037662396451045e-05, "loss": 1.0534, "step": 19180 }, { "epoch": 1.16, "learning_rate": 4.803615348633682e-05, "loss": 1.1694, "step": 19182 }, { "epoch": 1.16, "learning_rate": 4.80346445762226e-05, "loss": 1.2271, "step": 19184 }, { "epoch": 1.16, "learning_rate": 4.803313566610837e-05, "loss": 1.2385, "step": 19186 }, { "epoch": 1.16, "learning_rate": 4.8031626755994145e-05, "loss": 1.1402, "step": 19188 }, { "epoch": 1.16, "learning_rate": 4.8030117845879925e-05, "loss": 1.5405, "step": 19190 }, { "epoch": 1.16, "learning_rate": 4.80286089357657e-05, "loss": 0.9725, "step": 19192 }, { "epoch": 1.16, "learning_rate": 4.802710002565147e-05, "loss": 1.0135, "step": 19194 }, { "epoch": 1.16, "learning_rate": 4.802559111553725e-05, "loss": 1.1556, "step": 19196 }, { "epoch": 1.16, "learning_rate": 4.8024082205423024e-05, "loss": 1.3266, "step": 19198 }, { "epoch": 1.16, "learning_rate": 4.80225732953088e-05, "loss": 1.2681, "step": 19200 }, { "epoch": 1.16, "learning_rate": 4.802106438519458e-05, "loss": 1.7496, "step": 19202 }, { "epoch": 1.16, "learning_rate": 4.801955547508035e-05, "loss": 0.9209, "step": 19204 }, { "epoch": 1.16, "learning_rate": 4.801804656496613e-05, "loss": 0.9817, "step": 19206 }, { "epoch": 1.16, "learning_rate": 4.80165376548519e-05, "loss": 1.4336, "step": 19208 }, { "epoch": 1.16, "learning_rate": 4.8015028744737676e-05, "loss": 1.1166, "step": 19210 }, { "epoch": 1.16, "learning_rate": 4.8013519834623456e-05, "loss": 0.7578, "step": 19212 }, { "epoch": 1.16, "learning_rate": 4.801201092450923e-05, "loss": 0.8815, "step": 19214 }, { "epoch": 1.16, "learning_rate": 4.8010502014395e-05, "loss": 1.2512, "step": 19216 }, { "epoch": 1.16, "learning_rate": 4.800899310428078e-05, "loss": 1.4364, "step": 19218 }, { "epoch": 1.16, "learning_rate": 4.8007484194166555e-05, "loss": 1.1102, "step": 19220 }, { "epoch": 1.16, "learning_rate": 4.800597528405233e-05, "loss": 1.0927, "step": 19222 }, { "epoch": 1.16, "learning_rate": 4.800446637393811e-05, "loss": 1.3924, "step": 19224 }, { "epoch": 1.16, "learning_rate": 4.800295746382388e-05, "loss": 1.1032, "step": 19226 }, { "epoch": 1.16, "learning_rate": 4.800144855370966e-05, "loss": 1.8787, "step": 19228 }, { "epoch": 1.16, "learning_rate": 4.7999939643595435e-05, "loss": 0.9715, "step": 19230 }, { "epoch": 1.16, "learning_rate": 4.799843073348121e-05, "loss": 1.7385, "step": 19232 }, { "epoch": 1.16, "learning_rate": 4.799692182336699e-05, "loss": 0.9784, "step": 19234 }, { "epoch": 1.16, "learning_rate": 4.799541291325276e-05, "loss": 1.2279, "step": 19236 }, { "epoch": 1.16, "learning_rate": 4.7993904003138534e-05, "loss": 1.2581, "step": 19238 }, { "epoch": 1.16, "learning_rate": 4.7992395093024314e-05, "loss": 0.7395, "step": 19240 }, { "epoch": 1.16, "learning_rate": 4.799088618291009e-05, "loss": 0.8725, "step": 19242 }, { "epoch": 1.16, "learning_rate": 4.798937727279586e-05, "loss": 1.1529, "step": 19244 }, { "epoch": 1.16, "learning_rate": 4.798786836268164e-05, "loss": 1.1805, "step": 19246 }, { "epoch": 1.16, "learning_rate": 4.798635945256741e-05, "loss": 1.121, "step": 19248 }, { "epoch": 1.16, "learning_rate": 4.7984850542453186e-05, "loss": 0.8218, "step": 19250 }, { "epoch": 1.16, "learning_rate": 4.7983341632338966e-05, "loss": 0.7353, "step": 19252 }, { "epoch": 1.16, "learning_rate": 4.798183272222474e-05, "loss": 0.8588, "step": 19254 }, { "epoch": 1.16, "learning_rate": 4.798032381211052e-05, "loss": 0.9724, "step": 19256 }, { "epoch": 1.16, "learning_rate": 4.7978814901996285e-05, "loss": 0.9966, "step": 19258 }, { "epoch": 1.16, "learning_rate": 4.7977305991882065e-05, "loss": 1.2971, "step": 19260 }, { "epoch": 1.16, "learning_rate": 4.7975797081767845e-05, "loss": 1.4377, "step": 19262 }, { "epoch": 1.16, "learning_rate": 4.797428817165362e-05, "loss": 1.3234, "step": 19264 }, { "epoch": 1.16, "learning_rate": 4.797277926153939e-05, "loss": 1.2751, "step": 19266 }, { "epoch": 1.16, "learning_rate": 4.797127035142517e-05, "loss": 1.325, "step": 19268 }, { "epoch": 1.16, "learning_rate": 4.7969761441310945e-05, "loss": 1.1759, "step": 19270 }, { "epoch": 1.16, "learning_rate": 4.796825253119672e-05, "loss": 1.2288, "step": 19272 }, { "epoch": 1.16, "learning_rate": 4.796674362108249e-05, "loss": 1.0211, "step": 19274 }, { "epoch": 1.16, "learning_rate": 4.796523471096827e-05, "loss": 0.885, "step": 19276 }, { "epoch": 1.16, "learning_rate": 4.796372580085405e-05, "loss": 0.9731, "step": 19278 }, { "epoch": 1.16, "learning_rate": 4.796221689073982e-05, "loss": 0.9153, "step": 19280 }, { "epoch": 1.16, "learning_rate": 4.79607079806256e-05, "loss": 1.2742, "step": 19282 }, { "epoch": 1.16, "learning_rate": 4.795919907051138e-05, "loss": 1.7131, "step": 19284 }, { "epoch": 1.16, "learning_rate": 4.795769016039714e-05, "loss": 0.8864, "step": 19286 }, { "epoch": 1.16, "learning_rate": 4.795618125028292e-05, "loss": 1.1739, "step": 19288 }, { "epoch": 1.16, "learning_rate": 4.7954672340168696e-05, "loss": 0.892, "step": 19290 }, { "epoch": 1.16, "learning_rate": 4.7953163430054476e-05, "loss": 0.9322, "step": 19292 }, { "epoch": 1.16, "learning_rate": 4.795165451994025e-05, "loss": 1.0247, "step": 19294 }, { "epoch": 1.16, "learning_rate": 4.795014560982602e-05, "loss": 0.8877, "step": 19296 }, { "epoch": 1.16, "learning_rate": 4.79486366997118e-05, "loss": 0.9785, "step": 19298 }, { "epoch": 1.16, "learning_rate": 4.7947127789597575e-05, "loss": 1.3562, "step": 19300 }, { "epoch": 1.17, "learning_rate": 4.794561887948335e-05, "loss": 1.1731, "step": 19302 }, { "epoch": 1.17, "learning_rate": 4.794410996936913e-05, "loss": 1.0767, "step": 19304 }, { "epoch": 1.17, "learning_rate": 4.79426010592549e-05, "loss": 1.0632, "step": 19306 }, { "epoch": 1.17, "learning_rate": 4.7941092149140675e-05, "loss": 1.1704, "step": 19308 }, { "epoch": 1.17, "learning_rate": 4.7939583239026454e-05, "loss": 1.5169, "step": 19310 }, { "epoch": 1.17, "learning_rate": 4.793807432891223e-05, "loss": 0.7047, "step": 19312 }, { "epoch": 1.17, "learning_rate": 4.7936565418798e-05, "loss": 0.9908, "step": 19314 }, { "epoch": 1.17, "learning_rate": 4.793505650868378e-05, "loss": 1.1327, "step": 19316 }, { "epoch": 1.17, "learning_rate": 4.7933547598569554e-05, "loss": 0.9589, "step": 19318 }, { "epoch": 1.17, "learning_rate": 4.7932038688455334e-05, "loss": 0.6988, "step": 19320 }, { "epoch": 1.17, "learning_rate": 4.793052977834111e-05, "loss": 1.2028, "step": 19322 }, { "epoch": 1.17, "learning_rate": 4.792902086822688e-05, "loss": 1.2745, "step": 19324 }, { "epoch": 1.17, "learning_rate": 4.792751195811266e-05, "loss": 1.0414, "step": 19326 }, { "epoch": 1.17, "learning_rate": 4.792600304799843e-05, "loss": 0.9749, "step": 19328 }, { "epoch": 1.17, "learning_rate": 4.7924494137884206e-05, "loss": 1.0121, "step": 19330 }, { "epoch": 1.17, "learning_rate": 4.7922985227769986e-05, "loss": 1.0793, "step": 19332 }, { "epoch": 1.17, "learning_rate": 4.792147631765576e-05, "loss": 0.8378, "step": 19334 }, { "epoch": 1.17, "learning_rate": 4.791996740754153e-05, "loss": 0.9258, "step": 19336 }, { "epoch": 1.17, "learning_rate": 4.791845849742731e-05, "loss": 1.148, "step": 19338 }, { "epoch": 1.17, "learning_rate": 4.7916949587313085e-05, "loss": 0.7963, "step": 19340 }, { "epoch": 1.17, "learning_rate": 4.7915440677198865e-05, "loss": 1.3961, "step": 19342 }, { "epoch": 1.17, "learning_rate": 4.791393176708464e-05, "loss": 1.0415, "step": 19344 }, { "epoch": 1.17, "learning_rate": 4.791242285697041e-05, "loss": 1.0484, "step": 19346 }, { "epoch": 1.17, "learning_rate": 4.791091394685619e-05, "loss": 1.4199, "step": 19348 }, { "epoch": 1.17, "learning_rate": 4.7909405036741964e-05, "loss": 0.9758, "step": 19350 }, { "epoch": 1.17, "learning_rate": 4.790789612662774e-05, "loss": 1.063, "step": 19352 }, { "epoch": 1.17, "learning_rate": 4.790638721651352e-05, "loss": 0.9063, "step": 19354 }, { "epoch": 1.17, "learning_rate": 4.790487830639929e-05, "loss": 1.0212, "step": 19356 }, { "epoch": 1.17, "learning_rate": 4.7903369396285064e-05, "loss": 1.1546, "step": 19358 }, { "epoch": 1.17, "learning_rate": 4.7901860486170844e-05, "loss": 1.2161, "step": 19360 }, { "epoch": 1.17, "learning_rate": 4.790035157605662e-05, "loss": 1.0409, "step": 19362 }, { "epoch": 1.17, "learning_rate": 4.789884266594239e-05, "loss": 1.0599, "step": 19364 }, { "epoch": 1.17, "learning_rate": 4.789733375582817e-05, "loss": 1.0035, "step": 19366 }, { "epoch": 1.17, "learning_rate": 4.789582484571394e-05, "loss": 1.0324, "step": 19368 }, { "epoch": 1.17, "learning_rate": 4.789431593559972e-05, "loss": 1.3147, "step": 19370 }, { "epoch": 1.17, "learning_rate": 4.789280702548549e-05, "loss": 0.7605, "step": 19372 }, { "epoch": 1.17, "learning_rate": 4.789129811537127e-05, "loss": 0.9037, "step": 19374 }, { "epoch": 1.17, "learning_rate": 4.788978920525705e-05, "loss": 0.8749, "step": 19376 }, { "epoch": 1.17, "learning_rate": 4.788828029514282e-05, "loss": 1.1993, "step": 19378 }, { "epoch": 1.17, "learning_rate": 4.7886771385028595e-05, "loss": 1.2286, "step": 19380 }, { "epoch": 1.17, "learning_rate": 4.7885262474914375e-05, "loss": 1.0167, "step": 19382 }, { "epoch": 1.17, "learning_rate": 4.788375356480015e-05, "loss": 1.2897, "step": 19384 }, { "epoch": 1.17, "learning_rate": 4.788224465468592e-05, "loss": 0.9242, "step": 19386 }, { "epoch": 1.17, "learning_rate": 4.7880735744571694e-05, "loss": 0.8764, "step": 19388 }, { "epoch": 1.17, "learning_rate": 4.7879226834457474e-05, "loss": 1.1542, "step": 19390 }, { "epoch": 1.17, "learning_rate": 4.7877717924343254e-05, "loss": 1.1635, "step": 19392 }, { "epoch": 1.17, "learning_rate": 4.787620901422902e-05, "loss": 1.2216, "step": 19394 }, { "epoch": 1.17, "learning_rate": 4.78747001041148e-05, "loss": 1.137, "step": 19396 }, { "epoch": 1.17, "learning_rate": 4.787319119400058e-05, "loss": 1.1734, "step": 19398 }, { "epoch": 1.17, "learning_rate": 4.787168228388635e-05, "loss": 0.7742, "step": 19400 }, { "epoch": 1.17, "learning_rate": 4.787017337377213e-05, "loss": 1.1093, "step": 19402 }, { "epoch": 1.17, "learning_rate": 4.78686644636579e-05, "loss": 0.8706, "step": 19404 }, { "epoch": 1.17, "learning_rate": 4.786715555354368e-05, "loss": 1.1189, "step": 19406 }, { "epoch": 1.17, "learning_rate": 4.786564664342945e-05, "loss": 1.0241, "step": 19408 }, { "epoch": 1.17, "learning_rate": 4.7864137733315226e-05, "loss": 1.2413, "step": 19410 }, { "epoch": 1.17, "learning_rate": 4.7862628823201006e-05, "loss": 1.0014, "step": 19412 }, { "epoch": 1.17, "learning_rate": 4.786111991308678e-05, "loss": 0.9968, "step": 19414 }, { "epoch": 1.17, "learning_rate": 4.785961100297255e-05, "loss": 0.8884, "step": 19416 }, { "epoch": 1.17, "learning_rate": 4.785810209285833e-05, "loss": 1.003, "step": 19418 }, { "epoch": 1.17, "learning_rate": 4.7856593182744105e-05, "loss": 1.4118, "step": 19420 }, { "epoch": 1.17, "learning_rate": 4.785508427262988e-05, "loss": 0.9579, "step": 19422 }, { "epoch": 1.17, "learning_rate": 4.785357536251566e-05, "loss": 1.1681, "step": 19424 }, { "epoch": 1.17, "learning_rate": 4.785206645240143e-05, "loss": 1.3473, "step": 19426 }, { "epoch": 1.17, "learning_rate": 4.785055754228721e-05, "loss": 1.1677, "step": 19428 }, { "epoch": 1.17, "learning_rate": 4.7849048632172984e-05, "loss": 0.9947, "step": 19430 }, { "epoch": 1.17, "learning_rate": 4.784753972205876e-05, "loss": 1.5642, "step": 19432 }, { "epoch": 1.17, "learning_rate": 4.784603081194454e-05, "loss": 1.2723, "step": 19434 }, { "epoch": 1.17, "learning_rate": 4.784452190183031e-05, "loss": 0.6601, "step": 19436 }, { "epoch": 1.17, "learning_rate": 4.7843012991716084e-05, "loss": 0.8195, "step": 19438 }, { "epoch": 1.17, "learning_rate": 4.7841504081601863e-05, "loss": 1.1201, "step": 19440 }, { "epoch": 1.17, "learning_rate": 4.7839995171487637e-05, "loss": 0.8841, "step": 19442 }, { "epoch": 1.17, "learning_rate": 4.783848626137341e-05, "loss": 1.047, "step": 19444 }, { "epoch": 1.17, "learning_rate": 4.783697735125919e-05, "loss": 1.2159, "step": 19446 }, { "epoch": 1.17, "learning_rate": 4.783546844114496e-05, "loss": 1.2207, "step": 19448 }, { "epoch": 1.17, "learning_rate": 4.7833959531030736e-05, "loss": 0.9679, "step": 19450 }, { "epoch": 1.17, "learning_rate": 4.7832450620916516e-05, "loss": 0.9234, "step": 19452 }, { "epoch": 1.17, "learning_rate": 4.783094171080229e-05, "loss": 0.8634, "step": 19454 }, { "epoch": 1.17, "learning_rate": 4.782943280068807e-05, "loss": 1.063, "step": 19456 }, { "epoch": 1.17, "learning_rate": 4.782792389057384e-05, "loss": 0.8082, "step": 19458 }, { "epoch": 1.17, "learning_rate": 4.7826414980459615e-05, "loss": 1.5231, "step": 19460 }, { "epoch": 1.17, "learning_rate": 4.7824906070345395e-05, "loss": 1.2326, "step": 19462 }, { "epoch": 1.17, "learning_rate": 4.782339716023117e-05, "loss": 0.9667, "step": 19464 }, { "epoch": 1.17, "learning_rate": 4.782188825011694e-05, "loss": 1.0871, "step": 19466 }, { "epoch": 1.18, "learning_rate": 4.782037934000272e-05, "loss": 1.1278, "step": 19468 }, { "epoch": 1.18, "learning_rate": 4.7818870429888494e-05, "loss": 1.0026, "step": 19470 }, { "epoch": 1.18, "learning_rate": 4.781736151977427e-05, "loss": 1.0608, "step": 19472 }, { "epoch": 1.18, "learning_rate": 4.781585260966005e-05, "loss": 1.2687, "step": 19474 }, { "epoch": 1.18, "learning_rate": 4.781434369954582e-05, "loss": 1.1247, "step": 19476 }, { "epoch": 1.18, "learning_rate": 4.7812834789431594e-05, "loss": 1.0547, "step": 19478 }, { "epoch": 1.18, "learning_rate": 4.7811325879317373e-05, "loss": 0.875, "step": 19480 }, { "epoch": 1.18, "learning_rate": 4.7809816969203147e-05, "loss": 0.9811, "step": 19482 }, { "epoch": 1.18, "learning_rate": 4.7808308059088926e-05, "loss": 0.942, "step": 19484 }, { "epoch": 1.18, "learning_rate": 4.780679914897469e-05, "loss": 0.9562, "step": 19486 }, { "epoch": 1.18, "learning_rate": 4.780529023886047e-05, "loss": 0.8488, "step": 19488 }, { "epoch": 1.18, "learning_rate": 4.780378132874625e-05, "loss": 1.4598, "step": 19490 }, { "epoch": 1.18, "learning_rate": 4.7802272418632026e-05, "loss": 1.0423, "step": 19492 }, { "epoch": 1.18, "learning_rate": 4.78007635085178e-05, "loss": 1.5174, "step": 19494 }, { "epoch": 1.18, "learning_rate": 4.779925459840358e-05, "loss": 1.0799, "step": 19496 }, { "epoch": 1.18, "learning_rate": 4.779774568828935e-05, "loss": 0.9629, "step": 19498 }, { "epoch": 1.18, "learning_rate": 4.7796236778175125e-05, "loss": 0.9479, "step": 19500 }, { "epoch": 1.18, "learning_rate": 4.77947278680609e-05, "loss": 1.2688, "step": 19502 }, { "epoch": 1.18, "learning_rate": 4.779321895794668e-05, "loss": 0.9526, "step": 19504 }, { "epoch": 1.18, "learning_rate": 4.779171004783246e-05, "loss": 1.0384, "step": 19506 }, { "epoch": 1.18, "learning_rate": 4.7790201137718224e-05, "loss": 0.9727, "step": 19508 }, { "epoch": 1.18, "learning_rate": 4.7788692227604004e-05, "loss": 0.773, "step": 19510 }, { "epoch": 1.18, "learning_rate": 4.7787183317489784e-05, "loss": 0.8941, "step": 19512 }, { "epoch": 1.18, "learning_rate": 4.778567440737555e-05, "loss": 1.1224, "step": 19514 }, { "epoch": 1.18, "learning_rate": 4.778416549726133e-05, "loss": 0.9439, "step": 19516 }, { "epoch": 1.18, "learning_rate": 4.7782656587147103e-05, "loss": 1.1675, "step": 19518 }, { "epoch": 1.18, "learning_rate": 4.778114767703288e-05, "loss": 1.0031, "step": 19520 }, { "epoch": 1.18, "learning_rate": 4.7779638766918656e-05, "loss": 0.7329, "step": 19522 }, { "epoch": 1.18, "learning_rate": 4.777812985680443e-05, "loss": 1.2305, "step": 19524 }, { "epoch": 1.18, "learning_rate": 4.777662094669021e-05, "loss": 1.2598, "step": 19526 }, { "epoch": 1.18, "learning_rate": 4.777511203657598e-05, "loss": 1.3196, "step": 19528 }, { "epoch": 1.18, "learning_rate": 4.7773603126461756e-05, "loss": 1.3163, "step": 19530 }, { "epoch": 1.18, "learning_rate": 4.7772094216347536e-05, "loss": 1.0041, "step": 19532 }, { "epoch": 1.18, "learning_rate": 4.777058530623331e-05, "loss": 0.7757, "step": 19534 }, { "epoch": 1.18, "learning_rate": 4.776907639611908e-05, "loss": 1.1174, "step": 19536 }, { "epoch": 1.18, "learning_rate": 4.776756748600486e-05, "loss": 1.1752, "step": 19538 }, { "epoch": 1.18, "learning_rate": 4.7766058575890635e-05, "loss": 0.8662, "step": 19540 }, { "epoch": 1.18, "learning_rate": 4.7764549665776415e-05, "loss": 1.2269, "step": 19542 }, { "epoch": 1.18, "learning_rate": 4.776304075566219e-05, "loss": 1.4284, "step": 19544 }, { "epoch": 1.18, "learning_rate": 4.776153184554796e-05, "loss": 1.1575, "step": 19546 }, { "epoch": 1.18, "learning_rate": 4.776002293543374e-05, "loss": 1.0851, "step": 19548 }, { "epoch": 1.18, "learning_rate": 4.7758514025319514e-05, "loss": 1.3225, "step": 19550 }, { "epoch": 1.18, "learning_rate": 4.775700511520529e-05, "loss": 0.9044, "step": 19552 }, { "epoch": 1.18, "learning_rate": 4.775549620509107e-05, "loss": 1.23, "step": 19554 }, { "epoch": 1.18, "learning_rate": 4.775398729497684e-05, "loss": 1.227, "step": 19556 }, { "epoch": 1.18, "learning_rate": 4.7752478384862613e-05, "loss": 1.1793, "step": 19558 }, { "epoch": 1.18, "learning_rate": 4.775096947474839e-05, "loss": 1.1931, "step": 19560 }, { "epoch": 1.18, "learning_rate": 4.7749460564634166e-05, "loss": 0.9102, "step": 19562 }, { "epoch": 1.18, "learning_rate": 4.774795165451994e-05, "loss": 0.9156, "step": 19564 }, { "epoch": 1.18, "learning_rate": 4.774644274440572e-05, "loss": 1.2317, "step": 19566 }, { "epoch": 1.18, "learning_rate": 4.774493383429149e-05, "loss": 0.9999, "step": 19568 }, { "epoch": 1.18, "learning_rate": 4.774342492417727e-05, "loss": 1.0842, "step": 19570 }, { "epoch": 1.18, "learning_rate": 4.7741916014063046e-05, "loss": 0.8103, "step": 19572 }, { "epoch": 1.18, "learning_rate": 4.774040710394882e-05, "loss": 0.9099, "step": 19574 }, { "epoch": 1.18, "learning_rate": 4.77388981938346e-05, "loss": 1.4326, "step": 19576 }, { "epoch": 1.18, "learning_rate": 4.773738928372037e-05, "loss": 0.9637, "step": 19578 }, { "epoch": 1.18, "learning_rate": 4.7735880373606145e-05, "loss": 1.1535, "step": 19580 }, { "epoch": 1.18, "learning_rate": 4.7734371463491925e-05, "loss": 1.0588, "step": 19582 }, { "epoch": 1.18, "learning_rate": 4.77328625533777e-05, "loss": 0.8556, "step": 19584 }, { "epoch": 1.18, "learning_rate": 4.773135364326347e-05, "loss": 1.1037, "step": 19586 }, { "epoch": 1.18, "learning_rate": 4.772984473314925e-05, "loss": 1.2772, "step": 19588 }, { "epoch": 1.18, "learning_rate": 4.7728335823035024e-05, "loss": 1.4168, "step": 19590 }, { "epoch": 1.18, "learning_rate": 4.7726826912920804e-05, "loss": 1.0036, "step": 19592 }, { "epoch": 1.18, "learning_rate": 4.772531800280658e-05, "loss": 1.1669, "step": 19594 }, { "epoch": 1.18, "learning_rate": 4.772380909269235e-05, "loss": 1.6631, "step": 19596 }, { "epoch": 1.18, "learning_rate": 4.772230018257813e-05, "loss": 1.2536, "step": 19598 }, { "epoch": 1.18, "learning_rate": 4.7720791272463896e-05, "loss": 0.7764, "step": 19600 }, { "epoch": 1.18, "learning_rate": 4.7719282362349676e-05, "loss": 0.9616, "step": 19602 }, { "epoch": 1.18, "learning_rate": 4.7717773452235456e-05, "loss": 1.076, "step": 19604 }, { "epoch": 1.18, "learning_rate": 4.771626454212123e-05, "loss": 0.8949, "step": 19606 }, { "epoch": 1.18, "learning_rate": 4.7714755632007e-05, "loss": 0.9776, "step": 19608 }, { "epoch": 1.18, "learning_rate": 4.771324672189278e-05, "loss": 1.0715, "step": 19610 }, { "epoch": 1.18, "learning_rate": 4.7711737811778556e-05, "loss": 1.064, "step": 19612 }, { "epoch": 1.18, "learning_rate": 4.771022890166433e-05, "loss": 1.0293, "step": 19614 }, { "epoch": 1.18, "learning_rate": 4.77087199915501e-05, "loss": 0.9996, "step": 19616 }, { "epoch": 1.18, "learning_rate": 4.770721108143588e-05, "loss": 1.3389, "step": 19618 }, { "epoch": 1.18, "learning_rate": 4.770570217132166e-05, "loss": 0.923, "step": 19620 }, { "epoch": 1.18, "learning_rate": 4.770419326120743e-05, "loss": 1.1047, "step": 19622 }, { "epoch": 1.18, "learning_rate": 4.770268435109321e-05, "loss": 0.6947, "step": 19624 }, { "epoch": 1.18, "learning_rate": 4.770117544097899e-05, "loss": 0.9907, "step": 19626 }, { "epoch": 1.18, "learning_rate": 4.7699666530864754e-05, "loss": 0.9362, "step": 19628 }, { "epoch": 1.18, "learning_rate": 4.7698157620750534e-05, "loss": 0.9298, "step": 19630 }, { "epoch": 1.18, "learning_rate": 4.769664871063631e-05, "loss": 1.0081, "step": 19632 }, { "epoch": 1.19, "learning_rate": 4.769513980052209e-05, "loss": 1.0494, "step": 19634 }, { "epoch": 1.19, "learning_rate": 4.769363089040786e-05, "loss": 1.0918, "step": 19636 }, { "epoch": 1.19, "learning_rate": 4.769212198029363e-05, "loss": 1.3189, "step": 19638 }, { "epoch": 1.19, "learning_rate": 4.769061307017941e-05, "loss": 1.0841, "step": 19640 }, { "epoch": 1.19, "learning_rate": 4.7689104160065186e-05, "loss": 0.8475, "step": 19642 }, { "epoch": 1.19, "learning_rate": 4.768759524995096e-05, "loss": 1.0669, "step": 19644 }, { "epoch": 1.19, "learning_rate": 4.768608633983674e-05, "loss": 1.1272, "step": 19646 }, { "epoch": 1.19, "learning_rate": 4.768457742972251e-05, "loss": 1.0411, "step": 19648 }, { "epoch": 1.19, "learning_rate": 4.7683068519608286e-05, "loss": 1.4051, "step": 19650 }, { "epoch": 1.19, "learning_rate": 4.7681559609494066e-05, "loss": 0.976, "step": 19652 }, { "epoch": 1.19, "learning_rate": 4.768005069937984e-05, "loss": 1.0154, "step": 19654 }, { "epoch": 1.19, "learning_rate": 4.767854178926562e-05, "loss": 0.8842, "step": 19656 }, { "epoch": 1.19, "learning_rate": 4.767703287915139e-05, "loss": 0.8592, "step": 19658 }, { "epoch": 1.19, "learning_rate": 4.7675523969037165e-05, "loss": 0.6504, "step": 19660 }, { "epoch": 1.19, "learning_rate": 4.7674015058922945e-05, "loss": 1.029, "step": 19662 }, { "epoch": 1.19, "learning_rate": 4.767250614880872e-05, "loss": 1.2167, "step": 19664 }, { "epoch": 1.19, "learning_rate": 4.767099723869449e-05, "loss": 0.9926, "step": 19666 }, { "epoch": 1.19, "learning_rate": 4.766948832858027e-05, "loss": 1.0268, "step": 19668 }, { "epoch": 1.19, "learning_rate": 4.7667979418466044e-05, "loss": 1.2085, "step": 19670 }, { "epoch": 1.19, "learning_rate": 4.766647050835182e-05, "loss": 1.0634, "step": 19672 }, { "epoch": 1.19, "learning_rate": 4.76649615982376e-05, "loss": 1.1596, "step": 19674 }, { "epoch": 1.19, "learning_rate": 4.766345268812337e-05, "loss": 1.2863, "step": 19676 }, { "epoch": 1.19, "learning_rate": 4.766194377800914e-05, "loss": 0.7854, "step": 19678 }, { "epoch": 1.19, "learning_rate": 4.766043486789492e-05, "loss": 1.0829, "step": 19680 }, { "epoch": 1.19, "learning_rate": 4.7658925957780696e-05, "loss": 0.8499, "step": 19682 }, { "epoch": 1.19, "learning_rate": 4.7657417047666476e-05, "loss": 1.5106, "step": 19684 }, { "epoch": 1.19, "learning_rate": 4.765590813755225e-05, "loss": 0.9761, "step": 19686 }, { "epoch": 1.19, "learning_rate": 4.765439922743802e-05, "loss": 1.4077, "step": 19688 }, { "epoch": 1.19, "learning_rate": 4.76528903173238e-05, "loss": 0.9961, "step": 19690 }, { "epoch": 1.19, "learning_rate": 4.7651381407209575e-05, "loss": 1.01, "step": 19692 }, { "epoch": 1.19, "learning_rate": 4.764987249709535e-05, "loss": 0.9982, "step": 19694 }, { "epoch": 1.19, "learning_rate": 4.764836358698113e-05, "loss": 1.0292, "step": 19696 }, { "epoch": 1.19, "learning_rate": 4.76468546768669e-05, "loss": 1.6003, "step": 19698 }, { "epoch": 1.19, "learning_rate": 4.7645345766752675e-05, "loss": 1.1351, "step": 19700 }, { "epoch": 1.19, "learning_rate": 4.7643836856638455e-05, "loss": 1.4469, "step": 19702 }, { "epoch": 1.19, "learning_rate": 4.764232794652423e-05, "loss": 0.9816, "step": 19704 }, { "epoch": 1.19, "learning_rate": 4.764081903641001e-05, "loss": 1.3007, "step": 19706 }, { "epoch": 1.19, "learning_rate": 4.763931012629578e-05, "loss": 1.1027, "step": 19708 }, { "epoch": 1.19, "learning_rate": 4.7637801216181554e-05, "loss": 0.8906, "step": 19710 }, { "epoch": 1.19, "learning_rate": 4.7636292306067334e-05, "loss": 1.0621, "step": 19712 }, { "epoch": 1.19, "learning_rate": 4.76347833959531e-05, "loss": 1.3277, "step": 19714 }, { "epoch": 1.19, "learning_rate": 4.763327448583888e-05, "loss": 1.0559, "step": 19716 }, { "epoch": 1.19, "learning_rate": 4.763176557572466e-05, "loss": 1.1815, "step": 19718 }, { "epoch": 1.19, "learning_rate": 4.763025666561043e-05, "loss": 0.8762, "step": 19720 }, { "epoch": 1.19, "learning_rate": 4.7628747755496206e-05, "loss": 0.9463, "step": 19722 }, { "epoch": 1.19, "learning_rate": 4.7627238845381986e-05, "loss": 1.4546, "step": 19724 }, { "epoch": 1.19, "learning_rate": 4.762572993526776e-05, "loss": 0.8785, "step": 19726 }, { "epoch": 1.19, "learning_rate": 4.762422102515353e-05, "loss": 1.2263, "step": 19728 }, { "epoch": 1.19, "learning_rate": 4.7622712115039305e-05, "loss": 0.9918, "step": 19730 }, { "epoch": 1.19, "learning_rate": 4.7621203204925085e-05, "loss": 1.2078, "step": 19732 }, { "epoch": 1.19, "learning_rate": 4.7619694294810865e-05, "loss": 0.885, "step": 19734 }, { "epoch": 1.19, "learning_rate": 4.761818538469663e-05, "loss": 0.7213, "step": 19736 }, { "epoch": 1.19, "learning_rate": 4.761667647458241e-05, "loss": 0.9491, "step": 19738 }, { "epoch": 1.19, "learning_rate": 4.761516756446819e-05, "loss": 1.3551, "step": 19740 }, { "epoch": 1.19, "learning_rate": 4.761365865435396e-05, "loss": 0.9543, "step": 19742 }, { "epoch": 1.19, "learning_rate": 4.761214974423974e-05, "loss": 1.2043, "step": 19744 }, { "epoch": 1.19, "learning_rate": 4.761064083412551e-05, "loss": 0.9822, "step": 19746 }, { "epoch": 1.19, "learning_rate": 4.760913192401129e-05, "loss": 1.3697, "step": 19748 }, { "epoch": 1.19, "learning_rate": 4.7607623013897064e-05, "loss": 1.2659, "step": 19750 }, { "epoch": 1.19, "learning_rate": 4.760611410378284e-05, "loss": 1.2494, "step": 19752 }, { "epoch": 1.19, "learning_rate": 4.760460519366862e-05, "loss": 1.2723, "step": 19754 }, { "epoch": 1.19, "learning_rate": 4.76030962835544e-05, "loss": 1.4499, "step": 19756 }, { "epoch": 1.19, "learning_rate": 4.760158737344016e-05, "loss": 1.2098, "step": 19758 }, { "epoch": 1.19, "learning_rate": 4.760007846332594e-05, "loss": 1.0443, "step": 19760 }, { "epoch": 1.19, "learning_rate": 4.7598569553211716e-05, "loss": 1.3053, "step": 19762 }, { "epoch": 1.19, "learning_rate": 4.759706064309749e-05, "loss": 0.9235, "step": 19764 }, { "epoch": 1.19, "learning_rate": 4.759555173298327e-05, "loss": 0.9613, "step": 19766 }, { "epoch": 1.19, "learning_rate": 4.759404282286904e-05, "loss": 0.8609, "step": 19768 }, { "epoch": 1.19, "learning_rate": 4.759253391275482e-05, "loss": 0.8941, "step": 19770 }, { "epoch": 1.19, "learning_rate": 4.7591025002640595e-05, "loss": 1.0139, "step": 19772 }, { "epoch": 1.19, "learning_rate": 4.758951609252637e-05, "loss": 1.0774, "step": 19774 }, { "epoch": 1.19, "learning_rate": 4.758800718241215e-05, "loss": 1.3509, "step": 19776 }, { "epoch": 1.19, "learning_rate": 4.758649827229792e-05, "loss": 1.1081, "step": 19778 }, { "epoch": 1.19, "learning_rate": 4.7584989362183695e-05, "loss": 0.8286, "step": 19780 }, { "epoch": 1.19, "learning_rate": 4.7583480452069475e-05, "loss": 1.0608, "step": 19782 }, { "epoch": 1.19, "learning_rate": 4.758197154195525e-05, "loss": 1.2265, "step": 19784 }, { "epoch": 1.19, "learning_rate": 4.758046263184102e-05, "loss": 0.9647, "step": 19786 }, { "epoch": 1.19, "learning_rate": 4.75789537217268e-05, "loss": 1.0531, "step": 19788 }, { "epoch": 1.19, "learning_rate": 4.7577444811612574e-05, "loss": 1.19, "step": 19790 }, { "epoch": 1.19, "learning_rate": 4.757593590149835e-05, "loss": 1.136, "step": 19792 }, { "epoch": 1.19, "learning_rate": 4.757442699138413e-05, "loss": 0.8114, "step": 19794 }, { "epoch": 1.19, "learning_rate": 4.75729180812699e-05, "loss": 1.0881, "step": 19796 }, { "epoch": 1.19, "learning_rate": 4.757140917115568e-05, "loss": 1.3837, "step": 19798 }, { "epoch": 1.2, "learning_rate": 4.756990026104145e-05, "loss": 1.2592, "step": 19800 }, { "epoch": 1.2, "learning_rate": 4.7568391350927226e-05, "loss": 1.1233, "step": 19802 }, { "epoch": 1.2, "learning_rate": 4.7566882440813006e-05, "loss": 1.1811, "step": 19804 }, { "epoch": 1.2, "learning_rate": 4.756537353069878e-05, "loss": 1.0038, "step": 19806 }, { "epoch": 1.2, "learning_rate": 4.756386462058455e-05, "loss": 1.1095, "step": 19808 }, { "epoch": 1.2, "learning_rate": 4.756235571047033e-05, "loss": 1.15, "step": 19810 }, { "epoch": 1.2, "learning_rate": 4.7560846800356105e-05, "loss": 1.5052, "step": 19812 }, { "epoch": 1.2, "learning_rate": 4.755933789024188e-05, "loss": 0.779, "step": 19814 }, { "epoch": 1.2, "learning_rate": 4.755782898012766e-05, "loss": 1.0836, "step": 19816 }, { "epoch": 1.2, "learning_rate": 4.755632007001343e-05, "loss": 1.0169, "step": 19818 }, { "epoch": 1.2, "learning_rate": 4.755481115989921e-05, "loss": 0.948, "step": 19820 }, { "epoch": 1.2, "learning_rate": 4.7553302249784984e-05, "loss": 1.2699, "step": 19822 }, { "epoch": 1.2, "learning_rate": 4.755179333967076e-05, "loss": 1.0559, "step": 19824 }, { "epoch": 1.2, "learning_rate": 4.755028442955654e-05, "loss": 0.8385, "step": 19826 }, { "epoch": 1.2, "learning_rate": 4.7548775519442304e-05, "loss": 1.3081, "step": 19828 }, { "epoch": 1.2, "learning_rate": 4.7547266609328084e-05, "loss": 1.0813, "step": 19830 }, { "epoch": 1.2, "learning_rate": 4.7545757699213864e-05, "loss": 0.9396, "step": 19832 }, { "epoch": 1.2, "learning_rate": 4.754424878909964e-05, "loss": 1.2161, "step": 19834 }, { "epoch": 1.2, "learning_rate": 4.754273987898541e-05, "loss": 1.0576, "step": 19836 }, { "epoch": 1.2, "learning_rate": 4.754123096887119e-05, "loss": 1.3113, "step": 19838 }, { "epoch": 1.2, "learning_rate": 4.753972205875696e-05, "loss": 0.9344, "step": 19840 }, { "epoch": 1.2, "learning_rate": 4.7538213148642736e-05, "loss": 1.0545, "step": 19842 }, { "epoch": 1.2, "learning_rate": 4.753670423852851e-05, "loss": 1.0541, "step": 19844 }, { "epoch": 1.2, "learning_rate": 4.753519532841429e-05, "loss": 1.0898, "step": 19846 }, { "epoch": 1.2, "learning_rate": 4.753368641830007e-05, "loss": 0.9363, "step": 19848 }, { "epoch": 1.2, "learning_rate": 4.7532177508185835e-05, "loss": 1.1962, "step": 19850 }, { "epoch": 1.2, "learning_rate": 4.7530668598071615e-05, "loss": 1.24, "step": 19852 }, { "epoch": 1.2, "learning_rate": 4.7529159687957395e-05, "loss": 0.6808, "step": 19854 }, { "epoch": 1.2, "learning_rate": 4.752765077784316e-05, "loss": 0.7804, "step": 19856 }, { "epoch": 1.2, "learning_rate": 4.752614186772894e-05, "loss": 1.0222, "step": 19858 }, { "epoch": 1.2, "learning_rate": 4.7524632957614714e-05, "loss": 1.1614, "step": 19860 }, { "epoch": 1.2, "learning_rate": 4.7523124047500494e-05, "loss": 0.7262, "step": 19862 }, { "epoch": 1.2, "learning_rate": 4.752161513738627e-05, "loss": 1.0724, "step": 19864 }, { "epoch": 1.2, "learning_rate": 4.752010622727204e-05, "loss": 1.2841, "step": 19866 }, { "epoch": 1.2, "learning_rate": 4.751859731715782e-05, "loss": 1.4672, "step": 19868 }, { "epoch": 1.2, "learning_rate": 4.75170884070436e-05, "loss": 1.2123, "step": 19870 }, { "epoch": 1.2, "learning_rate": 4.751557949692937e-05, "loss": 1.1547, "step": 19872 }, { "epoch": 1.2, "learning_rate": 4.751407058681515e-05, "loss": 0.998, "step": 19874 }, { "epoch": 1.2, "learning_rate": 4.751256167670092e-05, "loss": 1.0547, "step": 19876 }, { "epoch": 1.2, "learning_rate": 4.751105276658669e-05, "loss": 1.2356, "step": 19878 }, { "epoch": 1.2, "learning_rate": 4.750954385647247e-05, "loss": 1.2616, "step": 19880 }, { "epoch": 1.2, "learning_rate": 4.7508034946358246e-05, "loss": 0.806, "step": 19882 }, { "epoch": 1.2, "learning_rate": 4.7506526036244026e-05, "loss": 1.2126, "step": 19884 }, { "epoch": 1.2, "learning_rate": 4.75050171261298e-05, "loss": 0.8673, "step": 19886 }, { "epoch": 1.2, "learning_rate": 4.750350821601557e-05, "loss": 1.0698, "step": 19888 }, { "epoch": 1.2, "learning_rate": 4.750199930590135e-05, "loss": 0.8033, "step": 19890 }, { "epoch": 1.2, "learning_rate": 4.7500490395787125e-05, "loss": 1.4215, "step": 19892 }, { "epoch": 1.2, "learning_rate": 4.74989814856729e-05, "loss": 0.9045, "step": 19894 }, { "epoch": 1.2, "learning_rate": 4.749747257555868e-05, "loss": 1.1704, "step": 19896 }, { "epoch": 1.2, "learning_rate": 4.749596366544445e-05, "loss": 1.116, "step": 19898 }, { "epoch": 1.2, "learning_rate": 4.7494454755330224e-05, "loss": 1.3646, "step": 19900 }, { "epoch": 1.2, "learning_rate": 4.7492945845216004e-05, "loss": 1.1984, "step": 19902 }, { "epoch": 1.2, "learning_rate": 4.749143693510178e-05, "loss": 1.2422, "step": 19904 }, { "epoch": 1.2, "learning_rate": 4.748992802498755e-05, "loss": 1.0296, "step": 19906 }, { "epoch": 1.2, "learning_rate": 4.748841911487333e-05, "loss": 1.149, "step": 19908 }, { "epoch": 1.2, "learning_rate": 4.7486910204759104e-05, "loss": 0.9476, "step": 19910 }, { "epoch": 1.2, "learning_rate": 4.7485401294644884e-05, "loss": 1.0709, "step": 19912 }, { "epoch": 1.2, "learning_rate": 4.748389238453066e-05, "loss": 1.0781, "step": 19914 }, { "epoch": 1.2, "learning_rate": 4.748238347441643e-05, "loss": 1.2249, "step": 19916 }, { "epoch": 1.2, "learning_rate": 4.748087456430221e-05, "loss": 1.3107, "step": 19918 }, { "epoch": 1.2, "learning_rate": 4.747936565418798e-05, "loss": 1.1335, "step": 19920 }, { "epoch": 1.2, "learning_rate": 4.7477856744073756e-05, "loss": 1.1338, "step": 19922 }, { "epoch": 1.2, "learning_rate": 4.7476347833959536e-05, "loss": 0.733, "step": 19924 }, { "epoch": 1.2, "learning_rate": 4.747483892384531e-05, "loss": 0.5739, "step": 19926 }, { "epoch": 1.2, "learning_rate": 4.747333001373108e-05, "loss": 1.1023, "step": 19928 }, { "epoch": 1.2, "learning_rate": 4.747182110361686e-05, "loss": 1.1496, "step": 19930 }, { "epoch": 1.2, "learning_rate": 4.7470312193502635e-05, "loss": 0.7713, "step": 19932 }, { "epoch": 1.2, "learning_rate": 4.7468803283388415e-05, "loss": 1.2727, "step": 19934 }, { "epoch": 1.2, "learning_rate": 4.746729437327419e-05, "loss": 0.9172, "step": 19936 }, { "epoch": 1.2, "learning_rate": 4.746578546315996e-05, "loss": 0.7373, "step": 19938 }, { "epoch": 1.2, "learning_rate": 4.746427655304574e-05, "loss": 0.8746, "step": 19940 }, { "epoch": 1.2, "learning_rate": 4.746276764293151e-05, "loss": 0.9633, "step": 19942 }, { "epoch": 1.2, "learning_rate": 4.746125873281729e-05, "loss": 1.0359, "step": 19944 }, { "epoch": 1.2, "learning_rate": 4.745974982270307e-05, "loss": 1.1956, "step": 19946 }, { "epoch": 1.2, "learning_rate": 4.745824091258884e-05, "loss": 0.9099, "step": 19948 }, { "epoch": 1.2, "learning_rate": 4.7456732002474614e-05, "loss": 0.9835, "step": 19950 }, { "epoch": 1.2, "learning_rate": 4.7455223092360393e-05, "loss": 1.0429, "step": 19952 }, { "epoch": 1.2, "learning_rate": 4.7453714182246167e-05, "loss": 1.0595, "step": 19954 }, { "epoch": 1.2, "learning_rate": 4.745220527213194e-05, "loss": 1.0506, "step": 19956 }, { "epoch": 1.2, "learning_rate": 4.745069636201771e-05, "loss": 0.8225, "step": 19958 }, { "epoch": 1.2, "learning_rate": 4.744918745190349e-05, "loss": 0.9291, "step": 19960 }, { "epoch": 1.2, "learning_rate": 4.744767854178927e-05, "loss": 1.0681, "step": 19962 }, { "epoch": 1.2, "learning_rate": 4.744616963167504e-05, "loss": 1.2939, "step": 19964 }, { "epoch": 1.21, "learning_rate": 4.744466072156082e-05, "loss": 0.8735, "step": 19966 }, { "epoch": 1.21, "learning_rate": 4.74431518114466e-05, "loss": 0.9648, "step": 19968 }, { "epoch": 1.21, "learning_rate": 4.744164290133237e-05, "loss": 1.2285, "step": 19970 }, { "epoch": 1.21, "learning_rate": 4.7440133991218145e-05, "loss": 1.1374, "step": 19972 }, { "epoch": 1.21, "learning_rate": 4.743862508110392e-05, "loss": 1.163, "step": 19974 }, { "epoch": 1.21, "learning_rate": 4.74371161709897e-05, "loss": 0.9595, "step": 19976 }, { "epoch": 1.21, "learning_rate": 4.743560726087547e-05, "loss": 0.8891, "step": 19978 }, { "epoch": 1.21, "learning_rate": 4.7434098350761244e-05, "loss": 0.8523, "step": 19980 }, { "epoch": 1.21, "learning_rate": 4.7432589440647024e-05, "loss": 1.4833, "step": 19982 }, { "epoch": 1.21, "learning_rate": 4.7431080530532804e-05, "loss": 0.8683, "step": 19984 }, { "epoch": 1.21, "learning_rate": 4.742957162041857e-05, "loss": 1.0457, "step": 19986 }, { "epoch": 1.21, "learning_rate": 4.742806271030435e-05, "loss": 1.1964, "step": 19988 }, { "epoch": 1.21, "learning_rate": 4.7426553800190124e-05, "loss": 0.986, "step": 19990 }, { "epoch": 1.21, "learning_rate": 4.74250448900759e-05, "loss": 1.1781, "step": 19992 }, { "epoch": 1.21, "learning_rate": 4.7423535979961677e-05, "loss": 0.9603, "step": 19994 }, { "epoch": 1.21, "learning_rate": 4.742202706984745e-05, "loss": 0.8174, "step": 19996 }, { "epoch": 1.21, "learning_rate": 4.742051815973323e-05, "loss": 1.6, "step": 19998 }, { "epoch": 1.21, "learning_rate": 4.7419009249619e-05, "loss": 1.3357, "step": 20000 }, { "epoch": 1.21, "learning_rate": 4.7417500339504776e-05, "loss": 1.0639, "step": 20002 }, { "epoch": 1.21, "learning_rate": 4.7415991429390556e-05, "loss": 0.9894, "step": 20004 }, { "epoch": 1.21, "learning_rate": 4.741448251927633e-05, "loss": 1.0616, "step": 20006 }, { "epoch": 1.21, "learning_rate": 4.74129736091621e-05, "loss": 1.0238, "step": 20008 }, { "epoch": 1.21, "learning_rate": 4.741146469904788e-05, "loss": 0.7033, "step": 20010 }, { "epoch": 1.21, "learning_rate": 4.7409955788933655e-05, "loss": 0.988, "step": 20012 }, { "epoch": 1.21, "learning_rate": 4.740844687881943e-05, "loss": 1.0377, "step": 20014 }, { "epoch": 1.21, "learning_rate": 4.740693796870521e-05, "loss": 0.9766, "step": 20016 }, { "epoch": 1.21, "learning_rate": 4.740542905859098e-05, "loss": 1.1257, "step": 20018 }, { "epoch": 1.21, "learning_rate": 4.7403920148476754e-05, "loss": 1.5392, "step": 20020 }, { "epoch": 1.21, "learning_rate": 4.7402411238362534e-05, "loss": 0.967, "step": 20022 }, { "epoch": 1.21, "learning_rate": 4.740090232824831e-05, "loss": 0.7356, "step": 20024 }, { "epoch": 1.21, "learning_rate": 4.739939341813409e-05, "loss": 0.7035, "step": 20026 }, { "epoch": 1.21, "learning_rate": 4.739788450801986e-05, "loss": 1.0202, "step": 20028 }, { "epoch": 1.21, "learning_rate": 4.7396375597905633e-05, "loss": 0.944, "step": 20030 }, { "epoch": 1.21, "learning_rate": 4.739486668779141e-05, "loss": 1.1592, "step": 20032 }, { "epoch": 1.21, "learning_rate": 4.7393357777677186e-05, "loss": 1.2714, "step": 20034 }, { "epoch": 1.21, "learning_rate": 4.739184886756296e-05, "loss": 0.7955, "step": 20036 }, { "epoch": 1.21, "learning_rate": 4.739033995744874e-05, "loss": 0.6567, "step": 20038 }, { "epoch": 1.21, "learning_rate": 4.738883104733451e-05, "loss": 1.4327, "step": 20040 }, { "epoch": 1.21, "learning_rate": 4.7387322137220286e-05, "loss": 0.9228, "step": 20042 }, { "epoch": 1.21, "learning_rate": 4.7385813227106066e-05, "loss": 1.1242, "step": 20044 }, { "epoch": 1.21, "learning_rate": 4.738430431699184e-05, "loss": 1.247, "step": 20046 }, { "epoch": 1.21, "learning_rate": 4.738279540687762e-05, "loss": 0.9736, "step": 20048 }, { "epoch": 1.21, "learning_rate": 4.738128649676339e-05, "loss": 0.9841, "step": 20050 }, { "epoch": 1.21, "learning_rate": 4.7379777586649165e-05, "loss": 1.0192, "step": 20052 }, { "epoch": 1.21, "learning_rate": 4.7378268676534945e-05, "loss": 1.2716, "step": 20054 }, { "epoch": 1.21, "learning_rate": 4.737675976642071e-05, "loss": 1.1314, "step": 20056 }, { "epoch": 1.21, "learning_rate": 4.737525085630649e-05, "loss": 0.9983, "step": 20058 }, { "epoch": 1.21, "learning_rate": 4.737374194619227e-05, "loss": 0.9996, "step": 20060 }, { "epoch": 1.21, "learning_rate": 4.7372233036078044e-05, "loss": 1.0752, "step": 20062 }, { "epoch": 1.21, "learning_rate": 4.737072412596382e-05, "loss": 1.1873, "step": 20064 }, { "epoch": 1.21, "learning_rate": 4.73692152158496e-05, "loss": 1.485, "step": 20066 }, { "epoch": 1.21, "learning_rate": 4.736770630573537e-05, "loss": 0.8587, "step": 20068 }, { "epoch": 1.21, "learning_rate": 4.7366197395621143e-05, "loss": 1.1082, "step": 20070 }, { "epoch": 1.21, "learning_rate": 4.7364688485506917e-05, "loss": 1.201, "step": 20072 }, { "epoch": 1.21, "learning_rate": 4.7363179575392696e-05, "loss": 1.043, "step": 20074 }, { "epoch": 1.21, "learning_rate": 4.7361670665278476e-05, "loss": 0.9232, "step": 20076 }, { "epoch": 1.21, "learning_rate": 4.736016175516424e-05, "loss": 1.0458, "step": 20078 }, { "epoch": 1.21, "learning_rate": 4.735865284505002e-05, "loss": 0.9659, "step": 20080 }, { "epoch": 1.21, "learning_rate": 4.73571439349358e-05, "loss": 0.9791, "step": 20082 }, { "epoch": 1.21, "learning_rate": 4.7355635024821576e-05, "loss": 0.9472, "step": 20084 }, { "epoch": 1.21, "learning_rate": 4.735412611470735e-05, "loss": 1.1294, "step": 20086 }, { "epoch": 1.21, "learning_rate": 4.735261720459312e-05, "loss": 0.9592, "step": 20088 }, { "epoch": 1.21, "learning_rate": 4.73511082944789e-05, "loss": 1.0079, "step": 20090 }, { "epoch": 1.21, "learning_rate": 4.7349599384364675e-05, "loss": 1.1876, "step": 20092 }, { "epoch": 1.21, "learning_rate": 4.734809047425045e-05, "loss": 0.9004, "step": 20094 }, { "epoch": 1.21, "learning_rate": 4.734658156413623e-05, "loss": 1.0819, "step": 20096 }, { "epoch": 1.21, "learning_rate": 4.734507265402201e-05, "loss": 0.9233, "step": 20098 }, { "epoch": 1.21, "learning_rate": 4.7343563743907774e-05, "loss": 1.0036, "step": 20100 }, { "epoch": 1.21, "learning_rate": 4.7342054833793554e-05, "loss": 1.0166, "step": 20102 }, { "epoch": 1.21, "learning_rate": 4.734054592367933e-05, "loss": 1.0704, "step": 20104 }, { "epoch": 1.21, "learning_rate": 4.73390370135651e-05, "loss": 1.0718, "step": 20106 }, { "epoch": 1.21, "learning_rate": 4.733752810345088e-05, "loss": 1.1163, "step": 20108 }, { "epoch": 1.21, "learning_rate": 4.733601919333665e-05, "loss": 1.173, "step": 20110 }, { "epoch": 1.21, "learning_rate": 4.733451028322243e-05, "loss": 1.3031, "step": 20112 }, { "epoch": 1.21, "learning_rate": 4.7333001373108206e-05, "loss": 1.092, "step": 20114 }, { "epoch": 1.21, "learning_rate": 4.733149246299398e-05, "loss": 0.9531, "step": 20116 }, { "epoch": 1.21, "learning_rate": 4.732998355287976e-05, "loss": 1.3551, "step": 20118 }, { "epoch": 1.21, "learning_rate": 4.732847464276553e-05, "loss": 1.0243, "step": 20120 }, { "epoch": 1.21, "learning_rate": 4.7326965732651306e-05, "loss": 0.9986, "step": 20122 }, { "epoch": 1.21, "learning_rate": 4.7325456822537086e-05, "loss": 1.2076, "step": 20124 }, { "epoch": 1.21, "learning_rate": 4.732394791242286e-05, "loss": 1.0571, "step": 20126 }, { "epoch": 1.21, "learning_rate": 4.732243900230863e-05, "loss": 0.8199, "step": 20128 }, { "epoch": 1.21, "learning_rate": 4.732093009219441e-05, "loss": 1.2312, "step": 20130 }, { "epoch": 1.22, "learning_rate": 4.7319421182080185e-05, "loss": 0.7931, "step": 20132 }, { "epoch": 1.22, "learning_rate": 4.7317912271965965e-05, "loss": 1.2018, "step": 20134 }, { "epoch": 1.22, "learning_rate": 4.731640336185174e-05, "loss": 0.9454, "step": 20136 }, { "epoch": 1.22, "learning_rate": 4.731489445173751e-05, "loss": 1.221, "step": 20138 }, { "epoch": 1.22, "learning_rate": 4.731338554162329e-05, "loss": 1.3348, "step": 20140 }, { "epoch": 1.22, "learning_rate": 4.7311876631509064e-05, "loss": 1.0225, "step": 20142 }, { "epoch": 1.22, "learning_rate": 4.731036772139484e-05, "loss": 1.2231, "step": 20144 }, { "epoch": 1.22, "learning_rate": 4.730885881128062e-05, "loss": 1.5499, "step": 20146 }, { "epoch": 1.22, "learning_rate": 4.730734990116639e-05, "loss": 1.181, "step": 20148 }, { "epoch": 1.22, "learning_rate": 4.730584099105216e-05, "loss": 1.0855, "step": 20150 }, { "epoch": 1.22, "learning_rate": 4.730433208093794e-05, "loss": 1.1072, "step": 20152 }, { "epoch": 1.22, "learning_rate": 4.7302823170823716e-05, "loss": 1.1231, "step": 20154 }, { "epoch": 1.22, "learning_rate": 4.730131426070949e-05, "loss": 1.1791, "step": 20156 }, { "epoch": 1.22, "learning_rate": 4.729980535059527e-05, "loss": 1.3944, "step": 20158 }, { "epoch": 1.22, "learning_rate": 4.729829644048104e-05, "loss": 1.2259, "step": 20160 }, { "epoch": 1.22, "learning_rate": 4.729678753036682e-05, "loss": 1.2868, "step": 20162 }, { "epoch": 1.22, "learning_rate": 4.7295278620252595e-05, "loss": 1.0395, "step": 20164 }, { "epoch": 1.22, "learning_rate": 4.729376971013837e-05, "loss": 1.1076, "step": 20166 }, { "epoch": 1.22, "learning_rate": 4.729226080002415e-05, "loss": 1.1746, "step": 20168 }, { "epoch": 1.22, "learning_rate": 4.7290751889909915e-05, "loss": 1.1585, "step": 20170 }, { "epoch": 1.22, "learning_rate": 4.7289242979795695e-05, "loss": 1.1087, "step": 20172 }, { "epoch": 1.22, "learning_rate": 4.7287734069681475e-05, "loss": 1.3974, "step": 20174 }, { "epoch": 1.22, "learning_rate": 4.728622515956725e-05, "loss": 1.207, "step": 20176 }, { "epoch": 1.22, "learning_rate": 4.728471624945302e-05, "loss": 0.9365, "step": 20178 }, { "epoch": 1.22, "learning_rate": 4.72832073393388e-05, "loss": 1.0152, "step": 20180 }, { "epoch": 1.22, "learning_rate": 4.7281698429224574e-05, "loss": 1.0162, "step": 20182 }, { "epoch": 1.22, "learning_rate": 4.728018951911035e-05, "loss": 1.1224, "step": 20184 }, { "epoch": 1.22, "learning_rate": 4.727868060899612e-05, "loss": 0.7953, "step": 20186 }, { "epoch": 1.22, "learning_rate": 4.72771716988819e-05, "loss": 0.8238, "step": 20188 }, { "epoch": 1.22, "learning_rate": 4.727566278876768e-05, "loss": 0.8511, "step": 20190 }, { "epoch": 1.22, "learning_rate": 4.7274153878653446e-05, "loss": 1.2893, "step": 20192 }, { "epoch": 1.22, "learning_rate": 4.7272644968539226e-05, "loss": 1.1678, "step": 20194 }, { "epoch": 1.22, "learning_rate": 4.7271136058425006e-05, "loss": 1.2252, "step": 20196 }, { "epoch": 1.22, "learning_rate": 4.726962714831078e-05, "loss": 0.6669, "step": 20198 }, { "epoch": 1.22, "learning_rate": 4.726811823819655e-05, "loss": 0.9417, "step": 20200 }, { "epoch": 1.22, "learning_rate": 4.7266609328082326e-05, "loss": 0.7343, "step": 20202 }, { "epoch": 1.22, "learning_rate": 4.7265100417968105e-05, "loss": 0.9142, "step": 20204 }, { "epoch": 1.22, "learning_rate": 4.726359150785388e-05, "loss": 0.9973, "step": 20206 }, { "epoch": 1.22, "learning_rate": 4.726208259773965e-05, "loss": 1.1917, "step": 20208 }, { "epoch": 1.22, "learning_rate": 4.726057368762543e-05, "loss": 0.9497, "step": 20210 }, { "epoch": 1.22, "learning_rate": 4.725906477751121e-05, "loss": 0.9358, "step": 20212 }, { "epoch": 1.22, "learning_rate": 4.725755586739698e-05, "loss": 1.0612, "step": 20214 }, { "epoch": 1.22, "learning_rate": 4.725604695728276e-05, "loss": 0.9144, "step": 20216 }, { "epoch": 1.22, "learning_rate": 4.725453804716853e-05, "loss": 0.965, "step": 20218 }, { "epoch": 1.22, "learning_rate": 4.7253029137054304e-05, "loss": 0.8975, "step": 20220 }, { "epoch": 1.22, "learning_rate": 4.7251520226940084e-05, "loss": 0.9244, "step": 20222 }, { "epoch": 1.22, "learning_rate": 4.725001131682586e-05, "loss": 1.2382, "step": 20224 }, { "epoch": 1.22, "learning_rate": 4.724850240671164e-05, "loss": 0.846, "step": 20226 }, { "epoch": 1.22, "learning_rate": 4.724699349659741e-05, "loss": 0.6726, "step": 20228 }, { "epoch": 1.22, "learning_rate": 4.724548458648318e-05, "loss": 0.9087, "step": 20230 }, { "epoch": 1.22, "learning_rate": 4.724397567636896e-05, "loss": 0.823, "step": 20232 }, { "epoch": 1.22, "learning_rate": 4.7242466766254736e-05, "loss": 1.3692, "step": 20234 }, { "epoch": 1.22, "learning_rate": 4.724095785614051e-05, "loss": 1.226, "step": 20236 }, { "epoch": 1.22, "learning_rate": 4.723944894602629e-05, "loss": 1.1727, "step": 20238 }, { "epoch": 1.22, "learning_rate": 4.723794003591206e-05, "loss": 1.077, "step": 20240 }, { "epoch": 1.22, "learning_rate": 4.7236431125797835e-05, "loss": 1.155, "step": 20242 }, { "epoch": 1.22, "learning_rate": 4.7234922215683615e-05, "loss": 1.8279, "step": 20244 }, { "epoch": 1.22, "learning_rate": 4.723341330556939e-05, "loss": 1.0919, "step": 20246 }, { "epoch": 1.22, "learning_rate": 4.723190439545517e-05, "loss": 1.3528, "step": 20248 }, { "epoch": 1.22, "learning_rate": 4.723039548534094e-05, "loss": 0.8978, "step": 20250 }, { "epoch": 1.22, "learning_rate": 4.7228886575226715e-05, "loss": 1.0412, "step": 20252 }, { "epoch": 1.22, "learning_rate": 4.7227377665112495e-05, "loss": 0.8293, "step": 20254 }, { "epoch": 1.22, "learning_rate": 4.722586875499827e-05, "loss": 1.0544, "step": 20256 }, { "epoch": 1.22, "learning_rate": 4.722435984488404e-05, "loss": 1.136, "step": 20258 }, { "epoch": 1.22, "learning_rate": 4.722285093476982e-05, "loss": 1.2259, "step": 20260 }, { "epoch": 1.22, "learning_rate": 4.7221342024655594e-05, "loss": 1.1383, "step": 20262 }, { "epoch": 1.22, "learning_rate": 4.721983311454137e-05, "loss": 1.3383, "step": 20264 }, { "epoch": 1.22, "learning_rate": 4.721832420442715e-05, "loss": 0.7341, "step": 20266 }, { "epoch": 1.22, "learning_rate": 4.721681529431292e-05, "loss": 1.05, "step": 20268 }, { "epoch": 1.22, "learning_rate": 4.721530638419869e-05, "loss": 1.2696, "step": 20270 }, { "epoch": 1.22, "learning_rate": 4.721379747408447e-05, "loss": 0.8577, "step": 20272 }, { "epoch": 1.22, "learning_rate": 4.7212288563970246e-05, "loss": 0.6969, "step": 20274 }, { "epoch": 1.22, "learning_rate": 4.7210779653856026e-05, "loss": 1.2048, "step": 20276 }, { "epoch": 1.22, "learning_rate": 4.72092707437418e-05, "loss": 0.8245, "step": 20278 }, { "epoch": 1.22, "learning_rate": 4.720776183362757e-05, "loss": 1.0171, "step": 20280 }, { "epoch": 1.22, "learning_rate": 4.720625292351335e-05, "loss": 0.6983, "step": 20282 }, { "epoch": 1.22, "learning_rate": 4.720474401339912e-05, "loss": 1.1214, "step": 20284 }, { "epoch": 1.22, "learning_rate": 4.72032351032849e-05, "loss": 1.1371, "step": 20286 }, { "epoch": 1.22, "learning_rate": 4.720172619317068e-05, "loss": 1.0938, "step": 20288 }, { "epoch": 1.22, "learning_rate": 4.720021728305645e-05, "loss": 0.8332, "step": 20290 }, { "epoch": 1.22, "learning_rate": 4.7198708372942225e-05, "loss": 1.3038, "step": 20292 }, { "epoch": 1.22, "learning_rate": 4.7197199462828005e-05, "loss": 1.1225, "step": 20294 }, { "epoch": 1.23, "learning_rate": 4.719569055271378e-05, "loss": 0.9988, "step": 20296 }, { "epoch": 1.23, "learning_rate": 4.719418164259956e-05, "loss": 0.926, "step": 20298 }, { "epoch": 1.23, "learning_rate": 4.7192672732485324e-05, "loss": 1.0386, "step": 20300 }, { "epoch": 1.23, "learning_rate": 4.7191163822371104e-05, "loss": 1.075, "step": 20302 }, { "epoch": 1.23, "learning_rate": 4.7189654912256884e-05, "loss": 0.9841, "step": 20304 }, { "epoch": 1.23, "learning_rate": 4.718814600214265e-05, "loss": 1.3254, "step": 20306 }, { "epoch": 1.23, "learning_rate": 4.718663709202843e-05, "loss": 0.8298, "step": 20308 }, { "epoch": 1.23, "learning_rate": 4.718512818191421e-05, "loss": 1.1133, "step": 20310 }, { "epoch": 1.23, "learning_rate": 4.718361927179998e-05, "loss": 1.1581, "step": 20312 }, { "epoch": 1.23, "learning_rate": 4.7182110361685756e-05, "loss": 0.9763, "step": 20314 }, { "epoch": 1.23, "learning_rate": 4.718060145157153e-05, "loss": 1.0155, "step": 20316 }, { "epoch": 1.23, "learning_rate": 4.717909254145731e-05, "loss": 0.8235, "step": 20318 }, { "epoch": 1.23, "learning_rate": 4.717758363134308e-05, "loss": 0.9249, "step": 20320 }, { "epoch": 1.23, "learning_rate": 4.7176074721228855e-05, "loss": 1.0691, "step": 20322 }, { "epoch": 1.23, "learning_rate": 4.7174565811114635e-05, "loss": 1.1108, "step": 20324 }, { "epoch": 1.23, "learning_rate": 4.7173056901000415e-05, "loss": 0.815, "step": 20326 }, { "epoch": 1.23, "learning_rate": 4.717154799088618e-05, "loss": 1.2219, "step": 20328 }, { "epoch": 1.23, "learning_rate": 4.717003908077196e-05, "loss": 0.7135, "step": 20330 }, { "epoch": 1.23, "learning_rate": 4.7168530170657735e-05, "loss": 1.1245, "step": 20332 }, { "epoch": 1.23, "learning_rate": 4.716702126054351e-05, "loss": 0.9492, "step": 20334 }, { "epoch": 1.23, "learning_rate": 4.716551235042929e-05, "loss": 1.0426, "step": 20336 }, { "epoch": 1.23, "learning_rate": 4.716400344031506e-05, "loss": 0.9109, "step": 20338 }, { "epoch": 1.23, "learning_rate": 4.716249453020084e-05, "loss": 1.1794, "step": 20340 }, { "epoch": 1.23, "learning_rate": 4.7160985620086614e-05, "loss": 0.9844, "step": 20342 }, { "epoch": 1.23, "learning_rate": 4.715947670997239e-05, "loss": 0.9308, "step": 20344 }, { "epoch": 1.23, "learning_rate": 4.715796779985817e-05, "loss": 0.8017, "step": 20346 }, { "epoch": 1.23, "learning_rate": 4.715645888974394e-05, "loss": 1.2476, "step": 20348 }, { "epoch": 1.23, "learning_rate": 4.715494997962971e-05, "loss": 1.0104, "step": 20350 }, { "epoch": 1.23, "learning_rate": 4.715344106951549e-05, "loss": 0.9872, "step": 20352 }, { "epoch": 1.23, "learning_rate": 4.7151932159401266e-05, "loss": 1.1071, "step": 20354 }, { "epoch": 1.23, "learning_rate": 4.715042324928704e-05, "loss": 0.9016, "step": 20356 }, { "epoch": 1.23, "learning_rate": 4.714891433917282e-05, "loss": 1.0557, "step": 20358 }, { "epoch": 1.23, "learning_rate": 4.714740542905859e-05, "loss": 1.1778, "step": 20360 }, { "epoch": 1.23, "learning_rate": 4.714589651894437e-05, "loss": 1.0124, "step": 20362 }, { "epoch": 1.23, "learning_rate": 4.7144387608830145e-05, "loss": 1.3366, "step": 20364 }, { "epoch": 1.23, "learning_rate": 4.714287869871592e-05, "loss": 0.9996, "step": 20366 }, { "epoch": 1.23, "learning_rate": 4.71413697886017e-05, "loss": 1.0677, "step": 20368 }, { "epoch": 1.23, "learning_rate": 4.713986087848747e-05, "loss": 1.0859, "step": 20370 }, { "epoch": 1.23, "learning_rate": 4.7138351968373244e-05, "loss": 0.9556, "step": 20372 }, { "epoch": 1.23, "learning_rate": 4.7136843058259024e-05, "loss": 0.9368, "step": 20374 }, { "epoch": 1.23, "learning_rate": 4.71353341481448e-05, "loss": 0.9576, "step": 20376 }, { "epoch": 1.23, "learning_rate": 4.713382523803057e-05, "loss": 0.8497, "step": 20378 }, { "epoch": 1.23, "learning_rate": 4.713231632791635e-05, "loss": 1.0706, "step": 20380 }, { "epoch": 1.23, "learning_rate": 4.7130807417802124e-05, "loss": 1.3109, "step": 20382 }, { "epoch": 1.23, "learning_rate": 4.71292985076879e-05, "loss": 1.2014, "step": 20384 }, { "epoch": 1.23, "learning_rate": 4.712778959757368e-05, "loss": 1.0813, "step": 20386 }, { "epoch": 1.23, "learning_rate": 4.712628068745945e-05, "loss": 1.152, "step": 20388 }, { "epoch": 1.23, "learning_rate": 4.712477177734523e-05, "loss": 1.1259, "step": 20390 }, { "epoch": 1.23, "learning_rate": 4.7123262867231e-05, "loss": 1.0255, "step": 20392 }, { "epoch": 1.23, "learning_rate": 4.7121753957116776e-05, "loss": 1.0575, "step": 20394 }, { "epoch": 1.23, "learning_rate": 4.7120245047002556e-05, "loss": 0.8987, "step": 20396 }, { "epoch": 1.23, "learning_rate": 4.711873613688832e-05, "loss": 1.0974, "step": 20398 }, { "epoch": 1.23, "learning_rate": 4.71172272267741e-05, "loss": 1.1955, "step": 20400 }, { "epoch": 1.23, "learning_rate": 4.711571831665988e-05, "loss": 1.0901, "step": 20402 }, { "epoch": 1.23, "learning_rate": 4.7114209406545655e-05, "loss": 0.9582, "step": 20404 }, { "epoch": 1.23, "learning_rate": 4.711270049643143e-05, "loss": 1.0745, "step": 20406 }, { "epoch": 1.23, "learning_rate": 4.711119158631721e-05, "loss": 1.2181, "step": 20408 }, { "epoch": 1.23, "learning_rate": 4.710968267620298e-05, "loss": 1.0931, "step": 20410 }, { "epoch": 1.23, "learning_rate": 4.710817376608876e-05, "loss": 0.9739, "step": 20412 }, { "epoch": 1.23, "learning_rate": 4.710666485597453e-05, "loss": 1.1011, "step": 20414 }, { "epoch": 1.23, "learning_rate": 4.710515594586031e-05, "loss": 0.8619, "step": 20416 }, { "epoch": 1.23, "learning_rate": 4.710364703574609e-05, "loss": 1.3276, "step": 20418 }, { "epoch": 1.23, "learning_rate": 4.7102138125631854e-05, "loss": 0.9484, "step": 20420 }, { "epoch": 1.23, "learning_rate": 4.7100629215517634e-05, "loss": 1.4373, "step": 20422 }, { "epoch": 1.23, "learning_rate": 4.7099120305403414e-05, "loss": 0.9597, "step": 20424 }, { "epoch": 1.23, "learning_rate": 4.709761139528919e-05, "loss": 1.1974, "step": 20426 }, { "epoch": 1.23, "learning_rate": 4.709610248517496e-05, "loss": 1.3626, "step": 20428 }, { "epoch": 1.23, "learning_rate": 4.709459357506073e-05, "loss": 0.8656, "step": 20430 }, { "epoch": 1.23, "learning_rate": 4.709308466494651e-05, "loss": 1.0942, "step": 20432 }, { "epoch": 1.23, "learning_rate": 4.7091575754832286e-05, "loss": 0.8508, "step": 20434 }, { "epoch": 1.23, "learning_rate": 4.709006684471806e-05, "loss": 1.1166, "step": 20436 }, { "epoch": 1.23, "learning_rate": 4.708855793460384e-05, "loss": 0.9868, "step": 20438 }, { "epoch": 1.23, "learning_rate": 4.708704902448962e-05, "loss": 1.1083, "step": 20440 }, { "epoch": 1.23, "learning_rate": 4.7085540114375385e-05, "loss": 1.0361, "step": 20442 }, { "epoch": 1.23, "learning_rate": 4.7084031204261165e-05, "loss": 1.086, "step": 20444 }, { "epoch": 1.23, "learning_rate": 4.708252229414694e-05, "loss": 1.0518, "step": 20446 }, { "epoch": 1.23, "learning_rate": 4.708101338403271e-05, "loss": 0.8322, "step": 20448 }, { "epoch": 1.23, "learning_rate": 4.707950447391849e-05, "loss": 0.696, "step": 20450 }, { "epoch": 1.23, "learning_rate": 4.7077995563804264e-05, "loss": 1.1573, "step": 20452 }, { "epoch": 1.23, "learning_rate": 4.7076486653690044e-05, "loss": 1.0257, "step": 20454 }, { "epoch": 1.23, "learning_rate": 4.707497774357582e-05, "loss": 1.451, "step": 20456 }, { "epoch": 1.23, "learning_rate": 4.707346883346159e-05, "loss": 0.994, "step": 20458 }, { "epoch": 1.23, "learning_rate": 4.707195992334737e-05, "loss": 0.8454, "step": 20460 }, { "epoch": 1.24, "learning_rate": 4.7070451013233144e-05, "loss": 0.9575, "step": 20462 }, { "epoch": 1.24, "learning_rate": 4.706894210311892e-05, "loss": 1.0221, "step": 20464 }, { "epoch": 1.24, "learning_rate": 4.7067433193004697e-05, "loss": 1.2347, "step": 20466 }, { "epoch": 1.24, "learning_rate": 4.706592428289047e-05, "loss": 0.7631, "step": 20468 }, { "epoch": 1.24, "learning_rate": 4.706441537277624e-05, "loss": 0.8896, "step": 20470 }, { "epoch": 1.24, "learning_rate": 4.706290646266202e-05, "loss": 1.1628, "step": 20472 }, { "epoch": 1.24, "learning_rate": 4.7061397552547796e-05, "loss": 1.2987, "step": 20474 }, { "epoch": 1.24, "learning_rate": 4.7059888642433576e-05, "loss": 1.3825, "step": 20476 }, { "epoch": 1.24, "learning_rate": 4.705837973231935e-05, "loss": 0.7929, "step": 20478 }, { "epoch": 1.24, "learning_rate": 4.705687082220512e-05, "loss": 0.8431, "step": 20480 }, { "epoch": 1.24, "learning_rate": 4.70553619120909e-05, "loss": 0.9704, "step": 20482 }, { "epoch": 1.24, "learning_rate": 4.7053853001976675e-05, "loss": 0.7706, "step": 20484 }, { "epoch": 1.24, "learning_rate": 4.705234409186245e-05, "loss": 1.1497, "step": 20486 }, { "epoch": 1.24, "learning_rate": 4.705083518174823e-05, "loss": 1.3915, "step": 20488 }, { "epoch": 1.24, "learning_rate": 4.7049326271634e-05, "loss": 0.9805, "step": 20490 }, { "epoch": 1.24, "learning_rate": 4.7047817361519774e-05, "loss": 1.2682, "step": 20492 }, { "epoch": 1.24, "learning_rate": 4.7046308451405554e-05, "loss": 0.9802, "step": 20494 }, { "epoch": 1.24, "learning_rate": 4.704479954129133e-05, "loss": 0.7741, "step": 20496 }, { "epoch": 1.24, "learning_rate": 4.70432906311771e-05, "loss": 1.1053, "step": 20498 }, { "epoch": 1.24, "learning_rate": 4.704178172106288e-05, "loss": 0.9433, "step": 20500 }, { "epoch": 1.24, "learning_rate": 4.7040272810948653e-05, "loss": 1.2472, "step": 20502 }, { "epoch": 1.24, "learning_rate": 4.7038763900834433e-05, "loss": 0.9301, "step": 20504 }, { "epoch": 1.24, "learning_rate": 4.7037254990720207e-05, "loss": 0.7013, "step": 20506 }, { "epoch": 1.24, "learning_rate": 4.703574608060598e-05, "loss": 1.1102, "step": 20508 }, { "epoch": 1.24, "learning_rate": 4.703423717049176e-05, "loss": 1.0886, "step": 20510 }, { "epoch": 1.24, "learning_rate": 4.703272826037753e-05, "loss": 1.4823, "step": 20512 }, { "epoch": 1.24, "learning_rate": 4.7031219350263306e-05, "loss": 1.0287, "step": 20514 }, { "epoch": 1.24, "learning_rate": 4.7029710440149086e-05, "loss": 1.033, "step": 20516 }, { "epoch": 1.24, "learning_rate": 4.702820153003486e-05, "loss": 1.1186, "step": 20518 }, { "epoch": 1.24, "learning_rate": 4.702669261992063e-05, "loss": 1.0118, "step": 20520 }, { "epoch": 1.24, "learning_rate": 4.702518370980641e-05, "loss": 1.0185, "step": 20522 }, { "epoch": 1.24, "learning_rate": 4.7023674799692185e-05, "loss": 1.1104, "step": 20524 }, { "epoch": 1.24, "learning_rate": 4.7022165889577965e-05, "loss": 1.0353, "step": 20526 }, { "epoch": 1.24, "learning_rate": 4.702065697946373e-05, "loss": 1.0506, "step": 20528 }, { "epoch": 1.24, "learning_rate": 4.701914806934951e-05, "loss": 0.9158, "step": 20530 }, { "epoch": 1.24, "learning_rate": 4.701763915923529e-05, "loss": 1.0118, "step": 20532 }, { "epoch": 1.24, "learning_rate": 4.701613024912106e-05, "loss": 0.8644, "step": 20534 }, { "epoch": 1.24, "learning_rate": 4.701462133900684e-05, "loss": 0.9985, "step": 20536 }, { "epoch": 1.24, "learning_rate": 4.701311242889262e-05, "loss": 1.0691, "step": 20538 }, { "epoch": 1.24, "learning_rate": 4.701160351877839e-05, "loss": 0.8037, "step": 20540 }, { "epoch": 1.24, "learning_rate": 4.7010094608664163e-05, "loss": 1.0275, "step": 20542 }, { "epoch": 1.24, "learning_rate": 4.7008585698549937e-05, "loss": 0.9953, "step": 20544 }, { "epoch": 1.24, "learning_rate": 4.7007076788435716e-05, "loss": 1.0417, "step": 20546 }, { "epoch": 1.24, "learning_rate": 4.700556787832149e-05, "loss": 1.0706, "step": 20548 }, { "epoch": 1.24, "learning_rate": 4.700405896820726e-05, "loss": 0.9584, "step": 20550 }, { "epoch": 1.24, "learning_rate": 4.700255005809304e-05, "loss": 1.2454, "step": 20552 }, { "epoch": 1.24, "learning_rate": 4.700104114797882e-05, "loss": 0.972, "step": 20554 }, { "epoch": 1.24, "learning_rate": 4.699953223786459e-05, "loss": 0.8132, "step": 20556 }, { "epoch": 1.24, "learning_rate": 4.699802332775037e-05, "loss": 0.7347, "step": 20558 }, { "epoch": 1.24, "learning_rate": 4.699651441763614e-05, "loss": 0.8747, "step": 20560 }, { "epoch": 1.24, "learning_rate": 4.6995005507521915e-05, "loss": 0.6795, "step": 20562 }, { "epoch": 1.24, "learning_rate": 4.6993496597407695e-05, "loss": 0.7756, "step": 20564 }, { "epoch": 1.24, "learning_rate": 4.699198768729347e-05, "loss": 0.8584, "step": 20566 }, { "epoch": 1.24, "learning_rate": 4.699047877717925e-05, "loss": 1.394, "step": 20568 }, { "epoch": 1.24, "learning_rate": 4.698896986706502e-05, "loss": 1.3761, "step": 20570 }, { "epoch": 1.24, "learning_rate": 4.6987460956950794e-05, "loss": 0.8473, "step": 20572 }, { "epoch": 1.24, "learning_rate": 4.6985952046836574e-05, "loss": 1.0578, "step": 20574 }, { "epoch": 1.24, "learning_rate": 4.698444313672235e-05, "loss": 1.4124, "step": 20576 }, { "epoch": 1.24, "learning_rate": 4.698293422660812e-05, "loss": 1.2668, "step": 20578 }, { "epoch": 1.24, "learning_rate": 4.69814253164939e-05, "loss": 1.0076, "step": 20580 }, { "epoch": 1.24, "learning_rate": 4.697991640637967e-05, "loss": 1.1434, "step": 20582 }, { "epoch": 1.24, "learning_rate": 4.6978407496265447e-05, "loss": 1.4793, "step": 20584 }, { "epoch": 1.24, "learning_rate": 4.6976898586151226e-05, "loss": 1.0533, "step": 20586 }, { "epoch": 1.24, "learning_rate": 4.6975389676037e-05, "loss": 1.1749, "step": 20588 }, { "epoch": 1.24, "learning_rate": 4.697388076592278e-05, "loss": 1.0633, "step": 20590 }, { "epoch": 1.24, "learning_rate": 4.697237185580855e-05, "loss": 1.4286, "step": 20592 }, { "epoch": 1.24, "learning_rate": 4.6970862945694326e-05, "loss": 1.1502, "step": 20594 }, { "epoch": 1.24, "learning_rate": 4.6969354035580106e-05, "loss": 0.8509, "step": 20596 }, { "epoch": 1.24, "learning_rate": 4.696784512546588e-05, "loss": 0.9264, "step": 20598 }, { "epoch": 1.24, "learning_rate": 4.696633621535165e-05, "loss": 1.056, "step": 20600 }, { "epoch": 1.24, "learning_rate": 4.696482730523743e-05, "loss": 1.0994, "step": 20602 }, { "epoch": 1.24, "learning_rate": 4.6963318395123205e-05, "loss": 0.8579, "step": 20604 }, { "epoch": 1.24, "learning_rate": 4.696180948500898e-05, "loss": 0.9788, "step": 20606 }, { "epoch": 1.24, "learning_rate": 4.696030057489476e-05, "loss": 1.1432, "step": 20608 }, { "epoch": 1.24, "learning_rate": 4.695879166478053e-05, "loss": 1.0235, "step": 20610 }, { "epoch": 1.24, "learning_rate": 4.6957282754666304e-05, "loss": 1.1887, "step": 20612 }, { "epoch": 1.24, "learning_rate": 4.6955773844552084e-05, "loss": 1.0579, "step": 20614 }, { "epoch": 1.24, "learning_rate": 4.695426493443786e-05, "loss": 0.9095, "step": 20616 }, { "epoch": 1.24, "learning_rate": 4.695275602432364e-05, "loss": 1.2032, "step": 20618 }, { "epoch": 1.24, "learning_rate": 4.6951247114209403e-05, "loss": 1.6559, "step": 20620 }, { "epoch": 1.24, "learning_rate": 4.694973820409518e-05, "loss": 1.0964, "step": 20622 }, { "epoch": 1.24, "learning_rate": 4.694822929398096e-05, "loss": 1.0144, "step": 20624 }, { "epoch": 1.24, "learning_rate": 4.6946720383866736e-05, "loss": 1.0492, "step": 20626 }, { "epoch": 1.25, "learning_rate": 4.694521147375251e-05, "loss": 0.7847, "step": 20628 }, { "epoch": 1.25, "learning_rate": 4.694370256363829e-05, "loss": 1.0965, "step": 20630 }, { "epoch": 1.25, "learning_rate": 4.694219365352406e-05, "loss": 1.4503, "step": 20632 }, { "epoch": 1.25, "learning_rate": 4.6940684743409836e-05, "loss": 0.9777, "step": 20634 }, { "epoch": 1.25, "learning_rate": 4.6939175833295616e-05, "loss": 0.9216, "step": 20636 }, { "epoch": 1.25, "learning_rate": 4.693766692318139e-05, "loss": 1.1575, "step": 20638 }, { "epoch": 1.25, "learning_rate": 4.693615801306717e-05, "loss": 1.3619, "step": 20640 }, { "epoch": 1.25, "learning_rate": 4.6934649102952935e-05, "loss": 0.8503, "step": 20642 }, { "epoch": 1.25, "learning_rate": 4.6933140192838715e-05, "loss": 1.0945, "step": 20644 }, { "epoch": 1.25, "learning_rate": 4.6931631282724495e-05, "loss": 0.8622, "step": 20646 }, { "epoch": 1.25, "learning_rate": 4.693012237261026e-05, "loss": 1.029, "step": 20648 }, { "epoch": 1.25, "learning_rate": 4.692861346249604e-05, "loss": 0.9849, "step": 20650 }, { "epoch": 1.25, "learning_rate": 4.692710455238182e-05, "loss": 0.6439, "step": 20652 }, { "epoch": 1.25, "learning_rate": 4.6925595642267594e-05, "loss": 1.0685, "step": 20654 }, { "epoch": 1.25, "learning_rate": 4.692408673215337e-05, "loss": 0.8471, "step": 20656 }, { "epoch": 1.25, "learning_rate": 4.692257782203914e-05, "loss": 0.8831, "step": 20658 }, { "epoch": 1.25, "learning_rate": 4.692106891192492e-05, "loss": 0.7369, "step": 20660 }, { "epoch": 1.25, "learning_rate": 4.691956000181069e-05, "loss": 1.019, "step": 20662 }, { "epoch": 1.25, "learning_rate": 4.6918051091696466e-05, "loss": 0.7894, "step": 20664 }, { "epoch": 1.25, "learning_rate": 4.6916542181582246e-05, "loss": 1.0039, "step": 20666 }, { "epoch": 1.25, "learning_rate": 4.6915033271468026e-05, "loss": 0.9825, "step": 20668 }, { "epoch": 1.25, "learning_rate": 4.691352436135379e-05, "loss": 1.2495, "step": 20670 }, { "epoch": 1.25, "learning_rate": 4.691201545123957e-05, "loss": 0.6338, "step": 20672 }, { "epoch": 1.25, "learning_rate": 4.6910506541125346e-05, "loss": 0.8648, "step": 20674 }, { "epoch": 1.25, "learning_rate": 4.6908997631011125e-05, "loss": 0.9041, "step": 20676 }, { "epoch": 1.25, "learning_rate": 4.69074887208969e-05, "loss": 0.7761, "step": 20678 }, { "epoch": 1.25, "learning_rate": 4.690597981078267e-05, "loss": 1.1179, "step": 20680 }, { "epoch": 1.25, "learning_rate": 4.690447090066845e-05, "loss": 0.8062, "step": 20682 }, { "epoch": 1.25, "learning_rate": 4.6902961990554225e-05, "loss": 0.8588, "step": 20684 }, { "epoch": 1.25, "learning_rate": 4.690145308044e-05, "loss": 1.0002, "step": 20686 }, { "epoch": 1.25, "learning_rate": 4.689994417032578e-05, "loss": 0.9215, "step": 20688 }, { "epoch": 1.25, "learning_rate": 4.689843526021155e-05, "loss": 1.0977, "step": 20690 }, { "epoch": 1.25, "learning_rate": 4.6896926350097324e-05, "loss": 0.8018, "step": 20692 }, { "epoch": 1.25, "learning_rate": 4.6895417439983104e-05, "loss": 0.7726, "step": 20694 }, { "epoch": 1.25, "learning_rate": 4.689390852986888e-05, "loss": 1.0077, "step": 20696 }, { "epoch": 1.25, "learning_rate": 4.689239961975465e-05, "loss": 1.3504, "step": 20698 }, { "epoch": 1.25, "learning_rate": 4.689089070964043e-05, "loss": 0.9097, "step": 20700 }, { "epoch": 1.25, "learning_rate": 4.68893817995262e-05, "loss": 0.9782, "step": 20702 }, { "epoch": 1.25, "learning_rate": 4.688787288941198e-05, "loss": 1.3208, "step": 20704 }, { "epoch": 1.25, "learning_rate": 4.6886363979297756e-05, "loss": 0.7765, "step": 20706 }, { "epoch": 1.25, "learning_rate": 4.688485506918353e-05, "loss": 0.987, "step": 20708 }, { "epoch": 1.25, "learning_rate": 4.688334615906931e-05, "loss": 0.9565, "step": 20710 }, { "epoch": 1.25, "learning_rate": 4.688183724895508e-05, "loss": 1.0862, "step": 20712 }, { "epoch": 1.25, "learning_rate": 4.6880328338840856e-05, "loss": 0.8432, "step": 20714 }, { "epoch": 1.25, "learning_rate": 4.6878819428726635e-05, "loss": 0.9619, "step": 20716 }, { "epoch": 1.25, "learning_rate": 4.687731051861241e-05, "loss": 1.0532, "step": 20718 }, { "epoch": 1.25, "learning_rate": 4.687580160849818e-05, "loss": 0.9616, "step": 20720 }, { "epoch": 1.25, "learning_rate": 4.687429269838396e-05, "loss": 1.2645, "step": 20722 }, { "epoch": 1.25, "learning_rate": 4.6872783788269735e-05, "loss": 1.1588, "step": 20724 }, { "epoch": 1.25, "learning_rate": 4.687127487815551e-05, "loss": 0.704, "step": 20726 }, { "epoch": 1.25, "learning_rate": 4.686976596804129e-05, "loss": 1.1585, "step": 20728 }, { "epoch": 1.25, "learning_rate": 4.686825705792706e-05, "loss": 1.142, "step": 20730 }, { "epoch": 1.25, "learning_rate": 4.686674814781284e-05, "loss": 1.28, "step": 20732 }, { "epoch": 1.25, "learning_rate": 4.686523923769861e-05, "loss": 1.2512, "step": 20734 }, { "epoch": 1.25, "learning_rate": 4.686373032758439e-05, "loss": 1.5165, "step": 20736 }, { "epoch": 1.25, "learning_rate": 4.686222141747017e-05, "loss": 1.0875, "step": 20738 }, { "epoch": 1.25, "learning_rate": 4.686071250735594e-05, "loss": 0.6459, "step": 20740 }, { "epoch": 1.25, "learning_rate": 4.685920359724171e-05, "loss": 0.9716, "step": 20742 }, { "epoch": 1.25, "learning_rate": 4.685769468712749e-05, "loss": 0.7226, "step": 20744 }, { "epoch": 1.25, "learning_rate": 4.6856185777013266e-05, "loss": 1.1499, "step": 20746 }, { "epoch": 1.25, "learning_rate": 4.685467686689904e-05, "loss": 0.9161, "step": 20748 }, { "epoch": 1.25, "learning_rate": 4.685316795678482e-05, "loss": 1.1715, "step": 20750 }, { "epoch": 1.25, "learning_rate": 4.685165904667059e-05, "loss": 0.8067, "step": 20752 }, { "epoch": 1.25, "learning_rate": 4.685015013655637e-05, "loss": 0.8872, "step": 20754 }, { "epoch": 1.25, "learning_rate": 4.684864122644214e-05, "loss": 0.9477, "step": 20756 }, { "epoch": 1.25, "learning_rate": 4.684713231632792e-05, "loss": 1.2319, "step": 20758 }, { "epoch": 1.25, "learning_rate": 4.68456234062137e-05, "loss": 1.1513, "step": 20760 }, { "epoch": 1.25, "learning_rate": 4.6844114496099465e-05, "loss": 0.9161, "step": 20762 }, { "epoch": 1.25, "learning_rate": 4.6842605585985245e-05, "loss": 1.3289, "step": 20764 }, { "epoch": 1.25, "learning_rate": 4.6841096675871025e-05, "loss": 0.9718, "step": 20766 }, { "epoch": 1.25, "learning_rate": 4.68395877657568e-05, "loss": 0.9635, "step": 20768 }, { "epoch": 1.25, "learning_rate": 4.683807885564257e-05, "loss": 1.1522, "step": 20770 }, { "epoch": 1.25, "learning_rate": 4.6836569945528344e-05, "loss": 0.9613, "step": 20772 }, { "epoch": 1.25, "learning_rate": 4.6835061035414124e-05, "loss": 0.9568, "step": 20774 }, { "epoch": 1.25, "learning_rate": 4.68335521252999e-05, "loss": 1.0493, "step": 20776 }, { "epoch": 1.25, "learning_rate": 4.683204321518567e-05, "loss": 1.0434, "step": 20778 }, { "epoch": 1.25, "learning_rate": 4.683053430507145e-05, "loss": 0.8999, "step": 20780 }, { "epoch": 1.25, "learning_rate": 4.682902539495723e-05, "loss": 0.8613, "step": 20782 }, { "epoch": 1.25, "learning_rate": 4.6827516484842996e-05, "loss": 0.7986, "step": 20784 }, { "epoch": 1.25, "learning_rate": 4.6826007574728776e-05, "loss": 1.2998, "step": 20786 }, { "epoch": 1.25, "learning_rate": 4.682449866461455e-05, "loss": 0.757, "step": 20788 }, { "epoch": 1.25, "learning_rate": 4.682298975450033e-05, "loss": 1.0435, "step": 20790 }, { "epoch": 1.25, "learning_rate": 4.68214808443861e-05, "loss": 1.0988, "step": 20792 }, { "epoch": 1.26, "learning_rate": 4.6819971934271875e-05, "loss": 1.2693, "step": 20794 }, { "epoch": 1.26, "learning_rate": 4.6818463024157655e-05, "loss": 1.5431, "step": 20796 }, { "epoch": 1.26, "learning_rate": 4.681695411404343e-05, "loss": 1.1398, "step": 20798 }, { "epoch": 1.26, "learning_rate": 4.68154452039292e-05, "loss": 0.8839, "step": 20800 }, { "epoch": 1.26, "learning_rate": 4.681393629381498e-05, "loss": 0.8836, "step": 20802 }, { "epoch": 1.26, "learning_rate": 4.6812427383700755e-05, "loss": 1.3605, "step": 20804 }, { "epoch": 1.26, "learning_rate": 4.681091847358653e-05, "loss": 1.0491, "step": 20806 }, { "epoch": 1.26, "learning_rate": 4.680940956347231e-05, "loss": 1.3695, "step": 20808 }, { "epoch": 1.26, "learning_rate": 4.680790065335808e-05, "loss": 1.0716, "step": 20810 }, { "epoch": 1.26, "learning_rate": 4.6806391743243854e-05, "loss": 1.0029, "step": 20812 }, { "epoch": 1.26, "learning_rate": 4.6804882833129634e-05, "loss": 0.7004, "step": 20814 }, { "epoch": 1.26, "learning_rate": 4.680337392301541e-05, "loss": 1.0758, "step": 20816 }, { "epoch": 1.26, "learning_rate": 4.680186501290119e-05, "loss": 1.2021, "step": 20818 }, { "epoch": 1.26, "learning_rate": 4.680035610278696e-05, "loss": 0.8992, "step": 20820 }, { "epoch": 1.26, "learning_rate": 4.679884719267273e-05, "loss": 0.9804, "step": 20822 }, { "epoch": 1.26, "learning_rate": 4.679733828255851e-05, "loss": 1.3945, "step": 20824 }, { "epoch": 1.26, "learning_rate": 4.6795829372444286e-05, "loss": 1.0134, "step": 20826 }, { "epoch": 1.26, "learning_rate": 4.679432046233006e-05, "loss": 1.0006, "step": 20828 }, { "epoch": 1.26, "learning_rate": 4.679281155221584e-05, "loss": 0.771, "step": 20830 }, { "epoch": 1.26, "learning_rate": 4.679130264210161e-05, "loss": 0.8623, "step": 20832 }, { "epoch": 1.26, "learning_rate": 4.6789793731987385e-05, "loss": 1.2908, "step": 20834 }, { "epoch": 1.26, "learning_rate": 4.6788284821873165e-05, "loss": 0.9167, "step": 20836 }, { "epoch": 1.26, "learning_rate": 4.678677591175894e-05, "loss": 0.9414, "step": 20838 }, { "epoch": 1.26, "learning_rate": 4.678526700164472e-05, "loss": 0.6875, "step": 20840 }, { "epoch": 1.26, "learning_rate": 4.678375809153049e-05, "loss": 0.9725, "step": 20842 }, { "epoch": 1.26, "learning_rate": 4.6782249181416265e-05, "loss": 0.8372, "step": 20844 }, { "epoch": 1.26, "learning_rate": 4.6780740271302044e-05, "loss": 1.077, "step": 20846 }, { "epoch": 1.26, "learning_rate": 4.677923136118781e-05, "loss": 1.0142, "step": 20848 }, { "epoch": 1.26, "learning_rate": 4.677772245107359e-05, "loss": 0.7178, "step": 20850 }, { "epoch": 1.26, "learning_rate": 4.677621354095937e-05, "loss": 0.8643, "step": 20852 }, { "epoch": 1.26, "learning_rate": 4.6774704630845144e-05, "loss": 1.6999, "step": 20854 }, { "epoch": 1.26, "learning_rate": 4.677319572073092e-05, "loss": 0.858, "step": 20856 }, { "epoch": 1.26, "learning_rate": 4.67716868106167e-05, "loss": 1.2541, "step": 20858 }, { "epoch": 1.26, "learning_rate": 4.677017790050247e-05, "loss": 1.0235, "step": 20860 }, { "epoch": 1.26, "learning_rate": 4.676866899038824e-05, "loss": 0.9574, "step": 20862 }, { "epoch": 1.26, "learning_rate": 4.676716008027402e-05, "loss": 1.3045, "step": 20864 }, { "epoch": 1.26, "learning_rate": 4.6765651170159796e-05, "loss": 1.0204, "step": 20866 }, { "epoch": 1.26, "learning_rate": 4.6764142260045576e-05, "loss": 1.2307, "step": 20868 }, { "epoch": 1.26, "learning_rate": 4.676263334993134e-05, "loss": 1.1627, "step": 20870 }, { "epoch": 1.26, "learning_rate": 4.676112443981712e-05, "loss": 1.0319, "step": 20872 }, { "epoch": 1.26, "learning_rate": 4.67596155297029e-05, "loss": 1.1448, "step": 20874 }, { "epoch": 1.26, "learning_rate": 4.675810661958867e-05, "loss": 1.167, "step": 20876 }, { "epoch": 1.26, "learning_rate": 4.675659770947445e-05, "loss": 1.0435, "step": 20878 }, { "epoch": 1.26, "learning_rate": 4.675508879936023e-05, "loss": 0.9008, "step": 20880 }, { "epoch": 1.26, "learning_rate": 4.6753579889246e-05, "loss": 0.9942, "step": 20882 }, { "epoch": 1.26, "learning_rate": 4.6752070979131774e-05, "loss": 1.1405, "step": 20884 }, { "epoch": 1.26, "learning_rate": 4.675056206901755e-05, "loss": 0.8886, "step": 20886 }, { "epoch": 1.26, "learning_rate": 4.674905315890333e-05, "loss": 0.9464, "step": 20888 }, { "epoch": 1.26, "learning_rate": 4.67475442487891e-05, "loss": 1.005, "step": 20890 }, { "epoch": 1.26, "learning_rate": 4.6746035338674874e-05, "loss": 1.1793, "step": 20892 }, { "epoch": 1.26, "learning_rate": 4.6744526428560654e-05, "loss": 1.1462, "step": 20894 }, { "epoch": 1.26, "learning_rate": 4.6743017518446434e-05, "loss": 1.0396, "step": 20896 }, { "epoch": 1.26, "learning_rate": 4.67415086083322e-05, "loss": 1.3011, "step": 20898 }, { "epoch": 1.26, "learning_rate": 4.673999969821798e-05, "loss": 1.4463, "step": 20900 }, { "epoch": 1.26, "learning_rate": 4.673849078810375e-05, "loss": 1.0004, "step": 20902 }, { "epoch": 1.26, "learning_rate": 4.673698187798953e-05, "loss": 0.975, "step": 20904 }, { "epoch": 1.26, "learning_rate": 4.6735472967875306e-05, "loss": 1.0443, "step": 20906 }, { "epoch": 1.26, "learning_rate": 4.673396405776108e-05, "loss": 1.5989, "step": 20908 }, { "epoch": 1.26, "learning_rate": 4.673245514764686e-05, "loss": 0.8738, "step": 20910 }, { "epoch": 1.26, "learning_rate": 4.673094623753263e-05, "loss": 1.0373, "step": 20912 }, { "epoch": 1.26, "learning_rate": 4.6729437327418405e-05, "loss": 1.0478, "step": 20914 }, { "epoch": 1.26, "learning_rate": 4.6727928417304185e-05, "loss": 1.1062, "step": 20916 }, { "epoch": 1.26, "learning_rate": 4.672641950718996e-05, "loss": 1.345, "step": 20918 }, { "epoch": 1.26, "learning_rate": 4.672491059707573e-05, "loss": 0.9383, "step": 20920 }, { "epoch": 1.26, "learning_rate": 4.672340168696151e-05, "loss": 1.2852, "step": 20922 }, { "epoch": 1.26, "learning_rate": 4.6721892776847284e-05, "loss": 0.9562, "step": 20924 }, { "epoch": 1.26, "learning_rate": 4.672038386673306e-05, "loss": 1.2262, "step": 20926 }, { "epoch": 1.26, "learning_rate": 4.671887495661884e-05, "loss": 0.9809, "step": 20928 }, { "epoch": 1.26, "learning_rate": 4.671736604650461e-05, "loss": 1.1438, "step": 20930 }, { "epoch": 1.26, "learning_rate": 4.671585713639039e-05, "loss": 0.9122, "step": 20932 }, { "epoch": 1.26, "learning_rate": 4.6714348226276164e-05, "loss": 1.1312, "step": 20934 }, { "epoch": 1.26, "learning_rate": 4.671283931616194e-05, "loss": 0.9861, "step": 20936 }, { "epoch": 1.26, "learning_rate": 4.671133040604772e-05, "loss": 0.9626, "step": 20938 }, { "epoch": 1.26, "learning_rate": 4.670982149593349e-05, "loss": 1.1004, "step": 20940 }, { "epoch": 1.26, "learning_rate": 4.670831258581926e-05, "loss": 1.1202, "step": 20942 }, { "epoch": 1.26, "learning_rate": 4.670680367570504e-05, "loss": 0.9231, "step": 20944 }, { "epoch": 1.26, "learning_rate": 4.6705294765590816e-05, "loss": 0.7797, "step": 20946 }, { "epoch": 1.26, "learning_rate": 4.670378585547659e-05, "loss": 1.4485, "step": 20948 }, { "epoch": 1.26, "learning_rate": 4.670227694536237e-05, "loss": 0.9407, "step": 20950 }, { "epoch": 1.26, "learning_rate": 4.670076803524814e-05, "loss": 0.8862, "step": 20952 }, { "epoch": 1.26, "learning_rate": 4.669925912513392e-05, "loss": 0.6564, "step": 20954 }, { "epoch": 1.26, "learning_rate": 4.6697750215019695e-05, "loss": 0.923, "step": 20956 }, { "epoch": 1.26, "learning_rate": 4.669624130490547e-05, "loss": 1.0247, "step": 20958 }, { "epoch": 1.27, "learning_rate": 4.669473239479125e-05, "loss": 0.9105, "step": 20960 }, { "epoch": 1.27, "learning_rate": 4.6693223484677014e-05, "loss": 0.9789, "step": 20962 }, { "epoch": 1.27, "learning_rate": 4.6691714574562794e-05, "loss": 0.7801, "step": 20964 }, { "epoch": 1.27, "learning_rate": 4.6690205664448574e-05, "loss": 0.9773, "step": 20966 }, { "epoch": 1.27, "learning_rate": 4.668869675433435e-05, "loss": 1.3039, "step": 20968 }, { "epoch": 1.27, "learning_rate": 4.668718784422012e-05, "loss": 0.9781, "step": 20970 }, { "epoch": 1.27, "learning_rate": 4.66856789341059e-05, "loss": 0.9768, "step": 20972 }, { "epoch": 1.27, "learning_rate": 4.6684170023991674e-05, "loss": 1.3326, "step": 20974 }, { "epoch": 1.27, "learning_rate": 4.668266111387745e-05, "loss": 0.9572, "step": 20976 }, { "epoch": 1.27, "learning_rate": 4.668115220376322e-05, "loss": 1.2869, "step": 20978 }, { "epoch": 1.27, "learning_rate": 4.6679643293649e-05, "loss": 0.8925, "step": 20980 }, { "epoch": 1.27, "learning_rate": 4.667813438353478e-05, "loss": 1.226, "step": 20982 }, { "epoch": 1.27, "learning_rate": 4.6676625473420546e-05, "loss": 1.1662, "step": 20984 }, { "epoch": 1.27, "learning_rate": 4.6675116563306326e-05, "loss": 0.9954, "step": 20986 }, { "epoch": 1.27, "learning_rate": 4.6673607653192106e-05, "loss": 1.0446, "step": 20988 }, { "epoch": 1.27, "learning_rate": 4.667209874307787e-05, "loss": 1.215, "step": 20990 }, { "epoch": 1.27, "learning_rate": 4.667058983296365e-05, "loss": 0.8617, "step": 20992 }, { "epoch": 1.27, "learning_rate": 4.666908092284943e-05, "loss": 0.9932, "step": 20994 }, { "epoch": 1.27, "learning_rate": 4.6667572012735205e-05, "loss": 0.9418, "step": 20996 }, { "epoch": 1.27, "learning_rate": 4.666606310262098e-05, "loss": 0.7501, "step": 20998 }, { "epoch": 1.27, "learning_rate": 4.666455419250675e-05, "loss": 0.8971, "step": 21000 }, { "epoch": 1.27, "learning_rate": 4.666304528239253e-05, "loss": 0.9632, "step": 21002 }, { "epoch": 1.27, "learning_rate": 4.666153637227831e-05, "loss": 1.3627, "step": 21004 }, { "epoch": 1.27, "learning_rate": 4.666002746216408e-05, "loss": 1.0381, "step": 21006 }, { "epoch": 1.27, "learning_rate": 4.665851855204986e-05, "loss": 1.4606, "step": 21008 }, { "epoch": 1.27, "learning_rate": 4.665700964193564e-05, "loss": 0.9346, "step": 21010 }, { "epoch": 1.27, "learning_rate": 4.6655500731821404e-05, "loss": 0.9421, "step": 21012 }, { "epoch": 1.27, "learning_rate": 4.6653991821707183e-05, "loss": 1.3393, "step": 21014 }, { "epoch": 1.27, "learning_rate": 4.6652482911592957e-05, "loss": 1.1961, "step": 21016 }, { "epoch": 1.27, "learning_rate": 4.6650974001478737e-05, "loss": 0.9987, "step": 21018 }, { "epoch": 1.27, "learning_rate": 4.664946509136451e-05, "loss": 1.1388, "step": 21020 }, { "epoch": 1.27, "learning_rate": 4.664795618125028e-05, "loss": 0.9831, "step": 21022 }, { "epoch": 1.27, "learning_rate": 4.664644727113606e-05, "loss": 0.9298, "step": 21024 }, { "epoch": 1.27, "learning_rate": 4.6644938361021836e-05, "loss": 0.8386, "step": 21026 }, { "epoch": 1.27, "learning_rate": 4.664342945090761e-05, "loss": 0.99, "step": 21028 }, { "epoch": 1.27, "learning_rate": 4.664192054079339e-05, "loss": 0.878, "step": 21030 }, { "epoch": 1.27, "learning_rate": 4.664041163067916e-05, "loss": 0.7728, "step": 21032 }, { "epoch": 1.27, "learning_rate": 4.6638902720564935e-05, "loss": 1.256, "step": 21034 }, { "epoch": 1.27, "learning_rate": 4.6637393810450715e-05, "loss": 1.3614, "step": 21036 }, { "epoch": 1.27, "learning_rate": 4.663588490033649e-05, "loss": 0.9447, "step": 21038 }, { "epoch": 1.27, "learning_rate": 4.663437599022226e-05, "loss": 1.07, "step": 21040 }, { "epoch": 1.27, "learning_rate": 4.663286708010804e-05, "loss": 1.115, "step": 21042 }, { "epoch": 1.27, "learning_rate": 4.6631358169993814e-05, "loss": 1.1336, "step": 21044 }, { "epoch": 1.27, "learning_rate": 4.6629849259879594e-05, "loss": 0.684, "step": 21046 }, { "epoch": 1.27, "learning_rate": 4.662834034976537e-05, "loss": 0.9464, "step": 21048 }, { "epoch": 1.27, "learning_rate": 4.662683143965114e-05, "loss": 0.7719, "step": 21050 }, { "epoch": 1.27, "learning_rate": 4.662532252953692e-05, "loss": 1.0137, "step": 21052 }, { "epoch": 1.27, "learning_rate": 4.6623813619422693e-05, "loss": 0.6002, "step": 21054 }, { "epoch": 1.27, "learning_rate": 4.6622304709308467e-05, "loss": 1.2558, "step": 21056 }, { "epoch": 1.27, "learning_rate": 4.6620795799194246e-05, "loss": 1.0407, "step": 21058 }, { "epoch": 1.27, "learning_rate": 4.661928688908002e-05, "loss": 1.0529, "step": 21060 }, { "epoch": 1.27, "learning_rate": 4.661777797896579e-05, "loss": 1.0664, "step": 21062 }, { "epoch": 1.27, "learning_rate": 4.661626906885157e-05, "loss": 1.1958, "step": 21064 }, { "epoch": 1.27, "learning_rate": 4.6614760158737346e-05, "loss": 1.0303, "step": 21066 }, { "epoch": 1.27, "learning_rate": 4.6613251248623126e-05, "loss": 1.0509, "step": 21068 }, { "epoch": 1.27, "learning_rate": 4.66117423385089e-05, "loss": 1.0976, "step": 21070 }, { "epoch": 1.27, "learning_rate": 4.661023342839467e-05, "loss": 0.863, "step": 21072 }, { "epoch": 1.27, "learning_rate": 4.660872451828045e-05, "loss": 0.7143, "step": 21074 }, { "epoch": 1.27, "learning_rate": 4.660721560816622e-05, "loss": 0.9103, "step": 21076 }, { "epoch": 1.27, "learning_rate": 4.6605706698052e-05, "loss": 1.0278, "step": 21078 }, { "epoch": 1.27, "learning_rate": 4.660419778793778e-05, "loss": 0.6936, "step": 21080 }, { "epoch": 1.27, "learning_rate": 4.660268887782355e-05, "loss": 1.1033, "step": 21082 }, { "epoch": 1.27, "learning_rate": 4.6601179967709324e-05, "loss": 0.6611, "step": 21084 }, { "epoch": 1.27, "learning_rate": 4.6599671057595104e-05, "loss": 0.8284, "step": 21086 }, { "epoch": 1.27, "learning_rate": 4.659816214748088e-05, "loss": 0.9643, "step": 21088 }, { "epoch": 1.27, "learning_rate": 4.659665323736665e-05, "loss": 1.0004, "step": 21090 }, { "epoch": 1.27, "learning_rate": 4.6595144327252423e-05, "loss": 1.0013, "step": 21092 }, { "epoch": 1.27, "learning_rate": 4.65936354171382e-05, "loss": 0.9762, "step": 21094 }, { "epoch": 1.27, "learning_rate": 4.659212650702398e-05, "loss": 1.0957, "step": 21096 }, { "epoch": 1.27, "learning_rate": 4.659061759690975e-05, "loss": 0.9506, "step": 21098 }, { "epoch": 1.27, "learning_rate": 4.658910868679553e-05, "loss": 1.0187, "step": 21100 }, { "epoch": 1.27, "learning_rate": 4.658759977668131e-05, "loss": 1.0581, "step": 21102 }, { "epoch": 1.27, "learning_rate": 4.6586090866567076e-05, "loss": 0.9335, "step": 21104 }, { "epoch": 1.27, "learning_rate": 4.6584581956452856e-05, "loss": 1.058, "step": 21106 }, { "epoch": 1.27, "learning_rate": 4.6583073046338636e-05, "loss": 1.1622, "step": 21108 }, { "epoch": 1.27, "learning_rate": 4.658156413622441e-05, "loss": 1.2176, "step": 21110 }, { "epoch": 1.27, "learning_rate": 4.658005522611018e-05, "loss": 1.2899, "step": 21112 }, { "epoch": 1.27, "learning_rate": 4.6578546315995955e-05, "loss": 1.2357, "step": 21114 }, { "epoch": 1.27, "learning_rate": 4.6577037405881735e-05, "loss": 1.2814, "step": 21116 }, { "epoch": 1.27, "learning_rate": 4.6575528495767515e-05, "loss": 1.0234, "step": 21118 }, { "epoch": 1.27, "learning_rate": 4.657401958565328e-05, "loss": 1.4146, "step": 21120 }, { "epoch": 1.27, "learning_rate": 4.657251067553906e-05, "loss": 0.9925, "step": 21122 }, { "epoch": 1.27, "learning_rate": 4.657100176542484e-05, "loss": 0.9765, "step": 21124 }, { "epoch": 1.28, "learning_rate": 4.656949285531061e-05, "loss": 1.1128, "step": 21126 }, { "epoch": 1.28, "learning_rate": 4.656798394519639e-05, "loss": 1.1153, "step": 21128 }, { "epoch": 1.28, "learning_rate": 4.656647503508216e-05, "loss": 1.2865, "step": 21130 }, { "epoch": 1.28, "learning_rate": 4.656496612496794e-05, "loss": 1.1519, "step": 21132 }, { "epoch": 1.28, "learning_rate": 4.656345721485371e-05, "loss": 1.1925, "step": 21134 }, { "epoch": 1.28, "learning_rate": 4.6561948304739486e-05, "loss": 0.8893, "step": 21136 }, { "epoch": 1.28, "learning_rate": 4.6560439394625266e-05, "loss": 1.0361, "step": 21138 }, { "epoch": 1.28, "learning_rate": 4.655893048451104e-05, "loss": 0.9726, "step": 21140 }, { "epoch": 1.28, "learning_rate": 4.655742157439681e-05, "loss": 1.0685, "step": 21142 }, { "epoch": 1.28, "learning_rate": 4.655591266428259e-05, "loss": 1.1832, "step": 21144 }, { "epoch": 1.28, "learning_rate": 4.6554403754168366e-05, "loss": 1.0231, "step": 21146 }, { "epoch": 1.28, "learning_rate": 4.655289484405414e-05, "loss": 0.861, "step": 21148 }, { "epoch": 1.28, "learning_rate": 4.655138593393992e-05, "loss": 0.9566, "step": 21150 }, { "epoch": 1.28, "learning_rate": 4.654987702382569e-05, "loss": 0.9523, "step": 21152 }, { "epoch": 1.28, "learning_rate": 4.6548368113711465e-05, "loss": 1.2639, "step": 21154 }, { "epoch": 1.28, "learning_rate": 4.6546859203597245e-05, "loss": 0.8296, "step": 21156 }, { "epoch": 1.28, "learning_rate": 4.654535029348302e-05, "loss": 0.9331, "step": 21158 }, { "epoch": 1.28, "learning_rate": 4.65438413833688e-05, "loss": 1.1244, "step": 21160 }, { "epoch": 1.28, "learning_rate": 4.654233247325457e-05, "loss": 1.2356, "step": 21162 }, { "epoch": 1.28, "learning_rate": 4.6540823563140344e-05, "loss": 1.0222, "step": 21164 }, { "epoch": 1.28, "learning_rate": 4.6539314653026124e-05, "loss": 1.0222, "step": 21166 }, { "epoch": 1.28, "learning_rate": 4.65378057429119e-05, "loss": 1.0443, "step": 21168 }, { "epoch": 1.28, "learning_rate": 4.653629683279767e-05, "loss": 0.836, "step": 21170 }, { "epoch": 1.28, "learning_rate": 4.653478792268345e-05, "loss": 1.289, "step": 21172 }, { "epoch": 1.28, "learning_rate": 4.653327901256922e-05, "loss": 0.9197, "step": 21174 }, { "epoch": 1.28, "learning_rate": 4.6531770102454996e-05, "loss": 0.8831, "step": 21176 }, { "epoch": 1.28, "learning_rate": 4.6530261192340776e-05, "loss": 0.9116, "step": 21178 }, { "epoch": 1.28, "learning_rate": 4.652875228222655e-05, "loss": 0.8337, "step": 21180 }, { "epoch": 1.28, "learning_rate": 4.652724337211233e-05, "loss": 0.9682, "step": 21182 }, { "epoch": 1.28, "learning_rate": 4.65257344619981e-05, "loss": 0.953, "step": 21184 }, { "epoch": 1.28, "learning_rate": 4.6524225551883876e-05, "loss": 0.8034, "step": 21186 }, { "epoch": 1.28, "learning_rate": 4.6522716641769655e-05, "loss": 1.1, "step": 21188 }, { "epoch": 1.28, "learning_rate": 4.652120773165542e-05, "loss": 1.0928, "step": 21190 }, { "epoch": 1.28, "learning_rate": 4.65196988215412e-05, "loss": 1.1156, "step": 21192 }, { "epoch": 1.28, "learning_rate": 4.651818991142698e-05, "loss": 1.0691, "step": 21194 }, { "epoch": 1.28, "learning_rate": 4.6516681001312755e-05, "loss": 1.1842, "step": 21196 }, { "epoch": 1.28, "learning_rate": 4.651517209119853e-05, "loss": 1.3082, "step": 21198 }, { "epoch": 1.28, "learning_rate": 4.651366318108431e-05, "loss": 1.0873, "step": 21200 }, { "epoch": 1.28, "learning_rate": 4.651215427097008e-05, "loss": 1.1327, "step": 21202 }, { "epoch": 1.28, "learning_rate": 4.6510645360855854e-05, "loss": 0.8216, "step": 21204 }, { "epoch": 1.28, "learning_rate": 4.650913645074163e-05, "loss": 0.8095, "step": 21206 }, { "epoch": 1.28, "learning_rate": 4.650762754062741e-05, "loss": 0.8807, "step": 21208 }, { "epoch": 1.28, "learning_rate": 4.650611863051319e-05, "loss": 0.8563, "step": 21210 }, { "epoch": 1.28, "learning_rate": 4.650460972039895e-05, "loss": 1.214, "step": 21212 }, { "epoch": 1.28, "learning_rate": 4.650310081028473e-05, "loss": 0.8661, "step": 21214 }, { "epoch": 1.28, "learning_rate": 4.650159190017051e-05, "loss": 1.1042, "step": 21216 }, { "epoch": 1.28, "learning_rate": 4.6500082990056286e-05, "loss": 0.9999, "step": 21218 }, { "epoch": 1.28, "learning_rate": 4.649857407994206e-05, "loss": 1.0814, "step": 21220 }, { "epoch": 1.28, "learning_rate": 4.649706516982784e-05, "loss": 1.0734, "step": 21222 }, { "epoch": 1.28, "learning_rate": 4.649555625971361e-05, "loss": 1.0776, "step": 21224 }, { "epoch": 1.28, "learning_rate": 4.6494047349599386e-05, "loss": 0.8968, "step": 21226 }, { "epoch": 1.28, "learning_rate": 4.649253843948516e-05, "loss": 1.0328, "step": 21228 }, { "epoch": 1.28, "learning_rate": 4.649102952937094e-05, "loss": 0.8916, "step": 21230 }, { "epoch": 1.28, "learning_rate": 4.648952061925672e-05, "loss": 0.9567, "step": 21232 }, { "epoch": 1.28, "learning_rate": 4.6488011709142485e-05, "loss": 1.1349, "step": 21234 }, { "epoch": 1.28, "learning_rate": 4.6486502799028265e-05, "loss": 1.0581, "step": 21236 }, { "epoch": 1.28, "learning_rate": 4.6484993888914045e-05, "loss": 0.7786, "step": 21238 }, { "epoch": 1.28, "learning_rate": 4.648348497879981e-05, "loss": 0.8128, "step": 21240 }, { "epoch": 1.28, "learning_rate": 4.648197606868559e-05, "loss": 0.816, "step": 21242 }, { "epoch": 1.28, "learning_rate": 4.6480467158571364e-05, "loss": 1.0287, "step": 21244 }, { "epoch": 1.28, "learning_rate": 4.6478958248457144e-05, "loss": 0.8585, "step": 21246 }, { "epoch": 1.28, "learning_rate": 4.647744933834292e-05, "loss": 1.0191, "step": 21248 }, { "epoch": 1.28, "learning_rate": 4.647594042822869e-05, "loss": 1.0373, "step": 21250 }, { "epoch": 1.28, "learning_rate": 4.647443151811447e-05, "loss": 0.9705, "step": 21252 }, { "epoch": 1.28, "learning_rate": 4.647292260800024e-05, "loss": 1.2843, "step": 21254 }, { "epoch": 1.28, "learning_rate": 4.6471413697886016e-05, "loss": 1.1056, "step": 21256 }, { "epoch": 1.28, "learning_rate": 4.6469904787771796e-05, "loss": 0.8551, "step": 21258 }, { "epoch": 1.28, "learning_rate": 4.646839587765757e-05, "loss": 0.8092, "step": 21260 }, { "epoch": 1.28, "learning_rate": 4.646688696754334e-05, "loss": 1.0946, "step": 21262 }, { "epoch": 1.28, "learning_rate": 4.646537805742912e-05, "loss": 1.1587, "step": 21264 }, { "epoch": 1.28, "learning_rate": 4.6463869147314895e-05, "loss": 1.0445, "step": 21266 }, { "epoch": 1.28, "learning_rate": 4.646236023720067e-05, "loss": 1.0613, "step": 21268 }, { "epoch": 1.28, "learning_rate": 4.646085132708645e-05, "loss": 0.7492, "step": 21270 }, { "epoch": 1.28, "learning_rate": 4.645934241697222e-05, "loss": 0.883, "step": 21272 }, { "epoch": 1.28, "learning_rate": 4.6457833506858e-05, "loss": 1.1557, "step": 21274 }, { "epoch": 1.28, "learning_rate": 4.6456324596743775e-05, "loss": 0.9268, "step": 21276 }, { "epoch": 1.28, "learning_rate": 4.645481568662955e-05, "loss": 1.2074, "step": 21278 }, { "epoch": 1.28, "learning_rate": 4.645330677651533e-05, "loss": 0.8723, "step": 21280 }, { "epoch": 1.28, "learning_rate": 4.64517978664011e-05, "loss": 0.8996, "step": 21282 }, { "epoch": 1.28, "learning_rate": 4.6450288956286874e-05, "loss": 0.9837, "step": 21284 }, { "epoch": 1.28, "learning_rate": 4.6448780046172654e-05, "loss": 1.1479, "step": 21286 }, { "epoch": 1.28, "learning_rate": 4.644727113605843e-05, "loss": 0.7219, "step": 21288 }, { "epoch": 1.29, "learning_rate": 4.64457622259442e-05, "loss": 0.8143, "step": 21290 }, { "epoch": 1.29, "learning_rate": 4.644425331582998e-05, "loss": 0.8481, "step": 21292 }, { "epoch": 1.29, "learning_rate": 4.644274440571575e-05, "loss": 0.8518, "step": 21294 }, { "epoch": 1.29, "learning_rate": 4.644123549560153e-05, "loss": 1.3462, "step": 21296 }, { "epoch": 1.29, "learning_rate": 4.6439726585487306e-05, "loss": 0.8415, "step": 21298 }, { "epoch": 1.29, "learning_rate": 4.643821767537308e-05, "loss": 1.29, "step": 21300 }, { "epoch": 1.29, "learning_rate": 4.643670876525886e-05, "loss": 0.8585, "step": 21302 }, { "epoch": 1.29, "learning_rate": 4.6435199855144625e-05, "loss": 0.8671, "step": 21304 }, { "epoch": 1.29, "learning_rate": 4.6433690945030405e-05, "loss": 1.1055, "step": 21306 }, { "epoch": 1.29, "learning_rate": 4.6432182034916185e-05, "loss": 0.9143, "step": 21308 }, { "epoch": 1.29, "learning_rate": 4.643067312480196e-05, "loss": 0.8596, "step": 21310 }, { "epoch": 1.29, "learning_rate": 4.642916421468773e-05, "loss": 1.3457, "step": 21312 }, { "epoch": 1.29, "learning_rate": 4.642765530457351e-05, "loss": 1.0831, "step": 21314 }, { "epoch": 1.29, "learning_rate": 4.6426146394459285e-05, "loss": 0.9254, "step": 21316 }, { "epoch": 1.29, "learning_rate": 4.642463748434506e-05, "loss": 1.0385, "step": 21318 }, { "epoch": 1.29, "learning_rate": 4.642312857423083e-05, "loss": 1.4046, "step": 21320 }, { "epoch": 1.29, "learning_rate": 4.642161966411661e-05, "loss": 0.9727, "step": 21322 }, { "epoch": 1.29, "learning_rate": 4.642011075400239e-05, "loss": 1.1826, "step": 21324 }, { "epoch": 1.29, "learning_rate": 4.641860184388816e-05, "loss": 0.9586, "step": 21326 }, { "epoch": 1.29, "learning_rate": 4.641709293377394e-05, "loss": 0.8531, "step": 21328 }, { "epoch": 1.29, "learning_rate": 4.641558402365972e-05, "loss": 0.9047, "step": 21330 }, { "epoch": 1.29, "learning_rate": 4.641407511354549e-05, "loss": 0.9951, "step": 21332 }, { "epoch": 1.29, "learning_rate": 4.641256620343126e-05, "loss": 0.7477, "step": 21334 }, { "epoch": 1.29, "learning_rate": 4.641105729331704e-05, "loss": 0.9054, "step": 21336 }, { "epoch": 1.29, "learning_rate": 4.6409548383202816e-05, "loss": 1.1019, "step": 21338 }, { "epoch": 1.29, "learning_rate": 4.640803947308859e-05, "loss": 1.001, "step": 21340 }, { "epoch": 1.29, "learning_rate": 4.640653056297436e-05, "loss": 0.9842, "step": 21342 }, { "epoch": 1.29, "learning_rate": 4.640502165286014e-05, "loss": 0.8949, "step": 21344 }, { "epoch": 1.29, "learning_rate": 4.640351274274592e-05, "loss": 1.0758, "step": 21346 }, { "epoch": 1.29, "learning_rate": 4.640200383263169e-05, "loss": 1.0144, "step": 21348 }, { "epoch": 1.29, "learning_rate": 4.640049492251747e-05, "loss": 0.7243, "step": 21350 }, { "epoch": 1.29, "learning_rate": 4.639898601240325e-05, "loss": 0.9061, "step": 21352 }, { "epoch": 1.29, "learning_rate": 4.6397477102289015e-05, "loss": 1.2493, "step": 21354 }, { "epoch": 1.29, "learning_rate": 4.6395968192174795e-05, "loss": 1.1399, "step": 21356 }, { "epoch": 1.29, "learning_rate": 4.639445928206057e-05, "loss": 0.7736, "step": 21358 }, { "epoch": 1.29, "learning_rate": 4.639295037194635e-05, "loss": 0.8758, "step": 21360 }, { "epoch": 1.29, "learning_rate": 4.639144146183212e-05, "loss": 0.9328, "step": 21362 }, { "epoch": 1.29, "learning_rate": 4.6389932551717894e-05, "loss": 0.9499, "step": 21364 }, { "epoch": 1.29, "learning_rate": 4.6388423641603674e-05, "loss": 0.9685, "step": 21366 }, { "epoch": 1.29, "learning_rate": 4.638691473148945e-05, "loss": 0.887, "step": 21368 }, { "epoch": 1.29, "learning_rate": 4.638540582137522e-05, "loss": 0.9277, "step": 21370 }, { "epoch": 1.29, "learning_rate": 4.6383896911261e-05, "loss": 0.9504, "step": 21372 }, { "epoch": 1.29, "learning_rate": 4.638238800114677e-05, "loss": 1.3733, "step": 21374 }, { "epoch": 1.29, "learning_rate": 4.6380879091032546e-05, "loss": 1.2418, "step": 21376 }, { "epoch": 1.29, "learning_rate": 4.6379370180918326e-05, "loss": 1.0932, "step": 21378 }, { "epoch": 1.29, "learning_rate": 4.63778612708041e-05, "loss": 1.253, "step": 21380 }, { "epoch": 1.29, "learning_rate": 4.637635236068988e-05, "loss": 1.1192, "step": 21382 }, { "epoch": 1.29, "learning_rate": 4.637484345057565e-05, "loss": 0.9613, "step": 21384 }, { "epoch": 1.29, "learning_rate": 4.6373334540461425e-05, "loss": 1.4801, "step": 21386 }, { "epoch": 1.29, "learning_rate": 4.6371825630347205e-05, "loss": 0.7086, "step": 21388 }, { "epoch": 1.29, "learning_rate": 4.637031672023298e-05, "loss": 1.0164, "step": 21390 }, { "epoch": 1.29, "learning_rate": 4.636880781011875e-05, "loss": 0.9152, "step": 21392 }, { "epoch": 1.29, "learning_rate": 4.636729890000453e-05, "loss": 1.0279, "step": 21394 }, { "epoch": 1.29, "learning_rate": 4.6365789989890304e-05, "loss": 0.7375, "step": 21396 }, { "epoch": 1.29, "learning_rate": 4.636428107977608e-05, "loss": 0.9731, "step": 21398 }, { "epoch": 1.29, "learning_rate": 4.636277216966186e-05, "loss": 0.8078, "step": 21400 }, { "epoch": 1.29, "learning_rate": 4.636126325954763e-05, "loss": 0.9043, "step": 21402 }, { "epoch": 1.29, "learning_rate": 4.6359754349433404e-05, "loss": 1.0446, "step": 21404 }, { "epoch": 1.29, "learning_rate": 4.6358245439319184e-05, "loss": 1.2862, "step": 21406 }, { "epoch": 1.29, "learning_rate": 4.635673652920496e-05, "loss": 1.2276, "step": 21408 }, { "epoch": 1.29, "learning_rate": 4.635522761909074e-05, "loss": 1.2918, "step": 21410 }, { "epoch": 1.29, "learning_rate": 4.635371870897651e-05, "loss": 1.1201, "step": 21412 }, { "epoch": 1.29, "learning_rate": 4.635220979886228e-05, "loss": 0.8177, "step": 21414 }, { "epoch": 1.29, "learning_rate": 4.635070088874806e-05, "loss": 1.1444, "step": 21416 }, { "epoch": 1.29, "learning_rate": 4.634919197863383e-05, "loss": 1.2271, "step": 21418 }, { "epoch": 1.29, "learning_rate": 4.634768306851961e-05, "loss": 1.0817, "step": 21420 }, { "epoch": 1.29, "learning_rate": 4.634617415840539e-05, "loss": 0.8766, "step": 21422 }, { "epoch": 1.29, "learning_rate": 4.634466524829116e-05, "loss": 0.8079, "step": 21424 }, { "epoch": 1.29, "learning_rate": 4.6343156338176935e-05, "loss": 1.077, "step": 21426 }, { "epoch": 1.29, "learning_rate": 4.6341647428062715e-05, "loss": 1.5567, "step": 21428 }, { "epoch": 1.29, "learning_rate": 4.634013851794849e-05, "loss": 0.8934, "step": 21430 }, { "epoch": 1.29, "learning_rate": 4.633862960783426e-05, "loss": 0.9272, "step": 21432 }, { "epoch": 1.29, "learning_rate": 4.6337120697720035e-05, "loss": 0.8636, "step": 21434 }, { "epoch": 1.29, "learning_rate": 4.6335611787605814e-05, "loss": 0.9337, "step": 21436 }, { "epoch": 1.29, "learning_rate": 4.6334102877491594e-05, "loss": 1.2933, "step": 21438 }, { "epoch": 1.29, "learning_rate": 4.633259396737736e-05, "loss": 1.1658, "step": 21440 }, { "epoch": 1.29, "learning_rate": 4.633108505726314e-05, "loss": 0.93, "step": 21442 }, { "epoch": 1.29, "learning_rate": 4.632957614714892e-05, "loss": 0.8846, "step": 21444 }, { "epoch": 1.29, "learning_rate": 4.6328067237034694e-05, "loss": 0.8371, "step": 21446 }, { "epoch": 1.29, "learning_rate": 4.632655832692047e-05, "loss": 1.0709, "step": 21448 }, { "epoch": 1.29, "learning_rate": 4.632504941680624e-05, "loss": 0.9279, "step": 21450 }, { "epoch": 1.29, "learning_rate": 4.632354050669202e-05, "loss": 1.2024, "step": 21452 }, { "epoch": 1.29, "learning_rate": 4.632203159657779e-05, "loss": 0.7724, "step": 21454 }, { "epoch": 1.3, "learning_rate": 4.6320522686463566e-05, "loss": 1.1337, "step": 21456 }, { "epoch": 1.3, "learning_rate": 4.6319013776349346e-05, "loss": 0.7431, "step": 21458 }, { "epoch": 1.3, "learning_rate": 4.6317504866235126e-05, "loss": 0.8989, "step": 21460 }, { "epoch": 1.3, "learning_rate": 4.631599595612089e-05, "loss": 1.085, "step": 21462 }, { "epoch": 1.3, "learning_rate": 4.631448704600667e-05, "loss": 0.6712, "step": 21464 }, { "epoch": 1.3, "learning_rate": 4.631297813589245e-05, "loss": 1.2567, "step": 21466 }, { "epoch": 1.3, "learning_rate": 4.631146922577822e-05, "loss": 1.3411, "step": 21468 }, { "epoch": 1.3, "learning_rate": 4.6309960315664e-05, "loss": 0.9454, "step": 21470 }, { "epoch": 1.3, "learning_rate": 4.630845140554977e-05, "loss": 1.1243, "step": 21472 }, { "epoch": 1.3, "learning_rate": 4.630694249543555e-05, "loss": 1.6282, "step": 21474 }, { "epoch": 1.3, "learning_rate": 4.6305433585321324e-05, "loss": 1.1732, "step": 21476 }, { "epoch": 1.3, "learning_rate": 4.63039246752071e-05, "loss": 0.7331, "step": 21478 }, { "epoch": 1.3, "learning_rate": 4.630241576509288e-05, "loss": 1.1253, "step": 21480 }, { "epoch": 1.3, "learning_rate": 4.630090685497865e-05, "loss": 0.7835, "step": 21482 }, { "epoch": 1.3, "learning_rate": 4.6299397944864424e-05, "loss": 0.911, "step": 21484 }, { "epoch": 1.3, "learning_rate": 4.6297889034750204e-05, "loss": 1.3813, "step": 21486 }, { "epoch": 1.3, "learning_rate": 4.629638012463598e-05, "loss": 1.0304, "step": 21488 }, { "epoch": 1.3, "learning_rate": 4.629487121452175e-05, "loss": 1.0409, "step": 21490 }, { "epoch": 1.3, "learning_rate": 4.629336230440753e-05, "loss": 0.6047, "step": 21492 }, { "epoch": 1.3, "learning_rate": 4.62918533942933e-05, "loss": 1.1516, "step": 21494 }, { "epoch": 1.3, "learning_rate": 4.629034448417908e-05, "loss": 0.9556, "step": 21496 }, { "epoch": 1.3, "learning_rate": 4.6288835574064856e-05, "loss": 0.9137, "step": 21498 }, { "epoch": 1.3, "learning_rate": 4.628732666395063e-05, "loss": 1.1629, "step": 21500 }, { "epoch": 1.3, "learning_rate": 4.628581775383641e-05, "loss": 1.0075, "step": 21502 }, { "epoch": 1.3, "learning_rate": 4.628430884372218e-05, "loss": 1.0048, "step": 21504 }, { "epoch": 1.3, "learning_rate": 4.6282799933607955e-05, "loss": 0.9427, "step": 21506 }, { "epoch": 1.3, "learning_rate": 4.6281291023493735e-05, "loss": 1.1689, "step": 21508 }, { "epoch": 1.3, "learning_rate": 4.627978211337951e-05, "loss": 1.1729, "step": 21510 }, { "epoch": 1.3, "learning_rate": 4.627827320326528e-05, "loss": 1.1587, "step": 21512 }, { "epoch": 1.3, "learning_rate": 4.627676429315106e-05, "loss": 0.8227, "step": 21514 }, { "epoch": 1.3, "learning_rate": 4.6275255383036834e-05, "loss": 0.8031, "step": 21516 }, { "epoch": 1.3, "learning_rate": 4.627374647292261e-05, "loss": 1.0377, "step": 21518 }, { "epoch": 1.3, "learning_rate": 4.627223756280839e-05, "loss": 0.6487, "step": 21520 }, { "epoch": 1.3, "learning_rate": 4.627072865269416e-05, "loss": 0.8715, "step": 21522 }, { "epoch": 1.3, "learning_rate": 4.626921974257994e-05, "loss": 0.9602, "step": 21524 }, { "epoch": 1.3, "learning_rate": 4.6267710832465713e-05, "loss": 1.1709, "step": 21526 }, { "epoch": 1.3, "learning_rate": 4.6266201922351487e-05, "loss": 1.0263, "step": 21528 }, { "epoch": 1.3, "learning_rate": 4.6264693012237267e-05, "loss": 1.1446, "step": 21530 }, { "epoch": 1.3, "learning_rate": 4.626318410212303e-05, "loss": 0.9092, "step": 21532 }, { "epoch": 1.3, "learning_rate": 4.626167519200881e-05, "loss": 0.8876, "step": 21534 }, { "epoch": 1.3, "learning_rate": 4.626016628189459e-05, "loss": 1.09, "step": 21536 }, { "epoch": 1.3, "learning_rate": 4.6258657371780366e-05, "loss": 0.9475, "step": 21538 }, { "epoch": 1.3, "learning_rate": 4.625714846166614e-05, "loss": 0.9783, "step": 21540 }, { "epoch": 1.3, "learning_rate": 4.625563955155192e-05, "loss": 0.9938, "step": 21542 }, { "epoch": 1.3, "learning_rate": 4.625413064143769e-05, "loss": 0.9993, "step": 21544 }, { "epoch": 1.3, "learning_rate": 4.625262173132347e-05, "loss": 0.9946, "step": 21546 }, { "epoch": 1.3, "learning_rate": 4.625111282120924e-05, "loss": 1.3898, "step": 21548 }, { "epoch": 1.3, "learning_rate": 4.624960391109502e-05, "loss": 1.2103, "step": 21550 }, { "epoch": 1.3, "learning_rate": 4.62480950009808e-05, "loss": 1.1371, "step": 21552 }, { "epoch": 1.3, "learning_rate": 4.6246586090866564e-05, "loss": 0.9655, "step": 21554 }, { "epoch": 1.3, "learning_rate": 4.6245077180752344e-05, "loss": 1.2757, "step": 21556 }, { "epoch": 1.3, "learning_rate": 4.6243568270638124e-05, "loss": 0.8259, "step": 21558 }, { "epoch": 1.3, "learning_rate": 4.62420593605239e-05, "loss": 1.2977, "step": 21560 }, { "epoch": 1.3, "learning_rate": 4.624055045040967e-05, "loss": 0.7255, "step": 21562 }, { "epoch": 1.3, "learning_rate": 4.6239041540295444e-05, "loss": 0.7779, "step": 21564 }, { "epoch": 1.3, "learning_rate": 4.6237532630181223e-05, "loss": 0.9189, "step": 21566 }, { "epoch": 1.3, "learning_rate": 4.6236023720066997e-05, "loss": 0.9483, "step": 21568 }, { "epoch": 1.3, "learning_rate": 4.623451480995277e-05, "loss": 1.1787, "step": 21570 }, { "epoch": 1.3, "learning_rate": 4.623300589983855e-05, "loss": 0.9966, "step": 21572 }, { "epoch": 1.3, "learning_rate": 4.623149698972433e-05, "loss": 1.2127, "step": 21574 }, { "epoch": 1.3, "learning_rate": 4.6229988079610096e-05, "loss": 0.5764, "step": 21576 }, { "epoch": 1.3, "learning_rate": 4.6228479169495876e-05, "loss": 1.4638, "step": 21578 }, { "epoch": 1.3, "learning_rate": 4.6226970259381656e-05, "loss": 0.7413, "step": 21580 }, { "epoch": 1.3, "learning_rate": 4.622546134926742e-05, "loss": 1.0918, "step": 21582 }, { "epoch": 1.3, "learning_rate": 4.62239524391532e-05, "loss": 1.1794, "step": 21584 }, { "epoch": 1.3, "learning_rate": 4.6222443529038975e-05, "loss": 0.9932, "step": 21586 }, { "epoch": 1.3, "learning_rate": 4.6220934618924755e-05, "loss": 1.2976, "step": 21588 }, { "epoch": 1.3, "learning_rate": 4.621942570881053e-05, "loss": 1.0672, "step": 21590 }, { "epoch": 1.3, "learning_rate": 4.62179167986963e-05, "loss": 1.3536, "step": 21592 }, { "epoch": 1.3, "learning_rate": 4.621640788858208e-05, "loss": 1.0773, "step": 21594 }, { "epoch": 1.3, "learning_rate": 4.621489897846786e-05, "loss": 0.9439, "step": 21596 }, { "epoch": 1.3, "learning_rate": 4.621339006835363e-05, "loss": 0.9157, "step": 21598 }, { "epoch": 1.3, "learning_rate": 4.621188115823941e-05, "loss": 1.0831, "step": 21600 }, { "epoch": 1.3, "learning_rate": 4.621037224812518e-05, "loss": 1.0335, "step": 21602 }, { "epoch": 1.3, "learning_rate": 4.6208863338010953e-05, "loss": 0.9866, "step": 21604 }, { "epoch": 1.3, "learning_rate": 4.620735442789673e-05, "loss": 0.8518, "step": 21606 }, { "epoch": 1.3, "learning_rate": 4.6205845517782506e-05, "loss": 0.9311, "step": 21608 }, { "epoch": 1.3, "learning_rate": 4.6204336607668286e-05, "loss": 0.6938, "step": 21610 }, { "epoch": 1.3, "learning_rate": 4.620282769755406e-05, "loss": 1.0427, "step": 21612 }, { "epoch": 1.3, "learning_rate": 4.620131878743983e-05, "loss": 1.0317, "step": 21614 }, { "epoch": 1.3, "learning_rate": 4.619980987732561e-05, "loss": 0.843, "step": 21616 }, { "epoch": 1.3, "learning_rate": 4.6198300967211386e-05, "loss": 1.3712, "step": 21618 }, { "epoch": 1.3, "learning_rate": 4.619679205709716e-05, "loss": 1.0963, "step": 21620 }, { "epoch": 1.31, "learning_rate": 4.619528314698294e-05, "loss": 0.9239, "step": 21622 }, { "epoch": 1.31, "learning_rate": 4.619377423686871e-05, "loss": 1.4481, "step": 21624 }, { "epoch": 1.31, "learning_rate": 4.6192265326754485e-05, "loss": 0.9653, "step": 21626 }, { "epoch": 1.31, "learning_rate": 4.6190756416640265e-05, "loss": 1.1018, "step": 21628 }, { "epoch": 1.31, "learning_rate": 4.618924750652604e-05, "loss": 0.8035, "step": 21630 }, { "epoch": 1.31, "learning_rate": 4.618773859641181e-05, "loss": 0.7875, "step": 21632 }, { "epoch": 1.31, "learning_rate": 4.618622968629759e-05, "loss": 1.201, "step": 21634 }, { "epoch": 1.31, "learning_rate": 4.6184720776183364e-05, "loss": 1.2145, "step": 21636 }, { "epoch": 1.31, "learning_rate": 4.6183211866069144e-05, "loss": 0.8369, "step": 21638 }, { "epoch": 1.31, "learning_rate": 4.618170295595492e-05, "loss": 0.9899, "step": 21640 }, { "epoch": 1.31, "learning_rate": 4.618019404584069e-05, "loss": 0.9354, "step": 21642 }, { "epoch": 1.31, "learning_rate": 4.617868513572647e-05, "loss": 0.9944, "step": 21644 }, { "epoch": 1.31, "learning_rate": 4.6177176225612237e-05, "loss": 0.8541, "step": 21646 }, { "epoch": 1.31, "learning_rate": 4.6175667315498016e-05, "loss": 0.8667, "step": 21648 }, { "epoch": 1.31, "learning_rate": 4.6174158405383796e-05, "loss": 0.725, "step": 21650 }, { "epoch": 1.31, "learning_rate": 4.617264949526957e-05, "loss": 0.9008, "step": 21652 }, { "epoch": 1.31, "learning_rate": 4.617114058515534e-05, "loss": 1.2127, "step": 21654 }, { "epoch": 1.31, "learning_rate": 4.616963167504112e-05, "loss": 0.9152, "step": 21656 }, { "epoch": 1.31, "learning_rate": 4.6168122764926896e-05, "loss": 0.9874, "step": 21658 }, { "epoch": 1.31, "learning_rate": 4.6166613854812676e-05, "loss": 0.9574, "step": 21660 }, { "epoch": 1.31, "learning_rate": 4.616510494469844e-05, "loss": 0.9768, "step": 21662 }, { "epoch": 1.31, "learning_rate": 4.616359603458422e-05, "loss": 0.7801, "step": 21664 }, { "epoch": 1.31, "learning_rate": 4.616208712447e-05, "loss": 0.9086, "step": 21666 }, { "epoch": 1.31, "learning_rate": 4.616057821435577e-05, "loss": 0.9144, "step": 21668 }, { "epoch": 1.31, "learning_rate": 4.615906930424155e-05, "loss": 0.7296, "step": 21670 }, { "epoch": 1.31, "learning_rate": 4.615756039412733e-05, "loss": 1.1187, "step": 21672 }, { "epoch": 1.31, "learning_rate": 4.61560514840131e-05, "loss": 0.8952, "step": 21674 }, { "epoch": 1.31, "learning_rate": 4.6154542573898874e-05, "loss": 0.8442, "step": 21676 }, { "epoch": 1.31, "learning_rate": 4.615303366378465e-05, "loss": 0.9306, "step": 21678 }, { "epoch": 1.31, "learning_rate": 4.615152475367043e-05, "loss": 0.9355, "step": 21680 }, { "epoch": 1.31, "learning_rate": 4.61500158435562e-05, "loss": 0.9966, "step": 21682 }, { "epoch": 1.31, "learning_rate": 4.614850693344197e-05, "loss": 1.2246, "step": 21684 }, { "epoch": 1.31, "learning_rate": 4.614699802332775e-05, "loss": 1.1557, "step": 21686 }, { "epoch": 1.31, "learning_rate": 4.614548911321353e-05, "loss": 1.2555, "step": 21688 }, { "epoch": 1.31, "learning_rate": 4.61439802030993e-05, "loss": 1.209, "step": 21690 }, { "epoch": 1.31, "learning_rate": 4.614247129298508e-05, "loss": 0.8315, "step": 21692 }, { "epoch": 1.31, "learning_rate": 4.614096238287086e-05, "loss": 0.8512, "step": 21694 }, { "epoch": 1.31, "learning_rate": 4.6139453472756626e-05, "loss": 1.0231, "step": 21696 }, { "epoch": 1.31, "learning_rate": 4.6137944562642406e-05, "loss": 0.9287, "step": 21698 }, { "epoch": 1.31, "learning_rate": 4.613643565252818e-05, "loss": 1.1571, "step": 21700 }, { "epoch": 1.31, "learning_rate": 4.613492674241396e-05, "loss": 0.9837, "step": 21702 }, { "epoch": 1.31, "learning_rate": 4.613341783229973e-05, "loss": 1.1954, "step": 21704 }, { "epoch": 1.31, "learning_rate": 4.6131908922185505e-05, "loss": 0.9046, "step": 21706 }, { "epoch": 1.31, "learning_rate": 4.6130400012071285e-05, "loss": 0.9918, "step": 21708 }, { "epoch": 1.31, "learning_rate": 4.6128891101957065e-05, "loss": 1.0239, "step": 21710 }, { "epoch": 1.31, "learning_rate": 4.612738219184283e-05, "loss": 0.9629, "step": 21712 }, { "epoch": 1.31, "learning_rate": 4.612587328172861e-05, "loss": 0.8286, "step": 21714 }, { "epoch": 1.31, "learning_rate": 4.6124364371614384e-05, "loss": 0.985, "step": 21716 }, { "epoch": 1.31, "learning_rate": 4.612285546150016e-05, "loss": 0.9522, "step": 21718 }, { "epoch": 1.31, "learning_rate": 4.612134655138594e-05, "loss": 0.9163, "step": 21720 }, { "epoch": 1.31, "learning_rate": 4.611983764127171e-05, "loss": 1.2174, "step": 21722 }, { "epoch": 1.31, "learning_rate": 4.611832873115749e-05, "loss": 1.0664, "step": 21724 }, { "epoch": 1.31, "learning_rate": 4.611681982104326e-05, "loss": 0.9819, "step": 21726 }, { "epoch": 1.31, "learning_rate": 4.6115310910929036e-05, "loss": 1.1533, "step": 21728 }, { "epoch": 1.31, "learning_rate": 4.6113802000814816e-05, "loss": 0.9138, "step": 21730 }, { "epoch": 1.31, "learning_rate": 4.611229309070059e-05, "loss": 0.8721, "step": 21732 }, { "epoch": 1.31, "learning_rate": 4.611078418058636e-05, "loss": 0.8118, "step": 21734 }, { "epoch": 1.31, "learning_rate": 4.610927527047214e-05, "loss": 1.3529, "step": 21736 }, { "epoch": 1.31, "learning_rate": 4.6107766360357915e-05, "loss": 1.119, "step": 21738 }, { "epoch": 1.31, "learning_rate": 4.610625745024369e-05, "loss": 1.1811, "step": 21740 }, { "epoch": 1.31, "learning_rate": 4.610474854012947e-05, "loss": 1.0031, "step": 21742 }, { "epoch": 1.31, "learning_rate": 4.610323963001524e-05, "loss": 1.1474, "step": 21744 }, { "epoch": 1.31, "learning_rate": 4.6101730719901015e-05, "loss": 1.3504, "step": 21746 }, { "epoch": 1.31, "learning_rate": 4.6100221809786795e-05, "loss": 1.0323, "step": 21748 }, { "epoch": 1.31, "learning_rate": 4.609871289967257e-05, "loss": 1.0258, "step": 21750 }, { "epoch": 1.31, "learning_rate": 4.609720398955835e-05, "loss": 0.844, "step": 21752 }, { "epoch": 1.31, "learning_rate": 4.609569507944412e-05, "loss": 0.816, "step": 21754 }, { "epoch": 1.31, "learning_rate": 4.6094186169329894e-05, "loss": 1.1102, "step": 21756 }, { "epoch": 1.31, "learning_rate": 4.6092677259215674e-05, "loss": 1.1278, "step": 21758 }, { "epoch": 1.31, "learning_rate": 4.609116834910145e-05, "loss": 0.6058, "step": 21760 }, { "epoch": 1.31, "learning_rate": 4.608965943898722e-05, "loss": 1.0136, "step": 21762 }, { "epoch": 1.31, "learning_rate": 4.6088150528873e-05, "loss": 0.9544, "step": 21764 }, { "epoch": 1.31, "learning_rate": 4.608664161875877e-05, "loss": 0.7685, "step": 21766 }, { "epoch": 1.31, "learning_rate": 4.6085132708644546e-05, "loss": 0.7047, "step": 21768 }, { "epoch": 1.31, "learning_rate": 4.6083623798530326e-05, "loss": 0.81, "step": 21770 }, { "epoch": 1.31, "learning_rate": 4.60821148884161e-05, "loss": 1.2221, "step": 21772 }, { "epoch": 1.31, "learning_rate": 4.608060597830188e-05, "loss": 0.7273, "step": 21774 }, { "epoch": 1.31, "learning_rate": 4.6079097068187646e-05, "loss": 1.2334, "step": 21776 }, { "epoch": 1.31, "learning_rate": 4.6077588158073425e-05, "loss": 1.0118, "step": 21778 }, { "epoch": 1.31, "learning_rate": 4.6076079247959205e-05, "loss": 1.0017, "step": 21780 }, { "epoch": 1.31, "learning_rate": 4.607457033784497e-05, "loss": 1.0471, "step": 21782 }, { "epoch": 1.31, "learning_rate": 4.607306142773075e-05, "loss": 0.752, "step": 21784 }, { "epoch": 1.31, "learning_rate": 4.607155251761653e-05, "loss": 1.1124, "step": 21786 }, { "epoch": 1.32, "learning_rate": 4.6070043607502305e-05, "loss": 0.9061, "step": 21788 }, { "epoch": 1.32, "learning_rate": 4.606853469738808e-05, "loss": 0.8998, "step": 21790 }, { "epoch": 1.32, "learning_rate": 4.606702578727385e-05, "loss": 0.8407, "step": 21792 }, { "epoch": 1.32, "learning_rate": 4.606551687715963e-05, "loss": 0.9595, "step": 21794 }, { "epoch": 1.32, "learning_rate": 4.6064007967045404e-05, "loss": 0.8024, "step": 21796 }, { "epoch": 1.32, "learning_rate": 4.606249905693118e-05, "loss": 1.0755, "step": 21798 }, { "epoch": 1.32, "learning_rate": 4.606099014681696e-05, "loss": 0.8237, "step": 21800 }, { "epoch": 1.32, "learning_rate": 4.605948123670274e-05, "loss": 0.9376, "step": 21802 }, { "epoch": 1.32, "learning_rate": 4.60579723265885e-05, "loss": 1.2644, "step": 21804 }, { "epoch": 1.32, "learning_rate": 4.605646341647428e-05, "loss": 0.9188, "step": 21806 }, { "epoch": 1.32, "learning_rate": 4.605495450636006e-05, "loss": 1.0106, "step": 21808 }, { "epoch": 1.32, "learning_rate": 4.605344559624583e-05, "loss": 1.2024, "step": 21810 }, { "epoch": 1.32, "learning_rate": 4.605193668613161e-05, "loss": 1.1093, "step": 21812 }, { "epoch": 1.32, "learning_rate": 4.605042777601738e-05, "loss": 1.2645, "step": 21814 }, { "epoch": 1.32, "learning_rate": 4.604891886590316e-05, "loss": 0.8738, "step": 21816 }, { "epoch": 1.32, "learning_rate": 4.6047409955788935e-05, "loss": 1.0158, "step": 21818 }, { "epoch": 1.32, "learning_rate": 4.604590104567471e-05, "loss": 1.0821, "step": 21820 }, { "epoch": 1.32, "learning_rate": 4.604439213556049e-05, "loss": 1.2935, "step": 21822 }, { "epoch": 1.32, "learning_rate": 4.604288322544627e-05, "loss": 1.0259, "step": 21824 }, { "epoch": 1.32, "learning_rate": 4.6041374315332035e-05, "loss": 0.7101, "step": 21826 }, { "epoch": 1.32, "learning_rate": 4.6039865405217815e-05, "loss": 1.0687, "step": 21828 }, { "epoch": 1.32, "learning_rate": 4.603835649510359e-05, "loss": 1.1438, "step": 21830 }, { "epoch": 1.32, "learning_rate": 4.603684758498936e-05, "loss": 1.0588, "step": 21832 }, { "epoch": 1.32, "learning_rate": 4.603533867487514e-05, "loss": 0.8797, "step": 21834 }, { "epoch": 1.32, "learning_rate": 4.6033829764760914e-05, "loss": 1.019, "step": 21836 }, { "epoch": 1.32, "learning_rate": 4.6032320854646694e-05, "loss": 0.9877, "step": 21838 }, { "epoch": 1.32, "learning_rate": 4.603081194453247e-05, "loss": 1.0701, "step": 21840 }, { "epoch": 1.32, "learning_rate": 4.602930303441824e-05, "loss": 1.2369, "step": 21842 }, { "epoch": 1.32, "learning_rate": 4.602779412430402e-05, "loss": 1.1605, "step": 21844 }, { "epoch": 1.32, "learning_rate": 4.602628521418979e-05, "loss": 0.8897, "step": 21846 }, { "epoch": 1.32, "learning_rate": 4.6024776304075566e-05, "loss": 1.095, "step": 21848 }, { "epoch": 1.32, "learning_rate": 4.6023267393961346e-05, "loss": 0.9876, "step": 21850 }, { "epoch": 1.32, "learning_rate": 4.602175848384712e-05, "loss": 1.0947, "step": 21852 }, { "epoch": 1.32, "learning_rate": 4.602024957373289e-05, "loss": 0.9102, "step": 21854 }, { "epoch": 1.32, "learning_rate": 4.601874066361867e-05, "loss": 1.005, "step": 21856 }, { "epoch": 1.32, "learning_rate": 4.6017231753504445e-05, "loss": 0.9997, "step": 21858 }, { "epoch": 1.32, "learning_rate": 4.601572284339022e-05, "loss": 0.7618, "step": 21860 }, { "epoch": 1.32, "learning_rate": 4.6014213933276e-05, "loss": 0.9469, "step": 21862 }, { "epoch": 1.32, "learning_rate": 4.601270502316177e-05, "loss": 1.2067, "step": 21864 }, { "epoch": 1.32, "learning_rate": 4.601119611304755e-05, "loss": 0.9752, "step": 21866 }, { "epoch": 1.32, "learning_rate": 4.6009687202933325e-05, "loss": 1.0741, "step": 21868 }, { "epoch": 1.32, "learning_rate": 4.60081782928191e-05, "loss": 0.9653, "step": 21870 }, { "epoch": 1.32, "learning_rate": 4.600666938270488e-05, "loss": 0.7713, "step": 21872 }, { "epoch": 1.32, "learning_rate": 4.600516047259065e-05, "loss": 1.0339, "step": 21874 }, { "epoch": 1.32, "learning_rate": 4.6003651562476424e-05, "loss": 1.0262, "step": 21876 }, { "epoch": 1.32, "learning_rate": 4.6002142652362204e-05, "loss": 0.9086, "step": 21878 }, { "epoch": 1.32, "learning_rate": 4.600063374224798e-05, "loss": 0.7149, "step": 21880 }, { "epoch": 1.32, "learning_rate": 4.599912483213375e-05, "loss": 0.8469, "step": 21882 }, { "epoch": 1.32, "learning_rate": 4.599761592201953e-05, "loss": 0.8921, "step": 21884 }, { "epoch": 1.32, "learning_rate": 4.59961070119053e-05, "loss": 1.0357, "step": 21886 }, { "epoch": 1.32, "learning_rate": 4.599459810179108e-05, "loss": 1.0671, "step": 21888 }, { "epoch": 1.32, "learning_rate": 4.599308919167685e-05, "loss": 1.0147, "step": 21890 }, { "epoch": 1.32, "learning_rate": 4.599158028156263e-05, "loss": 0.9665, "step": 21892 }, { "epoch": 1.32, "learning_rate": 4.599007137144841e-05, "loss": 1.2083, "step": 21894 }, { "epoch": 1.32, "learning_rate": 4.5988562461334175e-05, "loss": 0.9584, "step": 21896 }, { "epoch": 1.32, "learning_rate": 4.5987053551219955e-05, "loss": 0.9006, "step": 21898 }, { "epoch": 1.32, "learning_rate": 4.5985544641105735e-05, "loss": 0.7416, "step": 21900 }, { "epoch": 1.32, "learning_rate": 4.598403573099151e-05, "loss": 1.2004, "step": 21902 }, { "epoch": 1.32, "learning_rate": 4.598252682087728e-05, "loss": 0.7077, "step": 21904 }, { "epoch": 1.32, "learning_rate": 4.5981017910763055e-05, "loss": 1.4568, "step": 21906 }, { "epoch": 1.32, "learning_rate": 4.5979509000648834e-05, "loss": 0.7192, "step": 21908 }, { "epoch": 1.32, "learning_rate": 4.597800009053461e-05, "loss": 0.8692, "step": 21910 }, { "epoch": 1.32, "learning_rate": 4.597649118042038e-05, "loss": 0.9493, "step": 21912 }, { "epoch": 1.32, "learning_rate": 4.597498227030616e-05, "loss": 1.1329, "step": 21914 }, { "epoch": 1.32, "learning_rate": 4.597347336019194e-05, "loss": 1.1892, "step": 21916 }, { "epoch": 1.32, "learning_rate": 4.597196445007771e-05, "loss": 1.0423, "step": 21918 }, { "epoch": 1.32, "learning_rate": 4.597045553996349e-05, "loss": 0.9995, "step": 21920 }, { "epoch": 1.32, "learning_rate": 4.596894662984926e-05, "loss": 1.0464, "step": 21922 }, { "epoch": 1.32, "learning_rate": 4.596743771973504e-05, "loss": 1.2574, "step": 21924 }, { "epoch": 1.32, "learning_rate": 4.596592880962081e-05, "loss": 0.8744, "step": 21926 }, { "epoch": 1.32, "learning_rate": 4.5964419899506586e-05, "loss": 1.1468, "step": 21928 }, { "epoch": 1.32, "learning_rate": 4.5962910989392366e-05, "loss": 1.2026, "step": 21930 }, { "epoch": 1.32, "learning_rate": 4.596140207927814e-05, "loss": 0.7057, "step": 21932 }, { "epoch": 1.32, "learning_rate": 4.595989316916391e-05, "loss": 0.9493, "step": 21934 }, { "epoch": 1.32, "learning_rate": 4.595838425904969e-05, "loss": 0.7095, "step": 21936 }, { "epoch": 1.32, "learning_rate": 4.595687534893547e-05, "loss": 0.9662, "step": 21938 }, { "epoch": 1.32, "learning_rate": 4.595536643882124e-05, "loss": 0.7977, "step": 21940 }, { "epoch": 1.32, "learning_rate": 4.595385752870702e-05, "loss": 1.0171, "step": 21942 }, { "epoch": 1.32, "learning_rate": 4.595234861859279e-05, "loss": 0.9768, "step": 21944 }, { "epoch": 1.32, "learning_rate": 4.5950839708478564e-05, "loss": 1.0379, "step": 21946 }, { "epoch": 1.32, "learning_rate": 4.5949330798364344e-05, "loss": 1.0119, "step": 21948 }, { "epoch": 1.32, "learning_rate": 4.594782188825012e-05, "loss": 0.8801, "step": 21950 }, { "epoch": 1.32, "learning_rate": 4.59463129781359e-05, "loss": 1.2895, "step": 21952 }, { "epoch": 1.33, "learning_rate": 4.594480406802167e-05, "loss": 1.1608, "step": 21954 }, { "epoch": 1.33, "learning_rate": 4.5943295157907444e-05, "loss": 1.2881, "step": 21956 }, { "epoch": 1.33, "learning_rate": 4.5941786247793224e-05, "loss": 1.1152, "step": 21958 }, { "epoch": 1.33, "learning_rate": 4.5940277337679e-05, "loss": 1.1411, "step": 21960 }, { "epoch": 1.33, "learning_rate": 4.593876842756477e-05, "loss": 1.0071, "step": 21962 }, { "epoch": 1.33, "learning_rate": 4.593725951745055e-05, "loss": 1.2077, "step": 21964 }, { "epoch": 1.33, "learning_rate": 4.593575060733632e-05, "loss": 0.928, "step": 21966 }, { "epoch": 1.33, "learning_rate": 4.5934241697222096e-05, "loss": 1.0615, "step": 21968 }, { "epoch": 1.33, "learning_rate": 4.5932732787107876e-05, "loss": 0.7201, "step": 21970 }, { "epoch": 1.33, "learning_rate": 4.593122387699365e-05, "loss": 0.8632, "step": 21972 }, { "epoch": 1.33, "learning_rate": 4.592971496687942e-05, "loss": 1.3525, "step": 21974 }, { "epoch": 1.33, "learning_rate": 4.59282060567652e-05, "loss": 0.9677, "step": 21976 }, { "epoch": 1.33, "learning_rate": 4.5926697146650975e-05, "loss": 1.2575, "step": 21978 }, { "epoch": 1.33, "learning_rate": 4.5925188236536755e-05, "loss": 0.8503, "step": 21980 }, { "epoch": 1.33, "learning_rate": 4.592367932642253e-05, "loss": 0.9201, "step": 21982 }, { "epoch": 1.33, "learning_rate": 4.59221704163083e-05, "loss": 0.7716, "step": 21984 }, { "epoch": 1.33, "learning_rate": 4.592066150619408e-05, "loss": 1.1141, "step": 21986 }, { "epoch": 1.33, "learning_rate": 4.5919152596079854e-05, "loss": 1.35, "step": 21988 }, { "epoch": 1.33, "learning_rate": 4.591764368596563e-05, "loss": 1.0756, "step": 21990 }, { "epoch": 1.33, "learning_rate": 4.591613477585141e-05, "loss": 1.1745, "step": 21992 }, { "epoch": 1.33, "learning_rate": 4.591462586573718e-05, "loss": 0.8898, "step": 21994 }, { "epoch": 1.33, "learning_rate": 4.5913116955622954e-05, "loss": 0.7608, "step": 21996 }, { "epoch": 1.33, "learning_rate": 4.5911608045508734e-05, "loss": 0.9801, "step": 21998 }, { "epoch": 1.33, "learning_rate": 4.591009913539451e-05, "loss": 0.9958, "step": 22000 }, { "epoch": 1.33, "learning_rate": 4.5908590225280287e-05, "loss": 1.1703, "step": 22002 }, { "epoch": 1.33, "learning_rate": 4.590708131516605e-05, "loss": 0.9785, "step": 22004 }, { "epoch": 1.33, "learning_rate": 4.590557240505183e-05, "loss": 1.0137, "step": 22006 }, { "epoch": 1.33, "learning_rate": 4.590406349493761e-05, "loss": 0.8015, "step": 22008 }, { "epoch": 1.33, "learning_rate": 4.590255458482338e-05, "loss": 0.9286, "step": 22010 }, { "epoch": 1.33, "learning_rate": 4.590104567470916e-05, "loss": 0.7316, "step": 22012 }, { "epoch": 1.33, "learning_rate": 4.589953676459494e-05, "loss": 1.1099, "step": 22014 }, { "epoch": 1.33, "learning_rate": 4.589802785448071e-05, "loss": 1.2247, "step": 22016 }, { "epoch": 1.33, "learning_rate": 4.5896518944366485e-05, "loss": 0.7248, "step": 22018 }, { "epoch": 1.33, "learning_rate": 4.589501003425226e-05, "loss": 0.9803, "step": 22020 }, { "epoch": 1.33, "learning_rate": 4.589350112413804e-05, "loss": 0.7667, "step": 22022 }, { "epoch": 1.33, "learning_rate": 4.589199221402381e-05, "loss": 1.1963, "step": 22024 }, { "epoch": 1.33, "learning_rate": 4.5890483303909584e-05, "loss": 0.9039, "step": 22026 }, { "epoch": 1.33, "learning_rate": 4.5888974393795364e-05, "loss": 0.9664, "step": 22028 }, { "epoch": 1.33, "learning_rate": 4.5887465483681144e-05, "loss": 0.8258, "step": 22030 }, { "epoch": 1.33, "learning_rate": 4.588595657356691e-05, "loss": 0.949, "step": 22032 }, { "epoch": 1.33, "learning_rate": 4.588444766345269e-05, "loss": 0.6183, "step": 22034 }, { "epoch": 1.33, "learning_rate": 4.5882938753338464e-05, "loss": 1.1725, "step": 22036 }, { "epoch": 1.33, "learning_rate": 4.5881429843224243e-05, "loss": 0.7373, "step": 22038 }, { "epoch": 1.33, "learning_rate": 4.5879920933110017e-05, "loss": 1.1304, "step": 22040 }, { "epoch": 1.33, "learning_rate": 4.587841202299579e-05, "loss": 0.9271, "step": 22042 }, { "epoch": 1.33, "learning_rate": 4.587690311288157e-05, "loss": 1.1107, "step": 22044 }, { "epoch": 1.33, "learning_rate": 4.587539420276734e-05, "loss": 1.1251, "step": 22046 }, { "epoch": 1.33, "learning_rate": 4.5873885292653116e-05, "loss": 0.7958, "step": 22048 }, { "epoch": 1.33, "learning_rate": 4.5872376382538896e-05, "loss": 1.0712, "step": 22050 }, { "epoch": 1.33, "learning_rate": 4.5870867472424676e-05, "loss": 0.7604, "step": 22052 }, { "epoch": 1.33, "learning_rate": 4.586935856231044e-05, "loss": 1.3936, "step": 22054 }, { "epoch": 1.33, "learning_rate": 4.586784965219622e-05, "loss": 0.971, "step": 22056 }, { "epoch": 1.33, "learning_rate": 4.5866340742081995e-05, "loss": 0.7946, "step": 22058 }, { "epoch": 1.33, "learning_rate": 4.586483183196777e-05, "loss": 1.0872, "step": 22060 }, { "epoch": 1.33, "learning_rate": 4.586332292185355e-05, "loss": 0.9646, "step": 22062 }, { "epoch": 1.33, "learning_rate": 4.586181401173932e-05, "loss": 0.9035, "step": 22064 }, { "epoch": 1.33, "learning_rate": 4.58603051016251e-05, "loss": 0.7555, "step": 22066 }, { "epoch": 1.33, "learning_rate": 4.5858796191510874e-05, "loss": 0.8069, "step": 22068 }, { "epoch": 1.33, "learning_rate": 4.585728728139665e-05, "loss": 0.9436, "step": 22070 }, { "epoch": 1.33, "learning_rate": 4.585577837128243e-05, "loss": 0.9787, "step": 22072 }, { "epoch": 1.33, "learning_rate": 4.58542694611682e-05, "loss": 1.0988, "step": 22074 }, { "epoch": 1.33, "learning_rate": 4.5852760551053974e-05, "loss": 1.2271, "step": 22076 }, { "epoch": 1.33, "learning_rate": 4.5851251640939753e-05, "loss": 0.7529, "step": 22078 }, { "epoch": 1.33, "learning_rate": 4.5849742730825527e-05, "loss": 0.9613, "step": 22080 }, { "epoch": 1.33, "learning_rate": 4.58482338207113e-05, "loss": 1.0554, "step": 22082 }, { "epoch": 1.33, "learning_rate": 4.584672491059708e-05, "loss": 0.9774, "step": 22084 }, { "epoch": 1.33, "learning_rate": 4.584521600048285e-05, "loss": 0.8243, "step": 22086 }, { "epoch": 1.33, "learning_rate": 4.584370709036863e-05, "loss": 0.9178, "step": 22088 }, { "epoch": 1.33, "learning_rate": 4.5842198180254406e-05, "loss": 1.0498, "step": 22090 }, { "epoch": 1.33, "learning_rate": 4.584068927014018e-05, "loss": 0.9094, "step": 22092 }, { "epoch": 1.33, "learning_rate": 4.583918036002596e-05, "loss": 1.05, "step": 22094 }, { "epoch": 1.33, "learning_rate": 4.583767144991173e-05, "loss": 0.877, "step": 22096 }, { "epoch": 1.33, "learning_rate": 4.5836162539797505e-05, "loss": 0.6199, "step": 22098 }, { "epoch": 1.33, "learning_rate": 4.5834653629683285e-05, "loss": 0.9717, "step": 22100 }, { "epoch": 1.33, "learning_rate": 4.583314471956906e-05, "loss": 0.8081, "step": 22102 }, { "epoch": 1.33, "learning_rate": 4.583163580945483e-05, "loss": 0.7628, "step": 22104 }, { "epoch": 1.33, "learning_rate": 4.583012689934061e-05, "loss": 1.2039, "step": 22106 }, { "epoch": 1.33, "learning_rate": 4.5828617989226384e-05, "loss": 0.8696, "step": 22108 }, { "epoch": 1.33, "learning_rate": 4.582710907911216e-05, "loss": 0.8674, "step": 22110 }, { "epoch": 1.33, "learning_rate": 4.582560016899794e-05, "loss": 1.0677, "step": 22112 }, { "epoch": 1.33, "learning_rate": 4.582409125888371e-05, "loss": 1.0157, "step": 22114 }, { "epoch": 1.33, "learning_rate": 4.582258234876949e-05, "loss": 1.0342, "step": 22116 }, { "epoch": 1.33, "learning_rate": 4.5821073438655257e-05, "loss": 1.0377, "step": 22118 }, { "epoch": 1.34, "learning_rate": 4.5819564528541036e-05, "loss": 1.3721, "step": 22120 }, { "epoch": 1.34, "learning_rate": 4.5818055618426816e-05, "loss": 0.9943, "step": 22122 }, { "epoch": 1.34, "learning_rate": 4.581654670831258e-05, "loss": 0.7585, "step": 22124 }, { "epoch": 1.34, "learning_rate": 4.581503779819836e-05, "loss": 0.9311, "step": 22126 }, { "epoch": 1.34, "learning_rate": 4.581352888808414e-05, "loss": 0.6993, "step": 22128 }, { "epoch": 1.34, "learning_rate": 4.5812019977969916e-05, "loss": 0.9491, "step": 22130 }, { "epoch": 1.34, "learning_rate": 4.581051106785569e-05, "loss": 0.767, "step": 22132 }, { "epoch": 1.34, "learning_rate": 4.580900215774146e-05, "loss": 1.1869, "step": 22134 }, { "epoch": 1.34, "learning_rate": 4.580749324762724e-05, "loss": 1.0786, "step": 22136 }, { "epoch": 1.34, "learning_rate": 4.580598433751302e-05, "loss": 0.9158, "step": 22138 }, { "epoch": 1.34, "learning_rate": 4.580447542739879e-05, "loss": 1.1036, "step": 22140 }, { "epoch": 1.34, "learning_rate": 4.580296651728457e-05, "loss": 0.9672, "step": 22142 }, { "epoch": 1.34, "learning_rate": 4.580145760717035e-05, "loss": 1.1066, "step": 22144 }, { "epoch": 1.34, "learning_rate": 4.5799948697056114e-05, "loss": 0.7169, "step": 22146 }, { "epoch": 1.34, "learning_rate": 4.5798439786941894e-05, "loss": 1.0203, "step": 22148 }, { "epoch": 1.34, "learning_rate": 4.579693087682767e-05, "loss": 0.9399, "step": 22150 }, { "epoch": 1.34, "learning_rate": 4.579542196671345e-05, "loss": 0.884, "step": 22152 }, { "epoch": 1.34, "learning_rate": 4.579391305659922e-05, "loss": 1.107, "step": 22154 }, { "epoch": 1.34, "learning_rate": 4.5792404146484993e-05, "loss": 0.9014, "step": 22156 }, { "epoch": 1.34, "learning_rate": 4.579089523637077e-05, "loss": 0.9628, "step": 22158 }, { "epoch": 1.34, "learning_rate": 4.5789386326256546e-05, "loss": 0.8757, "step": 22160 }, { "epoch": 1.34, "learning_rate": 4.578787741614232e-05, "loss": 0.886, "step": 22162 }, { "epoch": 1.34, "learning_rate": 4.57863685060281e-05, "loss": 1.0346, "step": 22164 }, { "epoch": 1.34, "learning_rate": 4.578485959591388e-05, "loss": 1.0301, "step": 22166 }, { "epoch": 1.34, "learning_rate": 4.5783350685799646e-05, "loss": 1.3671, "step": 22168 }, { "epoch": 1.34, "learning_rate": 4.5781841775685426e-05, "loss": 1.289, "step": 22170 }, { "epoch": 1.34, "learning_rate": 4.57803328655712e-05, "loss": 0.9095, "step": 22172 }, { "epoch": 1.34, "learning_rate": 4.577882395545697e-05, "loss": 1.0609, "step": 22174 }, { "epoch": 1.34, "learning_rate": 4.577731504534275e-05, "loss": 0.7799, "step": 22176 }, { "epoch": 1.34, "learning_rate": 4.5775806135228525e-05, "loss": 0.9922, "step": 22178 }, { "epoch": 1.34, "learning_rate": 4.5774297225114305e-05, "loss": 0.7141, "step": 22180 }, { "epoch": 1.34, "learning_rate": 4.577278831500008e-05, "loss": 0.9264, "step": 22182 }, { "epoch": 1.34, "learning_rate": 4.577127940488585e-05, "loss": 0.9014, "step": 22184 }, { "epoch": 1.34, "learning_rate": 4.576977049477163e-05, "loss": 0.6803, "step": 22186 }, { "epoch": 1.34, "learning_rate": 4.5768261584657404e-05, "loss": 0.8563, "step": 22188 }, { "epoch": 1.34, "learning_rate": 4.576675267454318e-05, "loss": 0.9063, "step": 22190 }, { "epoch": 1.34, "learning_rate": 4.576524376442896e-05, "loss": 1.0299, "step": 22192 }, { "epoch": 1.34, "learning_rate": 4.576373485431473e-05, "loss": 0.9517, "step": 22194 }, { "epoch": 1.34, "learning_rate": 4.57622259442005e-05, "loss": 0.9909, "step": 22196 }, { "epoch": 1.34, "learning_rate": 4.576071703408628e-05, "loss": 1.0756, "step": 22198 }, { "epoch": 1.34, "learning_rate": 4.5759208123972056e-05, "loss": 0.8721, "step": 22200 }, { "epoch": 1.34, "learning_rate": 4.5757699213857836e-05, "loss": 0.8629, "step": 22202 }, { "epoch": 1.34, "learning_rate": 4.575619030374361e-05, "loss": 1.1554, "step": 22204 }, { "epoch": 1.34, "learning_rate": 4.575468139362938e-05, "loss": 0.7214, "step": 22206 }, { "epoch": 1.34, "learning_rate": 4.575317248351516e-05, "loss": 1.0107, "step": 22208 }, { "epoch": 1.34, "learning_rate": 4.5751663573400936e-05, "loss": 0.865, "step": 22210 }, { "epoch": 1.34, "learning_rate": 4.575015466328671e-05, "loss": 0.936, "step": 22212 }, { "epoch": 1.34, "learning_rate": 4.574864575317249e-05, "loss": 0.7381, "step": 22214 }, { "epoch": 1.34, "learning_rate": 4.574713684305826e-05, "loss": 0.8621, "step": 22216 }, { "epoch": 1.34, "learning_rate": 4.5745627932944035e-05, "loss": 1.4863, "step": 22218 }, { "epoch": 1.34, "learning_rate": 4.5744119022829815e-05, "loss": 0.9465, "step": 22220 }, { "epoch": 1.34, "learning_rate": 4.574261011271559e-05, "loss": 0.83, "step": 22222 }, { "epoch": 1.34, "learning_rate": 4.574110120260136e-05, "loss": 1.3047, "step": 22224 }, { "epoch": 1.34, "learning_rate": 4.573959229248714e-05, "loss": 0.8668, "step": 22226 }, { "epoch": 1.34, "learning_rate": 4.5738083382372914e-05, "loss": 0.9242, "step": 22228 }, { "epoch": 1.34, "learning_rate": 4.5736574472258694e-05, "loss": 0.9324, "step": 22230 }, { "epoch": 1.34, "learning_rate": 4.573506556214446e-05, "loss": 1.1222, "step": 22232 }, { "epoch": 1.34, "learning_rate": 4.573355665203024e-05, "loss": 0.8587, "step": 22234 }, { "epoch": 1.34, "learning_rate": 4.573204774191602e-05, "loss": 0.9324, "step": 22236 }, { "epoch": 1.34, "learning_rate": 4.5730538831801786e-05, "loss": 1.079, "step": 22238 }, { "epoch": 1.34, "learning_rate": 4.5729029921687566e-05, "loss": 0.8868, "step": 22240 }, { "epoch": 1.34, "learning_rate": 4.5727521011573346e-05, "loss": 0.8884, "step": 22242 }, { "epoch": 1.34, "learning_rate": 4.572601210145912e-05, "loss": 1.0309, "step": 22244 }, { "epoch": 1.34, "learning_rate": 4.572450319134489e-05, "loss": 0.9513, "step": 22246 }, { "epoch": 1.34, "learning_rate": 4.5722994281230666e-05, "loss": 0.9162, "step": 22248 }, { "epoch": 1.34, "learning_rate": 4.5721485371116445e-05, "loss": 1.0932, "step": 22250 }, { "epoch": 1.34, "learning_rate": 4.5719976461002225e-05, "loss": 0.7589, "step": 22252 }, { "epoch": 1.34, "learning_rate": 4.571846755088799e-05, "loss": 1.0583, "step": 22254 }, { "epoch": 1.34, "learning_rate": 4.571695864077377e-05, "loss": 1.0225, "step": 22256 }, { "epoch": 1.34, "learning_rate": 4.571544973065955e-05, "loss": 0.8367, "step": 22258 }, { "epoch": 1.34, "learning_rate": 4.571394082054532e-05, "loss": 0.9869, "step": 22260 }, { "epoch": 1.34, "learning_rate": 4.57124319104311e-05, "loss": 1.1656, "step": 22262 }, { "epoch": 1.34, "learning_rate": 4.571092300031687e-05, "loss": 1.1802, "step": 22264 }, { "epoch": 1.34, "learning_rate": 4.570941409020265e-05, "loss": 1.0923, "step": 22266 }, { "epoch": 1.34, "learning_rate": 4.5707905180088424e-05, "loss": 0.9522, "step": 22268 }, { "epoch": 1.34, "learning_rate": 4.57063962699742e-05, "loss": 0.7585, "step": 22270 }, { "epoch": 1.34, "learning_rate": 4.570488735985998e-05, "loss": 0.8526, "step": 22272 }, { "epoch": 1.34, "learning_rate": 4.570337844974575e-05, "loss": 0.9046, "step": 22274 }, { "epoch": 1.34, "learning_rate": 4.570186953963152e-05, "loss": 0.8631, "step": 22276 }, { "epoch": 1.34, "learning_rate": 4.57003606295173e-05, "loss": 0.9931, "step": 22278 }, { "epoch": 1.34, "learning_rate": 4.5698851719403076e-05, "loss": 0.8417, "step": 22280 }, { "epoch": 1.34, "learning_rate": 4.569734280928885e-05, "loss": 1.0778, "step": 22282 }, { "epoch": 1.35, "learning_rate": 4.569583389917463e-05, "loss": 0.8993, "step": 22284 }, { "epoch": 1.35, "learning_rate": 4.569507944411752e-05, "loss": 1.0334, "step": 22286 }, { "epoch": 1.35, "learning_rate": 4.569357053400329e-05, "loss": 0.8242, "step": 22288 }, { "epoch": 1.35, "learning_rate": 4.5692061623889065e-05, "loss": 1.06, "step": 22290 }, { "epoch": 1.35, "learning_rate": 4.5690552713774845e-05, "loss": 0.9524, "step": 22292 }, { "epoch": 1.35, "learning_rate": 4.568904380366062e-05, "loss": 1.1482, "step": 22294 }, { "epoch": 1.35, "learning_rate": 4.568753489354639e-05, "loss": 1.0902, "step": 22296 }, { "epoch": 1.35, "learning_rate": 4.5686025983432165e-05, "loss": 0.9668, "step": 22298 }, { "epoch": 1.35, "learning_rate": 4.5684517073317945e-05, "loss": 1.1495, "step": 22300 }, { "epoch": 1.35, "learning_rate": 4.5683008163203725e-05, "loss": 1.1678, "step": 22302 }, { "epoch": 1.35, "learning_rate": 4.568149925308949e-05, "loss": 0.8656, "step": 22304 }, { "epoch": 1.35, "learning_rate": 4.567999034297527e-05, "loss": 0.8208, "step": 22306 }, { "epoch": 1.35, "learning_rate": 4.567848143286105e-05, "loss": 1.0543, "step": 22308 }, { "epoch": 1.35, "learning_rate": 4.5676972522746824e-05, "loss": 1.2237, "step": 22310 }, { "epoch": 1.35, "learning_rate": 4.56754636126326e-05, "loss": 0.9899, "step": 22312 }, { "epoch": 1.35, "learning_rate": 4.567395470251837e-05, "loss": 1.1539, "step": 22314 }, { "epoch": 1.35, "learning_rate": 4.567244579240415e-05, "loss": 1.2133, "step": 22316 }, { "epoch": 1.35, "learning_rate": 4.567093688228992e-05, "loss": 0.8184, "step": 22318 }, { "epoch": 1.35, "learning_rate": 4.5669427972175696e-05, "loss": 1.0712, "step": 22320 }, { "epoch": 1.35, "learning_rate": 4.5667919062061476e-05, "loss": 0.8934, "step": 22322 }, { "epoch": 1.35, "learning_rate": 4.5666410151947256e-05, "loss": 0.7618, "step": 22324 }, { "epoch": 1.35, "learning_rate": 4.566490124183302e-05, "loss": 0.9814, "step": 22326 }, { "epoch": 1.35, "learning_rate": 4.56633923317188e-05, "loss": 0.9809, "step": 22328 }, { "epoch": 1.35, "learning_rate": 4.5661883421604575e-05, "loss": 1.3323, "step": 22330 }, { "epoch": 1.35, "learning_rate": 4.566037451149035e-05, "loss": 0.9322, "step": 22332 }, { "epoch": 1.35, "learning_rate": 4.565886560137613e-05, "loss": 0.9705, "step": 22334 }, { "epoch": 1.35, "learning_rate": 4.56573566912619e-05, "loss": 1.0902, "step": 22336 }, { "epoch": 1.35, "learning_rate": 4.565584778114768e-05, "loss": 0.9298, "step": 22338 }, { "epoch": 1.35, "learning_rate": 4.5654338871033455e-05, "loss": 1.0304, "step": 22340 }, { "epoch": 1.35, "learning_rate": 4.565282996091923e-05, "loss": 0.7535, "step": 22342 }, { "epoch": 1.35, "learning_rate": 4.565132105080501e-05, "loss": 0.9291, "step": 22344 }, { "epoch": 1.35, "learning_rate": 4.564981214069078e-05, "loss": 1.0167, "step": 22346 }, { "epoch": 1.35, "learning_rate": 4.5648303230576554e-05, "loss": 0.8211, "step": 22348 }, { "epoch": 1.35, "learning_rate": 4.5646794320462334e-05, "loss": 0.7009, "step": 22350 }, { "epoch": 1.35, "learning_rate": 4.564528541034811e-05, "loss": 0.8406, "step": 22352 }, { "epoch": 1.35, "learning_rate": 4.564377650023388e-05, "loss": 0.9445, "step": 22354 }, { "epoch": 1.35, "learning_rate": 4.564226759011966e-05, "loss": 1.1411, "step": 22356 }, { "epoch": 1.35, "learning_rate": 4.564075868000543e-05, "loss": 0.8881, "step": 22358 }, { "epoch": 1.35, "learning_rate": 4.563924976989121e-05, "loss": 0.9644, "step": 22360 }, { "epoch": 1.35, "learning_rate": 4.5637740859776986e-05, "loss": 1.3327, "step": 22362 }, { "epoch": 1.35, "learning_rate": 4.563623194966276e-05, "loss": 0.6393, "step": 22364 }, { "epoch": 1.35, "learning_rate": 4.563472303954854e-05, "loss": 0.7842, "step": 22366 }, { "epoch": 1.35, "learning_rate": 4.563321412943431e-05, "loss": 0.9112, "step": 22368 }, { "epoch": 1.35, "learning_rate": 4.5631705219320085e-05, "loss": 0.9011, "step": 22370 }, { "epoch": 1.35, "learning_rate": 4.5630196309205865e-05, "loss": 1.061, "step": 22372 }, { "epoch": 1.35, "learning_rate": 4.562868739909164e-05, "loss": 1.1655, "step": 22374 }, { "epoch": 1.35, "learning_rate": 4.562717848897741e-05, "loss": 0.862, "step": 22376 }, { "epoch": 1.35, "learning_rate": 4.562566957886319e-05, "loss": 0.9233, "step": 22378 }, { "epoch": 1.35, "learning_rate": 4.5624160668748965e-05, "loss": 0.9795, "step": 22380 }, { "epoch": 1.35, "learning_rate": 4.562265175863474e-05, "loss": 0.7425, "step": 22382 }, { "epoch": 1.35, "learning_rate": 4.562114284852052e-05, "loss": 0.6519, "step": 22384 }, { "epoch": 1.35, "learning_rate": 4.561963393840629e-05, "loss": 0.9696, "step": 22386 }, { "epoch": 1.35, "learning_rate": 4.561812502829207e-05, "loss": 1.0377, "step": 22388 }, { "epoch": 1.35, "learning_rate": 4.5616616118177844e-05, "loss": 0.7924, "step": 22390 }, { "epoch": 1.35, "learning_rate": 4.561510720806362e-05, "loss": 1.0981, "step": 22392 }, { "epoch": 1.35, "learning_rate": 4.56135982979494e-05, "loss": 1.2804, "step": 22394 }, { "epoch": 1.35, "learning_rate": 4.561208938783516e-05, "loss": 0.9517, "step": 22396 }, { "epoch": 1.35, "learning_rate": 4.561058047772094e-05, "loss": 0.7812, "step": 22398 }, { "epoch": 1.35, "learning_rate": 4.560907156760672e-05, "loss": 0.9496, "step": 22400 }, { "epoch": 1.35, "learning_rate": 4.5607562657492496e-05, "loss": 0.9784, "step": 22402 }, { "epoch": 1.35, "learning_rate": 4.560605374737827e-05, "loss": 0.9868, "step": 22404 }, { "epoch": 1.35, "learning_rate": 4.560454483726405e-05, "loss": 1.013, "step": 22406 }, { "epoch": 1.35, "learning_rate": 4.560303592714982e-05, "loss": 0.9673, "step": 22408 }, { "epoch": 1.35, "learning_rate": 4.5601527017035595e-05, "loss": 0.7397, "step": 22410 }, { "epoch": 1.35, "learning_rate": 4.560001810692137e-05, "loss": 1.0911, "step": 22412 }, { "epoch": 1.35, "learning_rate": 4.559850919680715e-05, "loss": 1.2455, "step": 22414 }, { "epoch": 1.35, "learning_rate": 4.559700028669293e-05, "loss": 1.1373, "step": 22416 }, { "epoch": 1.35, "learning_rate": 4.5595491376578695e-05, "loss": 1.1863, "step": 22418 }, { "epoch": 1.35, "learning_rate": 4.5593982466464474e-05, "loss": 1.1243, "step": 22420 }, { "epoch": 1.35, "learning_rate": 4.5592473556350254e-05, "loss": 0.7865, "step": 22422 }, { "epoch": 1.35, "learning_rate": 4.559096464623603e-05, "loss": 0.8047, "step": 22424 }, { "epoch": 1.35, "learning_rate": 4.55894557361218e-05, "loss": 0.7523, "step": 22426 }, { "epoch": 1.35, "learning_rate": 4.5587946826007574e-05, "loss": 1.1236, "step": 22428 }, { "epoch": 1.35, "learning_rate": 4.5586437915893354e-05, "loss": 0.9418, "step": 22430 }, { "epoch": 1.35, "learning_rate": 4.558492900577913e-05, "loss": 1.1305, "step": 22432 }, { "epoch": 1.35, "learning_rate": 4.55834200956649e-05, "loss": 0.7693, "step": 22434 }, { "epoch": 1.35, "learning_rate": 4.558191118555068e-05, "loss": 0.8153, "step": 22436 }, { "epoch": 1.35, "learning_rate": 4.558040227543646e-05, "loss": 0.9546, "step": 22438 }, { "epoch": 1.35, "learning_rate": 4.5578893365322226e-05, "loss": 0.9642, "step": 22440 }, { "epoch": 1.35, "learning_rate": 4.5577384455208006e-05, "loss": 1.1104, "step": 22442 }, { "epoch": 1.35, "learning_rate": 4.557587554509378e-05, "loss": 1.3187, "step": 22444 }, { "epoch": 1.35, "learning_rate": 4.557436663497955e-05, "loss": 0.7552, "step": 22446 }, { "epoch": 1.35, "learning_rate": 4.557285772486533e-05, "loss": 0.9114, "step": 22448 }, { "epoch": 1.36, "learning_rate": 4.5571348814751105e-05, "loss": 1.0469, "step": 22450 }, { "epoch": 1.36, "learning_rate": 4.5569839904636885e-05, "loss": 0.9436, "step": 22452 }, { "epoch": 1.36, "learning_rate": 4.556833099452266e-05, "loss": 0.9556, "step": 22454 }, { "epoch": 1.36, "learning_rate": 4.556682208440843e-05, "loss": 1.1106, "step": 22456 }, { "epoch": 1.36, "learning_rate": 4.556531317429421e-05, "loss": 1.0651, "step": 22458 }, { "epoch": 1.36, "learning_rate": 4.5563804264179984e-05, "loss": 1.1619, "step": 22460 }, { "epoch": 1.36, "learning_rate": 4.556229535406576e-05, "loss": 0.8413, "step": 22462 }, { "epoch": 1.36, "learning_rate": 4.556078644395154e-05, "loss": 0.8466, "step": 22464 }, { "epoch": 1.36, "learning_rate": 4.555927753383731e-05, "loss": 1.3408, "step": 22466 }, { "epoch": 1.36, "learning_rate": 4.5557768623723084e-05, "loss": 1.4884, "step": 22468 }, { "epoch": 1.36, "learning_rate": 4.5556259713608864e-05, "loss": 0.7505, "step": 22470 }, { "epoch": 1.36, "learning_rate": 4.555475080349464e-05, "loss": 0.9403, "step": 22472 }, { "epoch": 1.36, "learning_rate": 4.555324189338042e-05, "loss": 0.8732, "step": 22474 }, { "epoch": 1.36, "learning_rate": 4.555173298326619e-05, "loss": 1.0219, "step": 22476 }, { "epoch": 1.36, "learning_rate": 4.555022407315196e-05, "loss": 1.0208, "step": 22478 }, { "epoch": 1.36, "learning_rate": 4.554871516303774e-05, "loss": 0.9278, "step": 22480 }, { "epoch": 1.36, "learning_rate": 4.5547206252923516e-05, "loss": 1.2099, "step": 22482 }, { "epoch": 1.36, "learning_rate": 4.554569734280929e-05, "loss": 0.7284, "step": 22484 }, { "epoch": 1.36, "learning_rate": 4.554418843269507e-05, "loss": 0.7551, "step": 22486 }, { "epoch": 1.36, "learning_rate": 4.554267952258084e-05, "loss": 1.176, "step": 22488 }, { "epoch": 1.36, "learning_rate": 4.5541170612466615e-05, "loss": 1.0153, "step": 22490 }, { "epoch": 1.36, "learning_rate": 4.5539661702352395e-05, "loss": 0.9002, "step": 22492 }, { "epoch": 1.36, "learning_rate": 4.553815279223817e-05, "loss": 1.1912, "step": 22494 }, { "epoch": 1.36, "learning_rate": 4.553664388212394e-05, "loss": 0.9297, "step": 22496 }, { "epoch": 1.36, "learning_rate": 4.553513497200972e-05, "loss": 1.012, "step": 22498 }, { "epoch": 1.36, "learning_rate": 4.5533626061895494e-05, "loss": 0.7613, "step": 22500 }, { "epoch": 1.36, "learning_rate": 4.5532117151781274e-05, "loss": 0.6442, "step": 22502 }, { "epoch": 1.36, "learning_rate": 4.553060824166705e-05, "loss": 1.2775, "step": 22504 }, { "epoch": 1.36, "learning_rate": 4.552909933155282e-05, "loss": 1.0447, "step": 22506 }, { "epoch": 1.36, "learning_rate": 4.55275904214386e-05, "loss": 1.0008, "step": 22508 }, { "epoch": 1.36, "learning_rate": 4.552608151132437e-05, "loss": 0.9096, "step": 22510 }, { "epoch": 1.36, "learning_rate": 4.552457260121015e-05, "loss": 1.1695, "step": 22512 }, { "epoch": 1.36, "learning_rate": 4.5523063691095927e-05, "loss": 0.9554, "step": 22514 }, { "epoch": 1.36, "learning_rate": 4.55215547809817e-05, "loss": 0.9536, "step": 22516 }, { "epoch": 1.36, "learning_rate": 4.552004587086747e-05, "loss": 0.9604, "step": 22518 }, { "epoch": 1.36, "learning_rate": 4.551853696075325e-05, "loss": 0.7149, "step": 22520 }, { "epoch": 1.36, "learning_rate": 4.5517028050639026e-05, "loss": 0.857, "step": 22522 }, { "epoch": 1.36, "learning_rate": 4.5515519140524806e-05, "loss": 0.9907, "step": 22524 }, { "epoch": 1.36, "learning_rate": 4.551401023041057e-05, "loss": 0.8747, "step": 22526 }, { "epoch": 1.36, "learning_rate": 4.551250132029635e-05, "loss": 1.2486, "step": 22528 }, { "epoch": 1.36, "learning_rate": 4.551099241018213e-05, "loss": 1.0731, "step": 22530 }, { "epoch": 1.36, "learning_rate": 4.55094835000679e-05, "loss": 0.9021, "step": 22532 }, { "epoch": 1.36, "learning_rate": 4.550797458995368e-05, "loss": 1.1157, "step": 22534 }, { "epoch": 1.36, "learning_rate": 4.550646567983946e-05, "loss": 1.4418, "step": 22536 }, { "epoch": 1.36, "learning_rate": 4.550495676972523e-05, "loss": 0.9382, "step": 22538 }, { "epoch": 1.36, "learning_rate": 4.5503447859611004e-05, "loss": 1.34, "step": 22540 }, { "epoch": 1.36, "learning_rate": 4.550193894949678e-05, "loss": 0.8969, "step": 22542 }, { "epoch": 1.36, "learning_rate": 4.550043003938256e-05, "loss": 0.9031, "step": 22544 }, { "epoch": 1.36, "learning_rate": 4.549892112926833e-05, "loss": 1.0337, "step": 22546 }, { "epoch": 1.36, "learning_rate": 4.5497412219154104e-05, "loss": 0.9967, "step": 22548 }, { "epoch": 1.36, "learning_rate": 4.5495903309039884e-05, "loss": 1.0581, "step": 22550 }, { "epoch": 1.36, "learning_rate": 4.5494394398925663e-05, "loss": 1.1491, "step": 22552 }, { "epoch": 1.36, "learning_rate": 4.549288548881143e-05, "loss": 0.6504, "step": 22554 }, { "epoch": 1.36, "learning_rate": 4.549137657869721e-05, "loss": 1.1256, "step": 22556 }, { "epoch": 1.36, "learning_rate": 4.548986766858298e-05, "loss": 1.0249, "step": 22558 }, { "epoch": 1.36, "learning_rate": 4.5488358758468756e-05, "loss": 0.8719, "step": 22560 }, { "epoch": 1.36, "learning_rate": 4.5486849848354536e-05, "loss": 0.7447, "step": 22562 }, { "epoch": 1.36, "learning_rate": 4.548534093824031e-05, "loss": 1.0072, "step": 22564 }, { "epoch": 1.36, "learning_rate": 4.548383202812609e-05, "loss": 0.9658, "step": 22566 }, { "epoch": 1.36, "learning_rate": 4.548232311801186e-05, "loss": 1.1158, "step": 22568 }, { "epoch": 1.36, "learning_rate": 4.5480814207897635e-05, "loss": 1.0952, "step": 22570 }, { "epoch": 1.36, "learning_rate": 4.5479305297783415e-05, "loss": 1.0228, "step": 22572 }, { "epoch": 1.36, "learning_rate": 4.547779638766919e-05, "loss": 1.2461, "step": 22574 }, { "epoch": 1.36, "learning_rate": 4.547628747755496e-05, "loss": 0.7075, "step": 22576 }, { "epoch": 1.36, "learning_rate": 4.547477856744074e-05, "loss": 0.8747, "step": 22578 }, { "epoch": 1.36, "learning_rate": 4.5473269657326514e-05, "loss": 1.1908, "step": 22580 }, { "epoch": 1.36, "learning_rate": 4.547176074721229e-05, "loss": 1.0454, "step": 22582 }, { "epoch": 1.36, "learning_rate": 4.547025183709807e-05, "loss": 1.0321, "step": 22584 }, { "epoch": 1.36, "learning_rate": 4.546874292698384e-05, "loss": 1.1107, "step": 22586 }, { "epoch": 1.36, "learning_rate": 4.546723401686962e-05, "loss": 1.2416, "step": 22588 }, { "epoch": 1.36, "learning_rate": 4.5465725106755393e-05, "loss": 0.8562, "step": 22590 }, { "epoch": 1.36, "learning_rate": 4.5464216196641167e-05, "loss": 0.9532, "step": 22592 }, { "epoch": 1.36, "learning_rate": 4.5462707286526946e-05, "loss": 0.9917, "step": 22594 }, { "epoch": 1.36, "learning_rate": 4.546119837641272e-05, "loss": 0.9875, "step": 22596 }, { "epoch": 1.36, "learning_rate": 4.545968946629849e-05, "loss": 1.0063, "step": 22598 }, { "epoch": 1.36, "learning_rate": 4.545818055618427e-05, "loss": 1.0885, "step": 22600 }, { "epoch": 1.36, "learning_rate": 4.5456671646070046e-05, "loss": 1.0218, "step": 22602 }, { "epoch": 1.36, "learning_rate": 4.545516273595582e-05, "loss": 1.1467, "step": 22604 }, { "epoch": 1.36, "learning_rate": 4.54536538258416e-05, "loss": 1.1019, "step": 22606 }, { "epoch": 1.36, "learning_rate": 4.545214491572737e-05, "loss": 0.8775, "step": 22608 }, { "epoch": 1.36, "learning_rate": 4.5450636005613145e-05, "loss": 0.9409, "step": 22610 }, { "epoch": 1.36, "learning_rate": 4.5449127095498925e-05, "loss": 1.0832, "step": 22612 }, { "epoch": 1.36, "learning_rate": 4.54476181853847e-05, "loss": 0.9256, "step": 22614 }, { "epoch": 1.37, "learning_rate": 4.544610927527048e-05, "loss": 1.158, "step": 22616 }, { "epoch": 1.37, "learning_rate": 4.544460036515625e-05, "loss": 0.9662, "step": 22618 }, { "epoch": 1.37, "learning_rate": 4.5443091455042024e-05, "loss": 1.268, "step": 22620 }, { "epoch": 1.37, "learning_rate": 4.5441582544927804e-05, "loss": 0.9969, "step": 22622 }, { "epoch": 1.37, "learning_rate": 4.544007363481357e-05, "loss": 1.2808, "step": 22624 }, { "epoch": 1.37, "learning_rate": 4.543856472469935e-05, "loss": 1.1033, "step": 22626 }, { "epoch": 1.37, "learning_rate": 4.543705581458513e-05, "loss": 0.5445, "step": 22628 }, { "epoch": 1.37, "learning_rate": 4.54355469044709e-05, "loss": 1.0712, "step": 22630 }, { "epoch": 1.37, "learning_rate": 4.5434037994356677e-05, "loss": 0.9961, "step": 22632 }, { "epoch": 1.37, "learning_rate": 4.5432529084242456e-05, "loss": 1.0383, "step": 22634 }, { "epoch": 1.37, "learning_rate": 4.543102017412823e-05, "loss": 1.289, "step": 22636 }, { "epoch": 1.37, "learning_rate": 4.542951126401401e-05, "loss": 1.385, "step": 22638 }, { "epoch": 1.37, "learning_rate": 4.5428002353899776e-05, "loss": 0.9987, "step": 22640 }, { "epoch": 1.37, "learning_rate": 4.5426493443785556e-05, "loss": 0.9795, "step": 22642 }, { "epoch": 1.37, "learning_rate": 4.5424984533671336e-05, "loss": 1.2582, "step": 22644 }, { "epoch": 1.37, "learning_rate": 4.54234756235571e-05, "loss": 0.8241, "step": 22646 }, { "epoch": 1.37, "learning_rate": 4.542196671344288e-05, "loss": 0.9996, "step": 22648 }, { "epoch": 1.37, "learning_rate": 4.542045780332866e-05, "loss": 1.0336, "step": 22650 }, { "epoch": 1.37, "learning_rate": 4.5418948893214435e-05, "loss": 0.9101, "step": 22652 }, { "epoch": 1.37, "learning_rate": 4.541743998310021e-05, "loss": 1.1917, "step": 22654 }, { "epoch": 1.37, "learning_rate": 4.541593107298598e-05, "loss": 0.7892, "step": 22656 }, { "epoch": 1.37, "learning_rate": 4.541442216287176e-05, "loss": 0.9347, "step": 22658 }, { "epoch": 1.37, "learning_rate": 4.5412913252757534e-05, "loss": 0.9352, "step": 22660 }, { "epoch": 1.37, "learning_rate": 4.541140434264331e-05, "loss": 0.8211, "step": 22662 }, { "epoch": 1.37, "learning_rate": 4.540989543252909e-05, "loss": 1.2871, "step": 22664 }, { "epoch": 1.37, "learning_rate": 4.540838652241487e-05, "loss": 1.1646, "step": 22666 }, { "epoch": 1.37, "learning_rate": 4.5406877612300633e-05, "loss": 0.9756, "step": 22668 }, { "epoch": 1.37, "learning_rate": 4.540536870218641e-05, "loss": 0.8965, "step": 22670 }, { "epoch": 1.37, "learning_rate": 4.5403859792072186e-05, "loss": 1.0002, "step": 22672 }, { "epoch": 1.37, "learning_rate": 4.540235088195796e-05, "loss": 0.8113, "step": 22674 }, { "epoch": 1.37, "learning_rate": 4.540084197184374e-05, "loss": 0.8516, "step": 22676 }, { "epoch": 1.37, "learning_rate": 4.539933306172951e-05, "loss": 1.0758, "step": 22678 }, { "epoch": 1.37, "learning_rate": 4.539782415161529e-05, "loss": 1.1163, "step": 22680 }, { "epoch": 1.37, "learning_rate": 4.5396315241501066e-05, "loss": 0.9571, "step": 22682 }, { "epoch": 1.37, "learning_rate": 4.539480633138684e-05, "loss": 0.9863, "step": 22684 }, { "epoch": 1.37, "learning_rate": 4.539329742127262e-05, "loss": 1.1721, "step": 22686 }, { "epoch": 1.37, "learning_rate": 4.539178851115839e-05, "loss": 0.6884, "step": 22688 }, { "epoch": 1.37, "learning_rate": 4.5390279601044165e-05, "loss": 0.847, "step": 22690 }, { "epoch": 1.37, "learning_rate": 4.5388770690929945e-05, "loss": 0.7979, "step": 22692 }, { "epoch": 1.37, "learning_rate": 4.538726178081572e-05, "loss": 0.9216, "step": 22694 }, { "epoch": 1.37, "learning_rate": 4.538575287070149e-05, "loss": 0.8848, "step": 22696 }, { "epoch": 1.37, "learning_rate": 4.538424396058727e-05, "loss": 1.0107, "step": 22698 }, { "epoch": 1.37, "learning_rate": 4.5382735050473044e-05, "loss": 0.8876, "step": 22700 }, { "epoch": 1.37, "learning_rate": 4.5381226140358824e-05, "loss": 0.9277, "step": 22702 }, { "epoch": 1.37, "learning_rate": 4.53797172302446e-05, "loss": 0.9302, "step": 22704 }, { "epoch": 1.37, "learning_rate": 4.537820832013037e-05, "loss": 1.2076, "step": 22706 }, { "epoch": 1.37, "learning_rate": 4.537669941001615e-05, "loss": 1.0942, "step": 22708 }, { "epoch": 1.37, "learning_rate": 4.537519049990192e-05, "loss": 0.9408, "step": 22710 }, { "epoch": 1.37, "learning_rate": 4.5373681589787696e-05, "loss": 0.682, "step": 22712 }, { "epoch": 1.37, "learning_rate": 4.5372172679673476e-05, "loss": 0.9694, "step": 22714 }, { "epoch": 1.37, "learning_rate": 4.537066376955925e-05, "loss": 1.1594, "step": 22716 }, { "epoch": 1.37, "learning_rate": 4.536915485944502e-05, "loss": 0.763, "step": 22718 }, { "epoch": 1.37, "learning_rate": 4.53676459493308e-05, "loss": 0.8358, "step": 22720 }, { "epoch": 1.37, "learning_rate": 4.5366137039216576e-05, "loss": 0.9607, "step": 22722 }, { "epoch": 1.37, "learning_rate": 4.536462812910235e-05, "loss": 1.559, "step": 22724 }, { "epoch": 1.37, "learning_rate": 4.536311921898813e-05, "loss": 0.7415, "step": 22726 }, { "epoch": 1.37, "learning_rate": 4.53616103088739e-05, "loss": 1.2281, "step": 22728 }, { "epoch": 1.37, "learning_rate": 4.536010139875968e-05, "loss": 0.8161, "step": 22730 }, { "epoch": 1.37, "learning_rate": 4.5358592488645455e-05, "loss": 0.7525, "step": 22732 }, { "epoch": 1.37, "learning_rate": 4.535708357853123e-05, "loss": 0.8603, "step": 22734 }, { "epoch": 1.37, "learning_rate": 4.535557466841701e-05, "loss": 0.7882, "step": 22736 }, { "epoch": 1.37, "learning_rate": 4.535406575830278e-05, "loss": 0.7573, "step": 22738 }, { "epoch": 1.37, "learning_rate": 4.5352556848188554e-05, "loss": 0.8654, "step": 22740 }, { "epoch": 1.37, "learning_rate": 4.5351047938074334e-05, "loss": 0.9024, "step": 22742 }, { "epoch": 1.37, "learning_rate": 4.534953902796011e-05, "loss": 1.1141, "step": 22744 }, { "epoch": 1.37, "learning_rate": 4.534803011784588e-05, "loss": 1.0872, "step": 22746 }, { "epoch": 1.37, "learning_rate": 4.534652120773166e-05, "loss": 0.8417, "step": 22748 }, { "epoch": 1.37, "learning_rate": 4.534501229761743e-05, "loss": 1.0674, "step": 22750 }, { "epoch": 1.37, "learning_rate": 4.534350338750321e-05, "loss": 1.166, "step": 22752 }, { "epoch": 1.37, "learning_rate": 4.534199447738898e-05, "loss": 1.0153, "step": 22754 }, { "epoch": 1.37, "learning_rate": 4.534048556727476e-05, "loss": 0.857, "step": 22756 }, { "epoch": 1.37, "learning_rate": 4.533897665716054e-05, "loss": 1.195, "step": 22758 }, { "epoch": 1.37, "learning_rate": 4.5337467747046306e-05, "loss": 1.1425, "step": 22760 }, { "epoch": 1.37, "learning_rate": 4.5335958836932086e-05, "loss": 0.869, "step": 22762 }, { "epoch": 1.37, "learning_rate": 4.5334449926817865e-05, "loss": 0.8901, "step": 22764 }, { "epoch": 1.37, "learning_rate": 4.533294101670364e-05, "loss": 1.3424, "step": 22766 }, { "epoch": 1.37, "learning_rate": 4.533143210658941e-05, "loss": 0.8451, "step": 22768 }, { "epoch": 1.37, "learning_rate": 4.5329923196475185e-05, "loss": 1.0049, "step": 22770 }, { "epoch": 1.37, "learning_rate": 4.5328414286360965e-05, "loss": 1.0253, "step": 22772 }, { "epoch": 1.37, "learning_rate": 4.532690537624674e-05, "loss": 1.0475, "step": 22774 }, { "epoch": 1.37, "learning_rate": 4.532539646613251e-05, "loss": 0.7913, "step": 22776 }, { "epoch": 1.37, "learning_rate": 4.532388755601829e-05, "loss": 0.751, "step": 22778 }, { "epoch": 1.37, "learning_rate": 4.532237864590407e-05, "loss": 1.1865, "step": 22780 }, { "epoch": 1.38, "learning_rate": 4.532086973578984e-05, "loss": 1.0645, "step": 22782 }, { "epoch": 1.38, "learning_rate": 4.531936082567562e-05, "loss": 0.7979, "step": 22784 }, { "epoch": 1.38, "learning_rate": 4.531785191556139e-05, "loss": 1.1131, "step": 22786 }, { "epoch": 1.38, "learning_rate": 4.531634300544716e-05, "loss": 0.7866, "step": 22788 }, { "epoch": 1.38, "learning_rate": 4.531483409533294e-05, "loss": 0.7763, "step": 22790 }, { "epoch": 1.38, "learning_rate": 4.5313325185218716e-05, "loss": 0.7299, "step": 22792 }, { "epoch": 1.38, "learning_rate": 4.5311816275104496e-05, "loss": 0.9773, "step": 22794 }, { "epoch": 1.38, "learning_rate": 4.531030736499027e-05, "loss": 0.9904, "step": 22796 }, { "epoch": 1.38, "learning_rate": 4.530879845487604e-05, "loss": 0.8644, "step": 22798 }, { "epoch": 1.38, "learning_rate": 4.530728954476182e-05, "loss": 0.8998, "step": 22800 }, { "epoch": 1.38, "learning_rate": 4.5305780634647595e-05, "loss": 0.8826, "step": 22802 }, { "epoch": 1.38, "learning_rate": 4.530427172453337e-05, "loss": 1.1369, "step": 22804 }, { "epoch": 1.38, "learning_rate": 4.530276281441915e-05, "loss": 1.0451, "step": 22806 }, { "epoch": 1.38, "learning_rate": 4.530125390430492e-05, "loss": 0.9604, "step": 22808 }, { "epoch": 1.38, "learning_rate": 4.5299744994190695e-05, "loss": 0.8586, "step": 22810 }, { "epoch": 1.38, "learning_rate": 4.5298236084076475e-05, "loss": 1.3547, "step": 22812 }, { "epoch": 1.38, "learning_rate": 4.529672717396225e-05, "loss": 1.0244, "step": 22814 }, { "epoch": 1.38, "learning_rate": 4.529521826384803e-05, "loss": 1.0401, "step": 22816 }, { "epoch": 1.38, "learning_rate": 4.52937093537338e-05, "loss": 0.844, "step": 22818 }, { "epoch": 1.38, "learning_rate": 4.5292200443619574e-05, "loss": 1.3265, "step": 22820 }, { "epoch": 1.38, "learning_rate": 4.5290691533505354e-05, "loss": 1.4177, "step": 22822 }, { "epoch": 1.38, "learning_rate": 4.528918262339113e-05, "loss": 0.9093, "step": 22824 }, { "epoch": 1.38, "learning_rate": 4.52876737132769e-05, "loss": 1.0615, "step": 22826 }, { "epoch": 1.38, "learning_rate": 4.528616480316268e-05, "loss": 0.9182, "step": 22828 }, { "epoch": 1.38, "learning_rate": 4.528465589304845e-05, "loss": 1.3183, "step": 22830 }, { "epoch": 1.38, "learning_rate": 4.5283146982934226e-05, "loss": 1.2371, "step": 22832 }, { "epoch": 1.38, "learning_rate": 4.5281638072820006e-05, "loss": 0.7858, "step": 22834 }, { "epoch": 1.38, "learning_rate": 4.528012916270578e-05, "loss": 0.9045, "step": 22836 }, { "epoch": 1.38, "learning_rate": 4.527862025259155e-05, "loss": 0.7498, "step": 22838 }, { "epoch": 1.38, "learning_rate": 4.527711134247733e-05, "loss": 1.0124, "step": 22840 }, { "epoch": 1.38, "learning_rate": 4.5275602432363105e-05, "loss": 0.9203, "step": 22842 }, { "epoch": 1.38, "learning_rate": 4.5274093522248885e-05, "loss": 1.2019, "step": 22844 }, { "epoch": 1.38, "learning_rate": 4.527258461213466e-05, "loss": 1.1621, "step": 22846 }, { "epoch": 1.38, "learning_rate": 4.527107570202043e-05, "loss": 1.2042, "step": 22848 }, { "epoch": 1.38, "learning_rate": 4.526956679190621e-05, "loss": 1.0872, "step": 22850 }, { "epoch": 1.38, "learning_rate": 4.5268057881791985e-05, "loss": 0.9328, "step": 22852 }, { "epoch": 1.38, "learning_rate": 4.526654897167776e-05, "loss": 0.9156, "step": 22854 }, { "epoch": 1.38, "learning_rate": 4.526504006156354e-05, "loss": 0.9209, "step": 22856 }, { "epoch": 1.38, "learning_rate": 4.526353115144931e-05, "loss": 0.8103, "step": 22858 }, { "epoch": 1.38, "learning_rate": 4.5262022241335084e-05, "loss": 1.0789, "step": 22860 }, { "epoch": 1.38, "learning_rate": 4.5260513331220864e-05, "loss": 0.9092, "step": 22862 }, { "epoch": 1.38, "learning_rate": 4.525900442110664e-05, "loss": 0.8389, "step": 22864 }, { "epoch": 1.38, "learning_rate": 4.525749551099242e-05, "loss": 1.0265, "step": 22866 }, { "epoch": 1.38, "learning_rate": 4.525598660087818e-05, "loss": 1.3006, "step": 22868 }, { "epoch": 1.38, "learning_rate": 4.525447769076396e-05, "loss": 1.021, "step": 22870 }, { "epoch": 1.38, "learning_rate": 4.525296878064974e-05, "loss": 0.9931, "step": 22872 }, { "epoch": 1.38, "learning_rate": 4.525145987053551e-05, "loss": 0.7557, "step": 22874 }, { "epoch": 1.38, "learning_rate": 4.524995096042129e-05, "loss": 0.9786, "step": 22876 }, { "epoch": 1.38, "learning_rate": 4.524844205030707e-05, "loss": 1.1294, "step": 22878 }, { "epoch": 1.38, "learning_rate": 4.524693314019284e-05, "loss": 1.0899, "step": 22880 }, { "epoch": 1.38, "learning_rate": 4.5245424230078615e-05, "loss": 0.9618, "step": 22882 }, { "epoch": 1.38, "learning_rate": 4.524391531996439e-05, "loss": 0.7994, "step": 22884 }, { "epoch": 1.38, "learning_rate": 4.524240640985017e-05, "loss": 0.7307, "step": 22886 }, { "epoch": 1.38, "learning_rate": 4.524089749973594e-05, "loss": 1.168, "step": 22888 }, { "epoch": 1.38, "learning_rate": 4.5239388589621715e-05, "loss": 0.8654, "step": 22890 }, { "epoch": 1.38, "learning_rate": 4.5237879679507495e-05, "loss": 1.1902, "step": 22892 }, { "epoch": 1.38, "learning_rate": 4.5236370769393274e-05, "loss": 1.5737, "step": 22894 }, { "epoch": 1.38, "learning_rate": 4.523486185927904e-05, "loss": 0.8924, "step": 22896 }, { "epoch": 1.38, "learning_rate": 4.523335294916482e-05, "loss": 1.0551, "step": 22898 }, { "epoch": 1.38, "learning_rate": 4.5231844039050594e-05, "loss": 0.7646, "step": 22900 }, { "epoch": 1.38, "learning_rate": 4.5230335128936374e-05, "loss": 0.886, "step": 22902 }, { "epoch": 1.38, "learning_rate": 4.522882621882215e-05, "loss": 1.0166, "step": 22904 }, { "epoch": 1.38, "learning_rate": 4.522731730870792e-05, "loss": 0.8336, "step": 22906 }, { "epoch": 1.38, "learning_rate": 4.52258083985937e-05, "loss": 0.757, "step": 22908 }, { "epoch": 1.38, "learning_rate": 4.522429948847947e-05, "loss": 0.8112, "step": 22910 }, { "epoch": 1.38, "learning_rate": 4.5222790578365246e-05, "loss": 1.1105, "step": 22912 }, { "epoch": 1.38, "learning_rate": 4.5221281668251026e-05, "loss": 0.7683, "step": 22914 }, { "epoch": 1.38, "learning_rate": 4.52197727581368e-05, "loss": 1.0485, "step": 22916 }, { "epoch": 1.38, "learning_rate": 4.521826384802257e-05, "loss": 1.0078, "step": 22918 }, { "epoch": 1.38, "learning_rate": 4.521675493790835e-05, "loss": 0.7079, "step": 22920 }, { "epoch": 1.38, "learning_rate": 4.5215246027794125e-05, "loss": 1.172, "step": 22922 }, { "epoch": 1.38, "learning_rate": 4.52137371176799e-05, "loss": 1.1312, "step": 22924 }, { "epoch": 1.38, "learning_rate": 4.521222820756568e-05, "loss": 1.0723, "step": 22926 }, { "epoch": 1.38, "learning_rate": 4.521071929745145e-05, "loss": 0.7425, "step": 22928 }, { "epoch": 1.38, "learning_rate": 4.520921038733723e-05, "loss": 1.1384, "step": 22930 }, { "epoch": 1.38, "learning_rate": 4.5207701477223004e-05, "loss": 0.8998, "step": 22932 }, { "epoch": 1.38, "learning_rate": 4.520619256710878e-05, "loss": 0.7337, "step": 22934 }, { "epoch": 1.38, "learning_rate": 4.520468365699456e-05, "loss": 0.8019, "step": 22936 }, { "epoch": 1.38, "learning_rate": 4.520317474688033e-05, "loss": 0.9878, "step": 22938 }, { "epoch": 1.38, "learning_rate": 4.5201665836766104e-05, "loss": 1.0645, "step": 22940 }, { "epoch": 1.38, "learning_rate": 4.5200156926651884e-05, "loss": 0.8105, "step": 22942 }, { "epoch": 1.38, "learning_rate": 4.519864801653766e-05, "loss": 0.9311, "step": 22944 }, { "epoch": 1.38, "learning_rate": 4.519713910642343e-05, "loss": 0.8487, "step": 22946 }, { "epoch": 1.39, "learning_rate": 4.519563019630921e-05, "loss": 1.105, "step": 22948 }, { "epoch": 1.39, "learning_rate": 4.519412128619498e-05, "loss": 0.8957, "step": 22950 }, { "epoch": 1.39, "learning_rate": 4.5192612376080756e-05, "loss": 0.8274, "step": 22952 }, { "epoch": 1.39, "learning_rate": 4.5191103465966536e-05, "loss": 1.1506, "step": 22954 }, { "epoch": 1.39, "learning_rate": 4.518959455585231e-05, "loss": 0.995, "step": 22956 }, { "epoch": 1.39, "learning_rate": 4.518808564573809e-05, "loss": 1.2934, "step": 22958 }, { "epoch": 1.39, "learning_rate": 4.518657673562386e-05, "loss": 0.7279, "step": 22960 }, { "epoch": 1.39, "learning_rate": 4.5185067825509635e-05, "loss": 0.9014, "step": 22962 }, { "epoch": 1.39, "learning_rate": 4.5183558915395415e-05, "loss": 0.8966, "step": 22964 }, { "epoch": 1.39, "learning_rate": 4.518205000528119e-05, "loss": 0.9559, "step": 22966 }, { "epoch": 1.39, "learning_rate": 4.518054109516696e-05, "loss": 0.8274, "step": 22968 }, { "epoch": 1.39, "learning_rate": 4.517903218505274e-05, "loss": 0.8855, "step": 22970 }, { "epoch": 1.39, "learning_rate": 4.5177523274938514e-05, "loss": 1.4423, "step": 22972 }, { "epoch": 1.39, "learning_rate": 4.517601436482429e-05, "loss": 0.7824, "step": 22974 }, { "epoch": 1.39, "learning_rate": 4.517450545471007e-05, "loss": 1.0625, "step": 22976 }, { "epoch": 1.39, "learning_rate": 4.517299654459584e-05, "loss": 1.0695, "step": 22978 }, { "epoch": 1.39, "learning_rate": 4.517148763448162e-05, "loss": 0.9032, "step": 22980 }, { "epoch": 1.39, "learning_rate": 4.516997872436739e-05, "loss": 0.9847, "step": 22982 }, { "epoch": 1.39, "learning_rate": 4.516846981425317e-05, "loss": 0.9977, "step": 22984 }, { "epoch": 1.39, "learning_rate": 4.516696090413895e-05, "loss": 1.1457, "step": 22986 }, { "epoch": 1.39, "learning_rate": 4.516545199402471e-05, "loss": 0.8275, "step": 22988 }, { "epoch": 1.39, "learning_rate": 4.516394308391049e-05, "loss": 0.7119, "step": 22990 }, { "epoch": 1.39, "learning_rate": 4.516243417379627e-05, "loss": 0.8847, "step": 22992 }, { "epoch": 1.39, "learning_rate": 4.5160925263682046e-05, "loss": 0.9586, "step": 22994 }, { "epoch": 1.39, "learning_rate": 4.515941635356782e-05, "loss": 0.9969, "step": 22996 }, { "epoch": 1.39, "learning_rate": 4.515790744345359e-05, "loss": 0.8724, "step": 22998 }, { "epoch": 1.39, "learning_rate": 4.515639853333937e-05, "loss": 0.7412, "step": 23000 }, { "epoch": 1.39, "learning_rate": 4.5154889623225145e-05, "loss": 0.9558, "step": 23002 }, { "epoch": 1.39, "learning_rate": 4.515338071311092e-05, "loss": 0.7653, "step": 23004 }, { "epoch": 1.39, "learning_rate": 4.51518718029967e-05, "loss": 0.5667, "step": 23006 }, { "epoch": 1.39, "learning_rate": 4.515036289288248e-05, "loss": 1.0439, "step": 23008 }, { "epoch": 1.39, "learning_rate": 4.5148853982768244e-05, "loss": 0.8332, "step": 23010 }, { "epoch": 1.39, "learning_rate": 4.5147345072654024e-05, "loss": 0.8845, "step": 23012 }, { "epoch": 1.39, "learning_rate": 4.51458361625398e-05, "loss": 0.7897, "step": 23014 }, { "epoch": 1.39, "learning_rate": 4.514432725242558e-05, "loss": 0.9046, "step": 23016 }, { "epoch": 1.39, "learning_rate": 4.514281834231135e-05, "loss": 0.8828, "step": 23018 }, { "epoch": 1.39, "learning_rate": 4.5141309432197124e-05, "loss": 0.9001, "step": 23020 }, { "epoch": 1.39, "learning_rate": 4.5139800522082904e-05, "loss": 0.9361, "step": 23022 }, { "epoch": 1.39, "learning_rate": 4.513829161196868e-05, "loss": 0.967, "step": 23024 }, { "epoch": 1.39, "learning_rate": 4.513678270185445e-05, "loss": 0.9965, "step": 23026 }, { "epoch": 1.39, "learning_rate": 4.513527379174023e-05, "loss": 1.0889, "step": 23028 }, { "epoch": 1.39, "learning_rate": 4.5133764881626e-05, "loss": 0.8556, "step": 23030 }, { "epoch": 1.39, "learning_rate": 4.5132255971511776e-05, "loss": 0.8957, "step": 23032 }, { "epoch": 1.39, "learning_rate": 4.5130747061397556e-05, "loss": 1.1309, "step": 23034 }, { "epoch": 1.39, "learning_rate": 4.512923815128333e-05, "loss": 1.3045, "step": 23036 }, { "epoch": 1.39, "learning_rate": 4.51277292411691e-05, "loss": 0.7007, "step": 23038 }, { "epoch": 1.39, "learning_rate": 4.512622033105488e-05, "loss": 1.1326, "step": 23040 }, { "epoch": 1.39, "learning_rate": 4.5124711420940655e-05, "loss": 1.0781, "step": 23042 }, { "epoch": 1.39, "learning_rate": 4.5123202510826435e-05, "loss": 0.6319, "step": 23044 }, { "epoch": 1.39, "learning_rate": 4.512169360071221e-05, "loss": 0.878, "step": 23046 }, { "epoch": 1.39, "learning_rate": 4.512018469059798e-05, "loss": 0.9403, "step": 23048 }, { "epoch": 1.39, "learning_rate": 4.511867578048376e-05, "loss": 0.8543, "step": 23050 }, { "epoch": 1.39, "learning_rate": 4.5117166870369534e-05, "loss": 0.6947, "step": 23052 }, { "epoch": 1.39, "learning_rate": 4.511565796025531e-05, "loss": 1.2819, "step": 23054 }, { "epoch": 1.39, "learning_rate": 4.511414905014109e-05, "loss": 1.0303, "step": 23056 }, { "epoch": 1.39, "learning_rate": 4.511264014002686e-05, "loss": 1.0656, "step": 23058 }, { "epoch": 1.39, "learning_rate": 4.5111131229912634e-05, "loss": 0.8159, "step": 23060 }, { "epoch": 1.39, "learning_rate": 4.5109622319798413e-05, "loss": 1.0245, "step": 23062 }, { "epoch": 1.39, "learning_rate": 4.510811340968419e-05, "loss": 0.9874, "step": 23064 }, { "epoch": 1.39, "learning_rate": 4.5106604499569967e-05, "loss": 0.9736, "step": 23066 }, { "epoch": 1.39, "learning_rate": 4.510509558945574e-05, "loss": 0.8564, "step": 23068 }, { "epoch": 1.39, "learning_rate": 4.510358667934151e-05, "loss": 0.9038, "step": 23070 }, { "epoch": 1.39, "learning_rate": 4.510207776922729e-05, "loss": 1.0407, "step": 23072 }, { "epoch": 1.39, "learning_rate": 4.5100568859113066e-05, "loss": 1.1534, "step": 23074 }, { "epoch": 1.39, "learning_rate": 4.509905994899884e-05, "loss": 1.1159, "step": 23076 }, { "epoch": 1.39, "learning_rate": 4.509755103888462e-05, "loss": 0.9067, "step": 23078 }, { "epoch": 1.39, "learning_rate": 4.509604212877039e-05, "loss": 1.0171, "step": 23080 }, { "epoch": 1.39, "learning_rate": 4.5094533218656165e-05, "loss": 0.7376, "step": 23082 }, { "epoch": 1.39, "learning_rate": 4.5093024308541945e-05, "loss": 0.7542, "step": 23084 }, { "epoch": 1.39, "learning_rate": 4.509151539842772e-05, "loss": 0.8548, "step": 23086 }, { "epoch": 1.39, "learning_rate": 4.509000648831349e-05, "loss": 1.0857, "step": 23088 }, { "epoch": 1.39, "learning_rate": 4.508849757819927e-05, "loss": 1.2018, "step": 23090 }, { "epoch": 1.39, "learning_rate": 4.5086988668085044e-05, "loss": 0.9263, "step": 23092 }, { "epoch": 1.39, "learning_rate": 4.5085479757970824e-05, "loss": 1.106, "step": 23094 }, { "epoch": 1.39, "learning_rate": 4.508397084785659e-05, "loss": 1.1261, "step": 23096 }, { "epoch": 1.39, "learning_rate": 4.508246193774237e-05, "loss": 0.9502, "step": 23098 }, { "epoch": 1.39, "learning_rate": 4.508095302762815e-05, "loss": 0.8006, "step": 23100 }, { "epoch": 1.39, "learning_rate": 4.507944411751392e-05, "loss": 0.7492, "step": 23102 }, { "epoch": 1.39, "learning_rate": 4.5077935207399697e-05, "loss": 0.9647, "step": 23104 }, { "epoch": 1.39, "learning_rate": 4.5076426297285476e-05, "loss": 0.8598, "step": 23106 }, { "epoch": 1.39, "learning_rate": 4.507491738717125e-05, "loss": 0.909, "step": 23108 }, { "epoch": 1.39, "learning_rate": 4.507340847705702e-05, "loss": 0.9231, "step": 23110 }, { "epoch": 1.39, "learning_rate": 4.5071899566942796e-05, "loss": 0.7816, "step": 23112 }, { "epoch": 1.4, "learning_rate": 4.5070390656828576e-05, "loss": 1.0238, "step": 23114 }, { "epoch": 1.4, "learning_rate": 4.5068881746714356e-05, "loss": 0.9667, "step": 23116 }, { "epoch": 1.4, "learning_rate": 4.506737283660012e-05, "loss": 0.8552, "step": 23118 }, { "epoch": 1.4, "learning_rate": 4.50658639264859e-05, "loss": 1.0336, "step": 23120 }, { "epoch": 1.4, "learning_rate": 4.506435501637168e-05, "loss": 1.1959, "step": 23122 }, { "epoch": 1.4, "learning_rate": 4.506284610625745e-05, "loss": 0.9362, "step": 23124 }, { "epoch": 1.4, "learning_rate": 4.506133719614323e-05, "loss": 0.937, "step": 23126 }, { "epoch": 1.4, "learning_rate": 4.5059828286029e-05, "loss": 1.1932, "step": 23128 }, { "epoch": 1.4, "learning_rate": 4.505831937591478e-05, "loss": 0.7652, "step": 23130 }, { "epoch": 1.4, "learning_rate": 4.5056810465800554e-05, "loss": 1.182, "step": 23132 }, { "epoch": 1.4, "learning_rate": 4.505530155568633e-05, "loss": 0.9873, "step": 23134 }, { "epoch": 1.4, "learning_rate": 4.505379264557211e-05, "loss": 0.9966, "step": 23136 }, { "epoch": 1.4, "learning_rate": 4.505228373545788e-05, "loss": 0.7555, "step": 23138 }, { "epoch": 1.4, "learning_rate": 4.5050774825343653e-05, "loss": 0.9746, "step": 23140 }, { "epoch": 1.4, "learning_rate": 4.504926591522943e-05, "loss": 0.9677, "step": 23142 }, { "epoch": 1.4, "learning_rate": 4.5047757005115207e-05, "loss": 0.7068, "step": 23144 }, { "epoch": 1.4, "learning_rate": 4.504624809500098e-05, "loss": 1.1002, "step": 23146 }, { "epoch": 1.4, "learning_rate": 4.504473918488676e-05, "loss": 1.2118, "step": 23148 }, { "epoch": 1.4, "learning_rate": 4.504323027477253e-05, "loss": 1.0891, "step": 23150 }, { "epoch": 1.4, "learning_rate": 4.5041721364658306e-05, "loss": 1.0218, "step": 23152 }, { "epoch": 1.4, "learning_rate": 4.5040212454544086e-05, "loss": 0.7882, "step": 23154 }, { "epoch": 1.4, "learning_rate": 4.503870354442986e-05, "loss": 0.855, "step": 23156 }, { "epoch": 1.4, "learning_rate": 4.503719463431564e-05, "loss": 1.0251, "step": 23158 }, { "epoch": 1.4, "learning_rate": 4.503568572420141e-05, "loss": 1.0449, "step": 23160 }, { "epoch": 1.4, "learning_rate": 4.5034176814087185e-05, "loss": 0.8798, "step": 23162 }, { "epoch": 1.4, "learning_rate": 4.5032667903972965e-05, "loss": 0.8827, "step": 23164 }, { "epoch": 1.4, "learning_rate": 4.503115899385874e-05, "loss": 0.6273, "step": 23166 }, { "epoch": 1.4, "learning_rate": 4.502965008374451e-05, "loss": 1.2691, "step": 23168 }, { "epoch": 1.4, "learning_rate": 4.502814117363029e-05, "loss": 0.9356, "step": 23170 }, { "epoch": 1.4, "learning_rate": 4.5026632263516064e-05, "loss": 0.72, "step": 23172 }, { "epoch": 1.4, "learning_rate": 4.502512335340184e-05, "loss": 0.9907, "step": 23174 }, { "epoch": 1.4, "learning_rate": 4.502361444328762e-05, "loss": 0.8711, "step": 23176 }, { "epoch": 1.4, "learning_rate": 4.502210553317339e-05, "loss": 0.9146, "step": 23178 }, { "epoch": 1.4, "learning_rate": 4.502059662305917e-05, "loss": 1.0332, "step": 23180 }, { "epoch": 1.4, "learning_rate": 4.501908771294494e-05, "loss": 0.8056, "step": 23182 }, { "epoch": 1.4, "learning_rate": 4.5017578802830716e-05, "loss": 0.7239, "step": 23184 }, { "epoch": 1.4, "learning_rate": 4.5016069892716496e-05, "loss": 0.8115, "step": 23186 }, { "epoch": 1.4, "learning_rate": 4.501456098260227e-05, "loss": 1.6032, "step": 23188 }, { "epoch": 1.4, "learning_rate": 4.501305207248804e-05, "loss": 0.55, "step": 23190 }, { "epoch": 1.4, "learning_rate": 4.501154316237382e-05, "loss": 1.1211, "step": 23192 }, { "epoch": 1.4, "learning_rate": 4.5010034252259596e-05, "loss": 0.789, "step": 23194 }, { "epoch": 1.4, "learning_rate": 4.500852534214537e-05, "loss": 0.8046, "step": 23196 }, { "epoch": 1.4, "learning_rate": 4.500701643203115e-05, "loss": 1.2679, "step": 23198 }, { "epoch": 1.4, "learning_rate": 4.500550752191692e-05, "loss": 0.765, "step": 23200 }, { "epoch": 1.4, "learning_rate": 4.5003998611802695e-05, "loss": 0.831, "step": 23202 }, { "epoch": 1.4, "learning_rate": 4.5002489701688475e-05, "loss": 0.7332, "step": 23204 }, { "epoch": 1.4, "learning_rate": 4.500098079157425e-05, "loss": 1.1543, "step": 23206 }, { "epoch": 1.4, "learning_rate": 4.499947188146003e-05, "loss": 0.7997, "step": 23208 }, { "epoch": 1.4, "learning_rate": 4.4997962971345794e-05, "loss": 1.0976, "step": 23210 }, { "epoch": 1.4, "learning_rate": 4.4996454061231574e-05, "loss": 0.9601, "step": 23212 }, { "epoch": 1.4, "learning_rate": 4.4994945151117354e-05, "loss": 1.1206, "step": 23214 }, { "epoch": 1.4, "learning_rate": 4.499343624100312e-05, "loss": 1.3481, "step": 23216 }, { "epoch": 1.4, "learning_rate": 4.49919273308889e-05, "loss": 0.9186, "step": 23218 }, { "epoch": 1.4, "learning_rate": 4.499041842077468e-05, "loss": 0.763, "step": 23220 }, { "epoch": 1.4, "learning_rate": 4.498890951066045e-05, "loss": 0.7689, "step": 23222 }, { "epoch": 1.4, "learning_rate": 4.4987400600546226e-05, "loss": 0.9013, "step": 23224 }, { "epoch": 1.4, "learning_rate": 4.4985891690432e-05, "loss": 0.9872, "step": 23226 }, { "epoch": 1.4, "learning_rate": 4.498438278031778e-05, "loss": 0.8059, "step": 23228 }, { "epoch": 1.4, "learning_rate": 4.498287387020356e-05, "loss": 0.9972, "step": 23230 }, { "epoch": 1.4, "learning_rate": 4.4981364960089326e-05, "loss": 1.1121, "step": 23232 }, { "epoch": 1.4, "learning_rate": 4.4979856049975106e-05, "loss": 1.3704, "step": 23234 }, { "epoch": 1.4, "learning_rate": 4.4978347139860885e-05, "loss": 0.8268, "step": 23236 }, { "epoch": 1.4, "learning_rate": 4.497683822974665e-05, "loss": 0.6675, "step": 23238 }, { "epoch": 1.4, "learning_rate": 4.497532931963243e-05, "loss": 1.4035, "step": 23240 }, { "epoch": 1.4, "learning_rate": 4.4973820409518205e-05, "loss": 1.1552, "step": 23242 }, { "epoch": 1.4, "learning_rate": 4.4972311499403985e-05, "loss": 0.8269, "step": 23244 }, { "epoch": 1.4, "learning_rate": 4.497080258928976e-05, "loss": 0.9678, "step": 23246 }, { "epoch": 1.4, "learning_rate": 4.496929367917553e-05, "loss": 0.8889, "step": 23248 }, { "epoch": 1.4, "learning_rate": 4.496778476906131e-05, "loss": 0.9124, "step": 23250 }, { "epoch": 1.4, "learning_rate": 4.4966275858947084e-05, "loss": 0.9916, "step": 23252 }, { "epoch": 1.4, "learning_rate": 4.496476694883286e-05, "loss": 0.8366, "step": 23254 }, { "epoch": 1.4, "learning_rate": 4.496325803871864e-05, "loss": 1.2897, "step": 23256 }, { "epoch": 1.4, "learning_rate": 4.496174912860441e-05, "loss": 0.7249, "step": 23258 }, { "epoch": 1.4, "learning_rate": 4.496024021849018e-05, "loss": 1.1811, "step": 23260 }, { "epoch": 1.4, "learning_rate": 4.495873130837596e-05, "loss": 0.9397, "step": 23262 }, { "epoch": 1.4, "learning_rate": 4.4957222398261736e-05, "loss": 1.0171, "step": 23264 }, { "epoch": 1.4, "learning_rate": 4.495571348814751e-05, "loss": 1.4015, "step": 23266 }, { "epoch": 1.4, "learning_rate": 4.495420457803329e-05, "loss": 0.8078, "step": 23268 }, { "epoch": 1.4, "learning_rate": 4.495269566791906e-05, "loss": 1.2083, "step": 23270 }, { "epoch": 1.4, "learning_rate": 4.495118675780484e-05, "loss": 0.5859, "step": 23272 }, { "epoch": 1.4, "learning_rate": 4.4949677847690616e-05, "loss": 0.9273, "step": 23274 }, { "epoch": 1.4, "learning_rate": 4.494816893757639e-05, "loss": 1.0855, "step": 23276 }, { "epoch": 1.4, "learning_rate": 4.494666002746217e-05, "loss": 0.7944, "step": 23278 }, { "epoch": 1.41, "learning_rate": 4.494515111734794e-05, "loss": 0.6025, "step": 23280 }, { "epoch": 1.41, "learning_rate": 4.4943642207233715e-05, "loss": 1.0615, "step": 23282 }, { "epoch": 1.41, "learning_rate": 4.4942133297119495e-05, "loss": 0.8022, "step": 23284 }, { "epoch": 1.41, "learning_rate": 4.494062438700527e-05, "loss": 0.7414, "step": 23286 }, { "epoch": 1.41, "learning_rate": 4.493911547689104e-05, "loss": 1.154, "step": 23288 }, { "epoch": 1.41, "learning_rate": 4.493760656677682e-05, "loss": 0.915, "step": 23290 }, { "epoch": 1.41, "learning_rate": 4.4936097656662594e-05, "loss": 0.9034, "step": 23292 }, { "epoch": 1.41, "learning_rate": 4.4934588746548374e-05, "loss": 0.8355, "step": 23294 }, { "epoch": 1.41, "learning_rate": 4.493307983643415e-05, "loss": 0.5667, "step": 23296 }, { "epoch": 1.41, "learning_rate": 4.493157092631992e-05, "loss": 1.0711, "step": 23298 }, { "epoch": 1.41, "learning_rate": 4.49300620162057e-05, "loss": 1.0657, "step": 23300 }, { "epoch": 1.41, "learning_rate": 4.492855310609147e-05, "loss": 1.0879, "step": 23302 }, { "epoch": 1.41, "learning_rate": 4.4927044195977246e-05, "loss": 1.1842, "step": 23304 }, { "epoch": 1.41, "learning_rate": 4.4925535285863026e-05, "loss": 0.9861, "step": 23306 }, { "epoch": 1.41, "learning_rate": 4.49240263757488e-05, "loss": 0.8624, "step": 23308 }, { "epoch": 1.41, "learning_rate": 4.492251746563457e-05, "loss": 0.9179, "step": 23310 }, { "epoch": 1.41, "learning_rate": 4.492100855552035e-05, "loss": 0.8247, "step": 23312 }, { "epoch": 1.41, "learning_rate": 4.4919499645406125e-05, "loss": 1.1669, "step": 23314 }, { "epoch": 1.41, "learning_rate": 4.49179907352919e-05, "loss": 0.8457, "step": 23316 }, { "epoch": 1.41, "learning_rate": 4.491648182517768e-05, "loss": 1.0922, "step": 23318 }, { "epoch": 1.41, "learning_rate": 4.491497291506345e-05, "loss": 1.0612, "step": 23320 }, { "epoch": 1.41, "learning_rate": 4.491346400494923e-05, "loss": 0.9889, "step": 23322 }, { "epoch": 1.41, "learning_rate": 4.4911955094835e-05, "loss": 0.758, "step": 23324 }, { "epoch": 1.41, "learning_rate": 4.491044618472078e-05, "loss": 0.8071, "step": 23326 }, { "epoch": 1.41, "learning_rate": 4.490893727460656e-05, "loss": 0.8673, "step": 23328 }, { "epoch": 1.41, "learning_rate": 4.4907428364492324e-05, "loss": 1.3369, "step": 23330 }, { "epoch": 1.41, "learning_rate": 4.4905919454378104e-05, "loss": 1.1594, "step": 23332 }, { "epoch": 1.41, "learning_rate": 4.4904410544263884e-05, "loss": 0.9082, "step": 23334 }, { "epoch": 1.41, "learning_rate": 4.490290163414966e-05, "loss": 1.1087, "step": 23336 }, { "epoch": 1.41, "learning_rate": 4.490139272403543e-05, "loss": 0.8128, "step": 23338 }, { "epoch": 1.41, "learning_rate": 4.48998838139212e-05, "loss": 0.8302, "step": 23340 }, { "epoch": 1.41, "learning_rate": 4.489837490380698e-05, "loss": 1.0657, "step": 23342 }, { "epoch": 1.41, "learning_rate": 4.489686599369276e-05, "loss": 0.8551, "step": 23344 }, { "epoch": 1.41, "learning_rate": 4.489535708357853e-05, "loss": 1.3209, "step": 23346 }, { "epoch": 1.41, "learning_rate": 4.489384817346431e-05, "loss": 0.8559, "step": 23348 }, { "epoch": 1.41, "learning_rate": 4.489233926335009e-05, "loss": 0.8662, "step": 23350 }, { "epoch": 1.41, "learning_rate": 4.4890830353235855e-05, "loss": 0.946, "step": 23352 }, { "epoch": 1.41, "learning_rate": 4.4889321443121635e-05, "loss": 1.2627, "step": 23354 }, { "epoch": 1.41, "learning_rate": 4.488781253300741e-05, "loss": 0.7571, "step": 23356 }, { "epoch": 1.41, "learning_rate": 4.488630362289319e-05, "loss": 0.594, "step": 23358 }, { "epoch": 1.41, "learning_rate": 4.488479471277896e-05, "loss": 0.7828, "step": 23360 }, { "epoch": 1.41, "learning_rate": 4.4883285802664735e-05, "loss": 0.758, "step": 23362 }, { "epoch": 1.41, "learning_rate": 4.4881776892550515e-05, "loss": 0.7718, "step": 23364 }, { "epoch": 1.41, "learning_rate": 4.488026798243629e-05, "loss": 0.7638, "step": 23366 }, { "epoch": 1.41, "learning_rate": 4.487875907232206e-05, "loss": 0.8248, "step": 23368 }, { "epoch": 1.41, "learning_rate": 4.487725016220784e-05, "loss": 0.7177, "step": 23370 }, { "epoch": 1.41, "learning_rate": 4.4875741252093614e-05, "loss": 0.901, "step": 23372 }, { "epoch": 1.41, "learning_rate": 4.487423234197939e-05, "loss": 0.8906, "step": 23374 }, { "epoch": 1.41, "learning_rate": 4.487272343186517e-05, "loss": 0.8101, "step": 23376 }, { "epoch": 1.41, "learning_rate": 4.487121452175094e-05, "loss": 0.7972, "step": 23378 }, { "epoch": 1.41, "learning_rate": 4.486970561163671e-05, "loss": 0.9512, "step": 23380 }, { "epoch": 1.41, "learning_rate": 4.486819670152249e-05, "loss": 0.7241, "step": 23382 }, { "epoch": 1.41, "learning_rate": 4.4866687791408266e-05, "loss": 1.2575, "step": 23384 }, { "epoch": 1.41, "learning_rate": 4.4865178881294046e-05, "loss": 1.0853, "step": 23386 }, { "epoch": 1.41, "learning_rate": 4.486366997117982e-05, "loss": 0.8175, "step": 23388 }, { "epoch": 1.41, "learning_rate": 4.486216106106559e-05, "loss": 0.853, "step": 23390 }, { "epoch": 1.41, "learning_rate": 4.486065215095137e-05, "loss": 0.9435, "step": 23392 }, { "epoch": 1.41, "learning_rate": 4.4859143240837145e-05, "loss": 0.9879, "step": 23394 }, { "epoch": 1.41, "learning_rate": 4.485763433072292e-05, "loss": 0.7931, "step": 23396 }, { "epoch": 1.41, "learning_rate": 4.48561254206087e-05, "loss": 1.0348, "step": 23398 }, { "epoch": 1.41, "learning_rate": 4.485461651049447e-05, "loss": 0.8126, "step": 23400 }, { "epoch": 1.41, "learning_rate": 4.4853107600380245e-05, "loss": 1.1849, "step": 23402 }, { "epoch": 1.41, "learning_rate": 4.4851598690266025e-05, "loss": 1.1459, "step": 23404 }, { "epoch": 1.41, "learning_rate": 4.48500897801518e-05, "loss": 1.0415, "step": 23406 }, { "epoch": 1.41, "learning_rate": 4.484858087003758e-05, "loss": 0.7008, "step": 23408 }, { "epoch": 1.41, "learning_rate": 4.484707195992335e-05, "loss": 0.7922, "step": 23410 }, { "epoch": 1.41, "learning_rate": 4.4845563049809124e-05, "loss": 1.0842, "step": 23412 }, { "epoch": 1.41, "learning_rate": 4.4844054139694904e-05, "loss": 0.8454, "step": 23414 }, { "epoch": 1.41, "learning_rate": 4.484254522958068e-05, "loss": 0.9783, "step": 23416 }, { "epoch": 1.41, "learning_rate": 4.484103631946645e-05, "loss": 1.0004, "step": 23418 }, { "epoch": 1.41, "learning_rate": 4.483952740935223e-05, "loss": 1.2934, "step": 23420 }, { "epoch": 1.41, "learning_rate": 4.4838018499238e-05, "loss": 1.189, "step": 23422 }, { "epoch": 1.41, "learning_rate": 4.4836509589123776e-05, "loss": 0.797, "step": 23424 }, { "epoch": 1.41, "learning_rate": 4.4835000679009556e-05, "loss": 1.0361, "step": 23426 }, { "epoch": 1.41, "learning_rate": 4.483349176889533e-05, "loss": 1.07, "step": 23428 }, { "epoch": 1.41, "learning_rate": 4.48319828587811e-05, "loss": 1.0878, "step": 23430 }, { "epoch": 1.41, "learning_rate": 4.483047394866688e-05, "loss": 1.085, "step": 23432 }, { "epoch": 1.41, "learning_rate": 4.4828965038552655e-05, "loss": 1.2262, "step": 23434 }, { "epoch": 1.41, "learning_rate": 4.4827456128438435e-05, "loss": 0.8785, "step": 23436 }, { "epoch": 1.41, "learning_rate": 4.48259472183242e-05, "loss": 0.8262, "step": 23438 }, { "epoch": 1.41, "learning_rate": 4.482443830820998e-05, "loss": 0.8912, "step": 23440 }, { "epoch": 1.41, "learning_rate": 4.482292939809576e-05, "loss": 1.2307, "step": 23442 }, { "epoch": 1.42, "learning_rate": 4.4821420487981534e-05, "loss": 0.9342, "step": 23444 }, { "epoch": 1.42, "learning_rate": 4.481991157786731e-05, "loss": 0.7135, "step": 23446 }, { "epoch": 1.42, "learning_rate": 4.481840266775309e-05, "loss": 0.6984, "step": 23448 }, { "epoch": 1.42, "learning_rate": 4.481689375763886e-05, "loss": 1.0708, "step": 23450 }, { "epoch": 1.42, "learning_rate": 4.4815384847524634e-05, "loss": 0.859, "step": 23452 }, { "epoch": 1.42, "learning_rate": 4.481387593741041e-05, "loss": 0.9396, "step": 23454 }, { "epoch": 1.42, "learning_rate": 4.481236702729619e-05, "loss": 0.878, "step": 23456 }, { "epoch": 1.42, "learning_rate": 4.481085811718197e-05, "loss": 0.6474, "step": 23458 }, { "epoch": 1.42, "learning_rate": 4.480934920706773e-05, "loss": 0.6639, "step": 23460 }, { "epoch": 1.42, "learning_rate": 4.480784029695351e-05, "loss": 0.9817, "step": 23462 }, { "epoch": 1.42, "learning_rate": 4.480633138683929e-05, "loss": 0.8675, "step": 23464 }, { "epoch": 1.42, "learning_rate": 4.480482247672506e-05, "loss": 0.7958, "step": 23466 }, { "epoch": 1.42, "learning_rate": 4.480331356661084e-05, "loss": 1.0373, "step": 23468 }, { "epoch": 1.42, "learning_rate": 4.480180465649661e-05, "loss": 1.0847, "step": 23470 }, { "epoch": 1.42, "learning_rate": 4.480029574638239e-05, "loss": 0.9426, "step": 23472 }, { "epoch": 1.42, "learning_rate": 4.4798786836268165e-05, "loss": 0.955, "step": 23474 }, { "epoch": 1.42, "learning_rate": 4.479727792615394e-05, "loss": 0.7208, "step": 23476 }, { "epoch": 1.42, "learning_rate": 4.479576901603972e-05, "loss": 1.2346, "step": 23478 }, { "epoch": 1.42, "learning_rate": 4.479426010592549e-05, "loss": 1.0908, "step": 23480 }, { "epoch": 1.42, "learning_rate": 4.4792751195811265e-05, "loss": 0.9686, "step": 23482 }, { "epoch": 1.42, "learning_rate": 4.4791242285697044e-05, "loss": 0.9699, "step": 23484 }, { "epoch": 1.42, "learning_rate": 4.478973337558282e-05, "loss": 0.9769, "step": 23486 }, { "epoch": 1.42, "learning_rate": 4.478822446546859e-05, "loss": 0.5443, "step": 23488 }, { "epoch": 1.42, "learning_rate": 4.478671555535437e-05, "loss": 0.8088, "step": 23490 }, { "epoch": 1.42, "learning_rate": 4.4785206645240144e-05, "loss": 0.9626, "step": 23492 }, { "epoch": 1.42, "learning_rate": 4.478369773512592e-05, "loss": 0.7595, "step": 23494 }, { "epoch": 1.42, "learning_rate": 4.47821888250117e-05, "loss": 0.904, "step": 23496 }, { "epoch": 1.42, "learning_rate": 4.478067991489747e-05, "loss": 0.6246, "step": 23498 }, { "epoch": 1.42, "learning_rate": 4.477917100478325e-05, "loss": 0.9826, "step": 23500 }, { "epoch": 1.42, "learning_rate": 4.477766209466902e-05, "loss": 0.8169, "step": 23502 }, { "epoch": 1.42, "learning_rate": 4.4776153184554796e-05, "loss": 1.0224, "step": 23504 }, { "epoch": 1.42, "learning_rate": 4.4774644274440576e-05, "loss": 0.9367, "step": 23506 }, { "epoch": 1.42, "learning_rate": 4.477313536432635e-05, "loss": 0.8757, "step": 23508 }, { "epoch": 1.42, "learning_rate": 4.477162645421212e-05, "loss": 0.9752, "step": 23510 }, { "epoch": 1.42, "learning_rate": 4.47701175440979e-05, "loss": 0.9158, "step": 23512 }, { "epoch": 1.42, "learning_rate": 4.4768608633983675e-05, "loss": 0.8479, "step": 23514 }, { "epoch": 1.42, "learning_rate": 4.476709972386945e-05, "loss": 0.7519, "step": 23516 }, { "epoch": 1.42, "learning_rate": 4.476559081375523e-05, "loss": 0.7262, "step": 23518 }, { "epoch": 1.42, "learning_rate": 4.4764081903641e-05, "loss": 1.0548, "step": 23520 }, { "epoch": 1.42, "learning_rate": 4.476257299352678e-05, "loss": 0.7304, "step": 23522 }, { "epoch": 1.42, "learning_rate": 4.4761064083412554e-05, "loss": 1.061, "step": 23524 }, { "epoch": 1.42, "learning_rate": 4.475955517329833e-05, "loss": 1.0753, "step": 23526 }, { "epoch": 1.42, "learning_rate": 4.475804626318411e-05, "loss": 1.0201, "step": 23528 }, { "epoch": 1.42, "learning_rate": 4.475653735306988e-05, "loss": 0.8175, "step": 23530 }, { "epoch": 1.42, "learning_rate": 4.4755028442955654e-05, "loss": 0.964, "step": 23532 }, { "epoch": 1.42, "learning_rate": 4.4753519532841434e-05, "loss": 0.7889, "step": 23534 }, { "epoch": 1.42, "learning_rate": 4.475201062272721e-05, "loss": 0.6726, "step": 23536 }, { "epoch": 1.42, "learning_rate": 4.475050171261298e-05, "loss": 0.8019, "step": 23538 }, { "epoch": 1.42, "learning_rate": 4.474899280249876e-05, "loss": 1.0374, "step": 23540 }, { "epoch": 1.42, "learning_rate": 4.474748389238453e-05, "loss": 0.8923, "step": 23542 }, { "epoch": 1.42, "learning_rate": 4.4745974982270306e-05, "loss": 1.18, "step": 23544 }, { "epoch": 1.42, "learning_rate": 4.4744466072156086e-05, "loss": 0.9751, "step": 23546 }, { "epoch": 1.42, "learning_rate": 4.474295716204186e-05, "loss": 1.1219, "step": 23548 }, { "epoch": 1.42, "learning_rate": 4.474144825192764e-05, "loss": 1.15, "step": 23550 }, { "epoch": 1.42, "learning_rate": 4.4739939341813405e-05, "loss": 1.0975, "step": 23552 }, { "epoch": 1.42, "learning_rate": 4.4738430431699185e-05, "loss": 1.0384, "step": 23554 }, { "epoch": 1.42, "learning_rate": 4.4736921521584965e-05, "loss": 0.721, "step": 23556 }, { "epoch": 1.42, "learning_rate": 4.473541261147074e-05, "loss": 1.2023, "step": 23558 }, { "epoch": 1.42, "learning_rate": 4.473390370135651e-05, "loss": 0.8622, "step": 23560 }, { "epoch": 1.42, "learning_rate": 4.473239479124229e-05, "loss": 0.8243, "step": 23562 }, { "epoch": 1.42, "learning_rate": 4.4730885881128064e-05, "loss": 0.7326, "step": 23564 }, { "epoch": 1.42, "learning_rate": 4.472937697101384e-05, "loss": 1.1018, "step": 23566 }, { "epoch": 1.42, "learning_rate": 4.472786806089961e-05, "loss": 0.7687, "step": 23568 }, { "epoch": 1.42, "learning_rate": 4.472635915078539e-05, "loss": 0.8184, "step": 23570 }, { "epoch": 1.42, "learning_rate": 4.472485024067117e-05, "loss": 0.7358, "step": 23572 }, { "epoch": 1.42, "learning_rate": 4.472334133055694e-05, "loss": 0.9224, "step": 23574 }, { "epoch": 1.42, "learning_rate": 4.4721832420442717e-05, "loss": 0.9846, "step": 23576 }, { "epoch": 1.42, "learning_rate": 4.4720323510328497e-05, "loss": 1.2617, "step": 23578 }, { "epoch": 1.42, "learning_rate": 4.471881460021426e-05, "loss": 0.9207, "step": 23580 }, { "epoch": 1.42, "learning_rate": 4.471730569010004e-05, "loss": 0.9394, "step": 23582 }, { "epoch": 1.42, "learning_rate": 4.4715796779985816e-05, "loss": 0.6715, "step": 23584 }, { "epoch": 1.42, "learning_rate": 4.4714287869871596e-05, "loss": 1.3045, "step": 23586 }, { "epoch": 1.42, "learning_rate": 4.471277895975737e-05, "loss": 1.0995, "step": 23588 }, { "epoch": 1.42, "learning_rate": 4.471127004964314e-05, "loss": 0.8681, "step": 23590 }, { "epoch": 1.42, "learning_rate": 4.470976113952892e-05, "loss": 0.8294, "step": 23592 }, { "epoch": 1.42, "learning_rate": 4.4708252229414695e-05, "loss": 0.5291, "step": 23594 }, { "epoch": 1.42, "learning_rate": 4.470674331930047e-05, "loss": 0.9899, "step": 23596 }, { "epoch": 1.42, "learning_rate": 4.470523440918625e-05, "loss": 0.9624, "step": 23598 }, { "epoch": 1.42, "learning_rate": 4.470372549907202e-05, "loss": 1.1619, "step": 23600 }, { "epoch": 1.42, "learning_rate": 4.4702216588957794e-05, "loss": 0.9817, "step": 23602 }, { "epoch": 1.42, "learning_rate": 4.4700707678843574e-05, "loss": 0.8758, "step": 23604 }, { "epoch": 1.42, "learning_rate": 4.469919876872935e-05, "loss": 0.8051, "step": 23606 }, { "epoch": 1.42, "learning_rate": 4.469768985861513e-05, "loss": 1.0707, "step": 23608 }, { "epoch": 1.43, "learning_rate": 4.46961809485009e-05, "loss": 1.1357, "step": 23610 }, { "epoch": 1.43, "learning_rate": 4.4694672038386674e-05, "loss": 0.8854, "step": 23612 }, { "epoch": 1.43, "learning_rate": 4.4693163128272453e-05, "loss": 0.784, "step": 23614 }, { "epoch": 1.43, "learning_rate": 4.4691654218158227e-05, "loss": 0.7615, "step": 23616 }, { "epoch": 1.43, "learning_rate": 4.4690145308044e-05, "loss": 0.691, "step": 23618 }, { "epoch": 1.43, "learning_rate": 4.468863639792978e-05, "loss": 1.128, "step": 23620 }, { "epoch": 1.43, "learning_rate": 4.468712748781555e-05, "loss": 1.0382, "step": 23622 }, { "epoch": 1.43, "learning_rate": 4.4685618577701326e-05, "loss": 0.9101, "step": 23624 }, { "epoch": 1.43, "learning_rate": 4.4684109667587106e-05, "loss": 0.771, "step": 23626 }, { "epoch": 1.43, "learning_rate": 4.468260075747288e-05, "loss": 0.9034, "step": 23628 }, { "epoch": 1.43, "learning_rate": 4.468109184735865e-05, "loss": 0.9827, "step": 23630 }, { "epoch": 1.43, "learning_rate": 4.467958293724443e-05, "loss": 1.0892, "step": 23632 }, { "epoch": 1.43, "learning_rate": 4.4678074027130205e-05, "loss": 0.8, "step": 23634 }, { "epoch": 1.43, "learning_rate": 4.4676565117015985e-05, "loss": 1.0041, "step": 23636 }, { "epoch": 1.43, "learning_rate": 4.467505620690176e-05, "loss": 0.913, "step": 23638 }, { "epoch": 1.43, "learning_rate": 4.467354729678753e-05, "loss": 0.9192, "step": 23640 }, { "epoch": 1.43, "learning_rate": 4.467203838667331e-05, "loss": 0.9835, "step": 23642 }, { "epoch": 1.43, "learning_rate": 4.4670529476559084e-05, "loss": 0.9602, "step": 23644 }, { "epoch": 1.43, "learning_rate": 4.466902056644486e-05, "loss": 1.0491, "step": 23646 }, { "epoch": 1.43, "learning_rate": 4.466751165633064e-05, "loss": 1.1561, "step": 23648 }, { "epoch": 1.43, "learning_rate": 4.466600274621641e-05, "loss": 0.9745, "step": 23650 }, { "epoch": 1.43, "learning_rate": 4.4664493836102183e-05, "loss": 1.0887, "step": 23652 }, { "epoch": 1.43, "learning_rate": 4.466298492598796e-05, "loss": 1.1348, "step": 23654 }, { "epoch": 1.43, "learning_rate": 4.4661476015873736e-05, "loss": 0.8718, "step": 23656 }, { "epoch": 1.43, "learning_rate": 4.4659967105759516e-05, "loss": 0.7739, "step": 23658 }, { "epoch": 1.43, "learning_rate": 4.465845819564529e-05, "loss": 1.1235, "step": 23660 }, { "epoch": 1.43, "learning_rate": 4.465694928553106e-05, "loss": 0.958, "step": 23662 }, { "epoch": 1.43, "learning_rate": 4.465544037541684e-05, "loss": 0.9943, "step": 23664 }, { "epoch": 1.43, "learning_rate": 4.465393146530261e-05, "loss": 0.7777, "step": 23666 }, { "epoch": 1.43, "learning_rate": 4.465242255518839e-05, "loss": 0.9659, "step": 23668 }, { "epoch": 1.43, "learning_rate": 4.465091364507417e-05, "loss": 0.9772, "step": 23670 }, { "epoch": 1.43, "learning_rate": 4.464940473495994e-05, "loss": 0.9148, "step": 23672 }, { "epoch": 1.43, "learning_rate": 4.4647895824845715e-05, "loss": 0.8221, "step": 23674 }, { "epoch": 1.43, "learning_rate": 4.4646386914731495e-05, "loss": 0.7134, "step": 23676 }, { "epoch": 1.43, "learning_rate": 4.464487800461727e-05, "loss": 0.8507, "step": 23678 }, { "epoch": 1.43, "learning_rate": 4.464336909450304e-05, "loss": 0.9543, "step": 23680 }, { "epoch": 1.43, "learning_rate": 4.4641860184388814e-05, "loss": 0.8724, "step": 23682 }, { "epoch": 1.43, "learning_rate": 4.4640351274274594e-05, "loss": 0.9668, "step": 23684 }, { "epoch": 1.43, "learning_rate": 4.4638842364160374e-05, "loss": 1.1723, "step": 23686 }, { "epoch": 1.43, "learning_rate": 4.463733345404614e-05, "loss": 0.9735, "step": 23688 }, { "epoch": 1.43, "learning_rate": 4.463582454393192e-05, "loss": 1.0133, "step": 23690 }, { "epoch": 1.43, "learning_rate": 4.46343156338177e-05, "loss": 0.7149, "step": 23692 }, { "epoch": 1.43, "learning_rate": 4.4632806723703467e-05, "loss": 0.8979, "step": 23694 }, { "epoch": 1.43, "learning_rate": 4.4631297813589246e-05, "loss": 1.247, "step": 23696 }, { "epoch": 1.43, "learning_rate": 4.462978890347502e-05, "loss": 1.0021, "step": 23698 }, { "epoch": 1.43, "learning_rate": 4.46282799933608e-05, "loss": 1.1041, "step": 23700 }, { "epoch": 1.43, "learning_rate": 4.462677108324657e-05, "loss": 0.6248, "step": 23702 }, { "epoch": 1.43, "learning_rate": 4.4625262173132346e-05, "loss": 0.8975, "step": 23704 }, { "epoch": 1.43, "learning_rate": 4.4623753263018126e-05, "loss": 1.1424, "step": 23706 }, { "epoch": 1.43, "learning_rate": 4.46222443529039e-05, "loss": 1.2033, "step": 23708 }, { "epoch": 1.43, "learning_rate": 4.462073544278967e-05, "loss": 1.0562, "step": 23710 }, { "epoch": 1.43, "learning_rate": 4.461922653267545e-05, "loss": 0.7558, "step": 23712 }, { "epoch": 1.43, "learning_rate": 4.4617717622561225e-05, "loss": 1.1437, "step": 23714 }, { "epoch": 1.43, "learning_rate": 4.4616208712447e-05, "loss": 0.79, "step": 23716 }, { "epoch": 1.43, "learning_rate": 4.461469980233278e-05, "loss": 0.9103, "step": 23718 }, { "epoch": 1.43, "learning_rate": 4.461319089221855e-05, "loss": 1.1802, "step": 23720 }, { "epoch": 1.43, "learning_rate": 4.461168198210433e-05, "loss": 1.2744, "step": 23722 }, { "epoch": 1.43, "learning_rate": 4.4610173071990104e-05, "loss": 0.9197, "step": 23724 }, { "epoch": 1.43, "learning_rate": 4.460866416187588e-05, "loss": 0.9506, "step": 23726 }, { "epoch": 1.43, "learning_rate": 4.460715525176166e-05, "loss": 1.0178, "step": 23728 }, { "epoch": 1.43, "learning_rate": 4.460564634164743e-05, "loss": 1.1236, "step": 23730 }, { "epoch": 1.43, "learning_rate": 4.46041374315332e-05, "loss": 0.7753, "step": 23732 }, { "epoch": 1.43, "learning_rate": 4.460262852141898e-05, "loss": 1.4047, "step": 23734 }, { "epoch": 1.43, "learning_rate": 4.4601119611304756e-05, "loss": 0.947, "step": 23736 }, { "epoch": 1.43, "learning_rate": 4.459961070119053e-05, "loss": 1.0943, "step": 23738 }, { "epoch": 1.43, "learning_rate": 4.459810179107631e-05, "loss": 1.3761, "step": 23740 }, { "epoch": 1.43, "learning_rate": 4.459659288096208e-05, "loss": 0.8395, "step": 23742 }, { "epoch": 1.43, "learning_rate": 4.4595083970847856e-05, "loss": 0.7815, "step": 23744 }, { "epoch": 1.43, "learning_rate": 4.4593575060733636e-05, "loss": 0.9687, "step": 23746 }, { "epoch": 1.43, "learning_rate": 4.459206615061941e-05, "loss": 0.9362, "step": 23748 }, { "epoch": 1.43, "learning_rate": 4.459055724050519e-05, "loss": 1.2616, "step": 23750 }, { "epoch": 1.43, "learning_rate": 4.458904833039096e-05, "loss": 0.9064, "step": 23752 }, { "epoch": 1.43, "learning_rate": 4.4587539420276735e-05, "loss": 0.859, "step": 23754 }, { "epoch": 1.43, "learning_rate": 4.4586030510162515e-05, "loss": 0.7796, "step": 23756 }, { "epoch": 1.43, "learning_rate": 4.458452160004829e-05, "loss": 0.8271, "step": 23758 }, { "epoch": 1.43, "learning_rate": 4.458301268993406e-05, "loss": 0.8263, "step": 23760 }, { "epoch": 1.43, "learning_rate": 4.458150377981984e-05, "loss": 1.0198, "step": 23762 }, { "epoch": 1.43, "learning_rate": 4.4579994869705614e-05, "loss": 0.8474, "step": 23764 }, { "epoch": 1.43, "learning_rate": 4.457848595959139e-05, "loss": 0.6974, "step": 23766 }, { "epoch": 1.43, "learning_rate": 4.457697704947717e-05, "loss": 1.1593, "step": 23768 }, { "epoch": 1.43, "learning_rate": 4.457546813936294e-05, "loss": 0.864, "step": 23770 }, { "epoch": 1.43, "learning_rate": 4.457395922924872e-05, "loss": 0.9461, "step": 23772 }, { "epoch": 1.43, "learning_rate": 4.457245031913449e-05, "loss": 1.0371, "step": 23774 }, { "epoch": 1.44, "learning_rate": 4.4570941409020266e-05, "loss": 0.7668, "step": 23776 }, { "epoch": 1.44, "learning_rate": 4.4569432498906046e-05, "loss": 0.7147, "step": 23778 }, { "epoch": 1.44, "learning_rate": 4.456792358879181e-05, "loss": 1.0114, "step": 23780 }, { "epoch": 1.44, "learning_rate": 4.456641467867759e-05, "loss": 0.799, "step": 23782 }, { "epoch": 1.44, "learning_rate": 4.456490576856337e-05, "loss": 0.9985, "step": 23784 }, { "epoch": 1.44, "learning_rate": 4.4563396858449146e-05, "loss": 0.8114, "step": 23786 }, { "epoch": 1.44, "learning_rate": 4.456188794833492e-05, "loss": 0.8371, "step": 23788 }, { "epoch": 1.44, "learning_rate": 4.45603790382207e-05, "loss": 1.174, "step": 23790 }, { "epoch": 1.44, "learning_rate": 4.455887012810647e-05, "loss": 0.9002, "step": 23792 }, { "epoch": 1.44, "learning_rate": 4.4557361217992245e-05, "loss": 1.0573, "step": 23794 }, { "epoch": 1.44, "learning_rate": 4.455585230787802e-05, "loss": 0.8124, "step": 23796 }, { "epoch": 1.44, "learning_rate": 4.45543433977638e-05, "loss": 0.7857, "step": 23798 }, { "epoch": 1.44, "learning_rate": 4.455283448764958e-05, "loss": 1.2235, "step": 23800 }, { "epoch": 1.44, "learning_rate": 4.4551325577535344e-05, "loss": 0.7302, "step": 23802 }, { "epoch": 1.44, "learning_rate": 4.4549816667421124e-05, "loss": 1.1492, "step": 23804 }, { "epoch": 1.44, "learning_rate": 4.4548307757306904e-05, "loss": 0.7063, "step": 23806 }, { "epoch": 1.44, "learning_rate": 4.454679884719267e-05, "loss": 0.7804, "step": 23808 }, { "epoch": 1.44, "learning_rate": 4.454528993707845e-05, "loss": 0.6611, "step": 23810 }, { "epoch": 1.44, "learning_rate": 4.454378102696422e-05, "loss": 0.8128, "step": 23812 }, { "epoch": 1.44, "learning_rate": 4.454227211685e-05, "loss": 0.9626, "step": 23814 }, { "epoch": 1.44, "learning_rate": 4.4540763206735776e-05, "loss": 1.0188, "step": 23816 }, { "epoch": 1.44, "learning_rate": 4.453925429662155e-05, "loss": 1.2741, "step": 23818 }, { "epoch": 1.44, "learning_rate": 4.453774538650733e-05, "loss": 0.9908, "step": 23820 }, { "epoch": 1.44, "learning_rate": 4.453623647639311e-05, "loss": 0.9024, "step": 23822 }, { "epoch": 1.44, "learning_rate": 4.4534727566278876e-05, "loss": 1.046, "step": 23824 }, { "epoch": 1.44, "learning_rate": 4.4533218656164655e-05, "loss": 0.7548, "step": 23826 }, { "epoch": 1.44, "learning_rate": 4.453170974605043e-05, "loss": 0.9892, "step": 23828 }, { "epoch": 1.44, "learning_rate": 4.45302008359362e-05, "loss": 0.9461, "step": 23830 }, { "epoch": 1.44, "learning_rate": 4.452869192582198e-05, "loss": 0.7173, "step": 23832 }, { "epoch": 1.44, "learning_rate": 4.4527183015707755e-05, "loss": 1.073, "step": 23834 }, { "epoch": 1.44, "learning_rate": 4.4525674105593535e-05, "loss": 0.6902, "step": 23836 }, { "epoch": 1.44, "learning_rate": 4.452416519547931e-05, "loss": 1.0923, "step": 23838 }, { "epoch": 1.44, "learning_rate": 4.452265628536508e-05, "loss": 1.0279, "step": 23840 }, { "epoch": 1.44, "learning_rate": 4.452114737525086e-05, "loss": 1.1076, "step": 23842 }, { "epoch": 1.44, "learning_rate": 4.4519638465136634e-05, "loss": 1.0273, "step": 23844 }, { "epoch": 1.44, "learning_rate": 4.451812955502241e-05, "loss": 1.0009, "step": 23846 }, { "epoch": 1.44, "learning_rate": 4.451662064490819e-05, "loss": 1.0662, "step": 23848 }, { "epoch": 1.44, "learning_rate": 4.451511173479396e-05, "loss": 0.8954, "step": 23850 }, { "epoch": 1.44, "learning_rate": 4.451360282467973e-05, "loss": 0.7512, "step": 23852 }, { "epoch": 1.44, "learning_rate": 4.451209391456551e-05, "loss": 1.1778, "step": 23854 }, { "epoch": 1.44, "learning_rate": 4.4510585004451286e-05, "loss": 1.2008, "step": 23856 }, { "epoch": 1.44, "learning_rate": 4.450907609433706e-05, "loss": 0.9742, "step": 23858 }, { "epoch": 1.44, "learning_rate": 4.450756718422284e-05, "loss": 1.1248, "step": 23860 }, { "epoch": 1.44, "learning_rate": 4.450605827410861e-05, "loss": 1.1221, "step": 23862 }, { "epoch": 1.44, "learning_rate": 4.450454936399439e-05, "loss": 0.8501, "step": 23864 }, { "epoch": 1.44, "learning_rate": 4.4503040453880165e-05, "loss": 1.0665, "step": 23866 }, { "epoch": 1.44, "learning_rate": 4.450153154376594e-05, "loss": 0.8895, "step": 23868 }, { "epoch": 1.44, "learning_rate": 4.450002263365172e-05, "loss": 0.9394, "step": 23870 }, { "epoch": 1.44, "learning_rate": 4.449851372353749e-05, "loss": 0.8299, "step": 23872 }, { "epoch": 1.44, "learning_rate": 4.4497004813423265e-05, "loss": 0.7305, "step": 23874 }, { "epoch": 1.44, "learning_rate": 4.4495495903309045e-05, "loss": 1.1878, "step": 23876 }, { "epoch": 1.44, "learning_rate": 4.449398699319482e-05, "loss": 0.7619, "step": 23878 }, { "epoch": 1.44, "learning_rate": 4.449247808308059e-05, "loss": 1.1325, "step": 23880 }, { "epoch": 1.44, "learning_rate": 4.449096917296637e-05, "loss": 1.1189, "step": 23882 }, { "epoch": 1.44, "learning_rate": 4.4489460262852144e-05, "loss": 0.7589, "step": 23884 }, { "epoch": 1.44, "learning_rate": 4.4487951352737924e-05, "loss": 0.838, "step": 23886 }, { "epoch": 1.44, "learning_rate": 4.44864424426237e-05, "loss": 1.1317, "step": 23888 }, { "epoch": 1.44, "learning_rate": 4.448493353250947e-05, "loss": 1.1123, "step": 23890 }, { "epoch": 1.44, "learning_rate": 4.448342462239525e-05, "loss": 1.1716, "step": 23892 }, { "epoch": 1.44, "learning_rate": 4.4481915712281016e-05, "loss": 1.4122, "step": 23894 }, { "epoch": 1.44, "learning_rate": 4.4480406802166796e-05, "loss": 1.226, "step": 23896 }, { "epoch": 1.44, "learning_rate": 4.4478897892052576e-05, "loss": 0.9802, "step": 23898 }, { "epoch": 1.44, "learning_rate": 4.447738898193835e-05, "loss": 0.8232, "step": 23900 }, { "epoch": 1.44, "learning_rate": 4.447588007182412e-05, "loss": 0.8355, "step": 23902 }, { "epoch": 1.44, "learning_rate": 4.44743711617099e-05, "loss": 0.858, "step": 23904 }, { "epoch": 1.44, "learning_rate": 4.4472862251595675e-05, "loss": 0.8514, "step": 23906 }, { "epoch": 1.44, "learning_rate": 4.447135334148145e-05, "loss": 1.0749, "step": 23908 }, { "epoch": 1.44, "learning_rate": 4.446984443136722e-05, "loss": 0.8661, "step": 23910 }, { "epoch": 1.44, "learning_rate": 4.4468335521253e-05, "loss": 0.697, "step": 23912 }, { "epoch": 1.44, "learning_rate": 4.446682661113878e-05, "loss": 1.0315, "step": 23914 }, { "epoch": 1.44, "learning_rate": 4.446531770102455e-05, "loss": 0.855, "step": 23916 }, { "epoch": 1.44, "learning_rate": 4.446380879091033e-05, "loss": 1.2545, "step": 23918 }, { "epoch": 1.44, "learning_rate": 4.446229988079611e-05, "loss": 0.6901, "step": 23920 }, { "epoch": 1.44, "learning_rate": 4.4460790970681874e-05, "loss": 1.0044, "step": 23922 }, { "epoch": 1.44, "learning_rate": 4.4459282060567654e-05, "loss": 0.9986, "step": 23924 }, { "epoch": 1.44, "learning_rate": 4.445777315045343e-05, "loss": 0.8614, "step": 23926 }, { "epoch": 1.44, "learning_rate": 4.445626424033921e-05, "loss": 0.6872, "step": 23928 }, { "epoch": 1.44, "learning_rate": 4.445475533022498e-05, "loss": 1.289, "step": 23930 }, { "epoch": 1.44, "learning_rate": 4.445324642011075e-05, "loss": 0.6619, "step": 23932 }, { "epoch": 1.44, "learning_rate": 4.445173750999653e-05, "loss": 1.0207, "step": 23934 }, { "epoch": 1.44, "learning_rate": 4.445022859988231e-05, "loss": 0.7194, "step": 23936 }, { "epoch": 1.44, "learning_rate": 4.444871968976808e-05, "loss": 0.9641, "step": 23938 }, { "epoch": 1.44, "learning_rate": 4.444721077965386e-05, "loss": 0.8885, "step": 23940 }, { "epoch": 1.45, "learning_rate": 4.444570186953963e-05, "loss": 1.1794, "step": 23942 }, { "epoch": 1.45, "learning_rate": 4.4444192959425405e-05, "loss": 0.8037, "step": 23944 }, { "epoch": 1.45, "learning_rate": 4.4442684049311185e-05, "loss": 0.9468, "step": 23946 }, { "epoch": 1.45, "learning_rate": 4.444117513919696e-05, "loss": 0.7836, "step": 23948 }, { "epoch": 1.45, "learning_rate": 4.443966622908274e-05, "loss": 1.1325, "step": 23950 }, { "epoch": 1.45, "learning_rate": 4.443815731896851e-05, "loss": 0.9755, "step": 23952 }, { "epoch": 1.45, "learning_rate": 4.4436648408854285e-05, "loss": 0.93, "step": 23954 }, { "epoch": 1.45, "learning_rate": 4.4435139498740064e-05, "loss": 0.8836, "step": 23956 }, { "epoch": 1.45, "learning_rate": 4.443363058862584e-05, "loss": 0.9008, "step": 23958 }, { "epoch": 1.45, "learning_rate": 4.443212167851161e-05, "loss": 1.1233, "step": 23960 }, { "epoch": 1.45, "learning_rate": 4.443061276839739e-05, "loss": 1.0094, "step": 23962 }, { "epoch": 1.45, "learning_rate": 4.4429103858283164e-05, "loss": 1.1609, "step": 23964 }, { "epoch": 1.45, "learning_rate": 4.442759494816894e-05, "loss": 0.9899, "step": 23966 }, { "epoch": 1.45, "learning_rate": 4.442608603805472e-05, "loss": 1.1312, "step": 23968 }, { "epoch": 1.45, "learning_rate": 4.442457712794049e-05, "loss": 1.1551, "step": 23970 }, { "epoch": 1.45, "learning_rate": 4.442306821782626e-05, "loss": 0.8119, "step": 23972 }, { "epoch": 1.45, "learning_rate": 4.442155930771204e-05, "loss": 0.7582, "step": 23974 }, { "epoch": 1.45, "learning_rate": 4.4420050397597816e-05, "loss": 0.6699, "step": 23976 }, { "epoch": 1.45, "learning_rate": 4.4418541487483596e-05, "loss": 0.9375, "step": 23978 }, { "epoch": 1.45, "learning_rate": 4.441703257736937e-05, "loss": 0.7133, "step": 23980 }, { "epoch": 1.45, "learning_rate": 4.441552366725514e-05, "loss": 0.7351, "step": 23982 }, { "epoch": 1.45, "learning_rate": 4.441401475714092e-05, "loss": 0.784, "step": 23984 }, { "epoch": 1.45, "learning_rate": 4.4412505847026695e-05, "loss": 0.9427, "step": 23986 }, { "epoch": 1.45, "learning_rate": 4.441099693691247e-05, "loss": 0.6054, "step": 23988 }, { "epoch": 1.45, "learning_rate": 4.440948802679825e-05, "loss": 0.7459, "step": 23990 }, { "epoch": 1.45, "learning_rate": 4.440797911668402e-05, "loss": 0.853, "step": 23992 }, { "epoch": 1.45, "learning_rate": 4.4406470206569794e-05, "loss": 1.2475, "step": 23994 }, { "epoch": 1.45, "learning_rate": 4.4404961296455574e-05, "loss": 0.8699, "step": 23996 }, { "epoch": 1.45, "learning_rate": 4.440345238634135e-05, "loss": 0.8939, "step": 23998 }, { "epoch": 1.45, "learning_rate": 4.440194347622713e-05, "loss": 1.2811, "step": 24000 }, { "epoch": 1.45, "learning_rate": 4.44004345661129e-05, "loss": 0.6818, "step": 24002 }, { "epoch": 1.45, "learning_rate": 4.4398925655998674e-05, "loss": 0.9958, "step": 24004 }, { "epoch": 1.45, "learning_rate": 4.4397416745884454e-05, "loss": 0.949, "step": 24006 }, { "epoch": 1.45, "learning_rate": 4.439590783577022e-05, "loss": 0.9544, "step": 24008 }, { "epoch": 1.45, "learning_rate": 4.4394398925656e-05, "loss": 0.9128, "step": 24010 }, { "epoch": 1.45, "learning_rate": 4.439289001554178e-05, "loss": 0.8541, "step": 24012 }, { "epoch": 1.45, "learning_rate": 4.439138110542755e-05, "loss": 0.8122, "step": 24014 }, { "epoch": 1.45, "learning_rate": 4.4389872195313326e-05, "loss": 0.7949, "step": 24016 }, { "epoch": 1.45, "learning_rate": 4.4388363285199106e-05, "loss": 1.1803, "step": 24018 }, { "epoch": 1.45, "learning_rate": 4.438685437508488e-05, "loss": 0.9502, "step": 24020 }, { "epoch": 1.45, "learning_rate": 4.438534546497065e-05, "loss": 0.6982, "step": 24022 }, { "epoch": 1.45, "learning_rate": 4.4383836554856425e-05, "loss": 1.1551, "step": 24024 }, { "epoch": 1.45, "learning_rate": 4.4382327644742205e-05, "loss": 1.0465, "step": 24026 }, { "epoch": 1.45, "learning_rate": 4.4380818734627985e-05, "loss": 0.9388, "step": 24028 }, { "epoch": 1.45, "learning_rate": 4.437930982451375e-05, "loss": 0.8053, "step": 24030 }, { "epoch": 1.45, "learning_rate": 4.437780091439953e-05, "loss": 0.7615, "step": 24032 }, { "epoch": 1.45, "learning_rate": 4.437629200428531e-05, "loss": 1.063, "step": 24034 }, { "epoch": 1.45, "learning_rate": 4.437478309417108e-05, "loss": 0.7607, "step": 24036 }, { "epoch": 1.45, "learning_rate": 4.437327418405686e-05, "loss": 0.847, "step": 24038 }, { "epoch": 1.45, "learning_rate": 4.437176527394263e-05, "loss": 1.0942, "step": 24040 }, { "epoch": 1.45, "learning_rate": 4.437025636382841e-05, "loss": 1.0871, "step": 24042 }, { "epoch": 1.45, "learning_rate": 4.4368747453714184e-05, "loss": 1.0197, "step": 24044 }, { "epoch": 1.45, "learning_rate": 4.436723854359996e-05, "loss": 0.9551, "step": 24046 }, { "epoch": 1.45, "learning_rate": 4.436572963348574e-05, "loss": 1.1178, "step": 24048 }, { "epoch": 1.45, "learning_rate": 4.4364220723371517e-05, "loss": 0.7821, "step": 24050 }, { "epoch": 1.45, "learning_rate": 4.436271181325728e-05, "loss": 0.7652, "step": 24052 }, { "epoch": 1.45, "learning_rate": 4.436120290314306e-05, "loss": 1.0254, "step": 24054 }, { "epoch": 1.45, "learning_rate": 4.4359693993028836e-05, "loss": 0.8059, "step": 24056 }, { "epoch": 1.45, "learning_rate": 4.435818508291461e-05, "loss": 1.1549, "step": 24058 }, { "epoch": 1.45, "learning_rate": 4.435667617280039e-05, "loss": 0.8754, "step": 24060 }, { "epoch": 1.45, "learning_rate": 4.435516726268616e-05, "loss": 1.0697, "step": 24062 }, { "epoch": 1.45, "learning_rate": 4.435365835257194e-05, "loss": 1.1881, "step": 24064 }, { "epoch": 1.45, "learning_rate": 4.4352149442457715e-05, "loss": 1.0942, "step": 24066 }, { "epoch": 1.45, "learning_rate": 4.435064053234349e-05, "loss": 0.861, "step": 24068 }, { "epoch": 1.45, "learning_rate": 4.434913162222927e-05, "loss": 0.8656, "step": 24070 }, { "epoch": 1.45, "learning_rate": 4.434762271211504e-05, "loss": 0.9613, "step": 24072 }, { "epoch": 1.45, "learning_rate": 4.4346113802000814e-05, "loss": 0.8632, "step": 24074 }, { "epoch": 1.45, "learning_rate": 4.4344604891886594e-05, "loss": 0.8764, "step": 24076 }, { "epoch": 1.45, "learning_rate": 4.434309598177237e-05, "loss": 0.8148, "step": 24078 }, { "epoch": 1.45, "learning_rate": 4.434158707165814e-05, "loss": 0.8018, "step": 24080 }, { "epoch": 1.45, "learning_rate": 4.434007816154392e-05, "loss": 0.8633, "step": 24082 }, { "epoch": 1.45, "learning_rate": 4.4338569251429694e-05, "loss": 0.8151, "step": 24084 }, { "epoch": 1.45, "learning_rate": 4.433706034131547e-05, "loss": 0.7558, "step": 24086 }, { "epoch": 1.45, "learning_rate": 4.4335551431201247e-05, "loss": 0.9185, "step": 24088 }, { "epoch": 1.45, "learning_rate": 4.433404252108702e-05, "loss": 0.8067, "step": 24090 }, { "epoch": 1.45, "learning_rate": 4.43325336109728e-05, "loss": 1.2744, "step": 24092 }, { "epoch": 1.45, "learning_rate": 4.433102470085857e-05, "loss": 1.0797, "step": 24094 }, { "epoch": 1.45, "learning_rate": 4.4329515790744346e-05, "loss": 0.9299, "step": 24096 }, { "epoch": 1.45, "learning_rate": 4.4328006880630126e-05, "loss": 0.9263, "step": 24098 }, { "epoch": 1.45, "learning_rate": 4.43264979705159e-05, "loss": 1.0878, "step": 24100 }, { "epoch": 1.45, "learning_rate": 4.432498906040167e-05, "loss": 0.899, "step": 24102 }, { "epoch": 1.45, "learning_rate": 4.432348015028745e-05, "loss": 0.9999, "step": 24104 }, { "epoch": 1.45, "learning_rate": 4.4321971240173225e-05, "loss": 0.9293, "step": 24106 }, { "epoch": 1.46, "learning_rate": 4.4320462330059e-05, "loss": 0.9909, "step": 24108 }, { "epoch": 1.46, "learning_rate": 4.431895341994478e-05, "loss": 0.8402, "step": 24110 }, { "epoch": 1.46, "learning_rate": 4.431744450983055e-05, "loss": 1.0732, "step": 24112 }, { "epoch": 1.46, "learning_rate": 4.431593559971633e-05, "loss": 0.8514, "step": 24114 }, { "epoch": 1.46, "learning_rate": 4.4314426689602104e-05, "loss": 0.9938, "step": 24116 }, { "epoch": 1.46, "learning_rate": 4.431291777948788e-05, "loss": 0.8758, "step": 24118 }, { "epoch": 1.46, "learning_rate": 4.431140886937366e-05, "loss": 1.1743, "step": 24120 }, { "epoch": 1.46, "learning_rate": 4.4309899959259424e-05, "loss": 1.014, "step": 24122 }, { "epoch": 1.46, "learning_rate": 4.4308391049145204e-05, "loss": 0.9614, "step": 24124 }, { "epoch": 1.46, "learning_rate": 4.4306882139030983e-05, "loss": 0.9405, "step": 24126 }, { "epoch": 1.46, "learning_rate": 4.4305373228916757e-05, "loss": 0.7135, "step": 24128 }, { "epoch": 1.46, "learning_rate": 4.430386431880253e-05, "loss": 0.8729, "step": 24130 }, { "epoch": 1.46, "learning_rate": 4.430235540868831e-05, "loss": 1.0929, "step": 24132 }, { "epoch": 1.46, "learning_rate": 4.430084649857408e-05, "loss": 1.0368, "step": 24134 }, { "epoch": 1.46, "learning_rate": 4.4299337588459856e-05, "loss": 0.9174, "step": 24136 }, { "epoch": 1.46, "learning_rate": 4.429782867834563e-05, "loss": 0.7503, "step": 24138 }, { "epoch": 1.46, "learning_rate": 4.429631976823141e-05, "loss": 1.0102, "step": 24140 }, { "epoch": 1.46, "learning_rate": 4.429481085811719e-05, "loss": 0.8299, "step": 24142 }, { "epoch": 1.46, "learning_rate": 4.4293301948002955e-05, "loss": 0.9167, "step": 24144 }, { "epoch": 1.46, "learning_rate": 4.4291793037888735e-05, "loss": 0.7052, "step": 24146 }, { "epoch": 1.46, "learning_rate": 4.4290284127774515e-05, "loss": 0.9222, "step": 24148 }, { "epoch": 1.46, "learning_rate": 4.428877521766029e-05, "loss": 0.7198, "step": 24150 }, { "epoch": 1.46, "learning_rate": 4.428726630754606e-05, "loss": 0.9246, "step": 24152 }, { "epoch": 1.46, "learning_rate": 4.4285757397431834e-05, "loss": 0.786, "step": 24154 }, { "epoch": 1.46, "learning_rate": 4.4284248487317614e-05, "loss": 0.5899, "step": 24156 }, { "epoch": 1.46, "learning_rate": 4.428273957720339e-05, "loss": 0.8862, "step": 24158 }, { "epoch": 1.46, "learning_rate": 4.428123066708916e-05, "loss": 0.9744, "step": 24160 }, { "epoch": 1.46, "learning_rate": 4.427972175697494e-05, "loss": 0.8054, "step": 24162 }, { "epoch": 1.46, "learning_rate": 4.427821284686072e-05, "loss": 0.7918, "step": 24164 }, { "epoch": 1.46, "learning_rate": 4.4276703936746487e-05, "loss": 0.7644, "step": 24166 }, { "epoch": 1.46, "learning_rate": 4.4275195026632266e-05, "loss": 0.8367, "step": 24168 }, { "epoch": 1.46, "learning_rate": 4.427368611651804e-05, "loss": 1.148, "step": 24170 }, { "epoch": 1.46, "learning_rate": 4.427217720640381e-05, "loss": 1.2532, "step": 24172 }, { "epoch": 1.46, "learning_rate": 4.427066829628959e-05, "loss": 1.0249, "step": 24174 }, { "epoch": 1.46, "learning_rate": 4.4269159386175366e-05, "loss": 0.8884, "step": 24176 }, { "epoch": 1.46, "learning_rate": 4.4267650476061146e-05, "loss": 1.5488, "step": 24178 }, { "epoch": 1.46, "learning_rate": 4.426614156594692e-05, "loss": 1.1052, "step": 24180 }, { "epoch": 1.46, "learning_rate": 4.426463265583269e-05, "loss": 1.1474, "step": 24182 }, { "epoch": 1.46, "learning_rate": 4.426312374571847e-05, "loss": 0.8235, "step": 24184 }, { "epoch": 1.46, "learning_rate": 4.4261614835604245e-05, "loss": 0.7126, "step": 24186 }, { "epoch": 1.46, "learning_rate": 4.426010592549002e-05, "loss": 1.0195, "step": 24188 }, { "epoch": 1.46, "learning_rate": 4.42585970153758e-05, "loss": 1.1491, "step": 24190 }, { "epoch": 1.46, "learning_rate": 4.425708810526157e-05, "loss": 0.9864, "step": 24192 }, { "epoch": 1.46, "learning_rate": 4.4255579195147344e-05, "loss": 1.0013, "step": 24194 }, { "epoch": 1.46, "learning_rate": 4.4254070285033124e-05, "loss": 1.0028, "step": 24196 }, { "epoch": 1.46, "learning_rate": 4.42525613749189e-05, "loss": 0.9695, "step": 24198 }, { "epoch": 1.46, "learning_rate": 4.425105246480468e-05, "loss": 1.1936, "step": 24200 }, { "epoch": 1.46, "learning_rate": 4.424954355469045e-05, "loss": 1.2201, "step": 24202 }, { "epoch": 1.46, "learning_rate": 4.4248034644576223e-05, "loss": 0.7214, "step": 24204 }, { "epoch": 1.46, "learning_rate": 4.4246525734462e-05, "loss": 0.7261, "step": 24206 }, { "epoch": 1.46, "learning_rate": 4.4245016824347776e-05, "loss": 1.1221, "step": 24208 }, { "epoch": 1.46, "learning_rate": 4.424350791423355e-05, "loss": 0.8053, "step": 24210 }, { "epoch": 1.46, "learning_rate": 4.424199900411933e-05, "loss": 0.7612, "step": 24212 }, { "epoch": 1.46, "learning_rate": 4.42404900940051e-05, "loss": 0.9672, "step": 24214 }, { "epoch": 1.46, "learning_rate": 4.4238981183890876e-05, "loss": 0.9139, "step": 24216 }, { "epoch": 1.46, "learning_rate": 4.4237472273776656e-05, "loss": 0.7091, "step": 24218 }, { "epoch": 1.46, "learning_rate": 4.423596336366243e-05, "loss": 1.0252, "step": 24220 }, { "epoch": 1.46, "learning_rate": 4.42344544535482e-05, "loss": 0.9539, "step": 24222 }, { "epoch": 1.46, "learning_rate": 4.423294554343398e-05, "loss": 0.8138, "step": 24224 }, { "epoch": 1.46, "learning_rate": 4.4231436633319755e-05, "loss": 0.784, "step": 24226 }, { "epoch": 1.46, "learning_rate": 4.4229927723205535e-05, "loss": 0.6194, "step": 24228 }, { "epoch": 1.46, "learning_rate": 4.422841881309131e-05, "loss": 0.7536, "step": 24230 }, { "epoch": 1.46, "learning_rate": 4.422690990297708e-05, "loss": 0.9901, "step": 24232 }, { "epoch": 1.46, "learning_rate": 4.422540099286286e-05, "loss": 1.1158, "step": 24234 }, { "epoch": 1.46, "learning_rate": 4.422389208274863e-05, "loss": 0.9385, "step": 24236 }, { "epoch": 1.46, "learning_rate": 4.422238317263441e-05, "loss": 1.3934, "step": 24238 }, { "epoch": 1.46, "learning_rate": 4.422087426252019e-05, "loss": 0.7416, "step": 24240 }, { "epoch": 1.46, "learning_rate": 4.421936535240596e-05, "loss": 0.9061, "step": 24242 }, { "epoch": 1.46, "learning_rate": 4.421785644229173e-05, "loss": 0.7471, "step": 24244 }, { "epoch": 1.46, "learning_rate": 4.421634753217751e-05, "loss": 0.5746, "step": 24246 }, { "epoch": 1.46, "learning_rate": 4.4214838622063286e-05, "loss": 1.2217, "step": 24248 }, { "epoch": 1.46, "learning_rate": 4.421332971194906e-05, "loss": 0.9328, "step": 24250 }, { "epoch": 1.46, "learning_rate": 4.421257525689195e-05, "loss": 1.1292, "step": 24252 }, { "epoch": 1.46, "learning_rate": 4.421106634677772e-05, "loss": 1.2126, "step": 24254 }, { "epoch": 1.46, "learning_rate": 4.42095574366635e-05, "loss": 0.8532, "step": 24256 }, { "epoch": 1.46, "learning_rate": 4.4208048526549276e-05, "loss": 0.9809, "step": 24258 }, { "epoch": 1.46, "learning_rate": 4.420653961643505e-05, "loss": 0.8285, "step": 24260 }, { "epoch": 1.46, "learning_rate": 4.420503070632083e-05, "loss": 0.9138, "step": 24262 }, { "epoch": 1.46, "learning_rate": 4.42035217962066e-05, "loss": 1.0431, "step": 24264 }, { "epoch": 1.46, "learning_rate": 4.4202012886092375e-05, "loss": 0.9781, "step": 24266 }, { "epoch": 1.46, "learning_rate": 4.4200503975978155e-05, "loss": 0.6976, "step": 24268 }, { "epoch": 1.46, "learning_rate": 4.419899506586393e-05, "loss": 0.9923, "step": 24270 }, { "epoch": 1.46, "learning_rate": 4.419748615574971e-05, "loss": 0.9745, "step": 24272 }, { "epoch": 1.47, "learning_rate": 4.419597724563548e-05, "loss": 1.0194, "step": 24274 }, { "epoch": 1.47, "learning_rate": 4.4194468335521254e-05, "loss": 0.9862, "step": 24276 }, { "epoch": 1.47, "learning_rate": 4.4192959425407034e-05, "loss": 0.9003, "step": 24278 }, { "epoch": 1.47, "learning_rate": 4.419145051529281e-05, "loss": 0.7921, "step": 24280 }, { "epoch": 1.47, "learning_rate": 4.418994160517858e-05, "loss": 0.7818, "step": 24282 }, { "epoch": 1.47, "learning_rate": 4.418843269506436e-05, "loss": 0.9008, "step": 24284 }, { "epoch": 1.47, "learning_rate": 4.418692378495013e-05, "loss": 0.8728, "step": 24286 }, { "epoch": 1.47, "learning_rate": 4.4185414874835906e-05, "loss": 0.6878, "step": 24288 }, { "epoch": 1.47, "learning_rate": 4.4183905964721686e-05, "loss": 1.0655, "step": 24290 }, { "epoch": 1.47, "learning_rate": 4.418239705460746e-05, "loss": 0.9131, "step": 24292 }, { "epoch": 1.47, "learning_rate": 4.418088814449323e-05, "loss": 0.6582, "step": 24294 }, { "epoch": 1.47, "learning_rate": 4.417937923437901e-05, "loss": 0.7045, "step": 24296 }, { "epoch": 1.47, "learning_rate": 4.4177870324264786e-05, "loss": 1.1491, "step": 24298 }, { "epoch": 1.47, "learning_rate": 4.4176361414150565e-05, "loss": 0.9185, "step": 24300 }, { "epoch": 1.47, "learning_rate": 4.417485250403633e-05, "loss": 0.8917, "step": 24302 }, { "epoch": 1.47, "learning_rate": 4.417334359392211e-05, "loss": 0.7408, "step": 24304 }, { "epoch": 1.47, "learning_rate": 4.417183468380789e-05, "loss": 1.087, "step": 24306 }, { "epoch": 1.47, "learning_rate": 4.417032577369366e-05, "loss": 1.0106, "step": 24308 }, { "epoch": 1.47, "learning_rate": 4.416881686357944e-05, "loss": 0.935, "step": 24310 }, { "epoch": 1.47, "learning_rate": 4.416730795346522e-05, "loss": 1.0392, "step": 24312 }, { "epoch": 1.47, "learning_rate": 4.416579904335099e-05, "loss": 1.2844, "step": 24314 }, { "epoch": 1.47, "learning_rate": 4.4164290133236764e-05, "loss": 0.5857, "step": 24316 }, { "epoch": 1.47, "learning_rate": 4.416278122312254e-05, "loss": 0.9086, "step": 24318 }, { "epoch": 1.47, "learning_rate": 4.416127231300832e-05, "loss": 0.8838, "step": 24320 }, { "epoch": 1.47, "learning_rate": 4.41597634028941e-05, "loss": 0.8867, "step": 24322 }, { "epoch": 1.47, "learning_rate": 4.415825449277986e-05, "loss": 0.9176, "step": 24324 }, { "epoch": 1.47, "learning_rate": 4.415674558266564e-05, "loss": 1.1573, "step": 24326 }, { "epoch": 1.47, "learning_rate": 4.415523667255142e-05, "loss": 0.8789, "step": 24328 }, { "epoch": 1.47, "learning_rate": 4.415372776243719e-05, "loss": 0.8126, "step": 24330 }, { "epoch": 1.47, "learning_rate": 4.415221885232297e-05, "loss": 0.8665, "step": 24332 }, { "epoch": 1.47, "learning_rate": 4.415070994220874e-05, "loss": 0.8211, "step": 24334 }, { "epoch": 1.47, "learning_rate": 4.414920103209452e-05, "loss": 0.7097, "step": 24336 }, { "epoch": 1.47, "learning_rate": 4.4147692121980295e-05, "loss": 0.795, "step": 24338 }, { "epoch": 1.47, "learning_rate": 4.414618321186607e-05, "loss": 0.8774, "step": 24340 }, { "epoch": 1.47, "learning_rate": 4.414467430175185e-05, "loss": 0.676, "step": 24342 }, { "epoch": 1.47, "learning_rate": 4.414316539163762e-05, "loss": 0.7141, "step": 24344 }, { "epoch": 1.47, "learning_rate": 4.4141656481523395e-05, "loss": 0.8595, "step": 24346 }, { "epoch": 1.47, "learning_rate": 4.4140147571409175e-05, "loss": 0.9694, "step": 24348 }, { "epoch": 1.47, "learning_rate": 4.413863866129495e-05, "loss": 0.8523, "step": 24350 }, { "epoch": 1.47, "learning_rate": 4.413712975118072e-05, "loss": 0.8743, "step": 24352 }, { "epoch": 1.47, "learning_rate": 4.41356208410665e-05, "loss": 0.9139, "step": 24354 }, { "epoch": 1.47, "learning_rate": 4.4134111930952274e-05, "loss": 1.0102, "step": 24356 }, { "epoch": 1.47, "learning_rate": 4.413260302083805e-05, "loss": 1.1901, "step": 24358 }, { "epoch": 1.47, "learning_rate": 4.413109411072383e-05, "loss": 1.2313, "step": 24360 }, { "epoch": 1.47, "learning_rate": 4.41295852006096e-05, "loss": 0.7335, "step": 24362 }, { "epoch": 1.47, "learning_rate": 4.412807629049538e-05, "loss": 0.9766, "step": 24364 }, { "epoch": 1.47, "learning_rate": 4.412656738038115e-05, "loss": 0.8495, "step": 24366 }, { "epoch": 1.47, "learning_rate": 4.4125058470266926e-05, "loss": 0.9494, "step": 24368 }, { "epoch": 1.47, "learning_rate": 4.4123549560152706e-05, "loss": 0.9411, "step": 24370 }, { "epoch": 1.47, "learning_rate": 4.412204065003848e-05, "loss": 0.8778, "step": 24372 }, { "epoch": 1.47, "learning_rate": 4.412053173992425e-05, "loss": 1.0684, "step": 24374 }, { "epoch": 1.47, "learning_rate": 4.411902282981003e-05, "loss": 0.6365, "step": 24376 }, { "epoch": 1.47, "learning_rate": 4.4117513919695805e-05, "loss": 1.0929, "step": 24378 }, { "epoch": 1.47, "learning_rate": 4.411600500958158e-05, "loss": 1.1663, "step": 24380 }, { "epoch": 1.47, "learning_rate": 4.411449609946736e-05, "loss": 0.6826, "step": 24382 }, { "epoch": 1.47, "learning_rate": 4.411298718935313e-05, "loss": 1.1027, "step": 24384 }, { "epoch": 1.47, "learning_rate": 4.411147827923891e-05, "loss": 1.0916, "step": 24386 }, { "epoch": 1.47, "learning_rate": 4.4109969369124685e-05, "loss": 0.7162, "step": 24388 }, { "epoch": 1.47, "learning_rate": 4.410846045901046e-05, "loss": 0.7466, "step": 24390 }, { "epoch": 1.47, "learning_rate": 4.410695154889624e-05, "loss": 0.9007, "step": 24392 }, { "epoch": 1.47, "learning_rate": 4.410544263878201e-05, "loss": 0.8507, "step": 24394 }, { "epoch": 1.47, "learning_rate": 4.4103933728667784e-05, "loss": 1.0596, "step": 24396 }, { "epoch": 1.47, "learning_rate": 4.4102424818553564e-05, "loss": 1.1349, "step": 24398 }, { "epoch": 1.47, "learning_rate": 4.410091590843934e-05, "loss": 0.8445, "step": 24400 }, { "epoch": 1.47, "learning_rate": 4.409940699832511e-05, "loss": 0.7548, "step": 24402 }, { "epoch": 1.47, "learning_rate": 4.409789808821089e-05, "loss": 0.8253, "step": 24404 }, { "epoch": 1.47, "learning_rate": 4.409638917809666e-05, "loss": 0.8798, "step": 24406 }, { "epoch": 1.47, "learning_rate": 4.4094880267982436e-05, "loss": 0.8262, "step": 24408 }, { "epoch": 1.47, "learning_rate": 4.4093371357868216e-05, "loss": 0.6233, "step": 24410 }, { "epoch": 1.47, "learning_rate": 4.409186244775399e-05, "loss": 0.7844, "step": 24412 }, { "epoch": 1.47, "learning_rate": 4.409035353763977e-05, "loss": 0.5879, "step": 24414 }, { "epoch": 1.47, "learning_rate": 4.4088844627525535e-05, "loss": 0.784, "step": 24416 }, { "epoch": 1.47, "learning_rate": 4.4087335717411315e-05, "loss": 0.759, "step": 24418 }, { "epoch": 1.47, "learning_rate": 4.4085826807297095e-05, "loss": 0.9742, "step": 24420 }, { "epoch": 1.47, "learning_rate": 4.408431789718287e-05, "loss": 1.3462, "step": 24422 }, { "epoch": 1.47, "learning_rate": 4.408280898706864e-05, "loss": 1.0312, "step": 24424 }, { "epoch": 1.47, "learning_rate": 4.408130007695442e-05, "loss": 1.067, "step": 24426 }, { "epoch": 1.47, "learning_rate": 4.4079791166840195e-05, "loss": 0.8357, "step": 24428 }, { "epoch": 1.47, "learning_rate": 4.407828225672597e-05, "loss": 0.5941, "step": 24430 }, { "epoch": 1.47, "learning_rate": 4.407677334661174e-05, "loss": 0.916, "step": 24432 }, { "epoch": 1.47, "learning_rate": 4.407526443649752e-05, "loss": 0.9539, "step": 24434 }, { "epoch": 1.47, "learning_rate": 4.40737555263833e-05, "loss": 1.1431, "step": 24436 }, { "epoch": 1.48, "learning_rate": 4.407224661626907e-05, "loss": 0.9055, "step": 24438 }, { "epoch": 1.48, "learning_rate": 4.407073770615485e-05, "loss": 0.9479, "step": 24440 }, { "epoch": 1.48, "learning_rate": 4.406922879604063e-05, "loss": 0.8163, "step": 24442 }, { "epoch": 1.48, "learning_rate": 4.406771988592639e-05, "loss": 1.0682, "step": 24444 }, { "epoch": 1.48, "learning_rate": 4.406621097581217e-05, "loss": 0.9954, "step": 24446 }, { "epoch": 1.48, "learning_rate": 4.4064702065697946e-05, "loss": 0.996, "step": 24448 }, { "epoch": 1.48, "learning_rate": 4.4063193155583726e-05, "loss": 0.9218, "step": 24450 }, { "epoch": 1.48, "learning_rate": 4.40616842454695e-05, "loss": 1.2831, "step": 24452 }, { "epoch": 1.48, "learning_rate": 4.406017533535527e-05, "loss": 0.7375, "step": 24454 }, { "epoch": 1.48, "learning_rate": 4.405866642524105e-05, "loss": 0.8599, "step": 24456 }, { "epoch": 1.48, "learning_rate": 4.4057157515126825e-05, "loss": 0.7866, "step": 24458 }, { "epoch": 1.48, "learning_rate": 4.40556486050126e-05, "loss": 0.8118, "step": 24460 }, { "epoch": 1.48, "learning_rate": 4.405413969489838e-05, "loss": 1.0712, "step": 24462 }, { "epoch": 1.48, "learning_rate": 4.405263078478415e-05, "loss": 0.7921, "step": 24464 }, { "epoch": 1.48, "learning_rate": 4.4051121874669925e-05, "loss": 0.8603, "step": 24466 }, { "epoch": 1.48, "learning_rate": 4.4049612964555704e-05, "loss": 0.9528, "step": 24468 }, { "epoch": 1.48, "learning_rate": 4.404810405444148e-05, "loss": 0.7476, "step": 24470 }, { "epoch": 1.48, "learning_rate": 4.404659514432726e-05, "loss": 0.8046, "step": 24472 }, { "epoch": 1.48, "learning_rate": 4.404508623421303e-05, "loss": 0.982, "step": 24474 }, { "epoch": 1.48, "learning_rate": 4.4043577324098804e-05, "loss": 1.1434, "step": 24476 }, { "epoch": 1.48, "learning_rate": 4.4042068413984584e-05, "loss": 0.9609, "step": 24478 }, { "epoch": 1.48, "learning_rate": 4.404055950387036e-05, "loss": 1.0367, "step": 24480 }, { "epoch": 1.48, "learning_rate": 4.403905059375613e-05, "loss": 0.6892, "step": 24482 }, { "epoch": 1.48, "learning_rate": 4.403754168364191e-05, "loss": 1.0142, "step": 24484 }, { "epoch": 1.48, "learning_rate": 4.403603277352768e-05, "loss": 1.0263, "step": 24486 }, { "epoch": 1.48, "learning_rate": 4.4034523863413456e-05, "loss": 1.1763, "step": 24488 }, { "epoch": 1.48, "learning_rate": 4.4033014953299236e-05, "loss": 1.0743, "step": 24490 }, { "epoch": 1.48, "learning_rate": 4.403150604318501e-05, "loss": 1.0924, "step": 24492 }, { "epoch": 1.48, "learning_rate": 4.402999713307078e-05, "loss": 1.0871, "step": 24494 }, { "epoch": 1.48, "learning_rate": 4.402848822295656e-05, "loss": 0.8639, "step": 24496 }, { "epoch": 1.48, "learning_rate": 4.4026979312842335e-05, "loss": 0.8559, "step": 24498 }, { "epoch": 1.48, "learning_rate": 4.4025470402728115e-05, "loss": 0.7814, "step": 24500 }, { "epoch": 1.48, "learning_rate": 4.402396149261389e-05, "loss": 0.9728, "step": 24502 }, { "epoch": 1.48, "learning_rate": 4.402245258249966e-05, "loss": 0.8887, "step": 24504 }, { "epoch": 1.48, "learning_rate": 4.402094367238544e-05, "loss": 1.1685, "step": 24506 }, { "epoch": 1.48, "learning_rate": 4.4019434762271214e-05, "loss": 0.9691, "step": 24508 }, { "epoch": 1.48, "learning_rate": 4.401792585215699e-05, "loss": 0.8274, "step": 24510 }, { "epoch": 1.48, "learning_rate": 4.401641694204277e-05, "loss": 1.1319, "step": 24512 }, { "epoch": 1.48, "learning_rate": 4.401490803192854e-05, "loss": 1.0095, "step": 24514 }, { "epoch": 1.48, "learning_rate": 4.4013399121814314e-05, "loss": 0.7882, "step": 24516 }, { "epoch": 1.48, "learning_rate": 4.4011890211700094e-05, "loss": 0.8062, "step": 24518 }, { "epoch": 1.48, "learning_rate": 4.401038130158587e-05, "loss": 0.9268, "step": 24520 }, { "epoch": 1.48, "learning_rate": 4.400887239147164e-05, "loss": 0.797, "step": 24522 }, { "epoch": 1.48, "learning_rate": 4.400736348135742e-05, "loss": 0.6814, "step": 24524 }, { "epoch": 1.48, "learning_rate": 4.400585457124319e-05, "loss": 0.7705, "step": 24526 }, { "epoch": 1.48, "learning_rate": 4.400434566112897e-05, "loss": 1.3507, "step": 24528 }, { "epoch": 1.48, "learning_rate": 4.400283675101474e-05, "loss": 1.1194, "step": 24530 }, { "epoch": 1.48, "learning_rate": 4.400132784090052e-05, "loss": 0.8495, "step": 24532 }, { "epoch": 1.48, "learning_rate": 4.39998189307863e-05, "loss": 0.9413, "step": 24534 }, { "epoch": 1.48, "learning_rate": 4.399831002067207e-05, "loss": 0.9119, "step": 24536 }, { "epoch": 1.48, "learning_rate": 4.3996801110557845e-05, "loss": 0.9169, "step": 24538 }, { "epoch": 1.48, "learning_rate": 4.3995292200443625e-05, "loss": 0.7574, "step": 24540 }, { "epoch": 1.48, "learning_rate": 4.39937832903294e-05, "loss": 1.048, "step": 24542 }, { "epoch": 1.48, "learning_rate": 4.399227438021517e-05, "loss": 0.7916, "step": 24544 }, { "epoch": 1.48, "learning_rate": 4.3990765470100944e-05, "loss": 0.5896, "step": 24546 }, { "epoch": 1.48, "learning_rate": 4.3989256559986724e-05, "loss": 1.132, "step": 24548 }, { "epoch": 1.48, "learning_rate": 4.3987747649872504e-05, "loss": 1.0269, "step": 24550 }, { "epoch": 1.48, "learning_rate": 4.398623873975827e-05, "loss": 0.9714, "step": 24552 }, { "epoch": 1.48, "learning_rate": 4.398472982964405e-05, "loss": 1.3428, "step": 24554 }, { "epoch": 1.48, "learning_rate": 4.398322091952983e-05, "loss": 0.6199, "step": 24556 }, { "epoch": 1.48, "learning_rate": 4.39817120094156e-05, "loss": 0.5333, "step": 24558 }, { "epoch": 1.48, "learning_rate": 4.398020309930138e-05, "loss": 1.0341, "step": 24560 }, { "epoch": 1.48, "learning_rate": 4.397869418918715e-05, "loss": 0.9695, "step": 24562 }, { "epoch": 1.48, "learning_rate": 4.397718527907293e-05, "loss": 1.0643, "step": 24564 }, { "epoch": 1.48, "learning_rate": 4.39756763689587e-05, "loss": 1.1702, "step": 24566 }, { "epoch": 1.48, "learning_rate": 4.3974167458844476e-05, "loss": 0.7854, "step": 24568 }, { "epoch": 1.48, "learning_rate": 4.3972658548730256e-05, "loss": 0.6472, "step": 24570 }, { "epoch": 1.48, "learning_rate": 4.397114963861603e-05, "loss": 0.7663, "step": 24572 }, { "epoch": 1.48, "learning_rate": 4.39696407285018e-05, "loss": 1.0109, "step": 24574 }, { "epoch": 1.48, "learning_rate": 4.396813181838758e-05, "loss": 0.8989, "step": 24576 }, { "epoch": 1.48, "learning_rate": 4.3966622908273355e-05, "loss": 0.8809, "step": 24578 }, { "epoch": 1.48, "learning_rate": 4.396511399815913e-05, "loss": 1.0182, "step": 24580 }, { "epoch": 1.48, "learning_rate": 4.396360508804491e-05, "loss": 0.9585, "step": 24582 }, { "epoch": 1.48, "learning_rate": 4.396209617793068e-05, "loss": 1.0239, "step": 24584 }, { "epoch": 1.48, "learning_rate": 4.396058726781646e-05, "loss": 0.7785, "step": 24586 }, { "epoch": 1.48, "learning_rate": 4.3959078357702234e-05, "loss": 0.8736, "step": 24588 }, { "epoch": 1.48, "learning_rate": 4.395756944758801e-05, "loss": 1.0145, "step": 24590 }, { "epoch": 1.48, "learning_rate": 4.395606053747379e-05, "loss": 0.8236, "step": 24592 }, { "epoch": 1.48, "learning_rate": 4.395455162735956e-05, "loss": 0.8082, "step": 24594 }, { "epoch": 1.48, "learning_rate": 4.3953042717245334e-05, "loss": 0.6596, "step": 24596 }, { "epoch": 1.48, "learning_rate": 4.3951533807131114e-05, "loss": 0.9471, "step": 24598 }, { "epoch": 1.48, "learning_rate": 4.395002489701689e-05, "loss": 1.0304, "step": 24600 }, { "epoch": 1.48, "learning_rate": 4.394851598690266e-05, "loss": 0.8772, "step": 24602 }, { "epoch": 1.49, "learning_rate": 4.394700707678844e-05, "loss": 1.0703, "step": 24604 }, { "epoch": 1.49, "learning_rate": 4.394549816667421e-05, "loss": 1.0928, "step": 24606 }, { "epoch": 1.49, "learning_rate": 4.3943989256559986e-05, "loss": 0.4701, "step": 24608 }, { "epoch": 1.49, "learning_rate": 4.3942480346445766e-05, "loss": 1.0979, "step": 24610 }, { "epoch": 1.49, "learning_rate": 4.394097143633154e-05, "loss": 0.6817, "step": 24612 }, { "epoch": 1.49, "learning_rate": 4.393946252621732e-05, "loss": 0.853, "step": 24614 }, { "epoch": 1.49, "learning_rate": 4.393795361610309e-05, "loss": 0.8326, "step": 24616 }, { "epoch": 1.49, "learning_rate": 4.3936444705988865e-05, "loss": 0.9147, "step": 24618 }, { "epoch": 1.49, "learning_rate": 4.3934935795874645e-05, "loss": 1.1822, "step": 24620 }, { "epoch": 1.49, "learning_rate": 4.393342688576042e-05, "loss": 0.8862, "step": 24622 }, { "epoch": 1.49, "learning_rate": 4.393191797564619e-05, "loss": 0.7031, "step": 24624 }, { "epoch": 1.49, "learning_rate": 4.393040906553197e-05, "loss": 0.7927, "step": 24626 }, { "epoch": 1.49, "learning_rate": 4.3928900155417744e-05, "loss": 1.0778, "step": 24628 }, { "epoch": 1.49, "learning_rate": 4.392739124530352e-05, "loss": 0.7614, "step": 24630 }, { "epoch": 1.49, "learning_rate": 4.39258823351893e-05, "loss": 0.9401, "step": 24632 }, { "epoch": 1.49, "learning_rate": 4.392437342507507e-05, "loss": 0.8396, "step": 24634 }, { "epoch": 1.49, "learning_rate": 4.392286451496085e-05, "loss": 0.8616, "step": 24636 }, { "epoch": 1.49, "learning_rate": 4.3921355604846623e-05, "loss": 0.8017, "step": 24638 }, { "epoch": 1.49, "learning_rate": 4.3919846694732397e-05, "loss": 0.8091, "step": 24640 }, { "epoch": 1.49, "learning_rate": 4.3918337784618176e-05, "loss": 1.1698, "step": 24642 }, { "epoch": 1.49, "learning_rate": 4.391682887450394e-05, "loss": 0.8251, "step": 24644 }, { "epoch": 1.49, "learning_rate": 4.391531996438972e-05, "loss": 1.0691, "step": 24646 }, { "epoch": 1.49, "learning_rate": 4.39138110542755e-05, "loss": 1.0628, "step": 24648 }, { "epoch": 1.49, "learning_rate": 4.3912302144161276e-05, "loss": 0.6304, "step": 24650 }, { "epoch": 1.49, "learning_rate": 4.391079323404705e-05, "loss": 1.16, "step": 24652 }, { "epoch": 1.49, "learning_rate": 4.390928432393283e-05, "loss": 0.7253, "step": 24654 }, { "epoch": 1.49, "learning_rate": 4.39077754138186e-05, "loss": 1.0522, "step": 24656 }, { "epoch": 1.49, "learning_rate": 4.3906266503704375e-05, "loss": 1.2095, "step": 24658 }, { "epoch": 1.49, "learning_rate": 4.390475759359015e-05, "loss": 1.1115, "step": 24660 }, { "epoch": 1.49, "learning_rate": 4.390324868347593e-05, "loss": 0.8881, "step": 24662 }, { "epoch": 1.49, "learning_rate": 4.390173977336171e-05, "loss": 0.9107, "step": 24664 }, { "epoch": 1.49, "learning_rate": 4.3900230863247474e-05, "loss": 0.7932, "step": 24666 }, { "epoch": 1.49, "learning_rate": 4.3898721953133254e-05, "loss": 0.9653, "step": 24668 }, { "epoch": 1.49, "learning_rate": 4.3897213043019034e-05, "loss": 1.0563, "step": 24670 }, { "epoch": 1.49, "learning_rate": 4.38957041329048e-05, "loss": 0.8352, "step": 24672 }, { "epoch": 1.49, "learning_rate": 4.389419522279058e-05, "loss": 1.0685, "step": 24674 }, { "epoch": 1.49, "learning_rate": 4.3892686312676353e-05, "loss": 0.9798, "step": 24676 }, { "epoch": 1.49, "learning_rate": 4.3891177402562133e-05, "loss": 0.8645, "step": 24678 }, { "epoch": 1.49, "learning_rate": 4.3889668492447907e-05, "loss": 0.748, "step": 24680 }, { "epoch": 1.49, "learning_rate": 4.388815958233368e-05, "loss": 0.7521, "step": 24682 }, { "epoch": 1.49, "learning_rate": 4.388665067221946e-05, "loss": 1.2596, "step": 24684 }, { "epoch": 1.49, "learning_rate": 4.388514176210523e-05, "loss": 1.1324, "step": 24686 }, { "epoch": 1.49, "learning_rate": 4.3883632851991006e-05, "loss": 0.8193, "step": 24688 }, { "epoch": 1.49, "learning_rate": 4.3882123941876786e-05, "loss": 1.1454, "step": 24690 }, { "epoch": 1.49, "learning_rate": 4.388061503176256e-05, "loss": 1.3198, "step": 24692 }, { "epoch": 1.49, "learning_rate": 4.387910612164833e-05, "loss": 1.1341, "step": 24694 }, { "epoch": 1.49, "learning_rate": 4.387759721153411e-05, "loss": 0.7685, "step": 24696 }, { "epoch": 1.49, "learning_rate": 4.3876088301419885e-05, "loss": 0.9433, "step": 24698 }, { "epoch": 1.49, "learning_rate": 4.3874579391305665e-05, "loss": 0.8111, "step": 24700 }, { "epoch": 1.49, "learning_rate": 4.387307048119144e-05, "loss": 0.8507, "step": 24702 }, { "epoch": 1.49, "learning_rate": 4.387156157107721e-05, "loss": 1.0749, "step": 24704 }, { "epoch": 1.49, "learning_rate": 4.387005266096299e-05, "loss": 0.8822, "step": 24706 }, { "epoch": 1.49, "learning_rate": 4.3868543750848764e-05, "loss": 0.6428, "step": 24708 }, { "epoch": 1.49, "learning_rate": 4.386703484073454e-05, "loss": 0.8206, "step": 24710 }, { "epoch": 1.49, "learning_rate": 4.386552593062032e-05, "loss": 0.9426, "step": 24712 }, { "epoch": 1.49, "learning_rate": 4.386401702050609e-05, "loss": 0.8739, "step": 24714 }, { "epoch": 1.49, "learning_rate": 4.3862508110391863e-05, "loss": 0.8232, "step": 24716 }, { "epoch": 1.49, "learning_rate": 4.386099920027764e-05, "loss": 1.0713, "step": 24718 }, { "epoch": 1.49, "learning_rate": 4.3859490290163416e-05, "loss": 0.9737, "step": 24720 }, { "epoch": 1.49, "learning_rate": 4.385798138004919e-05, "loss": 1.1033, "step": 24722 }, { "epoch": 1.49, "learning_rate": 4.385647246993497e-05, "loss": 0.8391, "step": 24724 }, { "epoch": 1.49, "learning_rate": 4.385496355982074e-05, "loss": 0.7581, "step": 24726 }, { "epoch": 1.49, "learning_rate": 4.385345464970652e-05, "loss": 1.0405, "step": 24728 }, { "epoch": 1.49, "learning_rate": 4.3851945739592296e-05, "loss": 0.9283, "step": 24730 }, { "epoch": 1.49, "learning_rate": 4.385043682947807e-05, "loss": 0.9168, "step": 24732 }, { "epoch": 1.49, "learning_rate": 4.384892791936385e-05, "loss": 0.8931, "step": 24734 }, { "epoch": 1.49, "learning_rate": 4.384741900924962e-05, "loss": 0.8361, "step": 24736 }, { "epoch": 1.49, "learning_rate": 4.3845910099135395e-05, "loss": 0.9931, "step": 24738 }, { "epoch": 1.49, "learning_rate": 4.3844401189021175e-05, "loss": 0.8141, "step": 24740 }, { "epoch": 1.49, "learning_rate": 4.384289227890695e-05, "loss": 0.8321, "step": 24742 }, { "epoch": 1.49, "learning_rate": 4.384138336879272e-05, "loss": 0.7769, "step": 24744 }, { "epoch": 1.49, "learning_rate": 4.38398744586785e-05, "loss": 0.8094, "step": 24746 }, { "epoch": 1.49, "learning_rate": 4.3838365548564274e-05, "loss": 1.2583, "step": 24748 }, { "epoch": 1.49, "learning_rate": 4.3836856638450054e-05, "loss": 0.942, "step": 24750 }, { "epoch": 1.49, "learning_rate": 4.383534772833583e-05, "loss": 1.225, "step": 24752 }, { "epoch": 1.49, "learning_rate": 4.38338388182216e-05, "loss": 0.7213, "step": 24754 }, { "epoch": 1.49, "learning_rate": 4.383232990810738e-05, "loss": 0.9319, "step": 24756 }, { "epoch": 1.49, "learning_rate": 4.3830820997993147e-05, "loss": 0.9732, "step": 24758 }, { "epoch": 1.49, "learning_rate": 4.3829312087878926e-05, "loss": 0.8775, "step": 24760 }, { "epoch": 1.49, "learning_rate": 4.3827803177764706e-05, "loss": 1.0227, "step": 24762 }, { "epoch": 1.49, "learning_rate": 4.382629426765048e-05, "loss": 1.0441, "step": 24764 }, { "epoch": 1.49, "learning_rate": 4.382478535753625e-05, "loss": 0.5833, "step": 24766 }, { "epoch": 1.49, "learning_rate": 4.382327644742203e-05, "loss": 0.657, "step": 24768 }, { "epoch": 1.5, "learning_rate": 4.3821767537307806e-05, "loss": 0.9624, "step": 24770 }, { "epoch": 1.5, "learning_rate": 4.382025862719358e-05, "loss": 0.8417, "step": 24772 }, { "epoch": 1.5, "learning_rate": 4.381874971707935e-05, "loss": 0.9, "step": 24774 }, { "epoch": 1.5, "learning_rate": 4.381724080696513e-05, "loss": 0.6063, "step": 24776 }, { "epoch": 1.5, "learning_rate": 4.381573189685091e-05, "loss": 0.8349, "step": 24778 }, { "epoch": 1.5, "learning_rate": 4.381422298673668e-05, "loss": 0.8726, "step": 24780 }, { "epoch": 1.5, "learning_rate": 4.381271407662246e-05, "loss": 0.9041, "step": 24782 }, { "epoch": 1.5, "learning_rate": 4.381120516650824e-05, "loss": 0.8687, "step": 24784 }, { "epoch": 1.5, "learning_rate": 4.3809696256394004e-05, "loss": 0.6963, "step": 24786 }, { "epoch": 1.5, "learning_rate": 4.3808187346279784e-05, "loss": 0.7665, "step": 24788 }, { "epoch": 1.5, "learning_rate": 4.380667843616556e-05, "loss": 0.8063, "step": 24790 }, { "epoch": 1.5, "learning_rate": 4.380516952605134e-05, "loss": 0.764, "step": 24792 }, { "epoch": 1.5, "learning_rate": 4.380366061593711e-05, "loss": 0.5763, "step": 24794 }, { "epoch": 1.5, "learning_rate": 4.380215170582288e-05, "loss": 0.99, "step": 24796 }, { "epoch": 1.5, "learning_rate": 4.380064279570866e-05, "loss": 0.7912, "step": 24798 }, { "epoch": 1.5, "learning_rate": 4.379913388559444e-05, "loss": 0.6731, "step": 24800 }, { "epoch": 1.5, "learning_rate": 4.379762497548021e-05, "loss": 0.9354, "step": 24802 }, { "epoch": 1.5, "learning_rate": 4.379611606536599e-05, "loss": 0.7839, "step": 24804 }, { "epoch": 1.5, "learning_rate": 4.379460715525176e-05, "loss": 0.6969, "step": 24806 }, { "epoch": 1.5, "learning_rate": 4.3793098245137536e-05, "loss": 0.8942, "step": 24808 }, { "epoch": 1.5, "learning_rate": 4.3791589335023316e-05, "loss": 0.6939, "step": 24810 }, { "epoch": 1.5, "learning_rate": 4.379008042490909e-05, "loss": 0.7577, "step": 24812 }, { "epoch": 1.5, "learning_rate": 4.378857151479487e-05, "loss": 0.8552, "step": 24814 }, { "epoch": 1.5, "learning_rate": 4.378706260468064e-05, "loss": 1.2258, "step": 24816 }, { "epoch": 1.5, "learning_rate": 4.3785553694566415e-05, "loss": 0.7996, "step": 24818 }, { "epoch": 1.5, "learning_rate": 4.3784044784452195e-05, "loss": 0.9059, "step": 24820 }, { "epoch": 1.5, "learning_rate": 4.378253587433797e-05, "loss": 1.0361, "step": 24822 }, { "epoch": 1.5, "learning_rate": 4.378102696422374e-05, "loss": 1.0098, "step": 24824 }, { "epoch": 1.5, "learning_rate": 4.377951805410952e-05, "loss": 1.2355, "step": 24826 }, { "epoch": 1.5, "learning_rate": 4.3778009143995294e-05, "loss": 0.9843, "step": 24828 }, { "epoch": 1.5, "learning_rate": 4.377650023388107e-05, "loss": 1.025, "step": 24830 }, { "epoch": 1.5, "learning_rate": 4.377499132376685e-05, "loss": 0.8772, "step": 24832 }, { "epoch": 1.5, "learning_rate": 4.377348241365262e-05, "loss": 0.6386, "step": 24834 }, { "epoch": 1.5, "learning_rate": 4.377197350353839e-05, "loss": 1.0112, "step": 24836 }, { "epoch": 1.5, "learning_rate": 4.377046459342417e-05, "loss": 0.9325, "step": 24838 }, { "epoch": 1.5, "learning_rate": 4.3768955683309946e-05, "loss": 0.7567, "step": 24840 }, { "epoch": 1.5, "learning_rate": 4.3767446773195726e-05, "loss": 1.0207, "step": 24842 }, { "epoch": 1.5, "learning_rate": 4.37659378630815e-05, "loss": 1.0359, "step": 24844 }, { "epoch": 1.5, "learning_rate": 4.376442895296727e-05, "loss": 0.8772, "step": 24846 }, { "epoch": 1.5, "learning_rate": 4.376292004285305e-05, "loss": 0.9564, "step": 24848 }, { "epoch": 1.5, "learning_rate": 4.3761411132738825e-05, "loss": 0.8229, "step": 24850 }, { "epoch": 1.5, "learning_rate": 4.37599022226246e-05, "loss": 0.9041, "step": 24852 }, { "epoch": 1.5, "learning_rate": 4.375839331251038e-05, "loss": 0.9035, "step": 24854 }, { "epoch": 1.5, "learning_rate": 4.375688440239615e-05, "loss": 1.1855, "step": 24856 }, { "epoch": 1.5, "learning_rate": 4.3755375492281925e-05, "loss": 1.1525, "step": 24858 }, { "epoch": 1.5, "learning_rate": 4.3753866582167705e-05, "loss": 0.8547, "step": 24860 }, { "epoch": 1.5, "learning_rate": 4.375235767205348e-05, "loss": 1.031, "step": 24862 }, { "epoch": 1.5, "learning_rate": 4.375084876193926e-05, "loss": 0.7635, "step": 24864 }, { "epoch": 1.5, "learning_rate": 4.374933985182503e-05, "loss": 1.0695, "step": 24866 }, { "epoch": 1.5, "learning_rate": 4.3747830941710804e-05, "loss": 0.7799, "step": 24868 }, { "epoch": 1.5, "learning_rate": 4.3746322031596584e-05, "loss": 0.7886, "step": 24870 }, { "epoch": 1.5, "learning_rate": 4.374481312148235e-05, "loss": 0.8912, "step": 24872 }, { "epoch": 1.5, "learning_rate": 4.374330421136813e-05, "loss": 0.8585, "step": 24874 }, { "epoch": 1.5, "learning_rate": 4.374179530125391e-05, "loss": 0.9969, "step": 24876 }, { "epoch": 1.5, "learning_rate": 4.374028639113968e-05, "loss": 1.0412, "step": 24878 }, { "epoch": 1.5, "learning_rate": 4.3738777481025456e-05, "loss": 0.8902, "step": 24880 }, { "epoch": 1.5, "learning_rate": 4.3737268570911236e-05, "loss": 0.8088, "step": 24882 }, { "epoch": 1.5, "learning_rate": 4.373575966079701e-05, "loss": 1.0196, "step": 24884 }, { "epoch": 1.5, "learning_rate": 4.373425075068278e-05, "loss": 0.7602, "step": 24886 }, { "epoch": 1.5, "learning_rate": 4.3732741840568556e-05, "loss": 0.947, "step": 24888 }, { "epoch": 1.5, "learning_rate": 4.3731232930454335e-05, "loss": 0.8737, "step": 24890 }, { "epoch": 1.5, "learning_rate": 4.3729724020340115e-05, "loss": 0.9487, "step": 24892 }, { "epoch": 1.5, "learning_rate": 4.372821511022588e-05, "loss": 0.6959, "step": 24894 }, { "epoch": 1.5, "learning_rate": 4.372670620011166e-05, "loss": 0.6856, "step": 24896 }, { "epoch": 1.5, "learning_rate": 4.372519728999744e-05, "loss": 0.8478, "step": 24898 }, { "epoch": 1.5, "learning_rate": 4.372368837988321e-05, "loss": 0.9474, "step": 24900 }, { "epoch": 1.5, "learning_rate": 4.372217946976899e-05, "loss": 0.9684, "step": 24902 }, { "epoch": 1.5, "learning_rate": 4.372067055965476e-05, "loss": 0.9999, "step": 24904 }, { "epoch": 1.5, "learning_rate": 4.371916164954054e-05, "loss": 0.9359, "step": 24906 }, { "epoch": 1.5, "learning_rate": 4.3717652739426314e-05, "loss": 0.8566, "step": 24908 }, { "epoch": 1.5, "learning_rate": 4.371614382931209e-05, "loss": 0.9798, "step": 24910 }, { "epoch": 1.5, "learning_rate": 4.371463491919787e-05, "loss": 0.7621, "step": 24912 }, { "epoch": 1.5, "learning_rate": 4.371312600908365e-05, "loss": 0.7337, "step": 24914 }, { "epoch": 1.5, "learning_rate": 4.371161709896941e-05, "loss": 0.8464, "step": 24916 }, { "epoch": 1.5, "learning_rate": 4.371010818885519e-05, "loss": 1.1355, "step": 24918 }, { "epoch": 1.5, "learning_rate": 4.3708599278740966e-05, "loss": 0.7646, "step": 24920 }, { "epoch": 1.5, "learning_rate": 4.370709036862674e-05, "loss": 0.8329, "step": 24922 }, { "epoch": 1.5, "learning_rate": 4.370558145851252e-05, "loss": 0.9735, "step": 24924 }, { "epoch": 1.5, "learning_rate": 4.370407254839829e-05, "loss": 1.0171, "step": 24926 }, { "epoch": 1.5, "learning_rate": 4.370256363828407e-05, "loss": 0.7055, "step": 24928 }, { "epoch": 1.5, "learning_rate": 4.3701054728169845e-05, "loss": 1.0875, "step": 24930 }, { "epoch": 1.5, "learning_rate": 4.369954581805562e-05, "loss": 1.0576, "step": 24932 }, { "epoch": 1.5, "learning_rate": 4.36980369079414e-05, "loss": 0.8296, "step": 24934 }, { "epoch": 1.51, "learning_rate": 4.369652799782717e-05, "loss": 0.7788, "step": 24936 }, { "epoch": 1.51, "learning_rate": 4.3695019087712945e-05, "loss": 0.9078, "step": 24938 }, { "epoch": 1.51, "learning_rate": 4.3693510177598725e-05, "loss": 0.9313, "step": 24940 }, { "epoch": 1.51, "learning_rate": 4.36920012674845e-05, "loss": 1.1046, "step": 24942 }, { "epoch": 1.51, "learning_rate": 4.369049235737027e-05, "loss": 1.0835, "step": 24944 }, { "epoch": 1.51, "learning_rate": 4.368898344725605e-05, "loss": 0.9656, "step": 24946 }, { "epoch": 1.51, "learning_rate": 4.3687474537141824e-05, "loss": 1.2361, "step": 24948 }, { "epoch": 1.51, "learning_rate": 4.36859656270276e-05, "loss": 1.007, "step": 24950 }, { "epoch": 1.51, "learning_rate": 4.368445671691338e-05, "loss": 0.7622, "step": 24952 }, { "epoch": 1.51, "learning_rate": 4.368294780679915e-05, "loss": 0.7961, "step": 24954 }, { "epoch": 1.51, "learning_rate": 4.368143889668493e-05, "loss": 1.1174, "step": 24956 }, { "epoch": 1.51, "learning_rate": 4.36799299865707e-05, "loss": 1.0956, "step": 24958 }, { "epoch": 1.51, "learning_rate": 4.3678421076456476e-05, "loss": 1.0527, "step": 24960 }, { "epoch": 1.51, "learning_rate": 4.3676912166342256e-05, "loss": 0.956, "step": 24962 }, { "epoch": 1.51, "learning_rate": 4.367540325622803e-05, "loss": 0.6949, "step": 24964 }, { "epoch": 1.51, "learning_rate": 4.36738943461138e-05, "loss": 0.8391, "step": 24966 }, { "epoch": 1.51, "learning_rate": 4.367238543599958e-05, "loss": 1.0645, "step": 24968 }, { "epoch": 1.51, "learning_rate": 4.3670876525885355e-05, "loss": 0.9026, "step": 24970 }, { "epoch": 1.51, "learning_rate": 4.366936761577113e-05, "loss": 0.9463, "step": 24972 }, { "epoch": 1.51, "learning_rate": 4.366785870565691e-05, "loss": 0.9974, "step": 24974 }, { "epoch": 1.51, "learning_rate": 4.366634979554268e-05, "loss": 0.7702, "step": 24976 }, { "epoch": 1.51, "learning_rate": 4.366484088542846e-05, "loss": 1.079, "step": 24978 }, { "epoch": 1.51, "learning_rate": 4.3663331975314234e-05, "loss": 0.8973, "step": 24980 }, { "epoch": 1.51, "learning_rate": 4.366182306520001e-05, "loss": 0.9795, "step": 24982 }, { "epoch": 1.51, "learning_rate": 4.366031415508579e-05, "loss": 0.6852, "step": 24984 }, { "epoch": 1.51, "learning_rate": 4.3658805244971554e-05, "loss": 0.9751, "step": 24986 }, { "epoch": 1.51, "learning_rate": 4.3657296334857334e-05, "loss": 0.9373, "step": 24988 }, { "epoch": 1.51, "learning_rate": 4.3655787424743114e-05, "loss": 1.0196, "step": 24990 }, { "epoch": 1.51, "learning_rate": 4.365427851462889e-05, "loss": 0.9599, "step": 24992 }, { "epoch": 1.51, "learning_rate": 4.365276960451466e-05, "loss": 1.0327, "step": 24994 }, { "epoch": 1.51, "learning_rate": 4.365126069440044e-05, "loss": 0.841, "step": 24996 }, { "epoch": 1.51, "learning_rate": 4.364975178428621e-05, "loss": 0.9709, "step": 24998 }, { "epoch": 1.51, "learning_rate": 4.3648242874171986e-05, "loss": 1.0448, "step": 25000 }, { "epoch": 1.51, "learning_rate": 4.364673396405776e-05, "loss": 0.9295, "step": 25002 }, { "epoch": 1.51, "learning_rate": 4.364522505394354e-05, "loss": 0.9766, "step": 25004 }, { "epoch": 1.51, "learning_rate": 4.364371614382932e-05, "loss": 1.2681, "step": 25006 }, { "epoch": 1.51, "learning_rate": 4.3642207233715085e-05, "loss": 0.7213, "step": 25008 }, { "epoch": 1.51, "learning_rate": 4.3640698323600865e-05, "loss": 1.0299, "step": 25010 }, { "epoch": 1.51, "learning_rate": 4.3639189413486645e-05, "loss": 0.7861, "step": 25012 }, { "epoch": 1.51, "learning_rate": 4.363768050337241e-05, "loss": 0.9051, "step": 25014 }, { "epoch": 1.51, "learning_rate": 4.363617159325819e-05, "loss": 0.9767, "step": 25016 }, { "epoch": 1.51, "learning_rate": 4.3634662683143965e-05, "loss": 0.9935, "step": 25018 }, { "epoch": 1.51, "learning_rate": 4.3633153773029744e-05, "loss": 0.6962, "step": 25020 }, { "epoch": 1.51, "learning_rate": 4.363164486291552e-05, "loss": 0.782, "step": 25022 }, { "epoch": 1.51, "learning_rate": 4.363013595280129e-05, "loss": 1.0134, "step": 25024 }, { "epoch": 1.51, "learning_rate": 4.362862704268707e-05, "loss": 1.1664, "step": 25026 }, { "epoch": 1.51, "learning_rate": 4.362711813257285e-05, "loss": 1.0745, "step": 25028 }, { "epoch": 1.51, "learning_rate": 4.362560922245862e-05, "loss": 0.8753, "step": 25030 }, { "epoch": 1.51, "learning_rate": 4.36241003123444e-05, "loss": 0.7043, "step": 25032 }, { "epoch": 1.51, "learning_rate": 4.362259140223017e-05, "loss": 1.0073, "step": 25034 }, { "epoch": 1.51, "learning_rate": 4.362108249211594e-05, "loss": 0.6083, "step": 25036 }, { "epoch": 1.51, "learning_rate": 4.361957358200172e-05, "loss": 0.8632, "step": 25038 }, { "epoch": 1.51, "learning_rate": 4.3618064671887496e-05, "loss": 0.7868, "step": 25040 }, { "epoch": 1.51, "learning_rate": 4.3616555761773276e-05, "loss": 0.8741, "step": 25042 }, { "epoch": 1.51, "learning_rate": 4.361504685165905e-05, "loss": 1.3383, "step": 25044 }, { "epoch": 1.51, "learning_rate": 4.361353794154482e-05, "loss": 0.712, "step": 25046 }, { "epoch": 1.51, "learning_rate": 4.36120290314306e-05, "loss": 0.8312, "step": 25048 }, { "epoch": 1.51, "learning_rate": 4.3610520121316375e-05, "loss": 0.9757, "step": 25050 }, { "epoch": 1.51, "learning_rate": 4.360901121120215e-05, "loss": 1.0453, "step": 25052 }, { "epoch": 1.51, "learning_rate": 4.360750230108793e-05, "loss": 0.8118, "step": 25054 }, { "epoch": 1.51, "learning_rate": 4.36059933909737e-05, "loss": 0.8996, "step": 25056 }, { "epoch": 1.51, "learning_rate": 4.3604484480859474e-05, "loss": 0.8693, "step": 25058 }, { "epoch": 1.51, "learning_rate": 4.3602975570745254e-05, "loss": 0.9734, "step": 25060 }, { "epoch": 1.51, "learning_rate": 4.360146666063103e-05, "loss": 0.7987, "step": 25062 }, { "epoch": 1.51, "learning_rate": 4.35999577505168e-05, "loss": 1.027, "step": 25064 }, { "epoch": 1.51, "learning_rate": 4.359844884040258e-05, "loss": 0.6317, "step": 25066 }, { "epoch": 1.51, "learning_rate": 4.3596939930288354e-05, "loss": 1.053, "step": 25068 }, { "epoch": 1.51, "learning_rate": 4.3595431020174134e-05, "loss": 0.6936, "step": 25070 }, { "epoch": 1.51, "learning_rate": 4.359392211005991e-05, "loss": 0.7853, "step": 25072 }, { "epoch": 1.51, "learning_rate": 4.359241319994568e-05, "loss": 0.9794, "step": 25074 }, { "epoch": 1.51, "learning_rate": 4.359090428983146e-05, "loss": 0.7181, "step": 25076 }, { "epoch": 1.51, "learning_rate": 4.358939537971723e-05, "loss": 1.3087, "step": 25078 }, { "epoch": 1.51, "learning_rate": 4.3587886469603006e-05, "loss": 1.0519, "step": 25080 }, { "epoch": 1.51, "learning_rate": 4.3586377559488786e-05, "loss": 1.026, "step": 25082 }, { "epoch": 1.51, "learning_rate": 4.358486864937456e-05, "loss": 0.8987, "step": 25084 }, { "epoch": 1.51, "learning_rate": 4.358335973926033e-05, "loss": 0.7773, "step": 25086 }, { "epoch": 1.51, "learning_rate": 4.358185082914611e-05, "loss": 0.8064, "step": 25088 }, { "epoch": 1.51, "learning_rate": 4.3580341919031885e-05, "loss": 1.4718, "step": 25090 }, { "epoch": 1.51, "learning_rate": 4.3578833008917665e-05, "loss": 1.0472, "step": 25092 }, { "epoch": 1.51, "learning_rate": 4.357732409880344e-05, "loss": 1.271, "step": 25094 }, { "epoch": 1.51, "learning_rate": 4.357581518868921e-05, "loss": 0.8931, "step": 25096 }, { "epoch": 1.51, "learning_rate": 4.357430627857499e-05, "loss": 0.6711, "step": 25098 }, { "epoch": 1.51, "learning_rate": 4.357279736846076e-05, "loss": 0.9763, "step": 25100 }, { "epoch": 1.52, "learning_rate": 4.357128845834654e-05, "loss": 0.7212, "step": 25102 }, { "epoch": 1.52, "learning_rate": 4.356977954823232e-05, "loss": 0.6375, "step": 25104 }, { "epoch": 1.52, "learning_rate": 4.356827063811809e-05, "loss": 0.8264, "step": 25106 }, { "epoch": 1.52, "learning_rate": 4.3566761728003864e-05, "loss": 1.0728, "step": 25108 }, { "epoch": 1.52, "learning_rate": 4.3565252817889643e-05, "loss": 1.3865, "step": 25110 }, { "epoch": 1.52, "learning_rate": 4.356374390777542e-05, "loss": 1.0541, "step": 25112 }, { "epoch": 1.52, "learning_rate": 4.356223499766119e-05, "loss": 1.005, "step": 25114 }, { "epoch": 1.52, "learning_rate": 4.356072608754696e-05, "loss": 1.2122, "step": 25116 }, { "epoch": 1.52, "learning_rate": 4.355921717743274e-05, "loss": 0.7318, "step": 25118 }, { "epoch": 1.52, "learning_rate": 4.355770826731852e-05, "loss": 0.8879, "step": 25120 }, { "epoch": 1.52, "learning_rate": 4.355619935720429e-05, "loss": 1.1917, "step": 25122 }, { "epoch": 1.52, "learning_rate": 4.355469044709007e-05, "loss": 0.8704, "step": 25124 }, { "epoch": 1.52, "learning_rate": 4.355318153697585e-05, "loss": 0.8537, "step": 25126 }, { "epoch": 1.52, "learning_rate": 4.355167262686162e-05, "loss": 1.1712, "step": 25128 }, { "epoch": 1.52, "learning_rate": 4.3550163716747395e-05, "loss": 0.9047, "step": 25130 }, { "epoch": 1.52, "learning_rate": 4.354865480663317e-05, "loss": 0.8514, "step": 25132 }, { "epoch": 1.52, "learning_rate": 4.354714589651895e-05, "loss": 0.6787, "step": 25134 }, { "epoch": 1.52, "learning_rate": 4.354563698640472e-05, "loss": 1.2492, "step": 25136 }, { "epoch": 1.52, "learning_rate": 4.3544128076290494e-05, "loss": 0.909, "step": 25138 }, { "epoch": 1.52, "learning_rate": 4.3542619166176274e-05, "loss": 0.7813, "step": 25140 }, { "epoch": 1.52, "learning_rate": 4.3541110256062054e-05, "loss": 0.603, "step": 25142 }, { "epoch": 1.52, "learning_rate": 4.353960134594782e-05, "loss": 0.8401, "step": 25144 }, { "epoch": 1.52, "learning_rate": 4.35380924358336e-05, "loss": 1.3439, "step": 25146 }, { "epoch": 1.52, "learning_rate": 4.3536583525719374e-05, "loss": 0.75, "step": 25148 }, { "epoch": 1.52, "learning_rate": 4.353507461560515e-05, "loss": 1.0172, "step": 25150 }, { "epoch": 1.52, "learning_rate": 4.3533565705490927e-05, "loss": 0.9003, "step": 25152 }, { "epoch": 1.52, "learning_rate": 4.35320567953767e-05, "loss": 0.9253, "step": 25154 }, { "epoch": 1.52, "learning_rate": 4.353054788526248e-05, "loss": 0.9989, "step": 25156 }, { "epoch": 1.52, "learning_rate": 4.352903897514825e-05, "loss": 0.7565, "step": 25158 }, { "epoch": 1.52, "learning_rate": 4.3527530065034026e-05, "loss": 1.3389, "step": 25160 }, { "epoch": 1.52, "learning_rate": 4.3526021154919806e-05, "loss": 0.9301, "step": 25162 }, { "epoch": 1.52, "learning_rate": 4.352451224480558e-05, "loss": 0.6412, "step": 25164 }, { "epoch": 1.52, "learning_rate": 4.352300333469135e-05, "loss": 0.9726, "step": 25166 }, { "epoch": 1.52, "learning_rate": 4.352149442457713e-05, "loss": 1.0931, "step": 25168 }, { "epoch": 1.52, "learning_rate": 4.3519985514462905e-05, "loss": 0.9236, "step": 25170 }, { "epoch": 1.52, "learning_rate": 4.351847660434868e-05, "loss": 0.7664, "step": 25172 }, { "epoch": 1.52, "learning_rate": 4.351696769423446e-05, "loss": 0.8947, "step": 25174 }, { "epoch": 1.52, "learning_rate": 4.351545878412023e-05, "loss": 0.9086, "step": 25176 }, { "epoch": 1.52, "learning_rate": 4.351394987400601e-05, "loss": 0.7667, "step": 25178 }, { "epoch": 1.52, "learning_rate": 4.3512440963891784e-05, "loss": 0.7961, "step": 25180 }, { "epoch": 1.52, "learning_rate": 4.351093205377756e-05, "loss": 1.0254, "step": 25182 }, { "epoch": 1.52, "learning_rate": 4.350942314366334e-05, "loss": 0.7451, "step": 25184 }, { "epoch": 1.52, "learning_rate": 4.350791423354911e-05, "loss": 0.8286, "step": 25186 }, { "epoch": 1.52, "learning_rate": 4.3506405323434883e-05, "loss": 0.819, "step": 25188 }, { "epoch": 1.52, "learning_rate": 4.350489641332066e-05, "loss": 0.6421, "step": 25190 }, { "epoch": 1.52, "learning_rate": 4.3503387503206437e-05, "loss": 0.848, "step": 25192 }, { "epoch": 1.52, "learning_rate": 4.350187859309221e-05, "loss": 0.8251, "step": 25194 }, { "epoch": 1.52, "learning_rate": 4.350036968297799e-05, "loss": 1.0667, "step": 25196 }, { "epoch": 1.52, "learning_rate": 4.349886077286376e-05, "loss": 1.0542, "step": 25198 }, { "epoch": 1.52, "learning_rate": 4.3497351862749536e-05, "loss": 0.7141, "step": 25200 }, { "epoch": 1.52, "learning_rate": 4.3495842952635316e-05, "loss": 1.0014, "step": 25202 }, { "epoch": 1.52, "learning_rate": 4.349433404252109e-05, "loss": 0.8472, "step": 25204 }, { "epoch": 1.52, "learning_rate": 4.349282513240687e-05, "loss": 0.685, "step": 25206 }, { "epoch": 1.52, "learning_rate": 4.349131622229264e-05, "loss": 0.7399, "step": 25208 }, { "epoch": 1.52, "learning_rate": 4.3489807312178415e-05, "loss": 0.8676, "step": 25210 }, { "epoch": 1.52, "learning_rate": 4.3488298402064195e-05, "loss": 1.1024, "step": 25212 }, { "epoch": 1.52, "learning_rate": 4.348678949194996e-05, "loss": 0.9389, "step": 25214 }, { "epoch": 1.52, "learning_rate": 4.348528058183574e-05, "loss": 0.6415, "step": 25216 }, { "epoch": 1.52, "learning_rate": 4.348377167172152e-05, "loss": 0.8096, "step": 25218 }, { "epoch": 1.52, "learning_rate": 4.3482262761607294e-05, "loss": 0.9429, "step": 25220 }, { "epoch": 1.52, "learning_rate": 4.348075385149307e-05, "loss": 0.895, "step": 25222 }, { "epoch": 1.52, "learning_rate": 4.347924494137885e-05, "loss": 0.8282, "step": 25224 }, { "epoch": 1.52, "learning_rate": 4.347773603126462e-05, "loss": 0.859, "step": 25226 }, { "epoch": 1.52, "learning_rate": 4.3476227121150393e-05, "loss": 1.2129, "step": 25228 }, { "epoch": 1.52, "learning_rate": 4.3474718211036167e-05, "loss": 0.8804, "step": 25230 }, { "epoch": 1.52, "learning_rate": 4.3473209300921946e-05, "loss": 0.8871, "step": 25232 }, { "epoch": 1.52, "learning_rate": 4.3471700390807726e-05, "loss": 1.1944, "step": 25234 }, { "epoch": 1.52, "learning_rate": 4.347019148069349e-05, "loss": 0.8482, "step": 25236 }, { "epoch": 1.52, "learning_rate": 4.346868257057927e-05, "loss": 0.7093, "step": 25238 }, { "epoch": 1.52, "learning_rate": 4.346717366046505e-05, "loss": 1.0143, "step": 25240 }, { "epoch": 1.52, "learning_rate": 4.3465664750350826e-05, "loss": 0.8445, "step": 25242 }, { "epoch": 1.52, "learning_rate": 4.34641558402366e-05, "loss": 0.9149, "step": 25244 }, { "epoch": 1.52, "learning_rate": 4.346264693012237e-05, "loss": 0.9987, "step": 25246 }, { "epoch": 1.52, "learning_rate": 4.346113802000815e-05, "loss": 0.9629, "step": 25248 }, { "epoch": 1.52, "learning_rate": 4.3459629109893925e-05, "loss": 1.0191, "step": 25250 }, { "epoch": 1.52, "learning_rate": 4.34581201997797e-05, "loss": 0.7664, "step": 25252 }, { "epoch": 1.52, "learning_rate": 4.345661128966548e-05, "loss": 0.8995, "step": 25254 }, { "epoch": 1.52, "learning_rate": 4.345510237955126e-05, "loss": 0.8353, "step": 25256 }, { "epoch": 1.52, "learning_rate": 4.3453593469437024e-05, "loss": 0.9535, "step": 25258 }, { "epoch": 1.52, "learning_rate": 4.3452084559322804e-05, "loss": 0.9732, "step": 25260 }, { "epoch": 1.52, "learning_rate": 4.345057564920858e-05, "loss": 0.9872, "step": 25262 }, { "epoch": 1.52, "learning_rate": 4.344906673909435e-05, "loss": 0.8738, "step": 25264 }, { "epoch": 1.52, "learning_rate": 4.344755782898013e-05, "loss": 0.6268, "step": 25266 }, { "epoch": 1.53, "learning_rate": 4.34460489188659e-05, "loss": 0.9094, "step": 25268 }, { "epoch": 1.53, "learning_rate": 4.344454000875168e-05, "loss": 0.5677, "step": 25270 }, { "epoch": 1.53, "learning_rate": 4.3443031098637456e-05, "loss": 0.8338, "step": 25272 }, { "epoch": 1.53, "learning_rate": 4.344152218852323e-05, "loss": 0.7846, "step": 25274 }, { "epoch": 1.53, "learning_rate": 4.344001327840901e-05, "loss": 1.1603, "step": 25276 }, { "epoch": 1.53, "learning_rate": 4.343850436829478e-05, "loss": 0.9595, "step": 25278 }, { "epoch": 1.53, "learning_rate": 4.3436995458180556e-05, "loss": 0.7896, "step": 25280 }, { "epoch": 1.53, "learning_rate": 4.3435486548066336e-05, "loss": 0.9601, "step": 25282 }, { "epoch": 1.53, "learning_rate": 4.343397763795211e-05, "loss": 0.8325, "step": 25284 }, { "epoch": 1.53, "learning_rate": 4.343246872783788e-05, "loss": 0.9118, "step": 25286 }, { "epoch": 1.53, "learning_rate": 4.343095981772366e-05, "loss": 0.8687, "step": 25288 }, { "epoch": 1.53, "learning_rate": 4.3429450907609435e-05, "loss": 0.8669, "step": 25290 }, { "epoch": 1.53, "learning_rate": 4.3427941997495215e-05, "loss": 0.9607, "step": 25292 }, { "epoch": 1.53, "learning_rate": 4.342643308738099e-05, "loss": 0.7358, "step": 25294 }, { "epoch": 1.53, "learning_rate": 4.342492417726676e-05, "loss": 0.7314, "step": 25296 }, { "epoch": 1.53, "learning_rate": 4.342341526715254e-05, "loss": 1.016, "step": 25298 }, { "epoch": 1.53, "learning_rate": 4.3421906357038314e-05, "loss": 0.8788, "step": 25300 }, { "epoch": 1.53, "learning_rate": 4.342039744692409e-05, "loss": 1.0276, "step": 25302 }, { "epoch": 1.53, "learning_rate": 4.341888853680987e-05, "loss": 0.8705, "step": 25304 }, { "epoch": 1.53, "learning_rate": 4.341737962669564e-05, "loss": 0.6583, "step": 25306 }, { "epoch": 1.53, "learning_rate": 4.341587071658141e-05, "loss": 0.7522, "step": 25308 }, { "epoch": 1.53, "learning_rate": 4.341436180646719e-05, "loss": 0.8543, "step": 25310 }, { "epoch": 1.53, "learning_rate": 4.3412852896352966e-05, "loss": 0.8552, "step": 25312 }, { "epoch": 1.53, "learning_rate": 4.341134398623874e-05, "loss": 1.4113, "step": 25314 }, { "epoch": 1.53, "learning_rate": 4.340983507612452e-05, "loss": 0.7803, "step": 25316 }, { "epoch": 1.53, "learning_rate": 4.340832616601029e-05, "loss": 0.9259, "step": 25318 }, { "epoch": 1.53, "learning_rate": 4.340681725589607e-05, "loss": 0.7993, "step": 25320 }, { "epoch": 1.53, "learning_rate": 4.3405308345781846e-05, "loss": 0.8581, "step": 25322 }, { "epoch": 1.53, "learning_rate": 4.340379943566762e-05, "loss": 0.7705, "step": 25324 }, { "epoch": 1.53, "learning_rate": 4.34022905255534e-05, "loss": 0.8039, "step": 25326 }, { "epoch": 1.53, "learning_rate": 4.3400781615439165e-05, "loss": 0.7632, "step": 25328 }, { "epoch": 1.53, "learning_rate": 4.3399272705324945e-05, "loss": 0.9344, "step": 25330 }, { "epoch": 1.53, "learning_rate": 4.3397763795210725e-05, "loss": 0.8341, "step": 25332 }, { "epoch": 1.53, "learning_rate": 4.33962548850965e-05, "loss": 0.8342, "step": 25334 }, { "epoch": 1.53, "learning_rate": 4.339474597498227e-05, "loss": 1.131, "step": 25336 }, { "epoch": 1.53, "learning_rate": 4.339323706486805e-05, "loss": 0.7397, "step": 25338 }, { "epoch": 1.53, "learning_rate": 4.3391728154753824e-05, "loss": 0.7904, "step": 25340 }, { "epoch": 1.53, "learning_rate": 4.3390219244639604e-05, "loss": 0.7911, "step": 25342 }, { "epoch": 1.53, "learning_rate": 4.338871033452537e-05, "loss": 0.9537, "step": 25344 }, { "epoch": 1.53, "learning_rate": 4.338720142441115e-05, "loss": 0.9102, "step": 25346 }, { "epoch": 1.53, "learning_rate": 4.338569251429693e-05, "loss": 0.8779, "step": 25348 }, { "epoch": 1.53, "learning_rate": 4.3384183604182696e-05, "loss": 0.7531, "step": 25350 }, { "epoch": 1.53, "learning_rate": 4.3382674694068476e-05, "loss": 0.9006, "step": 25352 }, { "epoch": 1.53, "learning_rate": 4.3381165783954256e-05, "loss": 1.2929, "step": 25354 }, { "epoch": 1.53, "learning_rate": 4.337965687384003e-05, "loss": 1.1586, "step": 25356 }, { "epoch": 1.53, "learning_rate": 4.33781479637258e-05, "loss": 0.6693, "step": 25358 }, { "epoch": 1.53, "learning_rate": 4.3376639053611576e-05, "loss": 0.6422, "step": 25360 }, { "epoch": 1.53, "learning_rate": 4.3375130143497355e-05, "loss": 0.6645, "step": 25362 }, { "epoch": 1.53, "learning_rate": 4.337362123338313e-05, "loss": 0.722, "step": 25364 }, { "epoch": 1.53, "learning_rate": 4.33721123232689e-05, "loss": 0.9736, "step": 25366 }, { "epoch": 1.53, "learning_rate": 4.337060341315468e-05, "loss": 0.8053, "step": 25368 }, { "epoch": 1.53, "learning_rate": 4.336909450304046e-05, "loss": 0.9199, "step": 25370 }, { "epoch": 1.53, "learning_rate": 4.336758559292623e-05, "loss": 1.201, "step": 25372 }, { "epoch": 1.53, "learning_rate": 4.336607668281201e-05, "loss": 1.0651, "step": 25374 }, { "epoch": 1.53, "learning_rate": 4.336456777269778e-05, "loss": 0.8586, "step": 25376 }, { "epoch": 1.53, "learning_rate": 4.3363058862583554e-05, "loss": 0.9236, "step": 25378 }, { "epoch": 1.53, "learning_rate": 4.3361549952469334e-05, "loss": 1.0343, "step": 25380 }, { "epoch": 1.53, "learning_rate": 4.336004104235511e-05, "loss": 0.7095, "step": 25382 }, { "epoch": 1.53, "learning_rate": 4.335853213224089e-05, "loss": 0.9593, "step": 25384 }, { "epoch": 1.53, "learning_rate": 4.335702322212666e-05, "loss": 0.9924, "step": 25386 }, { "epoch": 1.53, "learning_rate": 4.335551431201243e-05, "loss": 0.864, "step": 25388 }, { "epoch": 1.53, "learning_rate": 4.335400540189821e-05, "loss": 0.8971, "step": 25390 }, { "epoch": 1.53, "learning_rate": 4.3352496491783986e-05, "loss": 0.9128, "step": 25392 }, { "epoch": 1.53, "learning_rate": 4.335098758166976e-05, "loss": 0.8692, "step": 25394 }, { "epoch": 1.53, "learning_rate": 4.334947867155554e-05, "loss": 0.9514, "step": 25396 }, { "epoch": 1.53, "learning_rate": 4.334796976144131e-05, "loss": 0.9454, "step": 25398 }, { "epoch": 1.53, "learning_rate": 4.3346460851327086e-05, "loss": 1.1295, "step": 25400 }, { "epoch": 1.53, "learning_rate": 4.3344951941212865e-05, "loss": 1.1691, "step": 25402 }, { "epoch": 1.53, "learning_rate": 4.334344303109864e-05, "loss": 1.0298, "step": 25404 }, { "epoch": 1.53, "learning_rate": 4.334193412098442e-05, "loss": 1.0666, "step": 25406 }, { "epoch": 1.53, "learning_rate": 4.334042521087019e-05, "loss": 0.7263, "step": 25408 }, { "epoch": 1.53, "learning_rate": 4.3338916300755965e-05, "loss": 1.1467, "step": 25410 }, { "epoch": 1.53, "learning_rate": 4.3337407390641745e-05, "loss": 0.7192, "step": 25412 }, { "epoch": 1.53, "learning_rate": 4.333589848052752e-05, "loss": 0.7104, "step": 25414 }, { "epoch": 1.53, "learning_rate": 4.333438957041329e-05, "loss": 1.0776, "step": 25416 }, { "epoch": 1.53, "learning_rate": 4.333288066029907e-05, "loss": 1.2921, "step": 25418 }, { "epoch": 1.53, "learning_rate": 4.3331371750184844e-05, "loss": 0.725, "step": 25420 }, { "epoch": 1.53, "learning_rate": 4.332986284007062e-05, "loss": 0.804, "step": 25422 }, { "epoch": 1.53, "learning_rate": 4.33283539299564e-05, "loss": 0.987, "step": 25424 }, { "epoch": 1.53, "learning_rate": 4.332684501984217e-05, "loss": 0.7278, "step": 25426 }, { "epoch": 1.53, "learning_rate": 4.332533610972794e-05, "loss": 0.9354, "step": 25428 }, { "epoch": 1.53, "learning_rate": 4.332382719961372e-05, "loss": 0.8593, "step": 25430 }, { "epoch": 1.54, "learning_rate": 4.3322318289499496e-05, "loss": 0.8681, "step": 25432 }, { "epoch": 1.54, "learning_rate": 4.3320809379385276e-05, "loss": 0.9041, "step": 25434 }, { "epoch": 1.54, "learning_rate": 4.331930046927104e-05, "loss": 1.0062, "step": 25436 }, { "epoch": 1.54, "learning_rate": 4.331779155915682e-05, "loss": 0.8872, "step": 25438 }, { "epoch": 1.54, "learning_rate": 4.33162826490426e-05, "loss": 0.6925, "step": 25440 }, { "epoch": 1.54, "learning_rate": 4.331477373892837e-05, "loss": 0.9328, "step": 25442 }, { "epoch": 1.54, "learning_rate": 4.331326482881415e-05, "loss": 0.9673, "step": 25444 }, { "epoch": 1.54, "learning_rate": 4.331175591869993e-05, "loss": 0.8387, "step": 25446 }, { "epoch": 1.54, "learning_rate": 4.33102470085857e-05, "loss": 0.8648, "step": 25448 }, { "epoch": 1.54, "learning_rate": 4.3308738098471475e-05, "loss": 0.8194, "step": 25450 }, { "epoch": 1.54, "learning_rate": 4.3307229188357255e-05, "loss": 0.9622, "step": 25452 }, { "epoch": 1.54, "learning_rate": 4.330572027824303e-05, "loss": 0.7481, "step": 25454 }, { "epoch": 1.54, "learning_rate": 4.330421136812881e-05, "loss": 1.189, "step": 25456 }, { "epoch": 1.54, "learning_rate": 4.3302702458014574e-05, "loss": 0.769, "step": 25458 }, { "epoch": 1.54, "learning_rate": 4.3301193547900354e-05, "loss": 0.8479, "step": 25460 }, { "epoch": 1.54, "learning_rate": 4.3299684637786134e-05, "loss": 1.0046, "step": 25462 }, { "epoch": 1.54, "learning_rate": 4.32981757276719e-05, "loss": 0.8828, "step": 25464 }, { "epoch": 1.54, "learning_rate": 4.329666681755768e-05, "loss": 1.0645, "step": 25466 }, { "epoch": 1.54, "learning_rate": 4.329515790744346e-05, "loss": 1.1292, "step": 25468 }, { "epoch": 1.54, "learning_rate": 4.329364899732923e-05, "loss": 0.9257, "step": 25470 }, { "epoch": 1.54, "learning_rate": 4.3292140087215006e-05, "loss": 0.9075, "step": 25472 }, { "epoch": 1.54, "learning_rate": 4.329063117710078e-05, "loss": 1.0772, "step": 25474 }, { "epoch": 1.54, "learning_rate": 4.328912226698656e-05, "loss": 0.9126, "step": 25476 }, { "epoch": 1.54, "learning_rate": 4.328761335687233e-05, "loss": 0.7627, "step": 25478 }, { "epoch": 1.54, "learning_rate": 4.3286104446758105e-05, "loss": 0.6488, "step": 25480 }, { "epoch": 1.54, "learning_rate": 4.3284595536643885e-05, "loss": 0.9319, "step": 25482 }, { "epoch": 1.54, "learning_rate": 4.3283086626529665e-05, "loss": 1.024, "step": 25484 }, { "epoch": 1.54, "learning_rate": 4.328157771641543e-05, "loss": 1.0792, "step": 25486 }, { "epoch": 1.54, "learning_rate": 4.328006880630121e-05, "loss": 1.0046, "step": 25488 }, { "epoch": 1.54, "learning_rate": 4.3278559896186985e-05, "loss": 0.9513, "step": 25490 }, { "epoch": 1.54, "learning_rate": 4.327705098607276e-05, "loss": 0.9479, "step": 25492 }, { "epoch": 1.54, "learning_rate": 4.327554207595854e-05, "loss": 0.8718, "step": 25494 }, { "epoch": 1.54, "learning_rate": 4.327403316584431e-05, "loss": 1.0762, "step": 25496 }, { "epoch": 1.54, "learning_rate": 4.327252425573009e-05, "loss": 0.7971, "step": 25498 }, { "epoch": 1.54, "learning_rate": 4.3271015345615864e-05, "loss": 1.1491, "step": 25500 }, { "epoch": 1.54, "learning_rate": 4.326950643550164e-05, "loss": 0.8497, "step": 25502 }, { "epoch": 1.54, "learning_rate": 4.326799752538742e-05, "loss": 0.6256, "step": 25504 }, { "epoch": 1.54, "learning_rate": 4.326648861527319e-05, "loss": 0.746, "step": 25506 }, { "epoch": 1.54, "learning_rate": 4.326497970515896e-05, "loss": 0.793, "step": 25508 }, { "epoch": 1.54, "learning_rate": 4.326347079504474e-05, "loss": 0.8353, "step": 25510 }, { "epoch": 1.54, "learning_rate": 4.3261961884930516e-05, "loss": 0.9812, "step": 25512 }, { "epoch": 1.54, "learning_rate": 4.326045297481629e-05, "loss": 1.0907, "step": 25514 }, { "epoch": 1.54, "learning_rate": 4.325894406470207e-05, "loss": 0.709, "step": 25516 }, { "epoch": 1.54, "learning_rate": 4.325743515458784e-05, "loss": 1.4602, "step": 25518 }, { "epoch": 1.54, "learning_rate": 4.325592624447362e-05, "loss": 0.8945, "step": 25520 }, { "epoch": 1.54, "learning_rate": 4.3254417334359395e-05, "loss": 0.8568, "step": 25522 }, { "epoch": 1.54, "learning_rate": 4.325290842424517e-05, "loss": 0.5338, "step": 25524 }, { "epoch": 1.54, "learning_rate": 4.325139951413095e-05, "loss": 0.7333, "step": 25526 }, { "epoch": 1.54, "learning_rate": 4.324989060401672e-05, "loss": 0.9663, "step": 25528 }, { "epoch": 1.54, "learning_rate": 4.3248381693902495e-05, "loss": 0.8778, "step": 25530 }, { "epoch": 1.54, "learning_rate": 4.3246872783788274e-05, "loss": 0.7234, "step": 25532 }, { "epoch": 1.54, "learning_rate": 4.324536387367405e-05, "loss": 1.0832, "step": 25534 }, { "epoch": 1.54, "learning_rate": 4.324385496355982e-05, "loss": 0.8185, "step": 25536 }, { "epoch": 1.54, "learning_rate": 4.32423460534456e-05, "loss": 0.5968, "step": 25538 }, { "epoch": 1.54, "learning_rate": 4.3240837143331374e-05, "loss": 0.8303, "step": 25540 }, { "epoch": 1.54, "learning_rate": 4.323932823321715e-05, "loss": 0.735, "step": 25542 }, { "epoch": 1.54, "learning_rate": 4.323781932310293e-05, "loss": 1.0753, "step": 25544 }, { "epoch": 1.54, "learning_rate": 4.32363104129887e-05, "loss": 0.9614, "step": 25546 }, { "epoch": 1.54, "learning_rate": 4.323480150287448e-05, "loss": 0.7924, "step": 25548 }, { "epoch": 1.54, "learning_rate": 4.3233292592760246e-05, "loss": 0.6837, "step": 25550 }, { "epoch": 1.54, "learning_rate": 4.3231783682646026e-05, "loss": 0.7533, "step": 25552 }, { "epoch": 1.54, "learning_rate": 4.3230274772531806e-05, "loss": 0.5605, "step": 25554 }, { "epoch": 1.54, "learning_rate": 4.322876586241757e-05, "loss": 0.9364, "step": 25556 }, { "epoch": 1.54, "learning_rate": 4.322725695230335e-05, "loss": 0.9932, "step": 25558 }, { "epoch": 1.54, "learning_rate": 4.322574804218913e-05, "loss": 1.0605, "step": 25560 }, { "epoch": 1.54, "learning_rate": 4.3224239132074905e-05, "loss": 1.026, "step": 25562 }, { "epoch": 1.54, "learning_rate": 4.322273022196068e-05, "loss": 0.7333, "step": 25564 }, { "epoch": 1.54, "learning_rate": 4.322122131184646e-05, "loss": 0.833, "step": 25566 }, { "epoch": 1.54, "learning_rate": 4.321971240173223e-05, "loss": 1.1862, "step": 25568 }, { "epoch": 1.54, "learning_rate": 4.321820349161801e-05, "loss": 0.7116, "step": 25570 }, { "epoch": 1.54, "learning_rate": 4.321669458150378e-05, "loss": 0.7409, "step": 25572 }, { "epoch": 1.54, "learning_rate": 4.321518567138956e-05, "loss": 0.6557, "step": 25574 }, { "epoch": 1.54, "learning_rate": 4.321367676127534e-05, "loss": 0.4509, "step": 25576 }, { "epoch": 1.54, "learning_rate": 4.3212167851161104e-05, "loss": 1.0425, "step": 25578 }, { "epoch": 1.54, "learning_rate": 4.3210658941046884e-05, "loss": 0.9367, "step": 25580 }, { "epoch": 1.54, "learning_rate": 4.3209150030932664e-05, "loss": 1.2579, "step": 25582 }, { "epoch": 1.54, "learning_rate": 4.320764112081844e-05, "loss": 0.945, "step": 25584 }, { "epoch": 1.54, "learning_rate": 4.320613221070421e-05, "loss": 1.1243, "step": 25586 }, { "epoch": 1.54, "learning_rate": 4.320462330058998e-05, "loss": 1.0109, "step": 25588 }, { "epoch": 1.54, "learning_rate": 4.320311439047576e-05, "loss": 0.68, "step": 25590 }, { "epoch": 1.54, "learning_rate": 4.3201605480361536e-05, "loss": 0.7195, "step": 25592 }, { "epoch": 1.54, "learning_rate": 4.320009657024731e-05, "loss": 1.0902, "step": 25594 }, { "epoch": 1.54, "learning_rate": 4.319858766013309e-05, "loss": 0.9907, "step": 25596 }, { "epoch": 1.55, "learning_rate": 4.319707875001887e-05, "loss": 1.0131, "step": 25598 }, { "epoch": 1.55, "learning_rate": 4.3195569839904635e-05, "loss": 1.0232, "step": 25600 }, { "epoch": 1.55, "learning_rate": 4.3194060929790415e-05, "loss": 0.7, "step": 25602 }, { "epoch": 1.55, "learning_rate": 4.319255201967619e-05, "loss": 0.9449, "step": 25604 }, { "epoch": 1.55, "learning_rate": 4.319104310956196e-05, "loss": 0.5632, "step": 25606 }, { "epoch": 1.55, "learning_rate": 4.318953419944774e-05, "loss": 0.9855, "step": 25608 }, { "epoch": 1.55, "learning_rate": 4.3188025289333514e-05, "loss": 1.0036, "step": 25610 }, { "epoch": 1.55, "learning_rate": 4.3186516379219294e-05, "loss": 1.0468, "step": 25612 }, { "epoch": 1.55, "learning_rate": 4.318500746910507e-05, "loss": 0.8258, "step": 25614 }, { "epoch": 1.55, "learning_rate": 4.318349855899084e-05, "loss": 1.3519, "step": 25616 }, { "epoch": 1.55, "learning_rate": 4.318198964887662e-05, "loss": 0.8656, "step": 25618 }, { "epoch": 1.55, "learning_rate": 4.3180480738762394e-05, "loss": 1.1239, "step": 25620 }, { "epoch": 1.55, "learning_rate": 4.317897182864817e-05, "loss": 0.9895, "step": 25622 }, { "epoch": 1.55, "learning_rate": 4.3177462918533947e-05, "loss": 0.8687, "step": 25624 }, { "epoch": 1.55, "learning_rate": 4.317595400841972e-05, "loss": 0.6546, "step": 25626 }, { "epoch": 1.55, "learning_rate": 4.317444509830549e-05, "loss": 0.8783, "step": 25628 }, { "epoch": 1.55, "learning_rate": 4.317293618819127e-05, "loss": 0.8207, "step": 25630 }, { "epoch": 1.55, "learning_rate": 4.3171427278077046e-05, "loss": 0.8833, "step": 25632 }, { "epoch": 1.55, "learning_rate": 4.3169918367962826e-05, "loss": 1.3168, "step": 25634 }, { "epoch": 1.55, "learning_rate": 4.31684094578486e-05, "loss": 0.7906, "step": 25636 }, { "epoch": 1.55, "learning_rate": 4.316690054773437e-05, "loss": 0.7056, "step": 25638 }, { "epoch": 1.55, "learning_rate": 4.316539163762015e-05, "loss": 0.8521, "step": 25640 }, { "epoch": 1.55, "learning_rate": 4.3163882727505925e-05, "loss": 0.8077, "step": 25642 }, { "epoch": 1.55, "learning_rate": 4.31623738173917e-05, "loss": 1.0906, "step": 25644 }, { "epoch": 1.55, "learning_rate": 4.316086490727748e-05, "loss": 0.8507, "step": 25646 }, { "epoch": 1.55, "learning_rate": 4.315935599716325e-05, "loss": 1.0685, "step": 25648 }, { "epoch": 1.55, "learning_rate": 4.3157847087049024e-05, "loss": 0.8877, "step": 25650 }, { "epoch": 1.55, "learning_rate": 4.3156338176934804e-05, "loss": 1.0068, "step": 25652 }, { "epoch": 1.55, "learning_rate": 4.315482926682058e-05, "loss": 1.0609, "step": 25654 }, { "epoch": 1.55, "learning_rate": 4.315332035670635e-05, "loss": 0.7502, "step": 25656 }, { "epoch": 1.55, "learning_rate": 4.315181144659213e-05, "loss": 0.9965, "step": 25658 }, { "epoch": 1.55, "learning_rate": 4.3150302536477904e-05, "loss": 1.0309, "step": 25660 }, { "epoch": 1.55, "learning_rate": 4.3148793626363683e-05, "loss": 0.5067, "step": 25662 }, { "epoch": 1.55, "learning_rate": 4.314728471624945e-05, "loss": 0.8816, "step": 25664 }, { "epoch": 1.55, "learning_rate": 4.314577580613523e-05, "loss": 0.9629, "step": 25666 }, { "epoch": 1.55, "learning_rate": 4.314426689602101e-05, "loss": 0.7222, "step": 25668 }, { "epoch": 1.55, "learning_rate": 4.314275798590678e-05, "loss": 0.8759, "step": 25670 }, { "epoch": 1.55, "learning_rate": 4.3141249075792556e-05, "loss": 0.5964, "step": 25672 }, { "epoch": 1.55, "learning_rate": 4.3139740165678336e-05, "loss": 0.7567, "step": 25674 }, { "epoch": 1.55, "learning_rate": 4.313823125556411e-05, "loss": 0.7193, "step": 25676 }, { "epoch": 1.55, "learning_rate": 4.313672234544988e-05, "loss": 0.9523, "step": 25678 }, { "epoch": 1.55, "learning_rate": 4.313521343533566e-05, "loss": 0.8731, "step": 25680 }, { "epoch": 1.55, "learning_rate": 4.3133704525221435e-05, "loss": 0.727, "step": 25682 }, { "epoch": 1.55, "learning_rate": 4.3132195615107215e-05, "loss": 0.9896, "step": 25684 }, { "epoch": 1.55, "learning_rate": 4.313068670499298e-05, "loss": 0.9592, "step": 25686 }, { "epoch": 1.55, "learning_rate": 4.312917779487876e-05, "loss": 0.6406, "step": 25688 }, { "epoch": 1.55, "learning_rate": 4.312766888476454e-05, "loss": 0.8469, "step": 25690 }, { "epoch": 1.55, "learning_rate": 4.312615997465031e-05, "loss": 1.2424, "step": 25692 }, { "epoch": 1.55, "learning_rate": 4.312465106453609e-05, "loss": 0.7933, "step": 25694 }, { "epoch": 1.55, "learning_rate": 4.312314215442187e-05, "loss": 0.71, "step": 25696 }, { "epoch": 1.55, "learning_rate": 4.312163324430764e-05, "loss": 0.9467, "step": 25698 }, { "epoch": 1.55, "learning_rate": 4.3120124334193413e-05, "loss": 0.8103, "step": 25700 }, { "epoch": 1.55, "learning_rate": 4.3118615424079187e-05, "loss": 0.9068, "step": 25702 }, { "epoch": 1.55, "learning_rate": 4.3117106513964966e-05, "loss": 0.9599, "step": 25704 }, { "epoch": 1.55, "learning_rate": 4.311559760385074e-05, "loss": 0.9211, "step": 25706 }, { "epoch": 1.55, "learning_rate": 4.311408869373651e-05, "loss": 1.2461, "step": 25708 }, { "epoch": 1.55, "learning_rate": 4.311257978362229e-05, "loss": 0.9834, "step": 25710 }, { "epoch": 1.55, "learning_rate": 4.311107087350807e-05, "loss": 0.7482, "step": 25712 }, { "epoch": 1.55, "learning_rate": 4.310956196339384e-05, "loss": 0.9604, "step": 25714 }, { "epoch": 1.55, "learning_rate": 4.310805305327962e-05, "loss": 0.6075, "step": 25716 }, { "epoch": 1.55, "learning_rate": 4.310654414316539e-05, "loss": 0.9718, "step": 25718 }, { "epoch": 1.55, "learning_rate": 4.310503523305117e-05, "loss": 0.6838, "step": 25720 }, { "epoch": 1.55, "learning_rate": 4.3103526322936945e-05, "loss": 0.6407, "step": 25722 }, { "epoch": 1.55, "learning_rate": 4.310201741282272e-05, "loss": 0.7527, "step": 25724 }, { "epoch": 1.55, "learning_rate": 4.31005085027085e-05, "loss": 1.07, "step": 25726 }, { "epoch": 1.55, "learning_rate": 4.309899959259427e-05, "loss": 0.7619, "step": 25728 }, { "epoch": 1.55, "learning_rate": 4.3097490682480044e-05, "loss": 1.0181, "step": 25730 }, { "epoch": 1.55, "learning_rate": 4.3095981772365824e-05, "loss": 0.6805, "step": 25732 }, { "epoch": 1.55, "learning_rate": 4.30944728622516e-05, "loss": 0.7154, "step": 25734 }, { "epoch": 1.55, "learning_rate": 4.309296395213737e-05, "loss": 0.8219, "step": 25736 }, { "epoch": 1.55, "learning_rate": 4.309145504202315e-05, "loss": 1.2446, "step": 25738 }, { "epoch": 1.55, "learning_rate": 4.3089946131908923e-05, "loss": 1.0041, "step": 25740 }, { "epoch": 1.55, "learning_rate": 4.3088437221794697e-05, "loss": 0.8688, "step": 25742 }, { "epoch": 1.55, "learning_rate": 4.3086928311680476e-05, "loss": 0.7281, "step": 25744 }, { "epoch": 1.55, "learning_rate": 4.308541940156625e-05, "loss": 0.9173, "step": 25746 }, { "epoch": 1.55, "learning_rate": 4.308391049145203e-05, "loss": 0.8896, "step": 25748 }, { "epoch": 1.55, "learning_rate": 4.30824015813378e-05, "loss": 0.8848, "step": 25750 }, { "epoch": 1.55, "learning_rate": 4.3080892671223576e-05, "loss": 1.1449, "step": 25752 }, { "epoch": 1.55, "learning_rate": 4.3079383761109356e-05, "loss": 0.8913, "step": 25754 }, { "epoch": 1.55, "learning_rate": 4.307787485099513e-05, "loss": 0.7444, "step": 25756 }, { "epoch": 1.55, "learning_rate": 4.30763659408809e-05, "loss": 0.8661, "step": 25758 }, { "epoch": 1.55, "learning_rate": 4.307485703076668e-05, "loss": 0.8676, "step": 25760 }, { "epoch": 1.55, "learning_rate": 4.3073348120652455e-05, "loss": 1.1551, "step": 25762 }, { "epoch": 1.56, "learning_rate": 4.307183921053823e-05, "loss": 0.9584, "step": 25764 }, { "epoch": 1.56, "learning_rate": 4.307033030042401e-05, "loss": 1.2812, "step": 25766 }, { "epoch": 1.56, "learning_rate": 4.306882139030978e-05, "loss": 0.7751, "step": 25768 }, { "epoch": 1.56, "learning_rate": 4.3067312480195554e-05, "loss": 0.622, "step": 25770 }, { "epoch": 1.56, "learning_rate": 4.3065803570081334e-05, "loss": 0.9671, "step": 25772 }, { "epoch": 1.56, "learning_rate": 4.306429465996711e-05, "loss": 1.1126, "step": 25774 }, { "epoch": 1.56, "learning_rate": 4.306278574985289e-05, "loss": 0.8489, "step": 25776 }, { "epoch": 1.56, "learning_rate": 4.3061276839738653e-05, "loss": 0.8973, "step": 25778 }, { "epoch": 1.56, "learning_rate": 4.305976792962443e-05, "loss": 0.7216, "step": 25780 }, { "epoch": 1.56, "learning_rate": 4.305825901951021e-05, "loss": 1.004, "step": 25782 }, { "epoch": 1.56, "learning_rate": 4.3056750109395986e-05, "loss": 1.1585, "step": 25784 }, { "epoch": 1.56, "learning_rate": 4.305524119928176e-05, "loss": 0.7938, "step": 25786 }, { "epoch": 1.56, "learning_rate": 4.305373228916754e-05, "loss": 0.8539, "step": 25788 }, { "epoch": 1.56, "learning_rate": 4.305222337905331e-05, "loss": 0.7995, "step": 25790 }, { "epoch": 1.56, "learning_rate": 4.3050714468939086e-05, "loss": 1.2157, "step": 25792 }, { "epoch": 1.56, "learning_rate": 4.304920555882486e-05, "loss": 1.0833, "step": 25794 }, { "epoch": 1.56, "learning_rate": 4.304769664871064e-05, "loss": 1.0022, "step": 25796 }, { "epoch": 1.56, "learning_rate": 4.304618773859642e-05, "loss": 0.9782, "step": 25798 }, { "epoch": 1.56, "learning_rate": 4.3044678828482185e-05, "loss": 0.8676, "step": 25800 }, { "epoch": 1.56, "learning_rate": 4.3043169918367965e-05, "loss": 0.7229, "step": 25802 }, { "epoch": 1.56, "learning_rate": 4.3041661008253745e-05, "loss": 0.8404, "step": 25804 }, { "epoch": 1.56, "learning_rate": 4.304015209813951e-05, "loss": 0.6498, "step": 25806 }, { "epoch": 1.56, "learning_rate": 4.303864318802529e-05, "loss": 0.9362, "step": 25808 }, { "epoch": 1.56, "learning_rate": 4.303713427791107e-05, "loss": 0.6683, "step": 25810 }, { "epoch": 1.56, "learning_rate": 4.3035625367796844e-05, "loss": 1.0783, "step": 25812 }, { "epoch": 1.56, "learning_rate": 4.303411645768262e-05, "loss": 0.7266, "step": 25814 }, { "epoch": 1.56, "learning_rate": 4.303260754756839e-05, "loss": 0.9697, "step": 25816 }, { "epoch": 1.56, "learning_rate": 4.303109863745417e-05, "loss": 0.8632, "step": 25818 }, { "epoch": 1.56, "learning_rate": 4.302958972733994e-05, "loss": 0.9304, "step": 25820 }, { "epoch": 1.56, "learning_rate": 4.3028080817225716e-05, "loss": 1.1056, "step": 25822 }, { "epoch": 1.56, "learning_rate": 4.3026571907111496e-05, "loss": 0.939, "step": 25824 }, { "epoch": 1.56, "learning_rate": 4.3025062996997276e-05, "loss": 1.0016, "step": 25826 }, { "epoch": 1.56, "learning_rate": 4.302355408688304e-05, "loss": 0.6661, "step": 25828 }, { "epoch": 1.56, "learning_rate": 4.302204517676882e-05, "loss": 0.694, "step": 25830 }, { "epoch": 1.56, "learning_rate": 4.3020536266654596e-05, "loss": 0.7581, "step": 25832 }, { "epoch": 1.56, "learning_rate": 4.3019027356540376e-05, "loss": 0.7758, "step": 25834 }, { "epoch": 1.56, "learning_rate": 4.301751844642615e-05, "loss": 1.0626, "step": 25836 }, { "epoch": 1.56, "learning_rate": 4.301600953631192e-05, "loss": 0.8242, "step": 25838 }, { "epoch": 1.56, "learning_rate": 4.30145006261977e-05, "loss": 0.7567, "step": 25840 }, { "epoch": 1.56, "learning_rate": 4.3012991716083475e-05, "loss": 0.8413, "step": 25842 }, { "epoch": 1.56, "learning_rate": 4.301148280596925e-05, "loss": 1.0529, "step": 25844 }, { "epoch": 1.56, "learning_rate": 4.300997389585503e-05, "loss": 0.8256, "step": 25846 }, { "epoch": 1.56, "learning_rate": 4.30084649857408e-05, "loss": 0.6951, "step": 25848 }, { "epoch": 1.56, "learning_rate": 4.3006956075626574e-05, "loss": 0.7938, "step": 25850 }, { "epoch": 1.56, "learning_rate": 4.3005447165512354e-05, "loss": 0.7733, "step": 25852 }, { "epoch": 1.56, "learning_rate": 4.300393825539813e-05, "loss": 0.8811, "step": 25854 }, { "epoch": 1.56, "learning_rate": 4.30024293452839e-05, "loss": 1.0561, "step": 25856 }, { "epoch": 1.56, "learning_rate": 4.300092043516968e-05, "loss": 0.8404, "step": 25858 }, { "epoch": 1.56, "learning_rate": 4.299941152505545e-05, "loss": 0.6706, "step": 25860 }, { "epoch": 1.56, "learning_rate": 4.299790261494123e-05, "loss": 1.0174, "step": 25862 }, { "epoch": 1.56, "learning_rate": 4.2996393704827006e-05, "loss": 1.1953, "step": 25864 }, { "epoch": 1.56, "learning_rate": 4.299488479471278e-05, "loss": 0.8581, "step": 25866 }, { "epoch": 1.56, "learning_rate": 4.299337588459856e-05, "loss": 1.1368, "step": 25868 }, { "epoch": 1.56, "learning_rate": 4.299186697448433e-05, "loss": 0.9052, "step": 25870 }, { "epoch": 1.56, "learning_rate": 4.2990358064370106e-05, "loss": 0.8187, "step": 25872 }, { "epoch": 1.56, "learning_rate": 4.2988849154255885e-05, "loss": 0.7742, "step": 25874 }, { "epoch": 1.56, "learning_rate": 4.298734024414166e-05, "loss": 0.7919, "step": 25876 }, { "epoch": 1.56, "learning_rate": 4.298583133402743e-05, "loss": 0.6669, "step": 25878 }, { "epoch": 1.56, "learning_rate": 4.298432242391321e-05, "loss": 0.8216, "step": 25880 }, { "epoch": 1.56, "learning_rate": 4.2982813513798985e-05, "loss": 0.9894, "step": 25882 }, { "epoch": 1.56, "learning_rate": 4.2981304603684765e-05, "loss": 0.756, "step": 25884 }, { "epoch": 1.56, "learning_rate": 4.297979569357054e-05, "loss": 0.9658, "step": 25886 }, { "epoch": 1.56, "learning_rate": 4.297828678345631e-05, "loss": 1.0115, "step": 25888 }, { "epoch": 1.56, "learning_rate": 4.297677787334209e-05, "loss": 0.799, "step": 25890 }, { "epoch": 1.56, "learning_rate": 4.297526896322786e-05, "loss": 0.6352, "step": 25892 }, { "epoch": 1.56, "learning_rate": 4.297376005311364e-05, "loss": 0.6161, "step": 25894 }, { "epoch": 1.56, "learning_rate": 4.297225114299942e-05, "loss": 0.9926, "step": 25896 }, { "epoch": 1.56, "learning_rate": 4.297074223288519e-05, "loss": 0.8372, "step": 25898 }, { "epoch": 1.56, "learning_rate": 4.296923332277096e-05, "loss": 0.5838, "step": 25900 }, { "epoch": 1.56, "learning_rate": 4.296772441265674e-05, "loss": 0.8752, "step": 25902 }, { "epoch": 1.56, "learning_rate": 4.2966215502542516e-05, "loss": 0.8053, "step": 25904 }, { "epoch": 1.56, "learning_rate": 4.296470659242829e-05, "loss": 0.64, "step": 25906 }, { "epoch": 1.56, "learning_rate": 4.296319768231406e-05, "loss": 1.0137, "step": 25908 }, { "epoch": 1.56, "learning_rate": 4.296168877219984e-05, "loss": 0.862, "step": 25910 }, { "epoch": 1.56, "learning_rate": 4.296017986208562e-05, "loss": 0.8725, "step": 25912 }, { "epoch": 1.56, "learning_rate": 4.295867095197139e-05, "loss": 0.9325, "step": 25914 }, { "epoch": 1.56, "learning_rate": 4.295716204185717e-05, "loss": 1.0582, "step": 25916 }, { "epoch": 1.56, "learning_rate": 4.295565313174295e-05, "loss": 0.6525, "step": 25918 }, { "epoch": 1.56, "learning_rate": 4.2954144221628715e-05, "loss": 1.1947, "step": 25920 }, { "epoch": 1.56, "learning_rate": 4.2952635311514495e-05, "loss": 0.8297, "step": 25922 }, { "epoch": 1.56, "learning_rate": 4.2951126401400275e-05, "loss": 1.0289, "step": 25924 }, { "epoch": 1.56, "learning_rate": 4.294961749128605e-05, "loss": 0.8918, "step": 25926 }, { "epoch": 1.56, "learning_rate": 4.294810858117182e-05, "loss": 0.97, "step": 25928 }, { "epoch": 1.57, "learning_rate": 4.2946599671057594e-05, "loss": 0.779, "step": 25930 }, { "epoch": 1.57, "learning_rate": 4.2945090760943374e-05, "loss": 0.8969, "step": 25932 }, { "epoch": 1.57, "learning_rate": 4.294358185082915e-05, "loss": 0.6402, "step": 25934 }, { "epoch": 1.57, "learning_rate": 4.294207294071492e-05, "loss": 0.6254, "step": 25936 }, { "epoch": 1.57, "learning_rate": 4.29405640306007e-05, "loss": 0.7907, "step": 25938 }, { "epoch": 1.57, "learning_rate": 4.293905512048648e-05, "loss": 0.9692, "step": 25940 }, { "epoch": 1.57, "learning_rate": 4.2937546210372246e-05, "loss": 0.7534, "step": 25942 }, { "epoch": 1.57, "learning_rate": 4.2936037300258026e-05, "loss": 0.8076, "step": 25944 }, { "epoch": 1.57, "learning_rate": 4.29345283901438e-05, "loss": 0.7295, "step": 25946 }, { "epoch": 1.57, "learning_rate": 4.293301948002958e-05, "loss": 1.0937, "step": 25948 }, { "epoch": 1.57, "learning_rate": 4.293151056991535e-05, "loss": 0.6785, "step": 25950 }, { "epoch": 1.57, "learning_rate": 4.2930001659801125e-05, "loss": 0.8263, "step": 25952 }, { "epoch": 1.57, "learning_rate": 4.2928492749686905e-05, "loss": 0.8146, "step": 25954 }, { "epoch": 1.57, "learning_rate": 4.292698383957268e-05, "loss": 0.7818, "step": 25956 }, { "epoch": 1.57, "learning_rate": 4.292547492945845e-05, "loss": 1.1341, "step": 25958 }, { "epoch": 1.57, "learning_rate": 4.292396601934423e-05, "loss": 1.0469, "step": 25960 }, { "epoch": 1.57, "learning_rate": 4.2922457109230005e-05, "loss": 1.2732, "step": 25962 }, { "epoch": 1.57, "learning_rate": 4.292094819911578e-05, "loss": 1.0411, "step": 25964 }, { "epoch": 1.57, "learning_rate": 4.291943928900156e-05, "loss": 0.775, "step": 25966 }, { "epoch": 1.57, "learning_rate": 4.291793037888733e-05, "loss": 1.3303, "step": 25968 }, { "epoch": 1.57, "learning_rate": 4.2916421468773104e-05, "loss": 0.95, "step": 25970 }, { "epoch": 1.57, "learning_rate": 4.2914912558658884e-05, "loss": 0.7511, "step": 25972 }, { "epoch": 1.57, "learning_rate": 4.291340364854466e-05, "loss": 0.8335, "step": 25974 }, { "epoch": 1.57, "learning_rate": 4.291189473843044e-05, "loss": 0.9345, "step": 25976 }, { "epoch": 1.57, "learning_rate": 4.291038582831621e-05, "loss": 0.5354, "step": 25978 }, { "epoch": 1.57, "learning_rate": 4.290887691820198e-05, "loss": 0.6633, "step": 25980 }, { "epoch": 1.57, "learning_rate": 4.290736800808776e-05, "loss": 0.7584, "step": 25982 }, { "epoch": 1.57, "learning_rate": 4.2905859097973536e-05, "loss": 0.7952, "step": 25984 }, { "epoch": 1.57, "learning_rate": 4.290435018785931e-05, "loss": 0.7545, "step": 25986 }, { "epoch": 1.57, "learning_rate": 4.290284127774509e-05, "loss": 0.7993, "step": 25988 }, { "epoch": 1.57, "learning_rate": 4.290133236763086e-05, "loss": 0.7922, "step": 25990 }, { "epoch": 1.57, "learning_rate": 4.2899823457516635e-05, "loss": 0.9385, "step": 25992 }, { "epoch": 1.57, "learning_rate": 4.2898314547402415e-05, "loss": 0.9459, "step": 25994 }, { "epoch": 1.57, "learning_rate": 4.289680563728819e-05, "loss": 0.8664, "step": 25996 }, { "epoch": 1.57, "learning_rate": 4.289529672717397e-05, "loss": 0.9166, "step": 25998 }, { "epoch": 1.57, "learning_rate": 4.289378781705974e-05, "loss": 0.976, "step": 26000 }, { "epoch": 1.57, "learning_rate": 4.2892278906945515e-05, "loss": 0.8027, "step": 26002 }, { "epoch": 1.57, "learning_rate": 4.2890769996831294e-05, "loss": 0.7535, "step": 26004 }, { "epoch": 1.57, "learning_rate": 4.288926108671706e-05, "loss": 0.8997, "step": 26006 }, { "epoch": 1.57, "learning_rate": 4.288775217660284e-05, "loss": 0.9551, "step": 26008 }, { "epoch": 1.57, "learning_rate": 4.288624326648862e-05, "loss": 0.7873, "step": 26010 }, { "epoch": 1.57, "learning_rate": 4.2884734356374394e-05, "loss": 0.7213, "step": 26012 }, { "epoch": 1.57, "learning_rate": 4.288322544626017e-05, "loss": 0.8726, "step": 26014 }, { "epoch": 1.57, "learning_rate": 4.288171653614595e-05, "loss": 0.8543, "step": 26016 }, { "epoch": 1.57, "learning_rate": 4.288020762603172e-05, "loss": 0.8862, "step": 26018 }, { "epoch": 1.57, "learning_rate": 4.287869871591749e-05, "loss": 0.9413, "step": 26020 }, { "epoch": 1.57, "learning_rate": 4.2877189805803266e-05, "loss": 0.9295, "step": 26022 }, { "epoch": 1.57, "learning_rate": 4.2875680895689046e-05, "loss": 0.7646, "step": 26024 }, { "epoch": 1.57, "learning_rate": 4.2874171985574826e-05, "loss": 0.7494, "step": 26026 }, { "epoch": 1.57, "learning_rate": 4.287266307546059e-05, "loss": 0.7971, "step": 26028 }, { "epoch": 1.57, "learning_rate": 4.287115416534637e-05, "loss": 0.628, "step": 26030 }, { "epoch": 1.57, "learning_rate": 4.286964525523215e-05, "loss": 0.8015, "step": 26032 }, { "epoch": 1.57, "learning_rate": 4.286813634511792e-05, "loss": 0.8413, "step": 26034 }, { "epoch": 1.57, "learning_rate": 4.28666274350037e-05, "loss": 0.7032, "step": 26036 }, { "epoch": 1.57, "learning_rate": 4.286511852488948e-05, "loss": 0.8143, "step": 26038 }, { "epoch": 1.57, "learning_rate": 4.286360961477525e-05, "loss": 0.9394, "step": 26040 }, { "epoch": 1.57, "learning_rate": 4.2862100704661025e-05, "loss": 0.91, "step": 26042 }, { "epoch": 1.57, "learning_rate": 4.28605917945468e-05, "loss": 0.9722, "step": 26044 }, { "epoch": 1.57, "learning_rate": 4.285908288443258e-05, "loss": 1.0108, "step": 26046 }, { "epoch": 1.57, "learning_rate": 4.285757397431836e-05, "loss": 0.7512, "step": 26048 }, { "epoch": 1.57, "learning_rate": 4.2856065064204124e-05, "loss": 0.9775, "step": 26050 }, { "epoch": 1.57, "learning_rate": 4.2854556154089904e-05, "loss": 0.7694, "step": 26052 }, { "epoch": 1.57, "learning_rate": 4.2853047243975684e-05, "loss": 0.939, "step": 26054 }, { "epoch": 1.57, "learning_rate": 4.285153833386145e-05, "loss": 0.893, "step": 26056 }, { "epoch": 1.57, "learning_rate": 4.285002942374723e-05, "loss": 0.8424, "step": 26058 }, { "epoch": 1.57, "learning_rate": 4.2848520513633e-05, "loss": 0.7991, "step": 26060 }, { "epoch": 1.57, "learning_rate": 4.284701160351878e-05, "loss": 1.0492, "step": 26062 }, { "epoch": 1.57, "learning_rate": 4.2845502693404556e-05, "loss": 1.041, "step": 26064 }, { "epoch": 1.57, "learning_rate": 4.284399378329033e-05, "loss": 0.9555, "step": 26066 }, { "epoch": 1.57, "learning_rate": 4.284248487317611e-05, "loss": 0.6845, "step": 26068 }, { "epoch": 1.57, "learning_rate": 4.284097596306188e-05, "loss": 0.7081, "step": 26070 }, { "epoch": 1.57, "learning_rate": 4.2839467052947655e-05, "loss": 0.8772, "step": 26072 }, { "epoch": 1.57, "learning_rate": 4.2837958142833435e-05, "loss": 0.9472, "step": 26074 }, { "epoch": 1.57, "learning_rate": 4.283644923271921e-05, "loss": 0.8586, "step": 26076 }, { "epoch": 1.57, "learning_rate": 4.283494032260498e-05, "loss": 0.9412, "step": 26078 }, { "epoch": 1.57, "learning_rate": 4.283343141249076e-05, "loss": 0.6528, "step": 26080 }, { "epoch": 1.57, "learning_rate": 4.2831922502376534e-05, "loss": 0.6486, "step": 26082 }, { "epoch": 1.57, "learning_rate": 4.283041359226231e-05, "loss": 0.6568, "step": 26084 }, { "epoch": 1.57, "learning_rate": 4.282890468214809e-05, "loss": 0.9016, "step": 26086 }, { "epoch": 1.57, "learning_rate": 4.282739577203386e-05, "loss": 0.7762, "step": 26088 }, { "epoch": 1.57, "learning_rate": 4.282588686191964e-05, "loss": 0.9194, "step": 26090 }, { "epoch": 1.57, "learning_rate": 4.2824377951805414e-05, "loss": 1.4046, "step": 26092 }, { "epoch": 1.57, "learning_rate": 4.282286904169119e-05, "loss": 1.1488, "step": 26094 }, { "epoch": 1.58, "learning_rate": 4.282136013157697e-05, "loss": 0.8141, "step": 26096 }, { "epoch": 1.58, "learning_rate": 4.281985122146274e-05, "loss": 0.9792, "step": 26098 }, { "epoch": 1.58, "learning_rate": 4.281834231134851e-05, "loss": 0.8946, "step": 26100 }, { "epoch": 1.58, "learning_rate": 4.281683340123429e-05, "loss": 0.9849, "step": 26102 }, { "epoch": 1.58, "learning_rate": 4.2815324491120066e-05, "loss": 1.0523, "step": 26104 }, { "epoch": 1.58, "learning_rate": 4.281381558100584e-05, "loss": 0.9129, "step": 26106 }, { "epoch": 1.58, "learning_rate": 4.281230667089162e-05, "loss": 1.2046, "step": 26108 }, { "epoch": 1.58, "learning_rate": 4.281079776077739e-05, "loss": 0.9143, "step": 26110 }, { "epoch": 1.58, "learning_rate": 4.280928885066317e-05, "loss": 0.9685, "step": 26112 }, { "epoch": 1.58, "learning_rate": 4.2807779940548945e-05, "loss": 0.8719, "step": 26114 }, { "epoch": 1.58, "learning_rate": 4.280627103043472e-05, "loss": 0.8189, "step": 26116 }, { "epoch": 1.58, "learning_rate": 4.28047621203205e-05, "loss": 0.8828, "step": 26118 }, { "epoch": 1.58, "learning_rate": 4.2803253210206264e-05, "loss": 0.5402, "step": 26120 }, { "epoch": 1.58, "learning_rate": 4.2801744300092044e-05, "loss": 0.9029, "step": 26122 }, { "epoch": 1.58, "learning_rate": 4.2800235389977824e-05, "loss": 0.816, "step": 26124 }, { "epoch": 1.58, "learning_rate": 4.27987264798636e-05, "loss": 0.9227, "step": 26126 }, { "epoch": 1.58, "learning_rate": 4.279721756974937e-05, "loss": 0.7748, "step": 26128 }, { "epoch": 1.58, "learning_rate": 4.279570865963515e-05, "loss": 1.2014, "step": 26130 }, { "epoch": 1.58, "learning_rate": 4.2794199749520924e-05, "loss": 0.5486, "step": 26132 }, { "epoch": 1.58, "learning_rate": 4.27926908394067e-05, "loss": 0.5054, "step": 26134 }, { "epoch": 1.58, "learning_rate": 4.279118192929247e-05, "loss": 0.8885, "step": 26136 }, { "epoch": 1.58, "learning_rate": 4.278967301917825e-05, "loss": 1.0165, "step": 26138 }, { "epoch": 1.58, "learning_rate": 4.278816410906403e-05, "loss": 0.8174, "step": 26140 }, { "epoch": 1.58, "learning_rate": 4.2786655198949796e-05, "loss": 0.949, "step": 26142 }, { "epoch": 1.58, "learning_rate": 4.2785146288835576e-05, "loss": 1.1878, "step": 26144 }, { "epoch": 1.58, "learning_rate": 4.2783637378721356e-05, "loss": 0.7774, "step": 26146 }, { "epoch": 1.58, "learning_rate": 4.278212846860712e-05, "loss": 0.6756, "step": 26148 }, { "epoch": 1.58, "learning_rate": 4.27806195584929e-05, "loss": 0.8568, "step": 26150 }, { "epoch": 1.58, "learning_rate": 4.277911064837868e-05, "loss": 1.117, "step": 26152 }, { "epoch": 1.58, "learning_rate": 4.2777601738264455e-05, "loss": 0.624, "step": 26154 }, { "epoch": 1.58, "learning_rate": 4.277609282815023e-05, "loss": 0.9464, "step": 26156 }, { "epoch": 1.58, "learning_rate": 4.2774583918036e-05, "loss": 0.9733, "step": 26158 }, { "epoch": 1.58, "learning_rate": 4.277307500792178e-05, "loss": 0.8709, "step": 26160 }, { "epoch": 1.58, "learning_rate": 4.277156609780756e-05, "loss": 1.2613, "step": 26162 }, { "epoch": 1.58, "learning_rate": 4.277005718769333e-05, "loss": 0.7583, "step": 26164 }, { "epoch": 1.58, "learning_rate": 4.276854827757911e-05, "loss": 1.0285, "step": 26166 }, { "epoch": 1.58, "learning_rate": 4.276703936746489e-05, "loss": 0.8721, "step": 26168 }, { "epoch": 1.58, "learning_rate": 4.2765530457350654e-05, "loss": 0.6781, "step": 26170 }, { "epoch": 1.58, "learning_rate": 4.2764021547236434e-05, "loss": 0.672, "step": 26172 }, { "epoch": 1.58, "learning_rate": 4.276251263712221e-05, "loss": 0.7787, "step": 26174 }, { "epoch": 1.58, "learning_rate": 4.2761003727007987e-05, "loss": 0.9914, "step": 26176 }, { "epoch": 1.58, "learning_rate": 4.275949481689376e-05, "loss": 1.0097, "step": 26178 }, { "epoch": 1.58, "learning_rate": 4.275798590677953e-05, "loss": 1.0332, "step": 26180 }, { "epoch": 1.58, "learning_rate": 4.275647699666531e-05, "loss": 1.2835, "step": 26182 }, { "epoch": 1.58, "learning_rate": 4.2754968086551086e-05, "loss": 0.8102, "step": 26184 }, { "epoch": 1.58, "learning_rate": 4.275345917643686e-05, "loss": 0.7632, "step": 26186 }, { "epoch": 1.58, "learning_rate": 4.275195026632264e-05, "loss": 1.0171, "step": 26188 }, { "epoch": 1.58, "learning_rate": 4.275044135620841e-05, "loss": 0.8799, "step": 26190 }, { "epoch": 1.58, "learning_rate": 4.2748932446094185e-05, "loss": 0.9194, "step": 26192 }, { "epoch": 1.58, "learning_rate": 4.2747423535979965e-05, "loss": 0.9854, "step": 26194 }, { "epoch": 1.58, "learning_rate": 4.274591462586574e-05, "loss": 1.0514, "step": 26196 }, { "epoch": 1.58, "learning_rate": 4.274440571575151e-05, "loss": 0.6779, "step": 26198 }, { "epoch": 1.58, "learning_rate": 4.274289680563729e-05, "loss": 1.0117, "step": 26200 }, { "epoch": 1.58, "learning_rate": 4.2741387895523064e-05, "loss": 0.7943, "step": 26202 }, { "epoch": 1.58, "learning_rate": 4.2739878985408844e-05, "loss": 1.07, "step": 26204 }, { "epoch": 1.58, "learning_rate": 4.273837007529462e-05, "loss": 0.8861, "step": 26206 }, { "epoch": 1.58, "learning_rate": 4.273686116518039e-05, "loss": 0.7006, "step": 26208 }, { "epoch": 1.58, "learning_rate": 4.273535225506617e-05, "loss": 0.7685, "step": 26210 }, { "epoch": 1.58, "learning_rate": 4.2733843344951943e-05, "loss": 0.7086, "step": 26212 }, { "epoch": 1.58, "learning_rate": 4.2732334434837717e-05, "loss": 1.1187, "step": 26214 }, { "epoch": 1.58, "learning_rate": 4.2730825524723496e-05, "loss": 0.7775, "step": 26216 }, { "epoch": 1.58, "learning_rate": 4.272931661460927e-05, "loss": 0.928, "step": 26218 }, { "epoch": 1.58, "learning_rate": 4.272780770449504e-05, "loss": 1.0718, "step": 26220 }, { "epoch": 1.58, "learning_rate": 4.272629879438082e-05, "loss": 0.9444, "step": 26222 }, { "epoch": 1.58, "learning_rate": 4.2724789884266596e-05, "loss": 0.9359, "step": 26224 }, { "epoch": 1.58, "learning_rate": 4.2723280974152376e-05, "loss": 0.9137, "step": 26226 }, { "epoch": 1.58, "learning_rate": 4.272177206403815e-05, "loss": 1.2499, "step": 26228 }, { "epoch": 1.58, "learning_rate": 4.272026315392392e-05, "loss": 1.5249, "step": 26230 }, { "epoch": 1.58, "learning_rate": 4.27187542438097e-05, "loss": 0.8324, "step": 26232 }, { "epoch": 1.58, "learning_rate": 4.271724533369547e-05, "loss": 0.7811, "step": 26234 }, { "epoch": 1.58, "learning_rate": 4.271573642358125e-05, "loss": 1.0722, "step": 26236 }, { "epoch": 1.58, "learning_rate": 4.271422751346703e-05, "loss": 0.7891, "step": 26238 }, { "epoch": 1.58, "learning_rate": 4.27127186033528e-05, "loss": 0.9701, "step": 26240 }, { "epoch": 1.58, "learning_rate": 4.2711209693238574e-05, "loss": 0.7927, "step": 26242 }, { "epoch": 1.58, "learning_rate": 4.2709700783124354e-05, "loss": 0.9066, "step": 26244 }, { "epoch": 1.58, "learning_rate": 4.270819187301013e-05, "loss": 0.9944, "step": 26246 }, { "epoch": 1.58, "learning_rate": 4.27066829628959e-05, "loss": 0.7334, "step": 26248 }, { "epoch": 1.58, "learning_rate": 4.2705174052781673e-05, "loss": 0.8877, "step": 26250 }, { "epoch": 1.58, "learning_rate": 4.2703665142667453e-05, "loss": 1.0117, "step": 26252 }, { "epoch": 1.58, "learning_rate": 4.270215623255323e-05, "loss": 1.0317, "step": 26254 }, { "epoch": 1.58, "learning_rate": 4.2700647322439e-05, "loss": 0.818, "step": 26256 }, { "epoch": 1.58, "learning_rate": 4.269913841232478e-05, "loss": 0.9846, "step": 26258 }, { "epoch": 1.58, "learning_rate": 4.269762950221056e-05, "loss": 0.9463, "step": 26260 }, { "epoch": 1.59, "learning_rate": 4.269612059209633e-05, "loss": 0.8049, "step": 26262 }, { "epoch": 1.59, "learning_rate": 4.2694611681982106e-05, "loss": 0.86, "step": 26264 }, { "epoch": 1.59, "learning_rate": 4.269310277186788e-05, "loss": 1.1681, "step": 26266 }, { "epoch": 1.59, "learning_rate": 4.269159386175366e-05, "loss": 0.6592, "step": 26268 }, { "epoch": 1.59, "learning_rate": 4.269008495163943e-05, "loss": 1.0746, "step": 26270 }, { "epoch": 1.59, "learning_rate": 4.2688576041525205e-05, "loss": 0.7547, "step": 26272 }, { "epoch": 1.59, "learning_rate": 4.2687067131410985e-05, "loss": 0.859, "step": 26274 }, { "epoch": 1.59, "learning_rate": 4.2685558221296765e-05, "loss": 0.8886, "step": 26276 }, { "epoch": 1.59, "learning_rate": 4.268404931118253e-05, "loss": 0.7184, "step": 26278 }, { "epoch": 1.59, "learning_rate": 4.268254040106831e-05, "loss": 1.0963, "step": 26280 }, { "epoch": 1.59, "learning_rate": 4.268103149095409e-05, "loss": 0.8921, "step": 26282 }, { "epoch": 1.59, "learning_rate": 4.267952258083986e-05, "loss": 1.1361, "step": 26284 }, { "epoch": 1.59, "learning_rate": 4.267801367072564e-05, "loss": 0.8646, "step": 26286 }, { "epoch": 1.59, "learning_rate": 4.267650476061141e-05, "loss": 1.2267, "step": 26288 }, { "epoch": 1.59, "learning_rate": 4.267499585049719e-05, "loss": 1.0431, "step": 26290 }, { "epoch": 1.59, "learning_rate": 4.267348694038296e-05, "loss": 0.9049, "step": 26292 }, { "epoch": 1.59, "learning_rate": 4.2671978030268736e-05, "loss": 0.7555, "step": 26294 }, { "epoch": 1.59, "learning_rate": 4.2670469120154516e-05, "loss": 1.0475, "step": 26296 }, { "epoch": 1.59, "learning_rate": 4.266896021004029e-05, "loss": 0.8755, "step": 26298 }, { "epoch": 1.59, "learning_rate": 4.266745129992606e-05, "loss": 0.6196, "step": 26300 }, { "epoch": 1.59, "learning_rate": 4.266594238981184e-05, "loss": 0.5827, "step": 26302 }, { "epoch": 1.59, "learning_rate": 4.2664433479697616e-05, "loss": 0.8131, "step": 26304 }, { "epoch": 1.59, "learning_rate": 4.266292456958339e-05, "loss": 0.8536, "step": 26306 }, { "epoch": 1.59, "learning_rate": 4.266141565946917e-05, "loss": 0.7446, "step": 26308 }, { "epoch": 1.59, "learning_rate": 4.265990674935494e-05, "loss": 0.7013, "step": 26310 }, { "epoch": 1.59, "learning_rate": 4.2658397839240715e-05, "loss": 0.8123, "step": 26312 }, { "epoch": 1.59, "learning_rate": 4.2656888929126495e-05, "loss": 0.7251, "step": 26314 }, { "epoch": 1.59, "learning_rate": 4.265538001901227e-05, "loss": 0.9201, "step": 26316 }, { "epoch": 1.59, "learning_rate": 4.265387110889805e-05, "loss": 0.7999, "step": 26318 }, { "epoch": 1.59, "learning_rate": 4.265236219878382e-05, "loss": 0.8651, "step": 26320 }, { "epoch": 1.59, "learning_rate": 4.2650853288669594e-05, "loss": 0.5612, "step": 26322 }, { "epoch": 1.59, "learning_rate": 4.2649344378555374e-05, "loss": 0.7621, "step": 26324 }, { "epoch": 1.59, "learning_rate": 4.264783546844115e-05, "loss": 0.9137, "step": 26326 }, { "epoch": 1.59, "learning_rate": 4.264632655832692e-05, "loss": 0.795, "step": 26328 }, { "epoch": 1.59, "learning_rate": 4.26448176482127e-05, "loss": 0.9363, "step": 26330 }, { "epoch": 1.59, "learning_rate": 4.264330873809847e-05, "loss": 1.0725, "step": 26332 }, { "epoch": 1.59, "learning_rate": 4.2641799827984246e-05, "loss": 0.8295, "step": 26334 }, { "epoch": 1.59, "learning_rate": 4.2640290917870026e-05, "loss": 0.9402, "step": 26336 }, { "epoch": 1.59, "learning_rate": 4.26387820077558e-05, "loss": 0.7911, "step": 26338 }, { "epoch": 1.59, "learning_rate": 4.263727309764158e-05, "loss": 0.7694, "step": 26340 }, { "epoch": 1.59, "learning_rate": 4.263576418752735e-05, "loss": 0.8689, "step": 26342 }, { "epoch": 1.59, "learning_rate": 4.2634255277413126e-05, "loss": 0.8628, "step": 26344 }, { "epoch": 1.59, "learning_rate": 4.2632746367298905e-05, "loss": 0.6699, "step": 26346 }, { "epoch": 1.59, "learning_rate": 4.263123745718467e-05, "loss": 0.9721, "step": 26348 }, { "epoch": 1.59, "learning_rate": 4.262972854707045e-05, "loss": 1.227, "step": 26350 }, { "epoch": 1.59, "learning_rate": 4.262821963695623e-05, "loss": 0.8612, "step": 26352 }, { "epoch": 1.59, "learning_rate": 4.2626710726842005e-05, "loss": 0.6764, "step": 26354 }, { "epoch": 1.59, "learning_rate": 4.262520181672778e-05, "loss": 0.8326, "step": 26356 }, { "epoch": 1.59, "learning_rate": 4.262369290661356e-05, "loss": 1.0793, "step": 26358 }, { "epoch": 1.59, "learning_rate": 4.262218399649933e-05, "loss": 0.9481, "step": 26360 }, { "epoch": 1.59, "learning_rate": 4.2620675086385104e-05, "loss": 0.6348, "step": 26362 }, { "epoch": 1.59, "learning_rate": 4.261916617627088e-05, "loss": 0.6062, "step": 26364 }, { "epoch": 1.59, "learning_rate": 4.261765726615666e-05, "loss": 0.7495, "step": 26366 }, { "epoch": 1.59, "learning_rate": 4.261614835604244e-05, "loss": 0.7808, "step": 26368 }, { "epoch": 1.59, "learning_rate": 4.26146394459282e-05, "loss": 0.9473, "step": 26370 }, { "epoch": 1.59, "learning_rate": 4.261313053581398e-05, "loss": 0.9501, "step": 26372 }, { "epoch": 1.59, "learning_rate": 4.261162162569976e-05, "loss": 0.8594, "step": 26374 }, { "epoch": 1.59, "learning_rate": 4.2610112715585536e-05, "loss": 0.8711, "step": 26376 }, { "epoch": 1.59, "learning_rate": 4.260860380547131e-05, "loss": 1.1179, "step": 26378 }, { "epoch": 1.59, "learning_rate": 4.260709489535708e-05, "loss": 1.0061, "step": 26380 }, { "epoch": 1.59, "learning_rate": 4.260558598524286e-05, "loss": 0.8452, "step": 26382 }, { "epoch": 1.59, "learning_rate": 4.2604077075128636e-05, "loss": 0.9557, "step": 26384 }, { "epoch": 1.59, "learning_rate": 4.260256816501441e-05, "loss": 0.9631, "step": 26386 }, { "epoch": 1.59, "learning_rate": 4.260105925490019e-05, "loss": 0.6695, "step": 26388 }, { "epoch": 1.59, "learning_rate": 4.259955034478597e-05, "loss": 0.84, "step": 26390 }, { "epoch": 1.59, "learning_rate": 4.2598041434671735e-05, "loss": 0.8311, "step": 26392 }, { "epoch": 1.59, "learning_rate": 4.2596532524557515e-05, "loss": 1.0515, "step": 26394 }, { "epoch": 1.59, "learning_rate": 4.2595023614443295e-05, "loss": 0.9011, "step": 26396 }, { "epoch": 1.59, "learning_rate": 4.259351470432906e-05, "loss": 0.645, "step": 26398 }, { "epoch": 1.59, "learning_rate": 4.259200579421484e-05, "loss": 1.0583, "step": 26400 }, { "epoch": 1.59, "learning_rate": 4.2590496884100614e-05, "loss": 0.6778, "step": 26402 }, { "epoch": 1.59, "learning_rate": 4.2588987973986394e-05, "loss": 0.9915, "step": 26404 }, { "epoch": 1.59, "learning_rate": 4.258747906387217e-05, "loss": 0.8417, "step": 26406 }, { "epoch": 1.59, "learning_rate": 4.258597015375794e-05, "loss": 0.8462, "step": 26408 }, { "epoch": 1.59, "learning_rate": 4.258446124364372e-05, "loss": 0.9862, "step": 26410 }, { "epoch": 1.59, "learning_rate": 4.258295233352949e-05, "loss": 0.7681, "step": 26412 }, { "epoch": 1.59, "learning_rate": 4.2581443423415266e-05, "loss": 0.6893, "step": 26414 }, { "epoch": 1.59, "learning_rate": 4.2579934513301046e-05, "loss": 1.0123, "step": 26416 }, { "epoch": 1.59, "learning_rate": 4.257842560318682e-05, "loss": 0.9396, "step": 26418 }, { "epoch": 1.59, "learning_rate": 4.257691669307259e-05, "loss": 0.7273, "step": 26420 }, { "epoch": 1.59, "learning_rate": 4.257540778295837e-05, "loss": 0.8008, "step": 26422 }, { "epoch": 1.59, "learning_rate": 4.2573898872844145e-05, "loss": 0.7604, "step": 26424 }, { "epoch": 1.6, "learning_rate": 4.2572389962729925e-05, "loss": 1.0016, "step": 26426 }, { "epoch": 1.6, "learning_rate": 4.25708810526157e-05, "loss": 0.9195, "step": 26428 }, { "epoch": 1.6, "learning_rate": 4.256937214250147e-05, "loss": 0.8844, "step": 26430 }, { "epoch": 1.6, "learning_rate": 4.256786323238725e-05, "loss": 0.8459, "step": 26432 }, { "epoch": 1.6, "learning_rate": 4.2566354322273025e-05, "loss": 1.0321, "step": 26434 }, { "epoch": 1.6, "learning_rate": 4.25648454121588e-05, "loss": 0.8778, "step": 26436 }, { "epoch": 1.6, "learning_rate": 4.256333650204458e-05, "loss": 0.7143, "step": 26438 }, { "epoch": 1.6, "learning_rate": 4.256182759193035e-05, "loss": 1.4303, "step": 26440 }, { "epoch": 1.6, "learning_rate": 4.2560318681816124e-05, "loss": 0.9666, "step": 26442 }, { "epoch": 1.6, "learning_rate": 4.2558809771701904e-05, "loss": 0.8126, "step": 26444 }, { "epoch": 1.6, "learning_rate": 4.255730086158768e-05, "loss": 0.8335, "step": 26446 }, { "epoch": 1.6, "learning_rate": 4.255579195147345e-05, "loss": 0.6935, "step": 26448 }, { "epoch": 1.6, "learning_rate": 4.255428304135923e-05, "loss": 0.883, "step": 26450 }, { "epoch": 1.6, "learning_rate": 4.2552774131245e-05, "loss": 0.8398, "step": 26452 }, { "epoch": 1.6, "learning_rate": 4.255126522113078e-05, "loss": 0.8517, "step": 26454 }, { "epoch": 1.6, "learning_rate": 4.2549756311016556e-05, "loss": 0.9121, "step": 26456 }, { "epoch": 1.6, "learning_rate": 4.254824740090233e-05, "loss": 0.7402, "step": 26458 }, { "epoch": 1.6, "learning_rate": 4.254673849078811e-05, "loss": 0.9012, "step": 26460 }, { "epoch": 1.6, "learning_rate": 4.2545229580673876e-05, "loss": 0.9602, "step": 26462 }, { "epoch": 1.6, "learning_rate": 4.2543720670559655e-05, "loss": 0.9633, "step": 26464 }, { "epoch": 1.6, "learning_rate": 4.2542211760445435e-05, "loss": 1.0124, "step": 26466 }, { "epoch": 1.6, "learning_rate": 4.254070285033121e-05, "loss": 0.8237, "step": 26468 }, { "epoch": 1.6, "learning_rate": 4.253919394021698e-05, "loss": 0.8541, "step": 26470 }, { "epoch": 1.6, "learning_rate": 4.253768503010276e-05, "loss": 0.9312, "step": 26472 }, { "epoch": 1.6, "learning_rate": 4.2536176119988535e-05, "loss": 0.8019, "step": 26474 }, { "epoch": 1.6, "learning_rate": 4.253466720987431e-05, "loss": 1.1708, "step": 26476 }, { "epoch": 1.6, "learning_rate": 4.253315829976008e-05, "loss": 1.2123, "step": 26478 }, { "epoch": 1.6, "learning_rate": 4.253164938964586e-05, "loss": 0.8358, "step": 26480 }, { "epoch": 1.6, "learning_rate": 4.253014047953164e-05, "loss": 1.069, "step": 26482 }, { "epoch": 1.6, "learning_rate": 4.252863156941741e-05, "loss": 0.6554, "step": 26484 }, { "epoch": 1.6, "learning_rate": 4.252712265930319e-05, "loss": 0.8719, "step": 26486 }, { "epoch": 1.6, "learning_rate": 4.252561374918897e-05, "loss": 0.9643, "step": 26488 }, { "epoch": 1.6, "learning_rate": 4.252410483907474e-05, "loss": 0.7409, "step": 26490 }, { "epoch": 1.6, "learning_rate": 4.252259592896051e-05, "loss": 1.1835, "step": 26492 }, { "epoch": 1.6, "learning_rate": 4.2521087018846286e-05, "loss": 1.0104, "step": 26494 }, { "epoch": 1.6, "learning_rate": 4.2519578108732066e-05, "loss": 0.861, "step": 26496 }, { "epoch": 1.6, "learning_rate": 4.251806919861784e-05, "loss": 0.7994, "step": 26498 }, { "epoch": 1.6, "learning_rate": 4.251656028850361e-05, "loss": 0.6968, "step": 26500 }, { "epoch": 1.6, "learning_rate": 4.251505137838939e-05, "loss": 0.6762, "step": 26502 }, { "epoch": 1.6, "learning_rate": 4.251354246827517e-05, "loss": 0.8656, "step": 26504 }, { "epoch": 1.6, "learning_rate": 4.251203355816094e-05, "loss": 0.9099, "step": 26506 }, { "epoch": 1.6, "learning_rate": 4.251052464804672e-05, "loss": 0.6638, "step": 26508 }, { "epoch": 1.6, "learning_rate": 4.25090157379325e-05, "loss": 0.7655, "step": 26510 }, { "epoch": 1.6, "learning_rate": 4.2507506827818265e-05, "loss": 0.8882, "step": 26512 }, { "epoch": 1.6, "learning_rate": 4.2505997917704045e-05, "loss": 1.021, "step": 26514 }, { "epoch": 1.6, "learning_rate": 4.250448900758982e-05, "loss": 1.0427, "step": 26516 }, { "epoch": 1.6, "learning_rate": 4.25029800974756e-05, "loss": 1.2376, "step": 26518 }, { "epoch": 1.6, "learning_rate": 4.250147118736137e-05, "loss": 0.8785, "step": 26520 }, { "epoch": 1.6, "learning_rate": 4.2499962277247144e-05, "loss": 1.223, "step": 26522 }, { "epoch": 1.6, "learning_rate": 4.2498453367132924e-05, "loss": 0.9218, "step": 26524 }, { "epoch": 1.6, "learning_rate": 4.24969444570187e-05, "loss": 1.3749, "step": 26526 }, { "epoch": 1.6, "learning_rate": 4.249543554690447e-05, "loss": 0.9672, "step": 26528 }, { "epoch": 1.6, "learning_rate": 4.249392663679025e-05, "loss": 0.6955, "step": 26530 }, { "epoch": 1.6, "learning_rate": 4.249241772667602e-05, "loss": 0.6601, "step": 26532 }, { "epoch": 1.6, "learning_rate": 4.2490908816561796e-05, "loss": 0.7674, "step": 26534 }, { "epoch": 1.6, "learning_rate": 4.2489399906447576e-05, "loss": 0.9222, "step": 26536 }, { "epoch": 1.6, "learning_rate": 4.248789099633335e-05, "loss": 1.0313, "step": 26538 }, { "epoch": 1.6, "learning_rate": 4.248638208621913e-05, "loss": 1.1362, "step": 26540 }, { "epoch": 1.6, "learning_rate": 4.24848731761049e-05, "loss": 1.105, "step": 26542 }, { "epoch": 1.6, "learning_rate": 4.2483364265990675e-05, "loss": 0.9347, "step": 26544 }, { "epoch": 1.6, "learning_rate": 4.2481855355876455e-05, "loss": 0.5487, "step": 26546 }, { "epoch": 1.6, "learning_rate": 4.248034644576223e-05, "loss": 1.1873, "step": 26548 }, { "epoch": 1.6, "learning_rate": 4.2478837535648e-05, "loss": 0.938, "step": 26550 }, { "epoch": 1.6, "learning_rate": 4.247732862553378e-05, "loss": 0.8984, "step": 26552 }, { "epoch": 1.6, "learning_rate": 4.2475819715419554e-05, "loss": 0.7867, "step": 26554 }, { "epoch": 1.6, "learning_rate": 4.247431080530533e-05, "loss": 0.7081, "step": 26556 }, { "epoch": 1.6, "learning_rate": 4.247280189519111e-05, "loss": 0.8754, "step": 26558 }, { "epoch": 1.6, "learning_rate": 4.247129298507688e-05, "loss": 1.0014, "step": 26560 }, { "epoch": 1.6, "learning_rate": 4.2469784074962654e-05, "loss": 0.8179, "step": 26562 }, { "epoch": 1.6, "learning_rate": 4.2468275164848434e-05, "loss": 0.8353, "step": 26564 }, { "epoch": 1.6, "learning_rate": 4.246676625473421e-05, "loss": 0.8848, "step": 26566 }, { "epoch": 1.6, "learning_rate": 4.246525734461999e-05, "loss": 1.2464, "step": 26568 }, { "epoch": 1.6, "learning_rate": 4.246374843450576e-05, "loss": 0.8767, "step": 26570 }, { "epoch": 1.6, "learning_rate": 4.246223952439153e-05, "loss": 0.8591, "step": 26572 }, { "epoch": 1.6, "learning_rate": 4.246073061427731e-05, "loss": 0.8095, "step": 26574 }, { "epoch": 1.6, "learning_rate": 4.245922170416308e-05, "loss": 0.9295, "step": 26576 }, { "epoch": 1.6, "learning_rate": 4.245771279404886e-05, "loss": 0.961, "step": 26578 }, { "epoch": 1.6, "learning_rate": 4.245620388393464e-05, "loss": 0.9574, "step": 26580 }, { "epoch": 1.6, "learning_rate": 4.245469497382041e-05, "loss": 0.7276, "step": 26582 }, { "epoch": 1.6, "learning_rate": 4.2453186063706185e-05, "loss": 0.9536, "step": 26584 }, { "epoch": 1.6, "learning_rate": 4.2451677153591965e-05, "loss": 0.9583, "step": 26586 }, { "epoch": 1.6, "learning_rate": 4.245016824347774e-05, "loss": 0.8724, "step": 26588 }, { "epoch": 1.6, "learning_rate": 4.244865933336352e-05, "loss": 0.7416, "step": 26590 }, { "epoch": 1.61, "learning_rate": 4.2447150423249285e-05, "loss": 0.7196, "step": 26592 }, { "epoch": 1.61, "learning_rate": 4.2445641513135064e-05, "loss": 0.7475, "step": 26594 }, { "epoch": 1.61, "learning_rate": 4.2444132603020844e-05, "loss": 0.7409, "step": 26596 }, { "epoch": 1.61, "learning_rate": 4.244262369290661e-05, "loss": 1.0661, "step": 26598 }, { "epoch": 1.61, "learning_rate": 4.244111478279239e-05, "loss": 0.8503, "step": 26600 }, { "epoch": 1.61, "learning_rate": 4.243960587267817e-05, "loss": 0.7014, "step": 26602 }, { "epoch": 1.61, "learning_rate": 4.2438096962563944e-05, "loss": 0.663, "step": 26604 }, { "epoch": 1.61, "learning_rate": 4.243658805244972e-05, "loss": 1.1284, "step": 26606 }, { "epoch": 1.61, "learning_rate": 4.243507914233549e-05, "loss": 0.5076, "step": 26608 }, { "epoch": 1.61, "learning_rate": 4.243357023222127e-05, "loss": 0.7141, "step": 26610 }, { "epoch": 1.61, "learning_rate": 4.243206132210704e-05, "loss": 0.9049, "step": 26612 }, { "epoch": 1.61, "learning_rate": 4.2430552411992816e-05, "loss": 0.9242, "step": 26614 }, { "epoch": 1.61, "learning_rate": 4.2429043501878596e-05, "loss": 0.9722, "step": 26616 }, { "epoch": 1.61, "learning_rate": 4.2427534591764376e-05, "loss": 0.5224, "step": 26618 }, { "epoch": 1.61, "learning_rate": 4.242602568165014e-05, "loss": 0.9361, "step": 26620 }, { "epoch": 1.61, "learning_rate": 4.242451677153592e-05, "loss": 0.6198, "step": 26622 }, { "epoch": 1.61, "learning_rate": 4.24230078614217e-05, "loss": 0.6889, "step": 26624 }, { "epoch": 1.61, "learning_rate": 4.242149895130747e-05, "loss": 0.5847, "step": 26626 }, { "epoch": 1.61, "learning_rate": 4.241999004119325e-05, "loss": 0.6681, "step": 26628 }, { "epoch": 1.61, "learning_rate": 4.241848113107902e-05, "loss": 0.7623, "step": 26630 }, { "epoch": 1.61, "learning_rate": 4.24169722209648e-05, "loss": 0.824, "step": 26632 }, { "epoch": 1.61, "learning_rate": 4.2415463310850574e-05, "loss": 1.0284, "step": 26634 }, { "epoch": 1.61, "learning_rate": 4.241395440073635e-05, "loss": 0.7015, "step": 26636 }, { "epoch": 1.61, "learning_rate": 4.241244549062213e-05, "loss": 0.8629, "step": 26638 }, { "epoch": 1.61, "learning_rate": 4.24109365805079e-05, "loss": 0.7781, "step": 26640 }, { "epoch": 1.61, "learning_rate": 4.2409427670393674e-05, "loss": 0.9103, "step": 26642 }, { "epoch": 1.61, "learning_rate": 4.2407918760279454e-05, "loss": 0.9457, "step": 26644 }, { "epoch": 1.61, "learning_rate": 4.240640985016523e-05, "loss": 0.8216, "step": 26646 }, { "epoch": 1.61, "learning_rate": 4.2404900940051e-05, "loss": 0.9583, "step": 26648 }, { "epoch": 1.61, "learning_rate": 4.240339202993678e-05, "loss": 0.6439, "step": 26650 }, { "epoch": 1.61, "learning_rate": 4.240188311982255e-05, "loss": 1.0298, "step": 26652 }, { "epoch": 1.61, "learning_rate": 4.240037420970833e-05, "loss": 0.6908, "step": 26654 }, { "epoch": 1.61, "learning_rate": 4.2398865299594106e-05, "loss": 0.8861, "step": 26656 }, { "epoch": 1.61, "learning_rate": 4.239735638947988e-05, "loss": 0.8353, "step": 26658 }, { "epoch": 1.61, "learning_rate": 4.239584747936566e-05, "loss": 1.0803, "step": 26660 }, { "epoch": 1.61, "learning_rate": 4.239433856925143e-05, "loss": 1.1364, "step": 26662 }, { "epoch": 1.61, "learning_rate": 4.2392829659137205e-05, "loss": 0.7859, "step": 26664 }, { "epoch": 1.61, "learning_rate": 4.2391320749022985e-05, "loss": 0.7369, "step": 26666 }, { "epoch": 1.61, "learning_rate": 4.238981183890876e-05, "loss": 0.6609, "step": 26668 }, { "epoch": 1.61, "learning_rate": 4.238830292879453e-05, "loss": 0.8694, "step": 26670 }, { "epoch": 1.61, "learning_rate": 4.238679401868031e-05, "loss": 0.9851, "step": 26672 }, { "epoch": 1.61, "learning_rate": 4.2385285108566084e-05, "loss": 0.7013, "step": 26674 }, { "epoch": 1.61, "learning_rate": 4.238377619845186e-05, "loss": 1.1251, "step": 26676 }, { "epoch": 1.61, "learning_rate": 4.238226728833764e-05, "loss": 0.9767, "step": 26678 }, { "epoch": 1.61, "learning_rate": 4.238075837822341e-05, "loss": 0.6034, "step": 26680 }, { "epoch": 1.61, "learning_rate": 4.237924946810919e-05, "loss": 0.6877, "step": 26682 }, { "epoch": 1.61, "learning_rate": 4.2377740557994963e-05, "loss": 0.69, "step": 26684 }, { "epoch": 1.61, "learning_rate": 4.237623164788074e-05, "loss": 0.8597, "step": 26686 }, { "epoch": 1.61, "learning_rate": 4.2374722737766517e-05, "loss": 0.8844, "step": 26688 }, { "epoch": 1.61, "learning_rate": 4.237321382765228e-05, "loss": 0.8447, "step": 26690 }, { "epoch": 1.61, "learning_rate": 4.237170491753806e-05, "loss": 0.7339, "step": 26692 }, { "epoch": 1.61, "learning_rate": 4.237019600742384e-05, "loss": 0.8076, "step": 26694 }, { "epoch": 1.61, "learning_rate": 4.2368687097309616e-05, "loss": 0.8649, "step": 26696 }, { "epoch": 1.61, "learning_rate": 4.236717818719539e-05, "loss": 0.7733, "step": 26698 }, { "epoch": 1.61, "learning_rate": 4.236566927708117e-05, "loss": 0.8786, "step": 26700 }, { "epoch": 1.61, "learning_rate": 4.236416036696694e-05, "loss": 0.6795, "step": 26702 }, { "epoch": 1.61, "learning_rate": 4.236265145685272e-05, "loss": 1.0593, "step": 26704 }, { "epoch": 1.61, "learning_rate": 4.236114254673849e-05, "loss": 0.891, "step": 26706 }, { "epoch": 1.61, "learning_rate": 4.235963363662427e-05, "loss": 0.8939, "step": 26708 }, { "epoch": 1.61, "learning_rate": 4.235812472651005e-05, "loss": 0.9444, "step": 26710 }, { "epoch": 1.61, "learning_rate": 4.2356615816395814e-05, "loss": 0.8676, "step": 26712 }, { "epoch": 1.61, "learning_rate": 4.2355106906281594e-05, "loss": 0.8156, "step": 26714 }, { "epoch": 1.61, "learning_rate": 4.2353597996167374e-05, "loss": 0.9248, "step": 26716 }, { "epoch": 1.61, "learning_rate": 4.235208908605315e-05, "loss": 0.8443, "step": 26718 }, { "epoch": 1.61, "learning_rate": 4.235058017593892e-05, "loss": 0.7196, "step": 26720 }, { "epoch": 1.61, "learning_rate": 4.2349071265824694e-05, "loss": 1.1434, "step": 26722 }, { "epoch": 1.61, "learning_rate": 4.2347562355710473e-05, "loss": 0.5279, "step": 26724 }, { "epoch": 1.61, "learning_rate": 4.2346053445596247e-05, "loss": 1.1792, "step": 26726 }, { "epoch": 1.61, "learning_rate": 4.234454453548202e-05, "loss": 1.0476, "step": 26728 }, { "epoch": 1.61, "learning_rate": 4.23430356253678e-05, "loss": 0.8975, "step": 26730 }, { "epoch": 1.61, "learning_rate": 4.234152671525358e-05, "loss": 0.6619, "step": 26732 }, { "epoch": 1.61, "learning_rate": 4.2340017805139346e-05, "loss": 0.7902, "step": 26734 }, { "epoch": 1.61, "learning_rate": 4.2338508895025126e-05, "loss": 1.009, "step": 26736 }, { "epoch": 1.61, "learning_rate": 4.23369999849109e-05, "loss": 0.6888, "step": 26738 }, { "epoch": 1.61, "learning_rate": 4.233549107479667e-05, "loss": 0.7916, "step": 26740 }, { "epoch": 1.61, "learning_rate": 4.233398216468245e-05, "loss": 0.9273, "step": 26742 }, { "epoch": 1.61, "learning_rate": 4.2332473254568225e-05, "loss": 0.9971, "step": 26744 }, { "epoch": 1.61, "learning_rate": 4.2330964344454005e-05, "loss": 0.6719, "step": 26746 }, { "epoch": 1.61, "learning_rate": 4.232945543433978e-05, "loss": 0.8581, "step": 26748 }, { "epoch": 1.61, "learning_rate": 4.232794652422555e-05, "loss": 0.8794, "step": 26750 }, { "epoch": 1.61, "learning_rate": 4.232643761411133e-05, "loss": 0.7053, "step": 26752 }, { "epoch": 1.61, "learning_rate": 4.232492870399711e-05, "loss": 0.9384, "step": 26754 }, { "epoch": 1.61, "learning_rate": 4.232341979388288e-05, "loss": 0.7811, "step": 26756 }, { "epoch": 1.62, "learning_rate": 4.232191088376866e-05, "loss": 0.7134, "step": 26758 }, { "epoch": 1.62, "learning_rate": 4.232040197365443e-05, "loss": 0.7172, "step": 26760 }, { "epoch": 1.62, "learning_rate": 4.2318893063540203e-05, "loss": 0.5376, "step": 26762 }, { "epoch": 1.62, "learning_rate": 4.231738415342598e-05, "loss": 0.9616, "step": 26764 }, { "epoch": 1.62, "learning_rate": 4.2315875243311757e-05, "loss": 1.0717, "step": 26766 }, { "epoch": 1.62, "learning_rate": 4.2314366333197536e-05, "loss": 0.7387, "step": 26768 }, { "epoch": 1.62, "learning_rate": 4.231285742308331e-05, "loss": 0.904, "step": 26770 }, { "epoch": 1.62, "learning_rate": 4.231134851296908e-05, "loss": 0.6347, "step": 26772 }, { "epoch": 1.62, "learning_rate": 4.230983960285486e-05, "loss": 0.9118, "step": 26774 }, { "epoch": 1.62, "learning_rate": 4.2308330692740636e-05, "loss": 0.7575, "step": 26776 }, { "epoch": 1.62, "learning_rate": 4.230682178262641e-05, "loss": 1.1625, "step": 26778 }, { "epoch": 1.62, "learning_rate": 4.230531287251219e-05, "loss": 0.56, "step": 26780 }, { "epoch": 1.62, "learning_rate": 4.230380396239796e-05, "loss": 0.9178, "step": 26782 }, { "epoch": 1.62, "learning_rate": 4.2302295052283735e-05, "loss": 0.8589, "step": 26784 }, { "epoch": 1.62, "learning_rate": 4.2300786142169515e-05, "loss": 0.9018, "step": 26786 }, { "epoch": 1.62, "learning_rate": 4.229927723205529e-05, "loss": 0.6896, "step": 26788 }, { "epoch": 1.62, "learning_rate": 4.229776832194106e-05, "loss": 0.9174, "step": 26790 }, { "epoch": 1.62, "learning_rate": 4.229625941182684e-05, "loss": 0.9252, "step": 26792 }, { "epoch": 1.62, "learning_rate": 4.2294750501712614e-05, "loss": 1.196, "step": 26794 }, { "epoch": 1.62, "learning_rate": 4.2293241591598394e-05, "loss": 0.8911, "step": 26796 }, { "epoch": 1.62, "learning_rate": 4.229173268148417e-05, "loss": 1.0049, "step": 26798 }, { "epoch": 1.62, "learning_rate": 4.229022377136994e-05, "loss": 0.8079, "step": 26800 }, { "epoch": 1.62, "learning_rate": 4.228871486125572e-05, "loss": 0.7326, "step": 26802 }, { "epoch": 1.62, "learning_rate": 4.228720595114149e-05, "loss": 0.915, "step": 26804 }, { "epoch": 1.62, "learning_rate": 4.2285697041027266e-05, "loss": 0.713, "step": 26806 }, { "epoch": 1.62, "learning_rate": 4.2284188130913046e-05, "loss": 0.7813, "step": 26808 }, { "epoch": 1.62, "learning_rate": 4.228267922079882e-05, "loss": 1.0236, "step": 26810 }, { "epoch": 1.62, "learning_rate": 4.228117031068459e-05, "loss": 0.7173, "step": 26812 }, { "epoch": 1.62, "learning_rate": 4.227966140057037e-05, "loss": 1.0835, "step": 26814 }, { "epoch": 1.62, "learning_rate": 4.2278152490456146e-05, "loss": 1.2204, "step": 26816 }, { "epoch": 1.62, "learning_rate": 4.2276643580341926e-05, "loss": 0.8467, "step": 26818 }, { "epoch": 1.62, "learning_rate": 4.227513467022769e-05, "loss": 0.819, "step": 26820 }, { "epoch": 1.62, "learning_rate": 4.227362576011347e-05, "loss": 0.6803, "step": 26822 }, { "epoch": 1.62, "learning_rate": 4.227211684999925e-05, "loss": 0.7761, "step": 26824 }, { "epoch": 1.62, "learning_rate": 4.227060793988502e-05, "loss": 0.9521, "step": 26826 }, { "epoch": 1.62, "learning_rate": 4.22690990297708e-05, "loss": 0.9262, "step": 26828 }, { "epoch": 1.62, "learning_rate": 4.226759011965658e-05, "loss": 0.9887, "step": 26830 }, { "epoch": 1.62, "learning_rate": 4.226608120954235e-05, "loss": 0.7806, "step": 26832 }, { "epoch": 1.62, "learning_rate": 4.2264572299428124e-05, "loss": 0.7546, "step": 26834 }, { "epoch": 1.62, "learning_rate": 4.22630633893139e-05, "loss": 1.0139, "step": 26836 }, { "epoch": 1.62, "learning_rate": 4.226155447919968e-05, "loss": 1.1092, "step": 26838 }, { "epoch": 1.62, "learning_rate": 4.226004556908545e-05, "loss": 0.914, "step": 26840 }, { "epoch": 1.62, "learning_rate": 4.225853665897122e-05, "loss": 0.8646, "step": 26842 }, { "epoch": 1.62, "learning_rate": 4.2257027748857e-05, "loss": 0.903, "step": 26844 }, { "epoch": 1.62, "learning_rate": 4.225551883874278e-05, "loss": 0.8105, "step": 26846 }, { "epoch": 1.62, "learning_rate": 4.225400992862855e-05, "loss": 1.0404, "step": 26848 }, { "epoch": 1.62, "learning_rate": 4.225250101851433e-05, "loss": 0.9506, "step": 26850 }, { "epoch": 1.62, "learning_rate": 4.22509921084001e-05, "loss": 1.0397, "step": 26852 }, { "epoch": 1.62, "learning_rate": 4.2249483198285876e-05, "loss": 0.664, "step": 26854 }, { "epoch": 1.62, "learning_rate": 4.2247974288171656e-05, "loss": 0.8223, "step": 26856 }, { "epoch": 1.62, "learning_rate": 4.224646537805743e-05, "loss": 0.956, "step": 26858 }, { "epoch": 1.62, "learning_rate": 4.224495646794321e-05, "loss": 0.9579, "step": 26860 }, { "epoch": 1.62, "learning_rate": 4.224344755782898e-05, "loss": 1.0189, "step": 26862 }, { "epoch": 1.62, "learning_rate": 4.2241938647714755e-05, "loss": 0.7468, "step": 26864 }, { "epoch": 1.62, "learning_rate": 4.2240429737600535e-05, "loss": 0.7667, "step": 26866 }, { "epoch": 1.62, "learning_rate": 4.2238920827486315e-05, "loss": 0.8825, "step": 26868 }, { "epoch": 1.62, "learning_rate": 4.223741191737208e-05, "loss": 0.8737, "step": 26870 }, { "epoch": 1.62, "learning_rate": 4.223590300725786e-05, "loss": 1.1907, "step": 26872 }, { "epoch": 1.62, "learning_rate": 4.2234394097143634e-05, "loss": 0.8083, "step": 26874 }, { "epoch": 1.62, "learning_rate": 4.223288518702941e-05, "loss": 0.8002, "step": 26876 }, { "epoch": 1.62, "learning_rate": 4.223137627691519e-05, "loss": 0.8567, "step": 26878 }, { "epoch": 1.62, "learning_rate": 4.222986736680096e-05, "loss": 0.9527, "step": 26880 }, { "epoch": 1.62, "learning_rate": 4.222835845668674e-05, "loss": 0.9664, "step": 26882 }, { "epoch": 1.62, "learning_rate": 4.222684954657251e-05, "loss": 0.936, "step": 26884 }, { "epoch": 1.62, "learning_rate": 4.2225340636458286e-05, "loss": 0.8946, "step": 26886 }, { "epoch": 1.62, "learning_rate": 4.2223831726344066e-05, "loss": 0.8374, "step": 26888 }, { "epoch": 1.62, "learning_rate": 4.222232281622984e-05, "loss": 0.6966, "step": 26890 }, { "epoch": 1.62, "learning_rate": 4.222081390611561e-05, "loss": 1.0863, "step": 26892 }, { "epoch": 1.62, "learning_rate": 4.221930499600139e-05, "loss": 0.8812, "step": 26894 }, { "epoch": 1.62, "learning_rate": 4.2217796085887166e-05, "loss": 0.9918, "step": 26896 }, { "epoch": 1.62, "learning_rate": 4.221628717577294e-05, "loss": 0.7527, "step": 26898 }, { "epoch": 1.62, "learning_rate": 4.221477826565872e-05, "loss": 0.9841, "step": 26900 }, { "epoch": 1.62, "learning_rate": 4.221326935554449e-05, "loss": 0.955, "step": 26902 }, { "epoch": 1.62, "learning_rate": 4.2211760445430265e-05, "loss": 0.6012, "step": 26904 }, { "epoch": 1.62, "learning_rate": 4.2210251535316045e-05, "loss": 0.9648, "step": 26906 }, { "epoch": 1.62, "learning_rate": 4.220874262520182e-05, "loss": 0.9814, "step": 26908 }, { "epoch": 1.62, "learning_rate": 4.22072337150876e-05, "loss": 1.1472, "step": 26910 }, { "epoch": 1.62, "learning_rate": 4.220572480497337e-05, "loss": 0.5154, "step": 26912 }, { "epoch": 1.62, "learning_rate": 4.2204215894859144e-05, "loss": 0.9462, "step": 26914 }, { "epoch": 1.62, "learning_rate": 4.2202706984744924e-05, "loss": 0.8353, "step": 26916 }, { "epoch": 1.62, "learning_rate": 4.22011980746307e-05, "loss": 0.7803, "step": 26918 }, { "epoch": 1.62, "learning_rate": 4.219968916451647e-05, "loss": 1.1715, "step": 26920 }, { "epoch": 1.62, "learning_rate": 4.219818025440225e-05, "loss": 1.0547, "step": 26922 }, { "epoch": 1.63, "learning_rate": 4.219667134428802e-05, "loss": 0.7275, "step": 26924 }, { "epoch": 1.63, "learning_rate": 4.2195162434173796e-05, "loss": 1.1141, "step": 26926 }, { "epoch": 1.63, "learning_rate": 4.2193653524059576e-05, "loss": 1.2213, "step": 26928 }, { "epoch": 1.63, "learning_rate": 4.219214461394535e-05, "loss": 0.7941, "step": 26930 }, { "epoch": 1.63, "learning_rate": 4.219063570383113e-05, "loss": 0.581, "step": 26932 }, { "epoch": 1.63, "learning_rate": 4.2189126793716896e-05, "loss": 1.1004, "step": 26934 }, { "epoch": 1.63, "learning_rate": 4.2187617883602675e-05, "loss": 0.674, "step": 26936 }, { "epoch": 1.63, "learning_rate": 4.2186108973488455e-05, "loss": 0.8822, "step": 26938 }, { "epoch": 1.63, "learning_rate": 4.218460006337422e-05, "loss": 0.6177, "step": 26940 }, { "epoch": 1.63, "learning_rate": 4.218309115326e-05, "loss": 0.8477, "step": 26942 }, { "epoch": 1.63, "learning_rate": 4.218158224314578e-05, "loss": 0.9154, "step": 26944 }, { "epoch": 1.63, "learning_rate": 4.2180073333031555e-05, "loss": 1.0355, "step": 26946 }, { "epoch": 1.63, "learning_rate": 4.217856442291733e-05, "loss": 0.8333, "step": 26948 }, { "epoch": 1.63, "learning_rate": 4.21770555128031e-05, "loss": 0.8882, "step": 26950 }, { "epoch": 1.63, "learning_rate": 4.217554660268888e-05, "loss": 0.9862, "step": 26952 }, { "epoch": 1.63, "learning_rate": 4.2174037692574654e-05, "loss": 1.1641, "step": 26954 }, { "epoch": 1.63, "learning_rate": 4.217252878246043e-05, "loss": 0.9026, "step": 26956 }, { "epoch": 1.63, "learning_rate": 4.217101987234621e-05, "loss": 0.6952, "step": 26958 }, { "epoch": 1.63, "learning_rate": 4.216951096223199e-05, "loss": 0.938, "step": 26960 }, { "epoch": 1.63, "learning_rate": 4.216800205211775e-05, "loss": 0.8781, "step": 26962 }, { "epoch": 1.63, "learning_rate": 4.216649314200353e-05, "loss": 1.1421, "step": 26964 }, { "epoch": 1.63, "learning_rate": 4.2164984231889306e-05, "loss": 0.8488, "step": 26966 }, { "epoch": 1.63, "learning_rate": 4.2163475321775086e-05, "loss": 0.8187, "step": 26968 }, { "epoch": 1.63, "learning_rate": 4.216196641166086e-05, "loss": 1.1828, "step": 26970 }, { "epoch": 1.63, "learning_rate": 4.216045750154663e-05, "loss": 0.9067, "step": 26972 }, { "epoch": 1.63, "learning_rate": 4.215894859143241e-05, "loss": 1.0674, "step": 26974 }, { "epoch": 1.63, "learning_rate": 4.2157439681318185e-05, "loss": 0.8614, "step": 26976 }, { "epoch": 1.63, "learning_rate": 4.215593077120396e-05, "loss": 0.9563, "step": 26978 }, { "epoch": 1.63, "learning_rate": 4.215442186108974e-05, "loss": 0.9619, "step": 26980 }, { "epoch": 1.63, "learning_rate": 4.215291295097552e-05, "loss": 0.6984, "step": 26982 }, { "epoch": 1.63, "learning_rate": 4.2151404040861285e-05, "loss": 0.9007, "step": 26984 }, { "epoch": 1.63, "learning_rate": 4.2149895130747065e-05, "loss": 0.539, "step": 26986 }, { "epoch": 1.63, "learning_rate": 4.214838622063284e-05, "loss": 1.0955, "step": 26988 }, { "epoch": 1.63, "learning_rate": 4.214687731051861e-05, "loss": 0.684, "step": 26990 }, { "epoch": 1.63, "learning_rate": 4.214536840040439e-05, "loss": 0.9894, "step": 26992 }, { "epoch": 1.63, "learning_rate": 4.2143859490290164e-05, "loss": 0.792, "step": 26994 }, { "epoch": 1.63, "learning_rate": 4.2142350580175944e-05, "loss": 1.1761, "step": 26996 }, { "epoch": 1.63, "learning_rate": 4.214084167006172e-05, "loss": 1.1703, "step": 26998 }, { "epoch": 1.63, "learning_rate": 4.213933275994749e-05, "loss": 1.0032, "step": 27000 }, { "epoch": 1.63, "learning_rate": 4.213782384983327e-05, "loss": 1.0312, "step": 27002 }, { "epoch": 1.63, "learning_rate": 4.213631493971904e-05, "loss": 0.888, "step": 27004 }, { "epoch": 1.63, "learning_rate": 4.2134806029604816e-05, "loss": 0.74, "step": 27006 }, { "epoch": 1.63, "learning_rate": 4.2133297119490596e-05, "loss": 1.118, "step": 27008 }, { "epoch": 1.63, "learning_rate": 4.213178820937637e-05, "loss": 0.8592, "step": 27010 }, { "epoch": 1.63, "learning_rate": 4.213027929926214e-05, "loss": 1.1023, "step": 27012 }, { "epoch": 1.63, "learning_rate": 4.212877038914792e-05, "loss": 0.8027, "step": 27014 }, { "epoch": 1.63, "learning_rate": 4.2127261479033695e-05, "loss": 0.8903, "step": 27016 }, { "epoch": 1.63, "learning_rate": 4.212575256891947e-05, "loss": 0.8243, "step": 27018 }, { "epoch": 1.63, "learning_rate": 4.212424365880525e-05, "loss": 0.8618, "step": 27020 }, { "epoch": 1.63, "learning_rate": 4.212273474869102e-05, "loss": 0.9084, "step": 27022 }, { "epoch": 1.63, "learning_rate": 4.21212258385768e-05, "loss": 1.0272, "step": 27024 }, { "epoch": 1.63, "learning_rate": 4.2119716928462575e-05, "loss": 0.6432, "step": 27026 }, { "epoch": 1.63, "learning_rate": 4.211820801834835e-05, "loss": 0.9858, "step": 27028 }, { "epoch": 1.63, "learning_rate": 4.211669910823413e-05, "loss": 0.7299, "step": 27030 }, { "epoch": 1.63, "learning_rate": 4.21151901981199e-05, "loss": 1.0344, "step": 27032 }, { "epoch": 1.63, "learning_rate": 4.2113681288005674e-05, "loss": 0.8194, "step": 27034 }, { "epoch": 1.63, "learning_rate": 4.2112172377891454e-05, "loss": 0.8065, "step": 27036 }, { "epoch": 1.63, "learning_rate": 4.211066346777723e-05, "loss": 0.5555, "step": 27038 }, { "epoch": 1.63, "learning_rate": 4.2109154557663e-05, "loss": 0.7749, "step": 27040 }, { "epoch": 1.63, "learning_rate": 4.210764564754878e-05, "loss": 1.0215, "step": 27042 }, { "epoch": 1.63, "learning_rate": 4.210613673743455e-05, "loss": 0.9194, "step": 27044 }, { "epoch": 1.63, "learning_rate": 4.210462782732033e-05, "loss": 0.637, "step": 27046 }, { "epoch": 1.63, "learning_rate": 4.21031189172061e-05, "loss": 0.9275, "step": 27048 }, { "epoch": 1.63, "learning_rate": 4.210161000709188e-05, "loss": 1.1619, "step": 27050 }, { "epoch": 1.63, "learning_rate": 4.210010109697766e-05, "loss": 1.0658, "step": 27052 }, { "epoch": 1.63, "learning_rate": 4.2098592186863425e-05, "loss": 1.032, "step": 27054 }, { "epoch": 1.63, "learning_rate": 4.2097083276749205e-05, "loss": 0.8937, "step": 27056 }, { "epoch": 1.63, "learning_rate": 4.2095574366634985e-05, "loss": 0.8771, "step": 27058 }, { "epoch": 1.63, "learning_rate": 4.209406545652076e-05, "loss": 0.8245, "step": 27060 }, { "epoch": 1.63, "learning_rate": 4.209255654640653e-05, "loss": 0.8236, "step": 27062 }, { "epoch": 1.63, "learning_rate": 4.2091047636292305e-05, "loss": 0.8133, "step": 27064 }, { "epoch": 1.63, "learning_rate": 4.2089538726178084e-05, "loss": 1.3947, "step": 27066 }, { "epoch": 1.63, "learning_rate": 4.208802981606386e-05, "loss": 0.6083, "step": 27068 }, { "epoch": 1.63, "learning_rate": 4.208652090594963e-05, "loss": 0.8946, "step": 27070 }, { "epoch": 1.63, "learning_rate": 4.208501199583541e-05, "loss": 0.914, "step": 27072 }, { "epoch": 1.63, "learning_rate": 4.208350308572119e-05, "loss": 0.6558, "step": 27074 }, { "epoch": 1.63, "learning_rate": 4.208199417560696e-05, "loss": 0.6686, "step": 27076 }, { "epoch": 1.63, "learning_rate": 4.208048526549274e-05, "loss": 0.7761, "step": 27078 }, { "epoch": 1.63, "learning_rate": 4.207897635537851e-05, "loss": 0.782, "step": 27080 }, { "epoch": 1.63, "learning_rate": 4.207746744526429e-05, "loss": 0.6664, "step": 27082 }, { "epoch": 1.63, "learning_rate": 4.207595853515006e-05, "loss": 0.752, "step": 27084 }, { "epoch": 1.63, "learning_rate": 4.2074449625035836e-05, "loss": 0.9171, "step": 27086 }, { "epoch": 1.63, "learning_rate": 4.2072940714921616e-05, "loss": 0.7462, "step": 27088 }, { "epoch": 1.64, "learning_rate": 4.207143180480739e-05, "loss": 1.0229, "step": 27090 }, { "epoch": 1.64, "learning_rate": 4.206992289469316e-05, "loss": 0.7938, "step": 27092 }, { "epoch": 1.64, "learning_rate": 4.206841398457894e-05, "loss": 0.6667, "step": 27094 }, { "epoch": 1.64, "learning_rate": 4.2066905074464715e-05, "loss": 0.9138, "step": 27096 }, { "epoch": 1.64, "learning_rate": 4.206539616435049e-05, "loss": 0.85, "step": 27098 }, { "epoch": 1.64, "learning_rate": 4.206388725423627e-05, "loss": 0.9018, "step": 27100 }, { "epoch": 1.64, "learning_rate": 4.206237834412204e-05, "loss": 0.9508, "step": 27102 }, { "epoch": 1.64, "learning_rate": 4.2060869434007815e-05, "loss": 0.7807, "step": 27104 }, { "epoch": 1.64, "learning_rate": 4.2059360523893594e-05, "loss": 0.9881, "step": 27106 }, { "epoch": 1.64, "learning_rate": 4.205785161377937e-05, "loss": 0.7584, "step": 27108 }, { "epoch": 1.64, "learning_rate": 4.205634270366515e-05, "loss": 1.1486, "step": 27110 }, { "epoch": 1.64, "learning_rate": 4.205483379355092e-05, "loss": 0.7438, "step": 27112 }, { "epoch": 1.64, "learning_rate": 4.2053324883436694e-05, "loss": 0.9802, "step": 27114 }, { "epoch": 1.64, "learning_rate": 4.2051815973322474e-05, "loss": 0.9327, "step": 27116 }, { "epoch": 1.64, "learning_rate": 4.205030706320825e-05, "loss": 0.7117, "step": 27118 }, { "epoch": 1.64, "learning_rate": 4.204879815309402e-05, "loss": 1.1679, "step": 27120 }, { "epoch": 1.64, "learning_rate": 4.20472892429798e-05, "loss": 0.8738, "step": 27122 }, { "epoch": 1.64, "learning_rate": 4.204578033286557e-05, "loss": 0.766, "step": 27124 }, { "epoch": 1.64, "learning_rate": 4.2044271422751346e-05, "loss": 1.0324, "step": 27126 }, { "epoch": 1.64, "learning_rate": 4.2042762512637126e-05, "loss": 0.8342, "step": 27128 }, { "epoch": 1.64, "learning_rate": 4.20412536025229e-05, "loss": 0.8796, "step": 27130 }, { "epoch": 1.64, "learning_rate": 4.203974469240868e-05, "loss": 0.7063, "step": 27132 }, { "epoch": 1.64, "learning_rate": 4.203823578229445e-05, "loss": 0.8017, "step": 27134 }, { "epoch": 1.64, "learning_rate": 4.2036726872180225e-05, "loss": 0.9029, "step": 27136 }, { "epoch": 1.64, "learning_rate": 4.2035217962066005e-05, "loss": 0.918, "step": 27138 }, { "epoch": 1.64, "learning_rate": 4.203370905195178e-05, "loss": 0.7551, "step": 27140 }, { "epoch": 1.64, "learning_rate": 4.203220014183755e-05, "loss": 0.9944, "step": 27142 }, { "epoch": 1.64, "learning_rate": 4.203069123172333e-05, "loss": 0.6373, "step": 27144 }, { "epoch": 1.64, "learning_rate": 4.2029182321609104e-05, "loss": 0.7671, "step": 27146 }, { "epoch": 1.64, "learning_rate": 4.202767341149488e-05, "loss": 1.1913, "step": 27148 }, { "epoch": 1.64, "learning_rate": 4.202616450138066e-05, "loss": 0.5308, "step": 27150 }, { "epoch": 1.64, "learning_rate": 4.202465559126643e-05, "loss": 0.8318, "step": 27152 }, { "epoch": 1.64, "learning_rate": 4.2023146681152204e-05, "loss": 0.6219, "step": 27154 }, { "epoch": 1.64, "learning_rate": 4.2021637771037984e-05, "loss": 0.8027, "step": 27156 }, { "epoch": 1.64, "learning_rate": 4.202012886092376e-05, "loss": 0.951, "step": 27158 }, { "epoch": 1.64, "learning_rate": 4.2018619950809537e-05, "loss": 0.6108, "step": 27160 }, { "epoch": 1.64, "learning_rate": 4.20171110406953e-05, "loss": 1.0312, "step": 27162 }, { "epoch": 1.64, "learning_rate": 4.201560213058108e-05, "loss": 1.1479, "step": 27164 }, { "epoch": 1.64, "learning_rate": 4.201409322046686e-05, "loss": 0.7458, "step": 27166 }, { "epoch": 1.64, "learning_rate": 4.201258431035263e-05, "loss": 0.9005, "step": 27168 }, { "epoch": 1.64, "learning_rate": 4.201107540023841e-05, "loss": 0.7959, "step": 27170 }, { "epoch": 1.64, "learning_rate": 4.200956649012419e-05, "loss": 0.9368, "step": 27172 }, { "epoch": 1.64, "learning_rate": 4.200805758000996e-05, "loss": 1.1526, "step": 27174 }, { "epoch": 1.64, "learning_rate": 4.2006548669895735e-05, "loss": 0.9983, "step": 27176 }, { "epoch": 1.64, "learning_rate": 4.200503975978151e-05, "loss": 0.8407, "step": 27178 }, { "epoch": 1.64, "learning_rate": 4.200353084966729e-05, "loss": 0.9303, "step": 27180 }, { "epoch": 1.64, "learning_rate": 4.200202193955306e-05, "loss": 1.0243, "step": 27182 }, { "epoch": 1.64, "learning_rate": 4.2000513029438834e-05, "loss": 0.8459, "step": 27184 }, { "epoch": 1.64, "learning_rate": 4.1999004119324614e-05, "loss": 0.8668, "step": 27186 }, { "epoch": 1.64, "learning_rate": 4.1997495209210394e-05, "loss": 0.8746, "step": 27188 }, { "epoch": 1.64, "learning_rate": 4.199598629909616e-05, "loss": 0.7517, "step": 27190 }, { "epoch": 1.64, "learning_rate": 4.199447738898194e-05, "loss": 0.8416, "step": 27192 }, { "epoch": 1.64, "learning_rate": 4.1992968478867714e-05, "loss": 0.8825, "step": 27194 }, { "epoch": 1.64, "learning_rate": 4.1991459568753493e-05, "loss": 0.5793, "step": 27196 }, { "epoch": 1.64, "learning_rate": 4.198995065863927e-05, "loss": 0.9992, "step": 27198 }, { "epoch": 1.64, "learning_rate": 4.198844174852504e-05, "loss": 0.9608, "step": 27200 }, { "epoch": 1.64, "learning_rate": 4.198693283841082e-05, "loss": 1.0619, "step": 27202 }, { "epoch": 1.64, "learning_rate": 4.198542392829659e-05, "loss": 0.7357, "step": 27204 }, { "epoch": 1.64, "learning_rate": 4.1983915018182366e-05, "loss": 0.7285, "step": 27206 }, { "epoch": 1.64, "learning_rate": 4.1982406108068146e-05, "loss": 0.6536, "step": 27208 }, { "epoch": 1.64, "learning_rate": 4.198089719795392e-05, "loss": 0.859, "step": 27210 }, { "epoch": 1.64, "learning_rate": 4.197938828783969e-05, "loss": 0.7547, "step": 27212 }, { "epoch": 1.64, "learning_rate": 4.197787937772547e-05, "loss": 0.7698, "step": 27214 }, { "epoch": 1.64, "learning_rate": 4.1976370467611245e-05, "loss": 1.1771, "step": 27216 }, { "epoch": 1.64, "learning_rate": 4.197486155749702e-05, "loss": 0.6802, "step": 27218 }, { "epoch": 1.64, "learning_rate": 4.19733526473828e-05, "loss": 1.0895, "step": 27220 }, { "epoch": 1.64, "learning_rate": 4.197184373726857e-05, "loss": 0.8153, "step": 27222 }, { "epoch": 1.64, "learning_rate": 4.197033482715435e-05, "loss": 0.5363, "step": 27224 }, { "epoch": 1.64, "learning_rate": 4.1968825917040124e-05, "loss": 0.7459, "step": 27226 }, { "epoch": 1.64, "learning_rate": 4.19673170069259e-05, "loss": 0.8934, "step": 27228 }, { "epoch": 1.64, "learning_rate": 4.196580809681168e-05, "loss": 0.7538, "step": 27230 }, { "epoch": 1.64, "learning_rate": 4.196429918669745e-05, "loss": 1.1879, "step": 27232 }, { "epoch": 1.64, "learning_rate": 4.1962790276583224e-05, "loss": 0.9016, "step": 27234 }, { "epoch": 1.64, "learning_rate": 4.1961281366469003e-05, "loss": 1.1288, "step": 27236 }, { "epoch": 1.64, "learning_rate": 4.1959772456354777e-05, "loss": 0.7651, "step": 27238 }, { "epoch": 1.64, "learning_rate": 4.195826354624055e-05, "loss": 0.8099, "step": 27240 }, { "epoch": 1.64, "learning_rate": 4.195675463612633e-05, "loss": 0.8824, "step": 27242 }, { "epoch": 1.64, "learning_rate": 4.19552457260121e-05, "loss": 0.9225, "step": 27244 }, { "epoch": 1.64, "learning_rate": 4.195373681589788e-05, "loss": 0.7842, "step": 27246 }, { "epoch": 1.64, "learning_rate": 4.1952227905783656e-05, "loss": 0.7949, "step": 27248 }, { "epoch": 1.64, "learning_rate": 4.195071899566943e-05, "loss": 0.6101, "step": 27250 }, { "epoch": 1.64, "learning_rate": 4.194921008555521e-05, "loss": 0.7592, "step": 27252 }, { "epoch": 1.64, "learning_rate": 4.194770117544098e-05, "loss": 0.4689, "step": 27254 }, { "epoch": 1.65, "learning_rate": 4.1946192265326755e-05, "loss": 0.9662, "step": 27256 }, { "epoch": 1.65, "learning_rate": 4.1944683355212535e-05, "loss": 0.7489, "step": 27258 }, { "epoch": 1.65, "learning_rate": 4.194317444509831e-05, "loss": 0.5942, "step": 27260 }, { "epoch": 1.65, "learning_rate": 4.194166553498408e-05, "loss": 0.9121, "step": 27262 }, { "epoch": 1.65, "learning_rate": 4.194015662486986e-05, "loss": 0.9696, "step": 27264 }, { "epoch": 1.65, "learning_rate": 4.1938647714755634e-05, "loss": 0.9728, "step": 27266 }, { "epoch": 1.65, "learning_rate": 4.193713880464141e-05, "loss": 0.6434, "step": 27268 }, { "epoch": 1.65, "learning_rate": 4.193562989452719e-05, "loss": 0.638, "step": 27270 }, { "epoch": 1.65, "learning_rate": 4.193412098441296e-05, "loss": 0.8005, "step": 27272 }, { "epoch": 1.65, "learning_rate": 4.193261207429874e-05, "loss": 0.6587, "step": 27274 }, { "epoch": 1.65, "learning_rate": 4.1931103164184507e-05, "loss": 0.6777, "step": 27276 }, { "epoch": 1.65, "learning_rate": 4.1929594254070287e-05, "loss": 0.7418, "step": 27278 }, { "epoch": 1.65, "learning_rate": 4.1928085343956066e-05, "loss": 0.73, "step": 27280 }, { "epoch": 1.65, "learning_rate": 4.192657643384183e-05, "loss": 0.8035, "step": 27282 }, { "epoch": 1.65, "learning_rate": 4.192506752372761e-05, "loss": 0.8001, "step": 27284 }, { "epoch": 1.65, "learning_rate": 4.192355861361339e-05, "loss": 1.0587, "step": 27286 }, { "epoch": 1.65, "learning_rate": 4.1922049703499166e-05, "loss": 0.8874, "step": 27288 }, { "epoch": 1.65, "learning_rate": 4.192054079338494e-05, "loss": 0.8283, "step": 27290 }, { "epoch": 1.65, "learning_rate": 4.191903188327071e-05, "loss": 0.8638, "step": 27292 }, { "epoch": 1.65, "learning_rate": 4.191752297315649e-05, "loss": 0.7942, "step": 27294 }, { "epoch": 1.65, "learning_rate": 4.191601406304227e-05, "loss": 0.9467, "step": 27296 }, { "epoch": 1.65, "learning_rate": 4.191450515292804e-05, "loss": 0.9333, "step": 27298 }, { "epoch": 1.65, "learning_rate": 4.191299624281382e-05, "loss": 0.8048, "step": 27300 }, { "epoch": 1.65, "learning_rate": 4.19114873326996e-05, "loss": 1.0373, "step": 27302 }, { "epoch": 1.65, "learning_rate": 4.1909978422585364e-05, "loss": 0.8556, "step": 27304 }, { "epoch": 1.65, "learning_rate": 4.1908469512471144e-05, "loss": 0.6789, "step": 27306 }, { "epoch": 1.65, "learning_rate": 4.190696060235692e-05, "loss": 0.6271, "step": 27308 }, { "epoch": 1.65, "learning_rate": 4.19054516922427e-05, "loss": 0.7501, "step": 27310 }, { "epoch": 1.65, "learning_rate": 4.190394278212847e-05, "loss": 0.8454, "step": 27312 }, { "epoch": 1.65, "learning_rate": 4.1902433872014243e-05, "loss": 0.5827, "step": 27314 }, { "epoch": 1.65, "learning_rate": 4.190092496190002e-05, "loss": 0.8407, "step": 27316 }, { "epoch": 1.65, "learning_rate": 4.1899416051785796e-05, "loss": 0.8402, "step": 27318 }, { "epoch": 1.65, "learning_rate": 4.189790714167157e-05, "loss": 0.9066, "step": 27320 }, { "epoch": 1.65, "learning_rate": 4.189639823155735e-05, "loss": 0.8107, "step": 27322 }, { "epoch": 1.65, "learning_rate": 4.189488932144312e-05, "loss": 1.0766, "step": 27324 }, { "epoch": 1.65, "learning_rate": 4.1893380411328896e-05, "loss": 0.9649, "step": 27326 }, { "epoch": 1.65, "learning_rate": 4.1891871501214676e-05, "loss": 0.6761, "step": 27328 }, { "epoch": 1.65, "learning_rate": 4.189036259110045e-05, "loss": 0.9092, "step": 27330 }, { "epoch": 1.65, "learning_rate": 4.188885368098622e-05, "loss": 0.8524, "step": 27332 }, { "epoch": 1.65, "learning_rate": 4.1887344770872e-05, "loss": 1.013, "step": 27334 }, { "epoch": 1.65, "learning_rate": 4.1885835860757775e-05, "loss": 0.9323, "step": 27336 }, { "epoch": 1.65, "learning_rate": 4.1884326950643555e-05, "loss": 0.6963, "step": 27338 }, { "epoch": 1.65, "learning_rate": 4.188281804052933e-05, "loss": 0.6543, "step": 27340 }, { "epoch": 1.65, "learning_rate": 4.18813091304151e-05, "loss": 0.719, "step": 27342 }, { "epoch": 1.65, "learning_rate": 4.187980022030088e-05, "loss": 1.2068, "step": 27344 }, { "epoch": 1.65, "learning_rate": 4.1878291310186654e-05, "loss": 0.7722, "step": 27346 }, { "epoch": 1.65, "learning_rate": 4.187678240007243e-05, "loss": 0.8138, "step": 27348 }, { "epoch": 1.65, "learning_rate": 4.187527348995821e-05, "loss": 0.6241, "step": 27350 }, { "epoch": 1.65, "learning_rate": 4.187376457984398e-05, "loss": 0.819, "step": 27352 }, { "epoch": 1.65, "learning_rate": 4.187225566972975e-05, "loss": 0.8929, "step": 27354 }, { "epoch": 1.65, "learning_rate": 4.187074675961553e-05, "loss": 0.8653, "step": 27356 }, { "epoch": 1.65, "learning_rate": 4.1869237849501306e-05, "loss": 0.8126, "step": 27358 }, { "epoch": 1.65, "learning_rate": 4.1867728939387086e-05, "loss": 0.8415, "step": 27360 }, { "epoch": 1.65, "learning_rate": 4.186622002927286e-05, "loss": 0.8949, "step": 27362 }, { "epoch": 1.65, "learning_rate": 4.186471111915863e-05, "loss": 0.9639, "step": 27364 }, { "epoch": 1.65, "learning_rate": 4.186320220904441e-05, "loss": 0.8346, "step": 27366 }, { "epoch": 1.65, "learning_rate": 4.1861693298930186e-05, "loss": 1.037, "step": 27368 }, { "epoch": 1.65, "learning_rate": 4.186018438881596e-05, "loss": 1.1315, "step": 27370 }, { "epoch": 1.65, "learning_rate": 4.185867547870174e-05, "loss": 0.777, "step": 27372 }, { "epoch": 1.65, "learning_rate": 4.185716656858751e-05, "loss": 0.8313, "step": 27374 }, { "epoch": 1.65, "learning_rate": 4.1855657658473285e-05, "loss": 0.7744, "step": 27376 }, { "epoch": 1.65, "learning_rate": 4.1854148748359065e-05, "loss": 0.6144, "step": 27378 }, { "epoch": 1.65, "learning_rate": 4.185263983824484e-05, "loss": 0.8171, "step": 27380 }, { "epoch": 1.65, "learning_rate": 4.185113092813061e-05, "loss": 0.925, "step": 27382 }, { "epoch": 1.65, "learning_rate": 4.184962201801639e-05, "loss": 0.8731, "step": 27384 }, { "epoch": 1.65, "learning_rate": 4.1848113107902164e-05, "loss": 0.8688, "step": 27386 }, { "epoch": 1.65, "learning_rate": 4.1846604197787944e-05, "loss": 0.8399, "step": 27388 }, { "epoch": 1.65, "learning_rate": 4.184509528767371e-05, "loss": 0.7149, "step": 27390 }, { "epoch": 1.65, "learning_rate": 4.184358637755949e-05, "loss": 1.2734, "step": 27392 }, { "epoch": 1.65, "learning_rate": 4.184207746744527e-05, "loss": 0.675, "step": 27394 }, { "epoch": 1.65, "learning_rate": 4.1840568557331036e-05, "loss": 0.8868, "step": 27396 }, { "epoch": 1.65, "learning_rate": 4.1839059647216816e-05, "loss": 0.8539, "step": 27398 }, { "epoch": 1.65, "learning_rate": 4.1837550737102596e-05, "loss": 0.8343, "step": 27400 }, { "epoch": 1.65, "learning_rate": 4.183604182698837e-05, "loss": 0.945, "step": 27402 }, { "epoch": 1.65, "learning_rate": 4.183453291687414e-05, "loss": 0.802, "step": 27404 }, { "epoch": 1.65, "learning_rate": 4.1833024006759916e-05, "loss": 0.825, "step": 27406 }, { "epoch": 1.65, "learning_rate": 4.1831515096645696e-05, "loss": 1.0396, "step": 27408 }, { "epoch": 1.65, "learning_rate": 4.1830006186531475e-05, "loss": 0.8972, "step": 27410 }, { "epoch": 1.65, "learning_rate": 4.182849727641724e-05, "loss": 0.8506, "step": 27412 }, { "epoch": 1.65, "learning_rate": 4.182698836630302e-05, "loss": 1.1858, "step": 27414 }, { "epoch": 1.65, "learning_rate": 4.18254794561888e-05, "loss": 0.8246, "step": 27416 }, { "epoch": 1.65, "learning_rate": 4.182397054607457e-05, "loss": 0.641, "step": 27418 }, { "epoch": 1.65, "learning_rate": 4.182246163596035e-05, "loss": 0.9618, "step": 27420 }, { "epoch": 1.66, "learning_rate": 4.182095272584612e-05, "loss": 0.8809, "step": 27422 }, { "epoch": 1.66, "learning_rate": 4.18194438157319e-05, "loss": 1.0817, "step": 27424 }, { "epoch": 1.66, "learning_rate": 4.1817934905617674e-05, "loss": 0.9896, "step": 27426 }, { "epoch": 1.66, "learning_rate": 4.181642599550345e-05, "loss": 0.6993, "step": 27428 }, { "epoch": 1.66, "learning_rate": 4.181491708538923e-05, "loss": 1.0438, "step": 27430 }, { "epoch": 1.66, "learning_rate": 4.1813408175275e-05, "loss": 0.7754, "step": 27432 }, { "epoch": 1.66, "learning_rate": 4.181189926516077e-05, "loss": 0.6527, "step": 27434 }, { "epoch": 1.66, "learning_rate": 4.181039035504655e-05, "loss": 0.9324, "step": 27436 }, { "epoch": 1.66, "learning_rate": 4.1808881444932326e-05, "loss": 0.9366, "step": 27438 }, { "epoch": 1.66, "learning_rate": 4.18073725348181e-05, "loss": 1.4241, "step": 27440 }, { "epoch": 1.66, "learning_rate": 4.180586362470388e-05, "loss": 0.8585, "step": 27442 }, { "epoch": 1.66, "learning_rate": 4.180435471458965e-05, "loss": 0.7454, "step": 27444 }, { "epoch": 1.66, "learning_rate": 4.1802845804475426e-05, "loss": 0.8445, "step": 27446 }, { "epoch": 1.66, "learning_rate": 4.1801336894361205e-05, "loss": 0.7977, "step": 27448 }, { "epoch": 1.66, "learning_rate": 4.179982798424698e-05, "loss": 0.6465, "step": 27450 }, { "epoch": 1.66, "learning_rate": 4.179831907413276e-05, "loss": 0.9587, "step": 27452 }, { "epoch": 1.66, "learning_rate": 4.179681016401853e-05, "loss": 1.0594, "step": 27454 }, { "epoch": 1.66, "learning_rate": 4.1795301253904305e-05, "loss": 0.9775, "step": 27456 }, { "epoch": 1.66, "learning_rate": 4.1793792343790085e-05, "loss": 0.7543, "step": 27458 }, { "epoch": 1.66, "learning_rate": 4.179228343367586e-05, "loss": 1.0205, "step": 27460 }, { "epoch": 1.66, "learning_rate": 4.179077452356163e-05, "loss": 1.0013, "step": 27462 }, { "epoch": 1.66, "learning_rate": 4.178926561344741e-05, "loss": 0.7986, "step": 27464 }, { "epoch": 1.66, "learning_rate": 4.1787756703333184e-05, "loss": 0.5512, "step": 27466 }, { "epoch": 1.66, "learning_rate": 4.178624779321896e-05, "loss": 0.7779, "step": 27468 }, { "epoch": 1.66, "learning_rate": 4.178473888310474e-05, "loss": 0.579, "step": 27470 }, { "epoch": 1.66, "learning_rate": 4.178322997299051e-05, "loss": 0.9379, "step": 27472 }, { "epoch": 1.66, "learning_rate": 4.178172106287629e-05, "loss": 0.8126, "step": 27474 }, { "epoch": 1.66, "learning_rate": 4.178021215276206e-05, "loss": 0.7327, "step": 27476 }, { "epoch": 1.66, "learning_rate": 4.1778703242647836e-05, "loss": 0.7114, "step": 27478 }, { "epoch": 1.66, "learning_rate": 4.1777194332533616e-05, "loss": 0.8399, "step": 27480 }, { "epoch": 1.66, "learning_rate": 4.177568542241939e-05, "loss": 0.7148, "step": 27482 }, { "epoch": 1.66, "learning_rate": 4.177417651230516e-05, "loss": 0.6545, "step": 27484 }, { "epoch": 1.66, "learning_rate": 4.177266760219094e-05, "loss": 0.7688, "step": 27486 }, { "epoch": 1.66, "learning_rate": 4.1771158692076715e-05, "loss": 0.8787, "step": 27488 }, { "epoch": 1.66, "learning_rate": 4.176964978196249e-05, "loss": 0.9551, "step": 27490 }, { "epoch": 1.66, "learning_rate": 4.176814087184827e-05, "loss": 0.7907, "step": 27492 }, { "epoch": 1.66, "learning_rate": 4.176663196173404e-05, "loss": 0.7116, "step": 27494 }, { "epoch": 1.66, "learning_rate": 4.1765123051619815e-05, "loss": 0.4361, "step": 27496 }, { "epoch": 1.66, "learning_rate": 4.1763614141505595e-05, "loss": 0.9262, "step": 27498 }, { "epoch": 1.66, "learning_rate": 4.176210523139137e-05, "loss": 0.8996, "step": 27500 }, { "epoch": 1.66, "learning_rate": 4.176059632127715e-05, "loss": 0.6869, "step": 27502 }, { "epoch": 1.66, "learning_rate": 4.1759087411162914e-05, "loss": 0.8173, "step": 27504 }, { "epoch": 1.66, "learning_rate": 4.1757578501048694e-05, "loss": 0.7762, "step": 27506 }, { "epoch": 1.66, "learning_rate": 4.1756069590934474e-05, "loss": 0.8427, "step": 27508 }, { "epoch": 1.66, "learning_rate": 4.175456068082025e-05, "loss": 0.7507, "step": 27510 }, { "epoch": 1.66, "learning_rate": 4.175305177070602e-05, "loss": 1.0282, "step": 27512 }, { "epoch": 1.66, "learning_rate": 4.17515428605918e-05, "loss": 1.3051, "step": 27514 }, { "epoch": 1.66, "learning_rate": 4.175003395047757e-05, "loss": 0.6456, "step": 27516 }, { "epoch": 1.66, "learning_rate": 4.1748525040363346e-05, "loss": 0.7783, "step": 27518 }, { "epoch": 1.66, "learning_rate": 4.174701613024912e-05, "loss": 0.7647, "step": 27520 }, { "epoch": 1.66, "learning_rate": 4.17455072201349e-05, "loss": 0.6758, "step": 27522 }, { "epoch": 1.66, "learning_rate": 4.174399831002068e-05, "loss": 0.8283, "step": 27524 }, { "epoch": 1.66, "learning_rate": 4.1742489399906445e-05, "loss": 0.9299, "step": 27526 }, { "epoch": 1.66, "learning_rate": 4.1740980489792225e-05, "loss": 0.8936, "step": 27528 }, { "epoch": 1.66, "learning_rate": 4.1739471579678005e-05, "loss": 0.7901, "step": 27530 }, { "epoch": 1.66, "learning_rate": 4.173796266956377e-05, "loss": 0.6552, "step": 27532 }, { "epoch": 1.66, "learning_rate": 4.173645375944955e-05, "loss": 0.842, "step": 27534 }, { "epoch": 1.66, "learning_rate": 4.1734944849335325e-05, "loss": 0.9934, "step": 27536 }, { "epoch": 1.66, "learning_rate": 4.1733435939221105e-05, "loss": 0.6826, "step": 27538 }, { "epoch": 1.66, "learning_rate": 4.173192702910688e-05, "loss": 0.8853, "step": 27540 }, { "epoch": 1.66, "learning_rate": 4.173041811899265e-05, "loss": 0.9071, "step": 27542 }, { "epoch": 1.66, "learning_rate": 4.172890920887843e-05, "loss": 1.3019, "step": 27544 }, { "epoch": 1.66, "learning_rate": 4.1727400298764204e-05, "loss": 0.9299, "step": 27546 }, { "epoch": 1.66, "learning_rate": 4.172589138864998e-05, "loss": 0.8215, "step": 27548 }, { "epoch": 1.66, "learning_rate": 4.172438247853576e-05, "loss": 0.9424, "step": 27550 }, { "epoch": 1.66, "learning_rate": 4.172287356842153e-05, "loss": 0.7322, "step": 27552 }, { "epoch": 1.66, "learning_rate": 4.17213646583073e-05, "loss": 0.6098, "step": 27554 }, { "epoch": 1.66, "learning_rate": 4.171985574819308e-05, "loss": 0.973, "step": 27556 }, { "epoch": 1.66, "learning_rate": 4.1718346838078856e-05, "loss": 0.9458, "step": 27558 }, { "epoch": 1.66, "learning_rate": 4.171683792796463e-05, "loss": 0.8334, "step": 27560 }, { "epoch": 1.66, "learning_rate": 4.171532901785041e-05, "loss": 0.7335, "step": 27562 }, { "epoch": 1.66, "learning_rate": 4.171382010773618e-05, "loss": 0.9681, "step": 27564 }, { "epoch": 1.66, "learning_rate": 4.171231119762196e-05, "loss": 0.7451, "step": 27566 }, { "epoch": 1.66, "learning_rate": 4.1710802287507735e-05, "loss": 0.8239, "step": 27568 }, { "epoch": 1.66, "learning_rate": 4.170929337739351e-05, "loss": 0.965, "step": 27570 }, { "epoch": 1.66, "learning_rate": 4.170778446727929e-05, "loss": 0.6804, "step": 27572 }, { "epoch": 1.66, "learning_rate": 4.170627555716506e-05, "loss": 0.7788, "step": 27574 }, { "epoch": 1.66, "learning_rate": 4.1704766647050835e-05, "loss": 0.7907, "step": 27576 }, { "epoch": 1.66, "learning_rate": 4.1703257736936614e-05, "loss": 0.8294, "step": 27578 }, { "epoch": 1.66, "learning_rate": 4.170174882682239e-05, "loss": 0.6481, "step": 27580 }, { "epoch": 1.66, "learning_rate": 4.170023991670816e-05, "loss": 1.1101, "step": 27582 }, { "epoch": 1.66, "learning_rate": 4.169873100659394e-05, "loss": 1.1232, "step": 27584 }, { "epoch": 1.67, "learning_rate": 4.1697222096479714e-05, "loss": 1.0287, "step": 27586 }, { "epoch": 1.67, "learning_rate": 4.1695713186365494e-05, "loss": 1.0488, "step": 27588 }, { "epoch": 1.67, "learning_rate": 4.169420427625127e-05, "loss": 0.9241, "step": 27590 }, { "epoch": 1.67, "learning_rate": 4.169269536613704e-05, "loss": 0.8368, "step": 27592 }, { "epoch": 1.67, "learning_rate": 4.169118645602282e-05, "loss": 0.8393, "step": 27594 }, { "epoch": 1.67, "learning_rate": 4.168967754590859e-05, "loss": 1.067, "step": 27596 }, { "epoch": 1.67, "learning_rate": 4.1688168635794366e-05, "loss": 0.7912, "step": 27598 }, { "epoch": 1.67, "learning_rate": 4.1686659725680146e-05, "loss": 0.8224, "step": 27600 }, { "epoch": 1.67, "learning_rate": 4.168515081556592e-05, "loss": 0.8295, "step": 27602 }, { "epoch": 1.67, "learning_rate": 4.168364190545169e-05, "loss": 0.6602, "step": 27604 }, { "epoch": 1.67, "learning_rate": 4.168213299533747e-05, "loss": 0.7149, "step": 27606 }, { "epoch": 1.67, "learning_rate": 4.1680624085223245e-05, "loss": 1.0712, "step": 27608 }, { "epoch": 1.67, "learning_rate": 4.167911517510902e-05, "loss": 0.6688, "step": 27610 }, { "epoch": 1.67, "learning_rate": 4.16776062649948e-05, "loss": 0.8004, "step": 27612 }, { "epoch": 1.67, "learning_rate": 4.167609735488057e-05, "loss": 0.7205, "step": 27614 }, { "epoch": 1.67, "learning_rate": 4.167458844476635e-05, "loss": 0.8311, "step": 27616 }, { "epoch": 1.67, "learning_rate": 4.167307953465212e-05, "loss": 0.6796, "step": 27618 }, { "epoch": 1.67, "learning_rate": 4.16715706245379e-05, "loss": 0.6956, "step": 27620 }, { "epoch": 1.67, "learning_rate": 4.167006171442368e-05, "loss": 0.629, "step": 27622 }, { "epoch": 1.67, "learning_rate": 4.166855280430945e-05, "loss": 0.5396, "step": 27624 }, { "epoch": 1.67, "learning_rate": 4.1667043894195224e-05, "loss": 0.9168, "step": 27626 }, { "epoch": 1.67, "learning_rate": 4.1665534984081004e-05, "loss": 1.0896, "step": 27628 }, { "epoch": 1.67, "learning_rate": 4.166402607396678e-05, "loss": 0.749, "step": 27630 }, { "epoch": 1.67, "learning_rate": 4.166251716385255e-05, "loss": 0.8081, "step": 27632 }, { "epoch": 1.67, "learning_rate": 4.166100825373832e-05, "loss": 0.9724, "step": 27634 }, { "epoch": 1.67, "learning_rate": 4.16594993436241e-05, "loss": 0.7525, "step": 27636 }, { "epoch": 1.67, "learning_rate": 4.165799043350988e-05, "loss": 0.8171, "step": 27638 }, { "epoch": 1.67, "learning_rate": 4.165648152339565e-05, "loss": 0.763, "step": 27640 }, { "epoch": 1.67, "learning_rate": 4.165497261328143e-05, "loss": 1.1759, "step": 27642 }, { "epoch": 1.67, "learning_rate": 4.165346370316721e-05, "loss": 0.7275, "step": 27644 }, { "epoch": 1.67, "learning_rate": 4.1651954793052975e-05, "loss": 0.8105, "step": 27646 }, { "epoch": 1.67, "learning_rate": 4.1650445882938755e-05, "loss": 0.9738, "step": 27648 }, { "epoch": 1.67, "learning_rate": 4.164893697282453e-05, "loss": 0.491, "step": 27650 }, { "epoch": 1.67, "learning_rate": 4.164742806271031e-05, "loss": 0.949, "step": 27652 }, { "epoch": 1.67, "learning_rate": 4.164591915259608e-05, "loss": 0.7532, "step": 27654 }, { "epoch": 1.67, "learning_rate": 4.1644410242481854e-05, "loss": 0.7432, "step": 27656 }, { "epoch": 1.67, "learning_rate": 4.1642901332367634e-05, "loss": 0.8944, "step": 27658 }, { "epoch": 1.67, "learning_rate": 4.164139242225341e-05, "loss": 1.0623, "step": 27660 }, { "epoch": 1.67, "learning_rate": 4.163988351213918e-05, "loss": 1.0401, "step": 27662 }, { "epoch": 1.67, "learning_rate": 4.163837460202496e-05, "loss": 0.6075, "step": 27664 }, { "epoch": 1.67, "learning_rate": 4.1636865691910734e-05, "loss": 0.7023, "step": 27666 }, { "epoch": 1.67, "learning_rate": 4.163535678179651e-05, "loss": 0.8648, "step": 27668 }, { "epoch": 1.67, "learning_rate": 4.163384787168229e-05, "loss": 1.0, "step": 27670 }, { "epoch": 1.67, "learning_rate": 4.163233896156806e-05, "loss": 0.5253, "step": 27672 }, { "epoch": 1.67, "learning_rate": 4.163083005145384e-05, "loss": 0.882, "step": 27674 }, { "epoch": 1.67, "learning_rate": 4.162932114133961e-05, "loss": 0.6928, "step": 27676 }, { "epoch": 1.67, "learning_rate": 4.1627812231225386e-05, "loss": 1.0043, "step": 27678 }, { "epoch": 1.67, "learning_rate": 4.1626303321111166e-05, "loss": 0.8705, "step": 27680 }, { "epoch": 1.67, "learning_rate": 4.162479441099694e-05, "loss": 0.8537, "step": 27682 }, { "epoch": 1.67, "learning_rate": 4.162328550088271e-05, "loss": 0.6903, "step": 27684 }, { "epoch": 1.67, "learning_rate": 4.162177659076849e-05, "loss": 0.5503, "step": 27686 }, { "epoch": 1.67, "learning_rate": 4.1620267680654265e-05, "loss": 0.7722, "step": 27688 }, { "epoch": 1.67, "learning_rate": 4.161875877054004e-05, "loss": 0.7076, "step": 27690 }, { "epoch": 1.67, "learning_rate": 4.161724986042582e-05, "loss": 0.9726, "step": 27692 }, { "epoch": 1.67, "learning_rate": 4.161574095031159e-05, "loss": 1.0684, "step": 27694 }, { "epoch": 1.67, "learning_rate": 4.1614232040197364e-05, "loss": 0.8232, "step": 27696 }, { "epoch": 1.67, "learning_rate": 4.1612723130083144e-05, "loss": 0.9135, "step": 27698 }, { "epoch": 1.67, "learning_rate": 4.161121421996892e-05, "loss": 0.9001, "step": 27700 }, { "epoch": 1.67, "learning_rate": 4.16097053098547e-05, "loss": 0.6929, "step": 27702 }, { "epoch": 1.67, "learning_rate": 4.160819639974047e-05, "loss": 0.663, "step": 27704 }, { "epoch": 1.67, "learning_rate": 4.1606687489626244e-05, "loss": 0.899, "step": 27706 }, { "epoch": 1.67, "learning_rate": 4.1605178579512023e-05, "loss": 0.9655, "step": 27708 }, { "epoch": 1.67, "learning_rate": 4.1603669669397797e-05, "loss": 0.6399, "step": 27710 }, { "epoch": 1.67, "learning_rate": 4.160216075928357e-05, "loss": 0.8947, "step": 27712 }, { "epoch": 1.67, "learning_rate": 4.160065184916935e-05, "loss": 0.934, "step": 27714 }, { "epoch": 1.67, "learning_rate": 4.159914293905512e-05, "loss": 0.8156, "step": 27716 }, { "epoch": 1.67, "learning_rate": 4.1597634028940896e-05, "loss": 0.8823, "step": 27718 }, { "epoch": 1.67, "learning_rate": 4.1596125118826676e-05, "loss": 0.6845, "step": 27720 }, { "epoch": 1.67, "learning_rate": 4.159461620871245e-05, "loss": 1.026, "step": 27722 }, { "epoch": 1.67, "learning_rate": 4.159310729859822e-05, "loss": 0.9844, "step": 27724 }, { "epoch": 1.67, "learning_rate": 4.1591598388484e-05, "loss": 0.5724, "step": 27726 }, { "epoch": 1.67, "learning_rate": 4.1590089478369775e-05, "loss": 0.8545, "step": 27728 }, { "epoch": 1.67, "learning_rate": 4.1588580568255555e-05, "loss": 1.2076, "step": 27730 }, { "epoch": 1.67, "learning_rate": 4.158707165814132e-05, "loss": 0.9801, "step": 27732 }, { "epoch": 1.67, "learning_rate": 4.15855627480271e-05, "loss": 0.8016, "step": 27734 }, { "epoch": 1.67, "learning_rate": 4.158405383791288e-05, "loss": 0.7837, "step": 27736 }, { "epoch": 1.67, "learning_rate": 4.1582544927798654e-05, "loss": 0.8945, "step": 27738 }, { "epoch": 1.67, "learning_rate": 4.158103601768443e-05, "loss": 1.0049, "step": 27740 }, { "epoch": 1.67, "learning_rate": 4.157952710757021e-05, "loss": 1.0581, "step": 27742 }, { "epoch": 1.67, "learning_rate": 4.157801819745598e-05, "loss": 0.7334, "step": 27744 }, { "epoch": 1.67, "learning_rate": 4.1576509287341754e-05, "loss": 1.0133, "step": 27746 }, { "epoch": 1.67, "learning_rate": 4.157500037722753e-05, "loss": 0.4234, "step": 27748 }, { "epoch": 1.67, "learning_rate": 4.1573491467113307e-05, "loss": 0.8376, "step": 27750 }, { "epoch": 1.68, "learning_rate": 4.1571982556999086e-05, "loss": 0.9248, "step": 27752 }, { "epoch": 1.68, "learning_rate": 4.157047364688485e-05, "loss": 1.1276, "step": 27754 }, { "epoch": 1.68, "learning_rate": 4.156896473677063e-05, "loss": 1.0472, "step": 27756 }, { "epoch": 1.68, "learning_rate": 4.156745582665641e-05, "loss": 0.8684, "step": 27758 }, { "epoch": 1.68, "learning_rate": 4.156594691654218e-05, "loss": 0.7782, "step": 27760 }, { "epoch": 1.68, "learning_rate": 4.156443800642796e-05, "loss": 0.7208, "step": 27762 }, { "epoch": 1.68, "learning_rate": 4.156292909631373e-05, "loss": 0.9855, "step": 27764 }, { "epoch": 1.68, "learning_rate": 4.156142018619951e-05, "loss": 0.6984, "step": 27766 }, { "epoch": 1.68, "learning_rate": 4.1559911276085285e-05, "loss": 0.6322, "step": 27768 }, { "epoch": 1.68, "learning_rate": 4.155840236597106e-05, "loss": 0.7603, "step": 27770 }, { "epoch": 1.68, "learning_rate": 4.155689345585684e-05, "loss": 0.9848, "step": 27772 }, { "epoch": 1.68, "learning_rate": 4.155538454574261e-05, "loss": 0.9265, "step": 27774 }, { "epoch": 1.68, "learning_rate": 4.1553875635628384e-05, "loss": 0.5406, "step": 27776 }, { "epoch": 1.68, "learning_rate": 4.1552366725514164e-05, "loss": 1.0591, "step": 27778 }, { "epoch": 1.68, "learning_rate": 4.155085781539994e-05, "loss": 1.082, "step": 27780 }, { "epoch": 1.68, "learning_rate": 4.154934890528571e-05, "loss": 0.7197, "step": 27782 }, { "epoch": 1.68, "learning_rate": 4.154783999517149e-05, "loss": 0.612, "step": 27784 }, { "epoch": 1.68, "learning_rate": 4.1546331085057263e-05, "loss": 0.9239, "step": 27786 }, { "epoch": 1.68, "learning_rate": 4.154482217494304e-05, "loss": 0.7546, "step": 27788 }, { "epoch": 1.68, "learning_rate": 4.1543313264828816e-05, "loss": 0.7424, "step": 27790 }, { "epoch": 1.68, "learning_rate": 4.154180435471459e-05, "loss": 0.8055, "step": 27792 }, { "epoch": 1.68, "learning_rate": 4.154029544460037e-05, "loss": 0.8801, "step": 27794 }, { "epoch": 1.68, "learning_rate": 4.153878653448614e-05, "loss": 0.7367, "step": 27796 }, { "epoch": 1.68, "learning_rate": 4.1537277624371916e-05, "loss": 1.0546, "step": 27798 }, { "epoch": 1.68, "learning_rate": 4.1535768714257696e-05, "loss": 0.6914, "step": 27800 }, { "epoch": 1.68, "learning_rate": 4.153425980414347e-05, "loss": 0.8542, "step": 27802 }, { "epoch": 1.68, "learning_rate": 4.153275089402924e-05, "loss": 0.7307, "step": 27804 }, { "epoch": 1.68, "learning_rate": 4.153124198391502e-05, "loss": 0.7712, "step": 27806 }, { "epoch": 1.68, "learning_rate": 4.1529733073800795e-05, "loss": 0.5845, "step": 27808 }, { "epoch": 1.68, "learning_rate": 4.152822416368657e-05, "loss": 1.1392, "step": 27810 }, { "epoch": 1.68, "learning_rate": 4.152671525357235e-05, "loss": 0.7162, "step": 27812 }, { "epoch": 1.68, "learning_rate": 4.152520634345812e-05, "loss": 0.9142, "step": 27814 }, { "epoch": 1.68, "learning_rate": 4.15236974333439e-05, "loss": 0.8122, "step": 27816 }, { "epoch": 1.68, "learning_rate": 4.1522188523229674e-05, "loss": 0.7374, "step": 27818 }, { "epoch": 1.68, "learning_rate": 4.152067961311545e-05, "loss": 0.8913, "step": 27820 }, { "epoch": 1.68, "learning_rate": 4.151917070300123e-05, "loss": 0.7421, "step": 27822 }, { "epoch": 1.68, "learning_rate": 4.1517661792887e-05, "loss": 0.9786, "step": 27824 }, { "epoch": 1.68, "learning_rate": 4.1516152882772773e-05, "loss": 1.012, "step": 27826 }, { "epoch": 1.68, "learning_rate": 4.151464397265855e-05, "loss": 0.719, "step": 27828 }, { "epoch": 1.68, "learning_rate": 4.1513135062544326e-05, "loss": 0.8328, "step": 27830 }, { "epoch": 1.68, "learning_rate": 4.15116261524301e-05, "loss": 0.6872, "step": 27832 }, { "epoch": 1.68, "learning_rate": 4.151011724231588e-05, "loss": 0.8262, "step": 27834 }, { "epoch": 1.68, "learning_rate": 4.150860833220165e-05, "loss": 0.6532, "step": 27836 }, { "epoch": 1.68, "learning_rate": 4.150709942208743e-05, "loss": 0.9784, "step": 27838 }, { "epoch": 1.68, "learning_rate": 4.1505590511973206e-05, "loss": 0.6656, "step": 27840 }, { "epoch": 1.68, "learning_rate": 4.150408160185898e-05, "loss": 0.7243, "step": 27842 }, { "epoch": 1.68, "learning_rate": 4.150257269174476e-05, "loss": 0.654, "step": 27844 }, { "epoch": 1.68, "learning_rate": 4.1501063781630525e-05, "loss": 0.7254, "step": 27846 }, { "epoch": 1.68, "learning_rate": 4.1499554871516305e-05, "loss": 0.7273, "step": 27848 }, { "epoch": 1.68, "learning_rate": 4.1498045961402085e-05, "loss": 1.0158, "step": 27850 }, { "epoch": 1.68, "learning_rate": 4.149653705128786e-05, "loss": 0.6319, "step": 27852 }, { "epoch": 1.68, "learning_rate": 4.149502814117363e-05, "loss": 0.7827, "step": 27854 }, { "epoch": 1.68, "learning_rate": 4.149351923105941e-05, "loss": 0.946, "step": 27856 }, { "epoch": 1.68, "learning_rate": 4.1492010320945184e-05, "loss": 1.0602, "step": 27858 }, { "epoch": 1.68, "learning_rate": 4.149050141083096e-05, "loss": 0.6873, "step": 27860 }, { "epoch": 1.68, "learning_rate": 4.148899250071673e-05, "loss": 0.5481, "step": 27862 }, { "epoch": 1.68, "learning_rate": 4.148748359060251e-05, "loss": 0.6888, "step": 27864 }, { "epoch": 1.68, "learning_rate": 4.148597468048829e-05, "loss": 0.7758, "step": 27866 }, { "epoch": 1.68, "learning_rate": 4.1484465770374056e-05, "loss": 0.4873, "step": 27868 }, { "epoch": 1.68, "learning_rate": 4.1482956860259836e-05, "loss": 0.5429, "step": 27870 }, { "epoch": 1.68, "learning_rate": 4.1481447950145616e-05, "loss": 0.7273, "step": 27872 }, { "epoch": 1.68, "learning_rate": 4.147993904003138e-05, "loss": 0.8957, "step": 27874 }, { "epoch": 1.68, "learning_rate": 4.147843012991716e-05, "loss": 0.9643, "step": 27876 }, { "epoch": 1.68, "learning_rate": 4.1476921219802936e-05, "loss": 0.7476, "step": 27878 }, { "epoch": 1.68, "learning_rate": 4.1475412309688716e-05, "loss": 0.7507, "step": 27880 }, { "epoch": 1.68, "learning_rate": 4.147390339957449e-05, "loss": 1.1186, "step": 27882 }, { "epoch": 1.68, "learning_rate": 4.147239448946026e-05, "loss": 0.7694, "step": 27884 }, { "epoch": 1.68, "learning_rate": 4.147088557934604e-05, "loss": 0.6017, "step": 27886 }, { "epoch": 1.68, "learning_rate": 4.1469376669231815e-05, "loss": 0.7897, "step": 27888 }, { "epoch": 1.68, "learning_rate": 4.146786775911759e-05, "loss": 0.864, "step": 27890 }, { "epoch": 1.68, "learning_rate": 4.146635884900337e-05, "loss": 0.7783, "step": 27892 }, { "epoch": 1.68, "learning_rate": 4.146484993888914e-05, "loss": 1.0321, "step": 27894 }, { "epoch": 1.68, "learning_rate": 4.1463341028774914e-05, "loss": 1.1331, "step": 27896 }, { "epoch": 1.68, "learning_rate": 4.1461832118660694e-05, "loss": 0.8565, "step": 27898 }, { "epoch": 1.68, "learning_rate": 4.146032320854647e-05, "loss": 0.7442, "step": 27900 }, { "epoch": 1.68, "learning_rate": 4.145881429843225e-05, "loss": 0.9162, "step": 27902 }, { "epoch": 1.68, "learning_rate": 4.145730538831802e-05, "loss": 0.8534, "step": 27904 }, { "epoch": 1.68, "learning_rate": 4.145579647820379e-05, "loss": 0.9063, "step": 27906 }, { "epoch": 1.68, "learning_rate": 4.145428756808957e-05, "loss": 0.6777, "step": 27908 }, { "epoch": 1.68, "learning_rate": 4.1452778657975346e-05, "loss": 0.6946, "step": 27910 }, { "epoch": 1.68, "learning_rate": 4.145126974786112e-05, "loss": 0.9568, "step": 27912 }, { "epoch": 1.68, "learning_rate": 4.14497608377469e-05, "loss": 0.7846, "step": 27914 }, { "epoch": 1.68, "learning_rate": 4.144825192763267e-05, "loss": 1.079, "step": 27916 }, { "epoch": 1.69, "learning_rate": 4.1446743017518446e-05, "loss": 0.9319, "step": 27918 }, { "epoch": 1.69, "learning_rate": 4.1445234107404226e-05, "loss": 1.104, "step": 27920 }, { "epoch": 1.69, "learning_rate": 4.144372519729e-05, "loss": 0.8369, "step": 27922 }, { "epoch": 1.69, "learning_rate": 4.144221628717577e-05, "loss": 0.8005, "step": 27924 }, { "epoch": 1.69, "learning_rate": 4.144070737706155e-05, "loss": 0.7485, "step": 27926 }, { "epoch": 1.69, "learning_rate": 4.1439198466947325e-05, "loss": 0.677, "step": 27928 }, { "epoch": 1.69, "learning_rate": 4.1437689556833105e-05, "loss": 1.0069, "step": 27930 }, { "epoch": 1.69, "learning_rate": 4.143618064671888e-05, "loss": 0.8144, "step": 27932 }, { "epoch": 1.69, "learning_rate": 4.143467173660465e-05, "loss": 0.8281, "step": 27934 }, { "epoch": 1.69, "learning_rate": 4.143316282649043e-05, "loss": 0.8067, "step": 27936 }, { "epoch": 1.69, "learning_rate": 4.1431653916376204e-05, "loss": 0.9369, "step": 27938 }, { "epoch": 1.69, "learning_rate": 4.143014500626198e-05, "loss": 0.9757, "step": 27940 }, { "epoch": 1.69, "learning_rate": 4.142863609614776e-05, "loss": 0.9617, "step": 27942 }, { "epoch": 1.69, "learning_rate": 4.142712718603353e-05, "loss": 0.6865, "step": 27944 }, { "epoch": 1.69, "learning_rate": 4.14256182759193e-05, "loss": 0.9244, "step": 27946 }, { "epoch": 1.69, "learning_rate": 4.142410936580508e-05, "loss": 1.0318, "step": 27948 }, { "epoch": 1.69, "learning_rate": 4.1422600455690856e-05, "loss": 0.6463, "step": 27950 }, { "epoch": 1.69, "learning_rate": 4.1421091545576636e-05, "loss": 0.7993, "step": 27952 }, { "epoch": 1.69, "learning_rate": 4.141958263546241e-05, "loss": 1.0428, "step": 27954 }, { "epoch": 1.69, "learning_rate": 4.141807372534818e-05, "loss": 0.9844, "step": 27956 }, { "epoch": 1.69, "learning_rate": 4.141656481523396e-05, "loss": 0.7514, "step": 27958 }, { "epoch": 1.69, "learning_rate": 4.141505590511973e-05, "loss": 0.8799, "step": 27960 }, { "epoch": 1.69, "learning_rate": 4.141354699500551e-05, "loss": 0.9389, "step": 27962 }, { "epoch": 1.69, "learning_rate": 4.141203808489129e-05, "loss": 0.742, "step": 27964 }, { "epoch": 1.69, "learning_rate": 4.141052917477706e-05, "loss": 0.8922, "step": 27966 }, { "epoch": 1.69, "learning_rate": 4.1409020264662835e-05, "loss": 0.7858, "step": 27968 }, { "epoch": 1.69, "learning_rate": 4.1407511354548615e-05, "loss": 0.772, "step": 27970 }, { "epoch": 1.69, "learning_rate": 4.140600244443439e-05, "loss": 0.6539, "step": 27972 }, { "epoch": 1.69, "learning_rate": 4.140449353432016e-05, "loss": 0.6887, "step": 27974 }, { "epoch": 1.69, "learning_rate": 4.1402984624205934e-05, "loss": 0.9575, "step": 27976 }, { "epoch": 1.69, "learning_rate": 4.1401475714091714e-05, "loss": 0.8557, "step": 27978 }, { "epoch": 1.69, "learning_rate": 4.1399966803977494e-05, "loss": 0.7786, "step": 27980 }, { "epoch": 1.69, "learning_rate": 4.139845789386326e-05, "loss": 1.0205, "step": 27982 }, { "epoch": 1.69, "learning_rate": 4.139694898374904e-05, "loss": 0.9864, "step": 27984 }, { "epoch": 1.69, "learning_rate": 4.139544007363482e-05, "loss": 1.0561, "step": 27986 }, { "epoch": 1.69, "learning_rate": 4.1393931163520586e-05, "loss": 1.0676, "step": 27988 }, { "epoch": 1.69, "learning_rate": 4.1392422253406366e-05, "loss": 0.6945, "step": 27990 }, { "epoch": 1.69, "learning_rate": 4.139091334329214e-05, "loss": 1.1913, "step": 27992 }, { "epoch": 1.69, "learning_rate": 4.138940443317792e-05, "loss": 1.0561, "step": 27994 }, { "epoch": 1.69, "learning_rate": 4.138789552306369e-05, "loss": 1.1324, "step": 27996 }, { "epoch": 1.69, "learning_rate": 4.1386386612949465e-05, "loss": 0.6319, "step": 27998 }, { "epoch": 1.69, "learning_rate": 4.1384877702835245e-05, "loss": 0.6374, "step": 28000 }, { "epoch": 1.69, "learning_rate": 4.1383368792721025e-05, "loss": 0.6801, "step": 28002 }, { "epoch": 1.69, "learning_rate": 4.138185988260679e-05, "loss": 1.0402, "step": 28004 }, { "epoch": 1.69, "learning_rate": 4.138035097249257e-05, "loss": 0.5586, "step": 28006 }, { "epoch": 1.69, "learning_rate": 4.1378842062378345e-05, "loss": 1.3677, "step": 28008 }, { "epoch": 1.69, "learning_rate": 4.137733315226412e-05, "loss": 0.7255, "step": 28010 }, { "epoch": 1.69, "learning_rate": 4.13758242421499e-05, "loss": 0.7842, "step": 28012 }, { "epoch": 1.69, "learning_rate": 4.137431533203567e-05, "loss": 0.9191, "step": 28014 }, { "epoch": 1.69, "learning_rate": 4.137280642192145e-05, "loss": 0.6969, "step": 28016 }, { "epoch": 1.69, "learning_rate": 4.1371297511807224e-05, "loss": 0.7003, "step": 28018 }, { "epoch": 1.69, "learning_rate": 4.1369788601693e-05, "loss": 0.8045, "step": 28020 }, { "epoch": 1.69, "learning_rate": 4.136827969157878e-05, "loss": 1.0449, "step": 28022 }, { "epoch": 1.69, "learning_rate": 4.136677078146455e-05, "loss": 0.6471, "step": 28024 }, { "epoch": 1.69, "learning_rate": 4.136526187135032e-05, "loss": 0.7538, "step": 28026 }, { "epoch": 1.69, "learning_rate": 4.13637529612361e-05, "loss": 0.8696, "step": 28028 }, { "epoch": 1.69, "learning_rate": 4.1362244051121876e-05, "loss": 0.8807, "step": 28030 }, { "epoch": 1.69, "learning_rate": 4.136073514100765e-05, "loss": 0.6375, "step": 28032 }, { "epoch": 1.69, "learning_rate": 4.135922623089343e-05, "loss": 0.9596, "step": 28034 }, { "epoch": 1.69, "learning_rate": 4.13577173207792e-05, "loss": 0.8353, "step": 28036 }, { "epoch": 1.69, "learning_rate": 4.1356208410664975e-05, "loss": 0.6158, "step": 28038 }, { "epoch": 1.69, "learning_rate": 4.1354699500550755e-05, "loss": 0.6663, "step": 28040 }, { "epoch": 1.69, "learning_rate": 4.135319059043653e-05, "loss": 0.7966, "step": 28042 }, { "epoch": 1.69, "learning_rate": 4.135168168032231e-05, "loss": 1.0284, "step": 28044 }, { "epoch": 1.69, "learning_rate": 4.135017277020808e-05, "loss": 0.8425, "step": 28046 }, { "epoch": 1.69, "learning_rate": 4.1348663860093855e-05, "loss": 0.5539, "step": 28048 }, { "epoch": 1.69, "learning_rate": 4.1347154949979635e-05, "loss": 0.6925, "step": 28050 }, { "epoch": 1.69, "learning_rate": 4.134564603986541e-05, "loss": 1.3428, "step": 28052 }, { "epoch": 1.69, "learning_rate": 4.134413712975118e-05, "loss": 1.1047, "step": 28054 }, { "epoch": 1.69, "learning_rate": 4.134262821963696e-05, "loss": 0.8138, "step": 28056 }, { "epoch": 1.69, "learning_rate": 4.1341119309522734e-05, "loss": 0.8018, "step": 28058 }, { "epoch": 1.69, "learning_rate": 4.133961039940851e-05, "loss": 0.7263, "step": 28060 }, { "epoch": 1.69, "learning_rate": 4.133810148929429e-05, "loss": 0.8339, "step": 28062 }, { "epoch": 1.69, "learning_rate": 4.133659257918006e-05, "loss": 1.0087, "step": 28064 }, { "epoch": 1.69, "learning_rate": 4.133508366906584e-05, "loss": 0.7142, "step": 28066 }, { "epoch": 1.69, "learning_rate": 4.133357475895161e-05, "loss": 0.9385, "step": 28068 }, { "epoch": 1.69, "learning_rate": 4.1332065848837386e-05, "loss": 1.0149, "step": 28070 }, { "epoch": 1.69, "learning_rate": 4.1330556938723166e-05, "loss": 0.8937, "step": 28072 }, { "epoch": 1.69, "learning_rate": 4.132904802860893e-05, "loss": 0.7277, "step": 28074 }, { "epoch": 1.69, "learning_rate": 4.132753911849471e-05, "loss": 0.7621, "step": 28076 }, { "epoch": 1.69, "learning_rate": 4.132603020838049e-05, "loss": 1.0254, "step": 28078 }, { "epoch": 1.69, "learning_rate": 4.1324521298266265e-05, "loss": 0.9338, "step": 28080 }, { "epoch": 1.69, "learning_rate": 4.132301238815204e-05, "loss": 0.675, "step": 28082 }, { "epoch": 1.7, "learning_rate": 4.132150347803782e-05, "loss": 0.7195, "step": 28084 }, { "epoch": 1.7, "learning_rate": 4.131999456792359e-05, "loss": 0.5161, "step": 28086 }, { "epoch": 1.7, "learning_rate": 4.1318485657809365e-05, "loss": 0.8828, "step": 28088 }, { "epoch": 1.7, "learning_rate": 4.131697674769514e-05, "loss": 0.6536, "step": 28090 }, { "epoch": 1.7, "learning_rate": 4.131546783758092e-05, "loss": 0.8832, "step": 28092 }, { "epoch": 1.7, "learning_rate": 4.13139589274667e-05, "loss": 0.6445, "step": 28094 }, { "epoch": 1.7, "learning_rate": 4.1312450017352464e-05, "loss": 0.6602, "step": 28096 }, { "epoch": 1.7, "learning_rate": 4.1310941107238244e-05, "loss": 0.928, "step": 28098 }, { "epoch": 1.7, "learning_rate": 4.1309432197124024e-05, "loss": 0.6757, "step": 28100 }, { "epoch": 1.7, "learning_rate": 4.130792328700979e-05, "loss": 0.8186, "step": 28102 }, { "epoch": 1.7, "learning_rate": 4.130641437689557e-05, "loss": 0.7511, "step": 28104 }, { "epoch": 1.7, "learning_rate": 4.130490546678134e-05, "loss": 1.0703, "step": 28106 }, { "epoch": 1.7, "learning_rate": 4.130339655666712e-05, "loss": 0.6353, "step": 28108 }, { "epoch": 1.7, "learning_rate": 4.1301887646552896e-05, "loss": 0.9346, "step": 28110 }, { "epoch": 1.7, "learning_rate": 4.130037873643867e-05, "loss": 0.9683, "step": 28112 }, { "epoch": 1.7, "learning_rate": 4.129886982632445e-05, "loss": 0.6582, "step": 28114 }, { "epoch": 1.7, "learning_rate": 4.129736091621023e-05, "loss": 0.9452, "step": 28116 }, { "epoch": 1.7, "learning_rate": 4.1295852006095995e-05, "loss": 0.9425, "step": 28118 }, { "epoch": 1.7, "learning_rate": 4.1294343095981775e-05, "loss": 0.7826, "step": 28120 }, { "epoch": 1.7, "learning_rate": 4.129283418586755e-05, "loss": 0.82, "step": 28122 }, { "epoch": 1.7, "learning_rate": 4.129132527575332e-05, "loss": 0.9638, "step": 28124 }, { "epoch": 1.7, "learning_rate": 4.12898163656391e-05, "loss": 0.86, "step": 28126 }, { "epoch": 1.7, "learning_rate": 4.1288307455524874e-05, "loss": 1.005, "step": 28128 }, { "epoch": 1.7, "learning_rate": 4.1286798545410654e-05, "loss": 0.7261, "step": 28130 }, { "epoch": 1.7, "learning_rate": 4.128528963529643e-05, "loss": 0.7427, "step": 28132 }, { "epoch": 1.7, "learning_rate": 4.12837807251822e-05, "loss": 0.7074, "step": 28134 }, { "epoch": 1.7, "learning_rate": 4.128227181506798e-05, "loss": 0.6453, "step": 28136 }, { "epoch": 1.7, "learning_rate": 4.1280762904953754e-05, "loss": 0.6708, "step": 28138 }, { "epoch": 1.7, "learning_rate": 4.127925399483953e-05, "loss": 0.9554, "step": 28140 }, { "epoch": 1.7, "learning_rate": 4.127774508472531e-05, "loss": 0.6998, "step": 28142 }, { "epoch": 1.7, "learning_rate": 4.127623617461108e-05, "loss": 1.0096, "step": 28144 }, { "epoch": 1.7, "learning_rate": 4.127472726449685e-05, "loss": 0.6412, "step": 28146 }, { "epoch": 1.7, "learning_rate": 4.127321835438263e-05, "loss": 0.8434, "step": 28148 }, { "epoch": 1.7, "learning_rate": 4.1271709444268406e-05, "loss": 0.9652, "step": 28150 }, { "epoch": 1.7, "learning_rate": 4.127020053415418e-05, "loss": 0.6121, "step": 28152 }, { "epoch": 1.7, "learning_rate": 4.126869162403996e-05, "loss": 0.9081, "step": 28154 }, { "epoch": 1.7, "learning_rate": 4.126718271392573e-05, "loss": 0.6835, "step": 28156 }, { "epoch": 1.7, "learning_rate": 4.126567380381151e-05, "loss": 0.9013, "step": 28158 }, { "epoch": 1.7, "learning_rate": 4.1264164893697285e-05, "loss": 0.8413, "step": 28160 }, { "epoch": 1.7, "learning_rate": 4.126265598358306e-05, "loss": 0.6651, "step": 28162 }, { "epoch": 1.7, "learning_rate": 4.126114707346884e-05, "loss": 0.7176, "step": 28164 }, { "epoch": 1.7, "learning_rate": 4.125963816335461e-05, "loss": 0.8572, "step": 28166 }, { "epoch": 1.7, "learning_rate": 4.1258129253240384e-05, "loss": 1.0086, "step": 28168 }, { "epoch": 1.7, "learning_rate": 4.1256620343126164e-05, "loss": 0.7908, "step": 28170 }, { "epoch": 1.7, "learning_rate": 4.125511143301194e-05, "loss": 1.1059, "step": 28172 }, { "epoch": 1.7, "learning_rate": 4.125360252289771e-05, "loss": 0.9206, "step": 28174 }, { "epoch": 1.7, "learning_rate": 4.125209361278349e-05, "loss": 0.9203, "step": 28176 }, { "epoch": 1.7, "learning_rate": 4.1250584702669264e-05, "loss": 1.1565, "step": 28178 }, { "epoch": 1.7, "learning_rate": 4.1249075792555044e-05, "loss": 0.635, "step": 28180 }, { "epoch": 1.7, "learning_rate": 4.124756688244082e-05, "loss": 1.0753, "step": 28182 }, { "epoch": 1.7, "learning_rate": 4.124605797232659e-05, "loss": 0.8692, "step": 28184 }, { "epoch": 1.7, "learning_rate": 4.124454906221237e-05, "loss": 0.8145, "step": 28186 }, { "epoch": 1.7, "learning_rate": 4.1243040152098136e-05, "loss": 0.9537, "step": 28188 }, { "epoch": 1.7, "learning_rate": 4.1241531241983916e-05, "loss": 0.915, "step": 28190 }, { "epoch": 1.7, "learning_rate": 4.1240022331869696e-05, "loss": 0.6695, "step": 28192 }, { "epoch": 1.7, "learning_rate": 4.123851342175547e-05, "loss": 0.8256, "step": 28194 }, { "epoch": 1.7, "learning_rate": 4.123700451164124e-05, "loss": 0.8059, "step": 28196 }, { "epoch": 1.7, "learning_rate": 4.123549560152702e-05, "loss": 0.9072, "step": 28198 }, { "epoch": 1.7, "learning_rate": 4.1233986691412795e-05, "loss": 0.6202, "step": 28200 }, { "epoch": 1.7, "learning_rate": 4.123247778129857e-05, "loss": 0.9643, "step": 28202 }, { "epoch": 1.7, "learning_rate": 4.123096887118434e-05, "loss": 0.6347, "step": 28204 }, { "epoch": 1.7, "learning_rate": 4.122945996107012e-05, "loss": 0.7965, "step": 28206 }, { "epoch": 1.7, "learning_rate": 4.12279510509559e-05, "loss": 0.7325, "step": 28208 }, { "epoch": 1.7, "learning_rate": 4.122644214084167e-05, "loss": 0.8693, "step": 28210 }, { "epoch": 1.7, "learning_rate": 4.122493323072745e-05, "loss": 1.1225, "step": 28212 }, { "epoch": 1.7, "learning_rate": 4.122342432061323e-05, "loss": 0.6073, "step": 28214 }, { "epoch": 1.7, "learning_rate": 4.1221915410499e-05, "loss": 0.9887, "step": 28216 }, { "epoch": 1.7, "learning_rate": 4.1220406500384774e-05, "loss": 0.8438, "step": 28218 }, { "epoch": 1.7, "learning_rate": 4.121889759027055e-05, "loss": 0.9242, "step": 28220 }, { "epoch": 1.7, "learning_rate": 4.1217388680156327e-05, "loss": 1.0021, "step": 28222 }, { "epoch": 1.7, "learning_rate": 4.12158797700421e-05, "loss": 0.8019, "step": 28224 }, { "epoch": 1.7, "learning_rate": 4.121437085992787e-05, "loss": 1.1681, "step": 28226 }, { "epoch": 1.7, "learning_rate": 4.121286194981365e-05, "loss": 0.843, "step": 28228 }, { "epoch": 1.7, "learning_rate": 4.121135303969943e-05, "loss": 0.7323, "step": 28230 }, { "epoch": 1.7, "learning_rate": 4.12098441295852e-05, "loss": 0.714, "step": 28232 }, { "epoch": 1.7, "learning_rate": 4.120833521947098e-05, "loss": 0.8694, "step": 28234 }, { "epoch": 1.7, "learning_rate": 4.120682630935675e-05, "loss": 0.9228, "step": 28236 }, { "epoch": 1.7, "learning_rate": 4.1205317399242525e-05, "loss": 0.8738, "step": 28238 }, { "epoch": 1.7, "learning_rate": 4.1203808489128305e-05, "loss": 0.9283, "step": 28240 }, { "epoch": 1.7, "learning_rate": 4.120229957901408e-05, "loss": 0.708, "step": 28242 }, { "epoch": 1.7, "learning_rate": 4.120079066889986e-05, "loss": 0.8177, "step": 28244 }, { "epoch": 1.7, "learning_rate": 4.119928175878563e-05, "loss": 0.8281, "step": 28246 }, { "epoch": 1.7, "learning_rate": 4.1197772848671404e-05, "loss": 0.6625, "step": 28248 }, { "epoch": 1.71, "learning_rate": 4.1196263938557184e-05, "loss": 0.9744, "step": 28250 }, { "epoch": 1.71, "learning_rate": 4.119475502844296e-05, "loss": 0.82, "step": 28252 }, { "epoch": 1.71, "learning_rate": 4.119324611832873e-05, "loss": 0.7078, "step": 28254 }, { "epoch": 1.71, "learning_rate": 4.119173720821451e-05, "loss": 0.6565, "step": 28256 }, { "epoch": 1.71, "learning_rate": 4.1190228298100284e-05, "loss": 0.9397, "step": 28258 }, { "epoch": 1.71, "learning_rate": 4.118871938798606e-05, "loss": 0.9546, "step": 28260 }, { "epoch": 1.71, "learning_rate": 4.1187210477871837e-05, "loss": 0.6217, "step": 28262 }, { "epoch": 1.71, "learning_rate": 4.118570156775761e-05, "loss": 0.9529, "step": 28264 }, { "epoch": 1.71, "learning_rate": 4.118419265764338e-05, "loss": 0.7372, "step": 28266 }, { "epoch": 1.71, "learning_rate": 4.118268374752916e-05, "loss": 0.7665, "step": 28268 }, { "epoch": 1.71, "learning_rate": 4.1181174837414936e-05, "loss": 0.6706, "step": 28270 }, { "epoch": 1.71, "learning_rate": 4.1179665927300716e-05, "loss": 1.2234, "step": 28272 }, { "epoch": 1.71, "learning_rate": 4.117815701718649e-05, "loss": 0.9879, "step": 28274 }, { "epoch": 1.71, "learning_rate": 4.117664810707226e-05, "loss": 0.7441, "step": 28276 }, { "epoch": 1.71, "learning_rate": 4.117513919695804e-05, "loss": 0.7168, "step": 28278 }, { "epoch": 1.71, "learning_rate": 4.1173630286843815e-05, "loss": 0.6275, "step": 28280 }, { "epoch": 1.71, "learning_rate": 4.117212137672959e-05, "loss": 0.8049, "step": 28282 }, { "epoch": 1.71, "learning_rate": 4.117061246661537e-05, "loss": 1.0383, "step": 28284 }, { "epoch": 1.71, "learning_rate": 4.116910355650114e-05, "loss": 1.0032, "step": 28286 }, { "epoch": 1.71, "learning_rate": 4.1167594646386914e-05, "loss": 0.701, "step": 28288 }, { "epoch": 1.71, "learning_rate": 4.1166085736272694e-05, "loss": 0.7201, "step": 28290 }, { "epoch": 1.71, "learning_rate": 4.116457682615847e-05, "loss": 0.8687, "step": 28292 }, { "epoch": 1.71, "learning_rate": 4.116306791604425e-05, "loss": 0.8995, "step": 28294 }, { "epoch": 1.71, "learning_rate": 4.116155900593002e-05, "loss": 0.8474, "step": 28296 }, { "epoch": 1.71, "learning_rate": 4.1160050095815793e-05, "loss": 0.8812, "step": 28298 }, { "epoch": 1.71, "learning_rate": 4.115854118570157e-05, "loss": 0.8128, "step": 28300 }, { "epoch": 1.71, "learning_rate": 4.115703227558734e-05, "loss": 0.9721, "step": 28302 }, { "epoch": 1.71, "learning_rate": 4.115552336547312e-05, "loss": 0.7384, "step": 28304 }, { "epoch": 1.71, "learning_rate": 4.11540144553589e-05, "loss": 0.6638, "step": 28306 }, { "epoch": 1.71, "learning_rate": 4.115250554524467e-05, "loss": 0.9021, "step": 28308 }, { "epoch": 1.71, "learning_rate": 4.1150996635130446e-05, "loss": 0.6575, "step": 28310 }, { "epoch": 1.71, "learning_rate": 4.1149487725016226e-05, "loss": 0.8698, "step": 28312 }, { "epoch": 1.71, "learning_rate": 4.1147978814902e-05, "loss": 0.7854, "step": 28314 }, { "epoch": 1.71, "learning_rate": 4.114646990478777e-05, "loss": 0.9861, "step": 28316 }, { "epoch": 1.71, "learning_rate": 4.1144960994673545e-05, "loss": 0.7081, "step": 28318 }, { "epoch": 1.71, "learning_rate": 4.1143452084559325e-05, "loss": 0.8528, "step": 28320 }, { "epoch": 1.71, "learning_rate": 4.1141943174445105e-05, "loss": 0.9238, "step": 28322 }, { "epoch": 1.71, "learning_rate": 4.114043426433087e-05, "loss": 0.7677, "step": 28324 }, { "epoch": 1.71, "learning_rate": 4.113892535421665e-05, "loss": 1.003, "step": 28326 }, { "epoch": 1.71, "learning_rate": 4.113741644410243e-05, "loss": 0.8191, "step": 28328 }, { "epoch": 1.71, "learning_rate": 4.1135907533988204e-05, "loss": 0.7995, "step": 28330 }, { "epoch": 1.71, "learning_rate": 4.113439862387398e-05, "loss": 0.6947, "step": 28332 }, { "epoch": 1.71, "learning_rate": 4.113288971375975e-05, "loss": 0.8397, "step": 28334 }, { "epoch": 1.71, "learning_rate": 4.113138080364553e-05, "loss": 1.0128, "step": 28336 }, { "epoch": 1.71, "learning_rate": 4.1129871893531303e-05, "loss": 1.269, "step": 28338 }, { "epoch": 1.71, "learning_rate": 4.1128362983417077e-05, "loss": 0.6757, "step": 28340 }, { "epoch": 1.71, "learning_rate": 4.1126854073302856e-05, "loss": 0.8557, "step": 28342 }, { "epoch": 1.71, "learning_rate": 4.1125345163188636e-05, "loss": 0.6857, "step": 28344 }, { "epoch": 1.71, "learning_rate": 4.11238362530744e-05, "loss": 0.9307, "step": 28346 }, { "epoch": 1.71, "learning_rate": 4.112232734296018e-05, "loss": 0.7839, "step": 28348 }, { "epoch": 1.71, "learning_rate": 4.1120818432845956e-05, "loss": 0.6141, "step": 28350 }, { "epoch": 1.71, "learning_rate": 4.111930952273173e-05, "loss": 0.7777, "step": 28352 }, { "epoch": 1.71, "learning_rate": 4.111780061261751e-05, "loss": 0.6775, "step": 28354 }, { "epoch": 1.71, "learning_rate": 4.111629170250328e-05, "loss": 1.0321, "step": 28356 }, { "epoch": 1.71, "learning_rate": 4.111478279238906e-05, "loss": 0.8346, "step": 28358 }, { "epoch": 1.71, "learning_rate": 4.1113273882274835e-05, "loss": 0.9852, "step": 28360 }, { "epoch": 1.71, "learning_rate": 4.111176497216061e-05, "loss": 1.0377, "step": 28362 }, { "epoch": 1.71, "learning_rate": 4.111025606204639e-05, "loss": 0.9941, "step": 28364 }, { "epoch": 1.71, "learning_rate": 4.110874715193216e-05, "loss": 0.8201, "step": 28366 }, { "epoch": 1.71, "learning_rate": 4.1107238241817934e-05, "loss": 0.5736, "step": 28368 }, { "epoch": 1.71, "learning_rate": 4.1105729331703714e-05, "loss": 0.9852, "step": 28370 }, { "epoch": 1.71, "learning_rate": 4.110422042158949e-05, "loss": 0.7418, "step": 28372 }, { "epoch": 1.71, "learning_rate": 4.110271151147526e-05, "loss": 0.6723, "step": 28374 }, { "epoch": 1.71, "learning_rate": 4.110120260136104e-05, "loss": 0.7144, "step": 28376 }, { "epoch": 1.71, "learning_rate": 4.109969369124681e-05, "loss": 0.5694, "step": 28378 }, { "epoch": 1.71, "learning_rate": 4.109818478113259e-05, "loss": 0.8513, "step": 28380 }, { "epoch": 1.71, "learning_rate": 4.1096675871018366e-05, "loss": 0.8625, "step": 28382 }, { "epoch": 1.71, "learning_rate": 4.109516696090414e-05, "loss": 0.834, "step": 28384 }, { "epoch": 1.71, "learning_rate": 4.109365805078992e-05, "loss": 0.8854, "step": 28386 }, { "epoch": 1.71, "learning_rate": 4.109214914067569e-05, "loss": 0.8035, "step": 28388 }, { "epoch": 1.71, "learning_rate": 4.1090640230561466e-05, "loss": 0.9216, "step": 28390 }, { "epoch": 1.71, "learning_rate": 4.1089131320447246e-05, "loss": 0.9224, "step": 28392 }, { "epoch": 1.71, "learning_rate": 4.108762241033302e-05, "loss": 0.7056, "step": 28394 }, { "epoch": 1.71, "learning_rate": 4.108611350021879e-05, "loss": 0.6867, "step": 28396 }, { "epoch": 1.71, "learning_rate": 4.108460459010457e-05, "loss": 0.7021, "step": 28398 }, { "epoch": 1.71, "learning_rate": 4.1083095679990345e-05, "loss": 0.7426, "step": 28400 }, { "epoch": 1.71, "learning_rate": 4.108158676987612e-05, "loss": 0.8732, "step": 28402 }, { "epoch": 1.71, "learning_rate": 4.10800778597619e-05, "loss": 0.855, "step": 28404 }, { "epoch": 1.71, "learning_rate": 4.107856894964767e-05, "loss": 0.8128, "step": 28406 }, { "epoch": 1.71, "learning_rate": 4.107706003953345e-05, "loss": 0.8032, "step": 28408 }, { "epoch": 1.71, "learning_rate": 4.1075551129419224e-05, "loss": 0.7356, "step": 28410 }, { "epoch": 1.71, "learning_rate": 4.1074042219305e-05, "loss": 0.9037, "step": 28412 }, { "epoch": 1.71, "learning_rate": 4.107253330919078e-05, "loss": 1.0267, "step": 28414 }, { "epoch": 1.72, "learning_rate": 4.107102439907654e-05, "loss": 0.5513, "step": 28416 }, { "epoch": 1.72, "learning_rate": 4.106951548896232e-05, "loss": 0.8956, "step": 28418 }, { "epoch": 1.72, "learning_rate": 4.10680065788481e-05, "loss": 0.8203, "step": 28420 }, { "epoch": 1.72, "learning_rate": 4.1066497668733876e-05, "loss": 0.7095, "step": 28422 }, { "epoch": 1.72, "learning_rate": 4.106498875861965e-05, "loss": 0.984, "step": 28424 }, { "epoch": 1.72, "learning_rate": 4.106347984850543e-05, "loss": 1.0139, "step": 28426 }, { "epoch": 1.72, "learning_rate": 4.10619709383912e-05, "loss": 0.9481, "step": 28428 }, { "epoch": 1.72, "learning_rate": 4.1060462028276976e-05, "loss": 0.8959, "step": 28430 }, { "epoch": 1.72, "learning_rate": 4.105895311816275e-05, "loss": 0.7011, "step": 28432 }, { "epoch": 1.72, "learning_rate": 4.105744420804853e-05, "loss": 0.6615, "step": 28434 }, { "epoch": 1.72, "learning_rate": 4.105593529793431e-05, "loss": 0.8995, "step": 28436 }, { "epoch": 1.72, "learning_rate": 4.1054426387820075e-05, "loss": 0.9988, "step": 28438 }, { "epoch": 1.72, "learning_rate": 4.1052917477705855e-05, "loss": 0.8877, "step": 28440 }, { "epoch": 1.72, "learning_rate": 4.1051408567591635e-05, "loss": 0.6467, "step": 28442 }, { "epoch": 1.72, "learning_rate": 4.104989965747741e-05, "loss": 0.855, "step": 28444 }, { "epoch": 1.72, "learning_rate": 4.104839074736318e-05, "loss": 0.7619, "step": 28446 }, { "epoch": 1.72, "learning_rate": 4.1046881837248954e-05, "loss": 0.9181, "step": 28448 }, { "epoch": 1.72, "learning_rate": 4.1045372927134734e-05, "loss": 0.7376, "step": 28450 }, { "epoch": 1.72, "learning_rate": 4.104386401702051e-05, "loss": 0.8345, "step": 28452 }, { "epoch": 1.72, "learning_rate": 4.104235510690628e-05, "loss": 0.8044, "step": 28454 }, { "epoch": 1.72, "learning_rate": 4.104084619679206e-05, "loss": 0.8923, "step": 28456 }, { "epoch": 1.72, "learning_rate": 4.103933728667784e-05, "loss": 0.7673, "step": 28458 }, { "epoch": 1.72, "learning_rate": 4.1037828376563606e-05, "loss": 0.7242, "step": 28460 }, { "epoch": 1.72, "learning_rate": 4.1036319466449386e-05, "loss": 0.7036, "step": 28462 }, { "epoch": 1.72, "learning_rate": 4.103481055633516e-05, "loss": 0.693, "step": 28464 }, { "epoch": 1.72, "learning_rate": 4.103330164622093e-05, "loss": 0.9188, "step": 28466 }, { "epoch": 1.72, "learning_rate": 4.103179273610671e-05, "loss": 0.7909, "step": 28468 }, { "epoch": 1.72, "learning_rate": 4.1030283825992486e-05, "loss": 1.0359, "step": 28470 }, { "epoch": 1.72, "learning_rate": 4.1028774915878265e-05, "loss": 0.7428, "step": 28472 }, { "epoch": 1.72, "learning_rate": 4.102726600576404e-05, "loss": 0.9606, "step": 28474 }, { "epoch": 1.72, "learning_rate": 4.102575709564981e-05, "loss": 0.984, "step": 28476 }, { "epoch": 1.72, "learning_rate": 4.102424818553559e-05, "loss": 0.9777, "step": 28478 }, { "epoch": 1.72, "learning_rate": 4.1022739275421365e-05, "loss": 0.8829, "step": 28480 }, { "epoch": 1.72, "learning_rate": 4.102123036530714e-05, "loss": 1.0271, "step": 28482 }, { "epoch": 1.72, "learning_rate": 4.101972145519292e-05, "loss": 1.1437, "step": 28484 }, { "epoch": 1.72, "learning_rate": 4.101821254507869e-05, "loss": 0.9966, "step": 28486 }, { "epoch": 1.72, "learning_rate": 4.1016703634964464e-05, "loss": 0.9251, "step": 28488 }, { "epoch": 1.72, "learning_rate": 4.1015194724850244e-05, "loss": 0.7939, "step": 28490 }, { "epoch": 1.72, "learning_rate": 4.101368581473602e-05, "loss": 0.7328, "step": 28492 }, { "epoch": 1.72, "learning_rate": 4.10121769046218e-05, "loss": 0.9524, "step": 28494 }, { "epoch": 1.72, "learning_rate": 4.101066799450757e-05, "loss": 0.9247, "step": 28496 }, { "epoch": 1.72, "learning_rate": 4.100915908439334e-05, "loss": 0.7378, "step": 28498 }, { "epoch": 1.72, "learning_rate": 4.100765017427912e-05, "loss": 1.1002, "step": 28500 }, { "epoch": 1.72, "learning_rate": 4.1006141264164896e-05, "loss": 0.9018, "step": 28502 }, { "epoch": 1.72, "learning_rate": 4.100463235405067e-05, "loss": 0.7351, "step": 28504 }, { "epoch": 1.72, "learning_rate": 4.100312344393645e-05, "loss": 1.0223, "step": 28506 }, { "epoch": 1.72, "learning_rate": 4.100161453382222e-05, "loss": 0.8889, "step": 28508 }, { "epoch": 1.72, "learning_rate": 4.1000105623707995e-05, "loss": 0.961, "step": 28510 }, { "epoch": 1.72, "learning_rate": 4.0998596713593775e-05, "loss": 0.9563, "step": 28512 }, { "epoch": 1.72, "learning_rate": 4.099708780347955e-05, "loss": 1.205, "step": 28514 }, { "epoch": 1.72, "learning_rate": 4.099557889336532e-05, "loss": 0.7039, "step": 28516 }, { "epoch": 1.72, "learning_rate": 4.09940699832511e-05, "loss": 0.7525, "step": 28518 }, { "epoch": 1.72, "learning_rate": 4.0992561073136875e-05, "loss": 0.7297, "step": 28520 }, { "epoch": 1.72, "learning_rate": 4.0991052163022655e-05, "loss": 0.9624, "step": 28522 }, { "epoch": 1.72, "learning_rate": 4.098954325290843e-05, "loss": 0.5724, "step": 28524 }, { "epoch": 1.72, "learning_rate": 4.09880343427942e-05, "loss": 0.7281, "step": 28526 }, { "epoch": 1.72, "learning_rate": 4.098652543267998e-05, "loss": 1.1958, "step": 28528 }, { "epoch": 1.72, "learning_rate": 4.098501652256575e-05, "loss": 0.8897, "step": 28530 }, { "epoch": 1.72, "learning_rate": 4.098350761245153e-05, "loss": 0.8848, "step": 28532 }, { "epoch": 1.72, "learning_rate": 4.098199870233731e-05, "loss": 0.7646, "step": 28534 }, { "epoch": 1.72, "learning_rate": 4.098048979222308e-05, "loss": 0.6634, "step": 28536 }, { "epoch": 1.72, "learning_rate": 4.097898088210885e-05, "loss": 1.1061, "step": 28538 }, { "epoch": 1.72, "learning_rate": 4.097747197199463e-05, "loss": 0.8786, "step": 28540 }, { "epoch": 1.72, "learning_rate": 4.0975963061880406e-05, "loss": 0.7948, "step": 28542 }, { "epoch": 1.72, "learning_rate": 4.0974454151766186e-05, "loss": 0.925, "step": 28544 }, { "epoch": 1.72, "learning_rate": 4.097294524165195e-05, "loss": 0.8508, "step": 28546 }, { "epoch": 1.72, "learning_rate": 4.097143633153773e-05, "loss": 0.714, "step": 28548 }, { "epoch": 1.72, "learning_rate": 4.096992742142351e-05, "loss": 0.6585, "step": 28550 }, { "epoch": 1.72, "learning_rate": 4.096841851130928e-05, "loss": 0.8032, "step": 28552 }, { "epoch": 1.72, "learning_rate": 4.096690960119506e-05, "loss": 1.1576, "step": 28554 }, { "epoch": 1.72, "learning_rate": 4.096540069108084e-05, "loss": 0.7811, "step": 28556 }, { "epoch": 1.72, "learning_rate": 4.096389178096661e-05, "loss": 0.7465, "step": 28558 }, { "epoch": 1.72, "learning_rate": 4.0962382870852385e-05, "loss": 0.877, "step": 28560 }, { "epoch": 1.72, "learning_rate": 4.096087396073816e-05, "loss": 0.9349, "step": 28562 }, { "epoch": 1.72, "learning_rate": 4.095936505062394e-05, "loss": 0.8875, "step": 28564 }, { "epoch": 1.72, "learning_rate": 4.095785614050971e-05, "loss": 0.8781, "step": 28566 }, { "epoch": 1.72, "learning_rate": 4.0956347230395484e-05, "loss": 0.872, "step": 28568 }, { "epoch": 1.72, "learning_rate": 4.0954838320281264e-05, "loss": 0.6869, "step": 28570 }, { "epoch": 1.72, "learning_rate": 4.0953329410167044e-05, "loss": 0.9357, "step": 28572 }, { "epoch": 1.72, "learning_rate": 4.095182050005281e-05, "loss": 0.7612, "step": 28574 }, { "epoch": 1.72, "learning_rate": 4.095031158993859e-05, "loss": 1.0922, "step": 28576 }, { "epoch": 1.72, "learning_rate": 4.094880267982436e-05, "loss": 0.6453, "step": 28578 }, { "epoch": 1.73, "learning_rate": 4.0947293769710136e-05, "loss": 0.7788, "step": 28580 }, { "epoch": 1.73, "learning_rate": 4.0945784859595916e-05, "loss": 0.7573, "step": 28582 }, { "epoch": 1.73, "learning_rate": 4.094427594948169e-05, "loss": 0.9676, "step": 28584 }, { "epoch": 1.73, "learning_rate": 4.094276703936747e-05, "loss": 0.817, "step": 28586 }, { "epoch": 1.73, "learning_rate": 4.094125812925324e-05, "loss": 0.8985, "step": 28588 }, { "epoch": 1.73, "learning_rate": 4.0939749219139015e-05, "loss": 0.8862, "step": 28590 }, { "epoch": 1.73, "learning_rate": 4.0938240309024795e-05, "loss": 0.6727, "step": 28592 }, { "epoch": 1.73, "learning_rate": 4.093673139891057e-05, "loss": 0.7986, "step": 28594 }, { "epoch": 1.73, "learning_rate": 4.093522248879634e-05, "loss": 0.7267, "step": 28596 }, { "epoch": 1.73, "learning_rate": 4.093371357868212e-05, "loss": 0.7141, "step": 28598 }, { "epoch": 1.73, "learning_rate": 4.0932204668567895e-05, "loss": 0.7567, "step": 28600 }, { "epoch": 1.73, "learning_rate": 4.093069575845367e-05, "loss": 0.8672, "step": 28602 }, { "epoch": 1.73, "learning_rate": 4.092918684833945e-05, "loss": 0.7376, "step": 28604 }, { "epoch": 1.73, "learning_rate": 4.092767793822522e-05, "loss": 0.4885, "step": 28606 }, { "epoch": 1.73, "learning_rate": 4.0926169028111e-05, "loss": 0.8685, "step": 28608 }, { "epoch": 1.73, "learning_rate": 4.0924660117996774e-05, "loss": 0.8645, "step": 28610 }, { "epoch": 1.73, "learning_rate": 4.092315120788255e-05, "loss": 0.6817, "step": 28612 }, { "epoch": 1.73, "learning_rate": 4.092164229776833e-05, "loss": 1.0786, "step": 28614 }, { "epoch": 1.73, "learning_rate": 4.09201333876541e-05, "loss": 1.0344, "step": 28616 }, { "epoch": 1.73, "learning_rate": 4.091862447753987e-05, "loss": 0.8127, "step": 28618 }, { "epoch": 1.73, "learning_rate": 4.091711556742565e-05, "loss": 0.9392, "step": 28620 }, { "epoch": 1.73, "learning_rate": 4.0915606657311426e-05, "loss": 0.8625, "step": 28622 }, { "epoch": 1.73, "learning_rate": 4.09140977471972e-05, "loss": 0.941, "step": 28624 }, { "epoch": 1.73, "learning_rate": 4.091258883708298e-05, "loss": 0.8544, "step": 28626 }, { "epoch": 1.73, "learning_rate": 4.091107992696875e-05, "loss": 0.5405, "step": 28628 }, { "epoch": 1.73, "learning_rate": 4.0909571016854525e-05, "loss": 1.0544, "step": 28630 }, { "epoch": 1.73, "learning_rate": 4.0908062106740305e-05, "loss": 0.9127, "step": 28632 }, { "epoch": 1.73, "learning_rate": 4.090655319662608e-05, "loss": 0.7485, "step": 28634 }, { "epoch": 1.73, "learning_rate": 4.090504428651186e-05, "loss": 1.2725, "step": 28636 }, { "epoch": 1.73, "learning_rate": 4.090353537639763e-05, "loss": 1.0989, "step": 28638 }, { "epoch": 1.73, "learning_rate": 4.0902026466283404e-05, "loss": 1.2735, "step": 28640 }, { "epoch": 1.73, "learning_rate": 4.0900517556169184e-05, "loss": 0.8348, "step": 28642 }, { "epoch": 1.73, "learning_rate": 4.089900864605495e-05, "loss": 0.6546, "step": 28644 }, { "epoch": 1.73, "learning_rate": 4.089749973594073e-05, "loss": 0.8983, "step": 28646 }, { "epoch": 1.73, "learning_rate": 4.089599082582651e-05, "loss": 0.7331, "step": 28648 }, { "epoch": 1.73, "learning_rate": 4.0894481915712284e-05, "loss": 1.1602, "step": 28650 }, { "epoch": 1.73, "learning_rate": 4.089297300559806e-05, "loss": 0.9626, "step": 28652 }, { "epoch": 1.73, "learning_rate": 4.089146409548384e-05, "loss": 0.914, "step": 28654 }, { "epoch": 1.73, "learning_rate": 4.088995518536961e-05, "loss": 0.9091, "step": 28656 }, { "epoch": 1.73, "learning_rate": 4.088844627525539e-05, "loss": 0.9956, "step": 28658 }, { "epoch": 1.73, "learning_rate": 4.0886937365141156e-05, "loss": 0.7894, "step": 28660 }, { "epoch": 1.73, "learning_rate": 4.0885428455026936e-05, "loss": 0.8935, "step": 28662 }, { "epoch": 1.73, "learning_rate": 4.0883919544912716e-05, "loss": 0.7926, "step": 28664 }, { "epoch": 1.73, "learning_rate": 4.088241063479848e-05, "loss": 0.8838, "step": 28666 }, { "epoch": 1.73, "learning_rate": 4.088090172468426e-05, "loss": 0.5182, "step": 28668 }, { "epoch": 1.73, "learning_rate": 4.087939281457004e-05, "loss": 0.672, "step": 28670 }, { "epoch": 1.73, "learning_rate": 4.0877883904455815e-05, "loss": 0.8854, "step": 28672 }, { "epoch": 1.73, "learning_rate": 4.087637499434159e-05, "loss": 0.7805, "step": 28674 }, { "epoch": 1.73, "learning_rate": 4.087486608422736e-05, "loss": 0.9661, "step": 28676 }, { "epoch": 1.73, "learning_rate": 4.087335717411314e-05, "loss": 0.8913, "step": 28678 }, { "epoch": 1.73, "learning_rate": 4.0871848263998914e-05, "loss": 0.6816, "step": 28680 }, { "epoch": 1.73, "learning_rate": 4.087033935388469e-05, "loss": 0.5764, "step": 28682 }, { "epoch": 1.73, "learning_rate": 4.086883044377047e-05, "loss": 0.7481, "step": 28684 }, { "epoch": 1.73, "learning_rate": 4.086732153365625e-05, "loss": 0.7514, "step": 28686 }, { "epoch": 1.73, "learning_rate": 4.0865812623542014e-05, "loss": 0.8114, "step": 28688 }, { "epoch": 1.73, "learning_rate": 4.0864303713427794e-05, "loss": 0.6445, "step": 28690 }, { "epoch": 1.73, "learning_rate": 4.086279480331357e-05, "loss": 0.5918, "step": 28692 }, { "epoch": 1.73, "learning_rate": 4.086128589319934e-05, "loss": 0.7512, "step": 28694 }, { "epoch": 1.73, "learning_rate": 4.085977698308512e-05, "loss": 0.5378, "step": 28696 }, { "epoch": 1.73, "learning_rate": 4.085826807297089e-05, "loss": 0.7577, "step": 28698 }, { "epoch": 1.73, "learning_rate": 4.085675916285667e-05, "loss": 0.6789, "step": 28700 }, { "epoch": 1.73, "learning_rate": 4.0855250252742446e-05, "loss": 0.861, "step": 28702 }, { "epoch": 1.73, "learning_rate": 4.085374134262822e-05, "loss": 0.7499, "step": 28704 }, { "epoch": 1.73, "learning_rate": 4.0852232432514e-05, "loss": 0.644, "step": 28706 }, { "epoch": 1.73, "learning_rate": 4.085072352239977e-05, "loss": 0.464, "step": 28708 }, { "epoch": 1.73, "learning_rate": 4.0849214612285545e-05, "loss": 0.8669, "step": 28710 }, { "epoch": 1.73, "learning_rate": 4.0847705702171325e-05, "loss": 1.0331, "step": 28712 }, { "epoch": 1.73, "learning_rate": 4.08461967920571e-05, "loss": 0.9961, "step": 28714 }, { "epoch": 1.73, "learning_rate": 4.084468788194287e-05, "loss": 0.7908, "step": 28716 }, { "epoch": 1.73, "learning_rate": 4.084317897182865e-05, "loss": 0.6701, "step": 28718 }, { "epoch": 1.73, "learning_rate": 4.0841670061714424e-05, "loss": 0.9885, "step": 28720 }, { "epoch": 1.73, "learning_rate": 4.0840161151600204e-05, "loss": 1.043, "step": 28722 }, { "epoch": 1.73, "learning_rate": 4.083865224148598e-05, "loss": 0.7974, "step": 28724 }, { "epoch": 1.73, "learning_rate": 4.083714333137175e-05, "loss": 0.7461, "step": 28726 }, { "epoch": 1.73, "learning_rate": 4.083563442125753e-05, "loss": 1.0709, "step": 28728 }, { "epoch": 1.73, "learning_rate": 4.0834125511143304e-05, "loss": 1.0086, "step": 28730 }, { "epoch": 1.73, "learning_rate": 4.083261660102908e-05, "loss": 0.7983, "step": 28732 }, { "epoch": 1.73, "learning_rate": 4.0831107690914857e-05, "loss": 0.6836, "step": 28734 }, { "epoch": 1.73, "learning_rate": 4.082959878080063e-05, "loss": 0.7855, "step": 28736 }, { "epoch": 1.73, "learning_rate": 4.08280898706864e-05, "loss": 0.9788, "step": 28738 }, { "epoch": 1.73, "learning_rate": 4.082658096057218e-05, "loss": 1.1042, "step": 28740 }, { "epoch": 1.73, "learning_rate": 4.0825072050457956e-05, "loss": 0.8762, "step": 28742 }, { "epoch": 1.73, "learning_rate": 4.082356314034373e-05, "loss": 0.9029, "step": 28744 }, { "epoch": 1.74, "learning_rate": 4.082205423022951e-05, "loss": 0.4928, "step": 28746 }, { "epoch": 1.74, "learning_rate": 4.082054532011528e-05, "loss": 0.7544, "step": 28748 }, { "epoch": 1.74, "learning_rate": 4.081903641000106e-05, "loss": 0.8604, "step": 28750 }, { "epoch": 1.74, "learning_rate": 4.0817527499886835e-05, "loss": 0.7737, "step": 28752 }, { "epoch": 1.74, "learning_rate": 4.081601858977261e-05, "loss": 0.7859, "step": 28754 }, { "epoch": 1.74, "learning_rate": 4.081450967965839e-05, "loss": 1.0664, "step": 28756 }, { "epoch": 1.74, "learning_rate": 4.081300076954416e-05, "loss": 0.6241, "step": 28758 }, { "epoch": 1.74, "learning_rate": 4.0811491859429934e-05, "loss": 0.6326, "step": 28760 }, { "epoch": 1.74, "learning_rate": 4.0809982949315714e-05, "loss": 0.7983, "step": 28762 }, { "epoch": 1.74, "learning_rate": 4.080847403920149e-05, "loss": 0.7585, "step": 28764 }, { "epoch": 1.74, "learning_rate": 4.080696512908726e-05, "loss": 0.8507, "step": 28766 }, { "epoch": 1.74, "learning_rate": 4.080545621897304e-05, "loss": 0.8003, "step": 28768 }, { "epoch": 1.74, "learning_rate": 4.0803947308858813e-05, "loss": 0.7348, "step": 28770 }, { "epoch": 1.74, "learning_rate": 4.0802438398744593e-05, "loss": 0.7746, "step": 28772 }, { "epoch": 1.74, "learning_rate": 4.080092948863036e-05, "loss": 0.6834, "step": 28774 }, { "epoch": 1.74, "learning_rate": 4.079942057851614e-05, "loss": 0.651, "step": 28776 }, { "epoch": 1.74, "learning_rate": 4.079791166840192e-05, "loss": 0.6385, "step": 28778 }, { "epoch": 1.74, "learning_rate": 4.0796402758287686e-05, "loss": 0.6338, "step": 28780 }, { "epoch": 1.74, "learning_rate": 4.0794893848173466e-05, "loss": 0.9091, "step": 28782 }, { "epoch": 1.74, "learning_rate": 4.0793384938059246e-05, "loss": 0.5999, "step": 28784 }, { "epoch": 1.74, "learning_rate": 4.079187602794502e-05, "loss": 0.8887, "step": 28786 }, { "epoch": 1.74, "learning_rate": 4.079036711783079e-05, "loss": 0.8692, "step": 28788 }, { "epoch": 1.74, "learning_rate": 4.0788858207716565e-05, "loss": 0.9486, "step": 28790 }, { "epoch": 1.74, "learning_rate": 4.0787349297602345e-05, "loss": 0.8544, "step": 28792 }, { "epoch": 1.74, "learning_rate": 4.078584038748812e-05, "loss": 0.7144, "step": 28794 }, { "epoch": 1.74, "learning_rate": 4.078433147737389e-05, "loss": 0.8803, "step": 28796 }, { "epoch": 1.74, "learning_rate": 4.078282256725967e-05, "loss": 0.7649, "step": 28798 }, { "epoch": 1.74, "learning_rate": 4.078131365714545e-05, "loss": 0.8987, "step": 28800 }, { "epoch": 1.74, "learning_rate": 4.077980474703122e-05, "loss": 0.7391, "step": 28802 }, { "epoch": 1.74, "learning_rate": 4.0778295836917e-05, "loss": 0.8344, "step": 28804 }, { "epoch": 1.74, "learning_rate": 4.077678692680277e-05, "loss": 0.7579, "step": 28806 }, { "epoch": 1.74, "learning_rate": 4.0775278016688544e-05, "loss": 0.6373, "step": 28808 }, { "epoch": 1.74, "learning_rate": 4.0773769106574323e-05, "loss": 0.6962, "step": 28810 }, { "epoch": 1.74, "learning_rate": 4.0772260196460097e-05, "loss": 0.9585, "step": 28812 }, { "epoch": 1.74, "learning_rate": 4.0770751286345876e-05, "loss": 0.7824, "step": 28814 }, { "epoch": 1.74, "learning_rate": 4.076924237623165e-05, "loss": 0.8813, "step": 28816 }, { "epoch": 1.74, "learning_rate": 4.076773346611742e-05, "loss": 1.0305, "step": 28818 }, { "epoch": 1.74, "learning_rate": 4.07662245560032e-05, "loss": 0.8013, "step": 28820 }, { "epoch": 1.74, "learning_rate": 4.0764715645888976e-05, "loss": 1.0158, "step": 28822 }, { "epoch": 1.74, "learning_rate": 4.076320673577475e-05, "loss": 0.7132, "step": 28824 }, { "epoch": 1.74, "learning_rate": 4.076169782566053e-05, "loss": 0.7822, "step": 28826 }, { "epoch": 1.74, "learning_rate": 4.07601889155463e-05, "loss": 0.9768, "step": 28828 }, { "epoch": 1.74, "learning_rate": 4.0758680005432075e-05, "loss": 0.7463, "step": 28830 }, { "epoch": 1.74, "learning_rate": 4.0757171095317855e-05, "loss": 0.6442, "step": 28832 }, { "epoch": 1.74, "learning_rate": 4.075566218520363e-05, "loss": 0.5813, "step": 28834 }, { "epoch": 1.74, "learning_rate": 4.075415327508941e-05, "loss": 0.8091, "step": 28836 }, { "epoch": 1.74, "learning_rate": 4.075264436497518e-05, "loss": 0.9247, "step": 28838 }, { "epoch": 1.74, "learning_rate": 4.0751135454860954e-05, "loss": 0.7065, "step": 28840 }, { "epoch": 1.74, "learning_rate": 4.0749626544746734e-05, "loss": 0.7663, "step": 28842 }, { "epoch": 1.74, "learning_rate": 4.074811763463251e-05, "loss": 0.7445, "step": 28844 }, { "epoch": 1.74, "learning_rate": 4.074660872451828e-05, "loss": 0.761, "step": 28846 }, { "epoch": 1.74, "learning_rate": 4.074509981440406e-05, "loss": 0.7043, "step": 28848 }, { "epoch": 1.74, "learning_rate": 4.074359090428983e-05, "loss": 0.9244, "step": 28850 }, { "epoch": 1.74, "learning_rate": 4.0742081994175607e-05, "loss": 1.0937, "step": 28852 }, { "epoch": 1.74, "learning_rate": 4.0740573084061386e-05, "loss": 0.8089, "step": 28854 }, { "epoch": 1.74, "learning_rate": 4.073906417394716e-05, "loss": 0.6445, "step": 28856 }, { "epoch": 1.74, "learning_rate": 4.073755526383293e-05, "loss": 0.9012, "step": 28858 }, { "epoch": 1.74, "learning_rate": 4.073604635371871e-05, "loss": 0.5683, "step": 28860 }, { "epoch": 1.74, "learning_rate": 4.0734537443604486e-05, "loss": 1.0479, "step": 28862 }, { "epoch": 1.74, "learning_rate": 4.0733028533490266e-05, "loss": 0.9029, "step": 28864 }, { "epoch": 1.74, "learning_rate": 4.073151962337604e-05, "loss": 0.8797, "step": 28866 }, { "epoch": 1.74, "learning_rate": 4.073001071326181e-05, "loss": 0.6464, "step": 28868 }, { "epoch": 1.74, "learning_rate": 4.072850180314759e-05, "loss": 0.9581, "step": 28870 }, { "epoch": 1.74, "learning_rate": 4.0726992893033365e-05, "loss": 0.7093, "step": 28872 }, { "epoch": 1.74, "learning_rate": 4.072548398291914e-05, "loss": 0.8139, "step": 28874 }, { "epoch": 1.74, "learning_rate": 4.072397507280492e-05, "loss": 0.8778, "step": 28876 }, { "epoch": 1.74, "learning_rate": 4.072246616269069e-05, "loss": 0.7523, "step": 28878 }, { "epoch": 1.74, "learning_rate": 4.0720957252576464e-05, "loss": 0.7153, "step": 28880 }, { "epoch": 1.74, "learning_rate": 4.0719448342462244e-05, "loss": 0.8428, "step": 28882 }, { "epoch": 1.74, "learning_rate": 4.071793943234802e-05, "loss": 0.8466, "step": 28884 }, { "epoch": 1.74, "learning_rate": 4.07164305222338e-05, "loss": 0.5836, "step": 28886 }, { "epoch": 1.74, "learning_rate": 4.0714921612119563e-05, "loss": 0.872, "step": 28888 }, { "epoch": 1.74, "learning_rate": 4.071341270200534e-05, "loss": 0.9679, "step": 28890 }, { "epoch": 1.74, "learning_rate": 4.071190379189112e-05, "loss": 0.6087, "step": 28892 }, { "epoch": 1.74, "learning_rate": 4.071039488177689e-05, "loss": 1.241, "step": 28894 }, { "epoch": 1.74, "learning_rate": 4.070888597166267e-05, "loss": 1.0523, "step": 28896 }, { "epoch": 1.74, "learning_rate": 4.070737706154845e-05, "loss": 1.0033, "step": 28898 }, { "epoch": 1.74, "learning_rate": 4.070586815143422e-05, "loss": 1.1039, "step": 28900 }, { "epoch": 1.74, "learning_rate": 4.0704359241319996e-05, "loss": 0.7075, "step": 28902 }, { "epoch": 1.74, "learning_rate": 4.070285033120577e-05, "loss": 0.9429, "step": 28904 }, { "epoch": 1.74, "learning_rate": 4.070134142109155e-05, "loss": 0.5625, "step": 28906 }, { "epoch": 1.74, "learning_rate": 4.069983251097732e-05, "loss": 0.9837, "step": 28908 }, { "epoch": 1.74, "learning_rate": 4.0698323600863095e-05, "loss": 1.0214, "step": 28910 }, { "epoch": 1.75, "learning_rate": 4.0696814690748875e-05, "loss": 0.7425, "step": 28912 }, { "epoch": 1.75, "learning_rate": 4.0695305780634655e-05, "loss": 0.7182, "step": 28914 }, { "epoch": 1.75, "learning_rate": 4.069379687052042e-05, "loss": 0.9897, "step": 28916 }, { "epoch": 1.75, "learning_rate": 4.06922879604062e-05, "loss": 1.0276, "step": 28918 }, { "epoch": 1.75, "learning_rate": 4.0690779050291974e-05, "loss": 0.955, "step": 28920 }, { "epoch": 1.75, "learning_rate": 4.0689270140177754e-05, "loss": 0.9261, "step": 28922 }, { "epoch": 1.75, "learning_rate": 4.068776123006353e-05, "loss": 0.7145, "step": 28924 }, { "epoch": 1.75, "learning_rate": 4.06862523199493e-05, "loss": 0.8081, "step": 28926 }, { "epoch": 1.75, "learning_rate": 4.068474340983508e-05, "loss": 0.6524, "step": 28928 }, { "epoch": 1.75, "learning_rate": 4.068323449972085e-05, "loss": 0.7187, "step": 28930 }, { "epoch": 1.75, "learning_rate": 4.0681725589606626e-05, "loss": 0.6773, "step": 28932 }, { "epoch": 1.75, "learning_rate": 4.0680216679492406e-05, "loss": 0.4992, "step": 28934 }, { "epoch": 1.75, "learning_rate": 4.067870776937818e-05, "loss": 0.733, "step": 28936 }, { "epoch": 1.75, "learning_rate": 4.067719885926395e-05, "loss": 0.6529, "step": 28938 }, { "epoch": 1.75, "learning_rate": 4.067568994914973e-05, "loss": 1.041, "step": 28940 }, { "epoch": 1.75, "learning_rate": 4.0674181039035506e-05, "loss": 0.5594, "step": 28942 }, { "epoch": 1.75, "learning_rate": 4.067267212892128e-05, "loss": 0.6825, "step": 28944 }, { "epoch": 1.75, "learning_rate": 4.067116321880706e-05, "loss": 0.8003, "step": 28946 }, { "epoch": 1.75, "learning_rate": 4.066965430869283e-05, "loss": 0.7156, "step": 28948 }, { "epoch": 1.75, "learning_rate": 4.066814539857861e-05, "loss": 0.8166, "step": 28950 }, { "epoch": 1.75, "learning_rate": 4.0666636488464385e-05, "loss": 0.7882, "step": 28952 }, { "epoch": 1.75, "learning_rate": 4.066512757835016e-05, "loss": 0.9353, "step": 28954 }, { "epoch": 1.75, "learning_rate": 4.066361866823594e-05, "loss": 0.9367, "step": 28956 }, { "epoch": 1.75, "learning_rate": 4.066210975812171e-05, "loss": 1.2869, "step": 28958 }, { "epoch": 1.75, "learning_rate": 4.0660600848007484e-05, "loss": 0.5806, "step": 28960 }, { "epoch": 1.75, "learning_rate": 4.0659091937893264e-05, "loss": 0.7444, "step": 28962 }, { "epoch": 1.75, "learning_rate": 4.065758302777904e-05, "loss": 0.7021, "step": 28964 }, { "epoch": 1.75, "learning_rate": 4.065607411766481e-05, "loss": 0.843, "step": 28966 }, { "epoch": 1.75, "learning_rate": 4.065456520755059e-05, "loss": 0.941, "step": 28968 }, { "epoch": 1.75, "learning_rate": 4.065305629743636e-05, "loss": 0.6862, "step": 28970 }, { "epoch": 1.75, "learning_rate": 4.0651547387322136e-05, "loss": 1.037, "step": 28972 }, { "epoch": 1.75, "learning_rate": 4.0650038477207916e-05, "loss": 1.225, "step": 28974 }, { "epoch": 1.75, "learning_rate": 4.064852956709369e-05, "loss": 0.9589, "step": 28976 }, { "epoch": 1.75, "learning_rate": 4.064702065697947e-05, "loss": 0.5876, "step": 28978 }, { "epoch": 1.75, "learning_rate": 4.064551174686524e-05, "loss": 0.551, "step": 28980 }, { "epoch": 1.75, "learning_rate": 4.0644002836751016e-05, "loss": 0.8566, "step": 28982 }, { "epoch": 1.75, "learning_rate": 4.0642493926636795e-05, "loss": 0.7282, "step": 28984 }, { "epoch": 1.75, "learning_rate": 4.064098501652257e-05, "loss": 0.6473, "step": 28986 }, { "epoch": 1.75, "learning_rate": 4.063947610640834e-05, "loss": 0.7007, "step": 28988 }, { "epoch": 1.75, "learning_rate": 4.063796719629412e-05, "loss": 0.5147, "step": 28990 }, { "epoch": 1.75, "learning_rate": 4.0636458286179895e-05, "loss": 0.717, "step": 28992 }, { "epoch": 1.75, "learning_rate": 4.063494937606567e-05, "loss": 0.9427, "step": 28994 }, { "epoch": 1.75, "learning_rate": 4.063344046595145e-05, "loss": 0.736, "step": 28996 }, { "epoch": 1.75, "learning_rate": 4.063193155583722e-05, "loss": 1.2989, "step": 28998 }, { "epoch": 1.75, "learning_rate": 4.0630422645723e-05, "loss": 1.1039, "step": 29000 }, { "epoch": 1.75, "learning_rate": 4.062891373560877e-05, "loss": 0.7364, "step": 29002 }, { "epoch": 1.75, "learning_rate": 4.062740482549455e-05, "loss": 0.884, "step": 29004 }, { "epoch": 1.75, "learning_rate": 4.062589591538033e-05, "loss": 0.8003, "step": 29006 }, { "epoch": 1.75, "learning_rate": 4.062438700526609e-05, "loss": 0.9367, "step": 29008 }, { "epoch": 1.75, "learning_rate": 4.062287809515187e-05, "loss": 0.6394, "step": 29010 }, { "epoch": 1.75, "learning_rate": 4.062136918503765e-05, "loss": 0.7406, "step": 29012 }, { "epoch": 1.75, "learning_rate": 4.0619860274923426e-05, "loss": 0.955, "step": 29014 }, { "epoch": 1.75, "learning_rate": 4.06183513648092e-05, "loss": 0.9198, "step": 29016 }, { "epoch": 1.75, "learning_rate": 4.061684245469497e-05, "loss": 0.6168, "step": 29018 }, { "epoch": 1.75, "learning_rate": 4.061533354458075e-05, "loss": 0.8992, "step": 29020 }, { "epoch": 1.75, "learning_rate": 4.0613824634466525e-05, "loss": 1.0691, "step": 29022 }, { "epoch": 1.75, "learning_rate": 4.06123157243523e-05, "loss": 0.8371, "step": 29024 }, { "epoch": 1.75, "learning_rate": 4.061080681423808e-05, "loss": 1.1669, "step": 29026 }, { "epoch": 1.75, "learning_rate": 4.060929790412386e-05, "loss": 0.7748, "step": 29028 }, { "epoch": 1.75, "learning_rate": 4.0607788994009625e-05, "loss": 0.85, "step": 29030 }, { "epoch": 1.75, "learning_rate": 4.0606280083895405e-05, "loss": 0.9756, "step": 29032 }, { "epoch": 1.75, "learning_rate": 4.060477117378118e-05, "loss": 0.6218, "step": 29034 }, { "epoch": 1.75, "learning_rate": 4.060326226366696e-05, "loss": 0.6254, "step": 29036 }, { "epoch": 1.75, "learning_rate": 4.060175335355273e-05, "loss": 0.6591, "step": 29038 }, { "epoch": 1.75, "learning_rate": 4.0600244443438504e-05, "loss": 0.7376, "step": 29040 }, { "epoch": 1.75, "learning_rate": 4.0598735533324284e-05, "loss": 1.0104, "step": 29042 }, { "epoch": 1.75, "learning_rate": 4.059722662321006e-05, "loss": 1.0477, "step": 29044 }, { "epoch": 1.75, "learning_rate": 4.059571771309583e-05, "loss": 0.7326, "step": 29046 }, { "epoch": 1.75, "learning_rate": 4.059420880298161e-05, "loss": 0.6417, "step": 29048 }, { "epoch": 1.75, "learning_rate": 4.059269989286738e-05, "loss": 0.7375, "step": 29050 }, { "epoch": 1.75, "learning_rate": 4.0591190982753156e-05, "loss": 0.9538, "step": 29052 }, { "epoch": 1.75, "learning_rate": 4.0589682072638936e-05, "loss": 0.6865, "step": 29054 }, { "epoch": 1.75, "learning_rate": 4.058817316252471e-05, "loss": 0.7536, "step": 29056 }, { "epoch": 1.75, "learning_rate": 4.058666425241048e-05, "loss": 0.9862, "step": 29058 }, { "epoch": 1.75, "learning_rate": 4.058515534229626e-05, "loss": 0.5737, "step": 29060 }, { "epoch": 1.75, "learning_rate": 4.0583646432182035e-05, "loss": 0.6359, "step": 29062 }, { "epoch": 1.75, "learning_rate": 4.0582137522067815e-05, "loss": 0.8717, "step": 29064 }, { "epoch": 1.75, "learning_rate": 4.058062861195359e-05, "loss": 0.8383, "step": 29066 }, { "epoch": 1.75, "learning_rate": 4.057911970183936e-05, "loss": 0.801, "step": 29068 }, { "epoch": 1.75, "learning_rate": 4.057761079172514e-05, "loss": 0.5051, "step": 29070 }, { "epoch": 1.75, "learning_rate": 4.0576101881610915e-05, "loss": 1.1896, "step": 29072 }, { "epoch": 1.75, "learning_rate": 4.057459297149669e-05, "loss": 1.0304, "step": 29074 }, { "epoch": 1.75, "learning_rate": 4.057308406138247e-05, "loss": 0.8913, "step": 29076 }, { "epoch": 1.76, "learning_rate": 4.057157515126824e-05, "loss": 0.9536, "step": 29078 }, { "epoch": 1.76, "learning_rate": 4.0570066241154014e-05, "loss": 0.7695, "step": 29080 }, { "epoch": 1.76, "learning_rate": 4.0568557331039794e-05, "loss": 0.8357, "step": 29082 }, { "epoch": 1.76, "learning_rate": 4.056704842092557e-05, "loss": 0.7191, "step": 29084 }, { "epoch": 1.76, "learning_rate": 4.056553951081135e-05, "loss": 0.5715, "step": 29086 }, { "epoch": 1.76, "learning_rate": 4.056403060069712e-05, "loss": 0.8242, "step": 29088 }, { "epoch": 1.76, "learning_rate": 4.056252169058289e-05, "loss": 1.0545, "step": 29090 }, { "epoch": 1.76, "learning_rate": 4.056101278046867e-05, "loss": 0.6074, "step": 29092 }, { "epoch": 1.76, "learning_rate": 4.0559503870354446e-05, "loss": 0.7636, "step": 29094 }, { "epoch": 1.76, "learning_rate": 4.055799496024022e-05, "loss": 0.8038, "step": 29096 }, { "epoch": 1.76, "learning_rate": 4.0556486050126e-05, "loss": 1.0989, "step": 29098 }, { "epoch": 1.76, "learning_rate": 4.055497714001177e-05, "loss": 0.6354, "step": 29100 }, { "epoch": 1.76, "learning_rate": 4.0553468229897545e-05, "loss": 0.8739, "step": 29102 }, { "epoch": 1.76, "learning_rate": 4.0551959319783325e-05, "loss": 0.6085, "step": 29104 }, { "epoch": 1.76, "learning_rate": 4.05504504096691e-05, "loss": 0.9824, "step": 29106 }, { "epoch": 1.76, "learning_rate": 4.054894149955487e-05, "loss": 0.7047, "step": 29108 }, { "epoch": 1.76, "learning_rate": 4.054743258944065e-05, "loss": 0.779, "step": 29110 }, { "epoch": 1.76, "learning_rate": 4.0545923679326425e-05, "loss": 0.8022, "step": 29112 }, { "epoch": 1.76, "learning_rate": 4.0544414769212204e-05, "loss": 0.6718, "step": 29114 }, { "epoch": 1.76, "learning_rate": 4.054290585909797e-05, "loss": 0.7038, "step": 29116 }, { "epoch": 1.76, "learning_rate": 4.054139694898375e-05, "loss": 0.8646, "step": 29118 }, { "epoch": 1.76, "learning_rate": 4.053988803886953e-05, "loss": 0.9132, "step": 29120 }, { "epoch": 1.76, "learning_rate": 4.05383791287553e-05, "loss": 0.8133, "step": 29122 }, { "epoch": 1.76, "learning_rate": 4.053687021864108e-05, "loss": 0.7433, "step": 29124 }, { "epoch": 1.76, "learning_rate": 4.053536130852686e-05, "loss": 0.8664, "step": 29126 }, { "epoch": 1.76, "learning_rate": 4.053385239841263e-05, "loss": 0.6725, "step": 29128 }, { "epoch": 1.76, "learning_rate": 4.05323434882984e-05, "loss": 0.7905, "step": 29130 }, { "epoch": 1.76, "learning_rate": 4.0530834578184176e-05, "loss": 0.9753, "step": 29132 }, { "epoch": 1.76, "learning_rate": 4.0529325668069956e-05, "loss": 0.7939, "step": 29134 }, { "epoch": 1.76, "learning_rate": 4.0527816757955736e-05, "loss": 0.7981, "step": 29136 }, { "epoch": 1.76, "learning_rate": 4.05263078478415e-05, "loss": 0.6772, "step": 29138 }, { "epoch": 1.76, "learning_rate": 4.052479893772728e-05, "loss": 0.7316, "step": 29140 }, { "epoch": 1.76, "learning_rate": 4.052329002761306e-05, "loss": 0.9753, "step": 29142 }, { "epoch": 1.76, "learning_rate": 4.052178111749883e-05, "loss": 0.7937, "step": 29144 }, { "epoch": 1.76, "learning_rate": 4.052027220738461e-05, "loss": 0.8346, "step": 29146 }, { "epoch": 1.76, "learning_rate": 4.051876329727038e-05, "loss": 0.9584, "step": 29148 }, { "epoch": 1.76, "learning_rate": 4.051725438715616e-05, "loss": 0.7701, "step": 29150 }, { "epoch": 1.76, "learning_rate": 4.0515745477041934e-05, "loss": 0.6535, "step": 29152 }, { "epoch": 1.76, "learning_rate": 4.051423656692771e-05, "loss": 0.8316, "step": 29154 }, { "epoch": 1.76, "learning_rate": 4.051272765681349e-05, "loss": 0.8319, "step": 29156 }, { "epoch": 1.76, "learning_rate": 4.051121874669926e-05, "loss": 0.6298, "step": 29158 }, { "epoch": 1.76, "learning_rate": 4.0509709836585034e-05, "loss": 1.0484, "step": 29160 }, { "epoch": 1.76, "learning_rate": 4.0508200926470814e-05, "loss": 0.806, "step": 29162 }, { "epoch": 1.76, "learning_rate": 4.050669201635659e-05, "loss": 0.7724, "step": 29164 }, { "epoch": 1.76, "learning_rate": 4.050518310624236e-05, "loss": 0.7275, "step": 29166 }, { "epoch": 1.76, "learning_rate": 4.050367419612814e-05, "loss": 0.7169, "step": 29168 }, { "epoch": 1.76, "learning_rate": 4.050216528601391e-05, "loss": 0.907, "step": 29170 }, { "epoch": 1.76, "learning_rate": 4.0500656375899686e-05, "loss": 0.7244, "step": 29172 }, { "epoch": 1.76, "learning_rate": 4.0499147465785466e-05, "loss": 0.9034, "step": 29174 }, { "epoch": 1.76, "learning_rate": 4.049763855567124e-05, "loss": 0.8911, "step": 29176 }, { "epoch": 1.76, "learning_rate": 4.049612964555702e-05, "loss": 0.502, "step": 29178 }, { "epoch": 1.76, "learning_rate": 4.049462073544279e-05, "loss": 0.8674, "step": 29180 }, { "epoch": 1.76, "learning_rate": 4.0493111825328565e-05, "loss": 0.7953, "step": 29182 }, { "epoch": 1.76, "learning_rate": 4.0491602915214345e-05, "loss": 0.8512, "step": 29184 }, { "epoch": 1.76, "learning_rate": 4.049009400510012e-05, "loss": 0.6532, "step": 29186 }, { "epoch": 1.76, "learning_rate": 4.048858509498589e-05, "loss": 0.7368, "step": 29188 }, { "epoch": 1.76, "learning_rate": 4.048707618487167e-05, "loss": 1.0598, "step": 29190 }, { "epoch": 1.76, "learning_rate": 4.0485567274757444e-05, "loss": 1.0391, "step": 29192 }, { "epoch": 1.76, "learning_rate": 4.048405836464322e-05, "loss": 0.5734, "step": 29194 }, { "epoch": 1.76, "learning_rate": 4.0482549454529e-05, "loss": 0.9372, "step": 29196 }, { "epoch": 1.76, "learning_rate": 4.048104054441477e-05, "loss": 0.6243, "step": 29198 }, { "epoch": 1.76, "learning_rate": 4.047953163430055e-05, "loss": 1.0705, "step": 29200 }, { "epoch": 1.76, "learning_rate": 4.0478022724186324e-05, "loss": 0.7401, "step": 29202 }, { "epoch": 1.76, "learning_rate": 4.04765138140721e-05, "loss": 0.734, "step": 29204 }, { "epoch": 1.76, "learning_rate": 4.047500490395788e-05, "loss": 0.6016, "step": 29206 }, { "epoch": 1.76, "learning_rate": 4.047349599384365e-05, "loss": 0.6985, "step": 29208 }, { "epoch": 1.76, "learning_rate": 4.047198708372942e-05, "loss": 0.7249, "step": 29210 }, { "epoch": 1.76, "learning_rate": 4.04704781736152e-05, "loss": 0.7742, "step": 29212 }, { "epoch": 1.76, "learning_rate": 4.0468969263500976e-05, "loss": 0.9072, "step": 29214 }, { "epoch": 1.76, "learning_rate": 4.046746035338675e-05, "loss": 0.7961, "step": 29216 }, { "epoch": 1.76, "learning_rate": 4.046595144327253e-05, "loss": 0.901, "step": 29218 }, { "epoch": 1.76, "learning_rate": 4.04644425331583e-05, "loss": 0.8525, "step": 29220 }, { "epoch": 1.76, "learning_rate": 4.0462933623044075e-05, "loss": 1.0516, "step": 29222 }, { "epoch": 1.76, "learning_rate": 4.0461424712929855e-05, "loss": 0.6436, "step": 29224 }, { "epoch": 1.76, "learning_rate": 4.045991580281563e-05, "loss": 0.6424, "step": 29226 }, { "epoch": 1.76, "learning_rate": 4.045840689270141e-05, "loss": 0.9295, "step": 29228 }, { "epoch": 1.76, "learning_rate": 4.0456897982587174e-05, "loss": 0.6736, "step": 29230 }, { "epoch": 1.76, "learning_rate": 4.0455389072472954e-05, "loss": 0.7156, "step": 29232 }, { "epoch": 1.76, "learning_rate": 4.0453880162358734e-05, "loss": 1.0161, "step": 29234 }, { "epoch": 1.76, "learning_rate": 4.04523712522445e-05, "loss": 1.0364, "step": 29236 }, { "epoch": 1.76, "learning_rate": 4.045086234213028e-05, "loss": 0.8912, "step": 29238 }, { "epoch": 1.76, "learning_rate": 4.044935343201606e-05, "loss": 0.829, "step": 29240 }, { "epoch": 1.76, "learning_rate": 4.0447844521901834e-05, "loss": 0.797, "step": 29242 }, { "epoch": 1.77, "learning_rate": 4.044633561178761e-05, "loss": 0.9557, "step": 29244 }, { "epoch": 1.77, "learning_rate": 4.044482670167338e-05, "loss": 0.5792, "step": 29246 }, { "epoch": 1.77, "learning_rate": 4.044331779155916e-05, "loss": 0.6258, "step": 29248 }, { "epoch": 1.77, "learning_rate": 4.044180888144494e-05, "loss": 0.9008, "step": 29250 }, { "epoch": 1.77, "learning_rate": 4.0440299971330706e-05, "loss": 0.9139, "step": 29252 }, { "epoch": 1.77, "learning_rate": 4.0438791061216486e-05, "loss": 0.7473, "step": 29254 }, { "epoch": 1.77, "learning_rate": 4.0437282151102266e-05, "loss": 0.7458, "step": 29256 }, { "epoch": 1.77, "learning_rate": 4.043577324098803e-05, "loss": 0.7484, "step": 29258 }, { "epoch": 1.77, "learning_rate": 4.043426433087381e-05, "loss": 0.8617, "step": 29260 }, { "epoch": 1.77, "learning_rate": 4.0432755420759585e-05, "loss": 0.9463, "step": 29262 }, { "epoch": 1.77, "learning_rate": 4.0431246510645365e-05, "loss": 0.9422, "step": 29264 }, { "epoch": 1.77, "learning_rate": 4.042973760053114e-05, "loss": 0.7688, "step": 29266 }, { "epoch": 1.77, "learning_rate": 4.042822869041691e-05, "loss": 0.7902, "step": 29268 }, { "epoch": 1.77, "learning_rate": 4.042671978030269e-05, "loss": 0.7269, "step": 29270 }, { "epoch": 1.77, "learning_rate": 4.0425210870188464e-05, "loss": 0.9628, "step": 29272 }, { "epoch": 1.77, "learning_rate": 4.042370196007424e-05, "loss": 0.7452, "step": 29274 }, { "epoch": 1.77, "learning_rate": 4.042219304996002e-05, "loss": 0.8595, "step": 29276 }, { "epoch": 1.77, "learning_rate": 4.042068413984579e-05, "loss": 0.709, "step": 29278 }, { "epoch": 1.77, "learning_rate": 4.0419175229731564e-05, "loss": 0.7913, "step": 29280 }, { "epoch": 1.77, "learning_rate": 4.0417666319617343e-05, "loss": 0.7728, "step": 29282 }, { "epoch": 1.77, "learning_rate": 4.0416157409503117e-05, "loss": 1.1718, "step": 29284 }, { "epoch": 1.77, "learning_rate": 4.041464849938889e-05, "loss": 0.9427, "step": 29286 }, { "epoch": 1.77, "learning_rate": 4.041313958927467e-05, "loss": 1.0258, "step": 29288 }, { "epoch": 1.77, "learning_rate": 4.041163067916044e-05, "loss": 0.7908, "step": 29290 }, { "epoch": 1.77, "learning_rate": 4.041012176904622e-05, "loss": 0.8119, "step": 29292 }, { "epoch": 1.77, "learning_rate": 4.0408612858931996e-05, "loss": 0.7731, "step": 29294 }, { "epoch": 1.77, "learning_rate": 4.040710394881777e-05, "loss": 1.0147, "step": 29296 }, { "epoch": 1.77, "learning_rate": 4.040559503870355e-05, "loss": 0.5732, "step": 29298 }, { "epoch": 1.77, "learning_rate": 4.040408612858932e-05, "loss": 1.1166, "step": 29300 }, { "epoch": 1.77, "learning_rate": 4.0402577218475095e-05, "loss": 0.6319, "step": 29302 }, { "epoch": 1.77, "learning_rate": 4.0401068308360875e-05, "loss": 0.8829, "step": 29304 }, { "epoch": 1.77, "learning_rate": 4.039955939824665e-05, "loss": 0.8653, "step": 29306 }, { "epoch": 1.77, "learning_rate": 4.039805048813242e-05, "loss": 0.8461, "step": 29308 }, { "epoch": 1.77, "learning_rate": 4.03965415780182e-05, "loss": 0.481, "step": 29310 }, { "epoch": 1.77, "learning_rate": 4.0395032667903974e-05, "loss": 0.8392, "step": 29312 }, { "epoch": 1.77, "learning_rate": 4.0393523757789754e-05, "loss": 0.8473, "step": 29314 }, { "epoch": 1.77, "learning_rate": 4.039201484767553e-05, "loss": 0.7095, "step": 29316 }, { "epoch": 1.77, "learning_rate": 4.03905059375613e-05, "loss": 0.7828, "step": 29318 }, { "epoch": 1.77, "learning_rate": 4.038899702744708e-05, "loss": 0.591, "step": 29320 }, { "epoch": 1.77, "learning_rate": 4.0387488117332853e-05, "loss": 1.1545, "step": 29322 }, { "epoch": 1.77, "learning_rate": 4.0385979207218627e-05, "loss": 0.6953, "step": 29324 }, { "epoch": 1.77, "learning_rate": 4.0384470297104406e-05, "loss": 0.9529, "step": 29326 }, { "epoch": 1.77, "learning_rate": 4.038296138699018e-05, "loss": 0.5772, "step": 29328 }, { "epoch": 1.77, "learning_rate": 4.038145247687595e-05, "loss": 0.6548, "step": 29330 }, { "epoch": 1.77, "learning_rate": 4.037994356676173e-05, "loss": 0.7137, "step": 29332 }, { "epoch": 1.77, "learning_rate": 4.0378434656647506e-05, "loss": 1.0069, "step": 29334 }, { "epoch": 1.77, "learning_rate": 4.037692574653328e-05, "loss": 0.5248, "step": 29336 }, { "epoch": 1.77, "learning_rate": 4.037541683641906e-05, "loss": 0.9086, "step": 29338 }, { "epoch": 1.77, "learning_rate": 4.037390792630483e-05, "loss": 0.7236, "step": 29340 }, { "epoch": 1.77, "learning_rate": 4.037239901619061e-05, "loss": 1.1155, "step": 29342 }, { "epoch": 1.77, "learning_rate": 4.037089010607638e-05, "loss": 0.601, "step": 29344 }, { "epoch": 1.77, "learning_rate": 4.036938119596216e-05, "loss": 0.7032, "step": 29346 }, { "epoch": 1.77, "learning_rate": 4.036787228584794e-05, "loss": 0.7951, "step": 29348 }, { "epoch": 1.77, "learning_rate": 4.0366363375733704e-05, "loss": 0.945, "step": 29350 }, { "epoch": 1.77, "learning_rate": 4.0364854465619484e-05, "loss": 0.8019, "step": 29352 }, { "epoch": 1.77, "learning_rate": 4.0363345555505264e-05, "loss": 0.8491, "step": 29354 }, { "epoch": 1.77, "learning_rate": 4.036183664539104e-05, "loss": 0.6064, "step": 29356 }, { "epoch": 1.77, "learning_rate": 4.036032773527681e-05, "loss": 0.8187, "step": 29358 }, { "epoch": 1.77, "learning_rate": 4.0358818825162583e-05, "loss": 0.9879, "step": 29360 }, { "epoch": 1.77, "learning_rate": 4.035730991504836e-05, "loss": 0.8177, "step": 29362 }, { "epoch": 1.77, "learning_rate": 4.035580100493414e-05, "loss": 0.8615, "step": 29364 }, { "epoch": 1.77, "learning_rate": 4.035429209481991e-05, "loss": 0.626, "step": 29366 }, { "epoch": 1.77, "learning_rate": 4.035278318470569e-05, "loss": 0.7679, "step": 29368 }, { "epoch": 1.77, "learning_rate": 4.035127427459147e-05, "loss": 0.8557, "step": 29370 }, { "epoch": 1.77, "learning_rate": 4.0349765364477236e-05, "loss": 0.5465, "step": 29372 }, { "epoch": 1.77, "learning_rate": 4.0348256454363016e-05, "loss": 0.7461, "step": 29374 }, { "epoch": 1.77, "learning_rate": 4.034674754424879e-05, "loss": 0.8837, "step": 29376 }, { "epoch": 1.77, "learning_rate": 4.034523863413457e-05, "loss": 0.9615, "step": 29378 }, { "epoch": 1.77, "learning_rate": 4.034372972402034e-05, "loss": 1.0499, "step": 29380 }, { "epoch": 1.77, "learning_rate": 4.0342220813906115e-05, "loss": 0.7088, "step": 29382 }, { "epoch": 1.77, "learning_rate": 4.0340711903791895e-05, "loss": 0.746, "step": 29384 }, { "epoch": 1.77, "learning_rate": 4.033920299367767e-05, "loss": 0.731, "step": 29386 }, { "epoch": 1.77, "learning_rate": 4.033769408356344e-05, "loss": 0.9279, "step": 29388 }, { "epoch": 1.77, "learning_rate": 4.033618517344922e-05, "loss": 0.5059, "step": 29390 }, { "epoch": 1.77, "learning_rate": 4.0334676263334994e-05, "loss": 0.7038, "step": 29392 }, { "epoch": 1.77, "learning_rate": 4.033316735322077e-05, "loss": 0.8038, "step": 29394 }, { "epoch": 1.77, "learning_rate": 4.033165844310655e-05, "loss": 0.8447, "step": 29396 }, { "epoch": 1.77, "learning_rate": 4.033014953299232e-05, "loss": 0.865, "step": 29398 }, { "epoch": 1.77, "learning_rate": 4.0328640622878093e-05, "loss": 0.7943, "step": 29400 }, { "epoch": 1.77, "learning_rate": 4.032713171276387e-05, "loss": 0.7821, "step": 29402 }, { "epoch": 1.77, "learning_rate": 4.0325622802649646e-05, "loss": 0.8911, "step": 29404 }, { "epoch": 1.77, "learning_rate": 4.0324113892535426e-05, "loss": 0.8056, "step": 29406 }, { "epoch": 1.77, "learning_rate": 4.03226049824212e-05, "loss": 0.9161, "step": 29408 }, { "epoch": 1.78, "learning_rate": 4.032109607230697e-05, "loss": 0.9629, "step": 29410 }, { "epoch": 1.78, "learning_rate": 4.031958716219275e-05, "loss": 0.9629, "step": 29412 }, { "epoch": 1.78, "learning_rate": 4.0318078252078526e-05, "loss": 0.6492, "step": 29414 }, { "epoch": 1.78, "learning_rate": 4.03165693419643e-05, "loss": 0.8116, "step": 29416 }, { "epoch": 1.78, "learning_rate": 4.031506043185008e-05, "loss": 0.9113, "step": 29418 }, { "epoch": 1.78, "learning_rate": 4.031355152173585e-05, "loss": 1.0083, "step": 29420 }, { "epoch": 1.78, "learning_rate": 4.0312042611621625e-05, "loss": 0.5548, "step": 29422 }, { "epoch": 1.78, "learning_rate": 4.0310533701507405e-05, "loss": 0.7838, "step": 29424 }, { "epoch": 1.78, "learning_rate": 4.030902479139318e-05, "loss": 0.8264, "step": 29426 }, { "epoch": 1.78, "learning_rate": 4.030751588127896e-05, "loss": 0.8767, "step": 29428 }, { "epoch": 1.78, "learning_rate": 4.030600697116473e-05, "loss": 0.4314, "step": 29430 }, { "epoch": 1.78, "learning_rate": 4.0304498061050504e-05, "loss": 0.9724, "step": 29432 }, { "epoch": 1.78, "learning_rate": 4.0302989150936284e-05, "loss": 0.8992, "step": 29434 }, { "epoch": 1.78, "learning_rate": 4.030148024082206e-05, "loss": 0.7783, "step": 29436 }, { "epoch": 1.78, "learning_rate": 4.029997133070783e-05, "loss": 0.6602, "step": 29438 }, { "epoch": 1.78, "learning_rate": 4.029846242059361e-05, "loss": 1.0067, "step": 29440 }, { "epoch": 1.78, "learning_rate": 4.029695351047938e-05, "loss": 0.6705, "step": 29442 }, { "epoch": 1.78, "learning_rate": 4.0295444600365156e-05, "loss": 1.1416, "step": 29444 }, { "epoch": 1.78, "learning_rate": 4.0293935690250936e-05, "loss": 0.5606, "step": 29446 }, { "epoch": 1.78, "learning_rate": 4.029242678013671e-05, "loss": 0.6621, "step": 29448 }, { "epoch": 1.78, "learning_rate": 4.029091787002248e-05, "loss": 0.7485, "step": 29450 }, { "epoch": 1.78, "learning_rate": 4.028940895990826e-05, "loss": 0.8647, "step": 29452 }, { "epoch": 1.78, "learning_rate": 4.0287900049794036e-05, "loss": 0.8439, "step": 29454 }, { "epoch": 1.78, "learning_rate": 4.0286391139679815e-05, "loss": 1.2228, "step": 29456 }, { "epoch": 1.78, "learning_rate": 4.028488222956558e-05, "loss": 0.9577, "step": 29458 }, { "epoch": 1.78, "learning_rate": 4.028337331945136e-05, "loss": 0.5139, "step": 29460 }, { "epoch": 1.78, "learning_rate": 4.028186440933714e-05, "loss": 0.8823, "step": 29462 }, { "epoch": 1.78, "learning_rate": 4.0280355499222915e-05, "loss": 1.0366, "step": 29464 }, { "epoch": 1.78, "learning_rate": 4.027884658910869e-05, "loss": 1.1382, "step": 29466 }, { "epoch": 1.78, "learning_rate": 4.027733767899447e-05, "loss": 0.9188, "step": 29468 }, { "epoch": 1.78, "learning_rate": 4.027582876888024e-05, "loss": 0.8972, "step": 29470 }, { "epoch": 1.78, "learning_rate": 4.0274319858766014e-05, "loss": 0.9615, "step": 29472 }, { "epoch": 1.78, "learning_rate": 4.027281094865179e-05, "loss": 0.8915, "step": 29474 }, { "epoch": 1.78, "learning_rate": 4.027130203853757e-05, "loss": 0.6882, "step": 29476 }, { "epoch": 1.78, "learning_rate": 4.026979312842335e-05, "loss": 0.6415, "step": 29478 }, { "epoch": 1.78, "learning_rate": 4.026828421830911e-05, "loss": 0.8358, "step": 29480 }, { "epoch": 1.78, "learning_rate": 4.026677530819489e-05, "loss": 0.7055, "step": 29482 }, { "epoch": 1.78, "learning_rate": 4.026526639808067e-05, "loss": 0.6654, "step": 29484 }, { "epoch": 1.78, "learning_rate": 4.026375748796644e-05, "loss": 0.7167, "step": 29486 }, { "epoch": 1.78, "learning_rate": 4.026224857785222e-05, "loss": 0.7972, "step": 29488 }, { "epoch": 1.78, "learning_rate": 4.026073966773799e-05, "loss": 0.7522, "step": 29490 }, { "epoch": 1.78, "learning_rate": 4.025923075762377e-05, "loss": 0.7356, "step": 29492 }, { "epoch": 1.78, "learning_rate": 4.0257721847509546e-05, "loss": 0.8522, "step": 29494 }, { "epoch": 1.78, "learning_rate": 4.025621293739532e-05, "loss": 0.6945, "step": 29496 }, { "epoch": 1.78, "learning_rate": 4.02547040272811e-05, "loss": 0.8076, "step": 29498 }, { "epoch": 1.78, "learning_rate": 4.025319511716687e-05, "loss": 0.8082, "step": 29500 }, { "epoch": 1.78, "learning_rate": 4.0251686207052645e-05, "loss": 0.9777, "step": 29502 }, { "epoch": 1.78, "learning_rate": 4.0250177296938425e-05, "loss": 0.6416, "step": 29504 }, { "epoch": 1.78, "learning_rate": 4.02486683868242e-05, "loss": 0.7721, "step": 29506 }, { "epoch": 1.78, "learning_rate": 4.024715947670997e-05, "loss": 0.7619, "step": 29508 }, { "epoch": 1.78, "learning_rate": 4.024565056659575e-05, "loss": 0.7306, "step": 29510 }, { "epoch": 1.78, "learning_rate": 4.0244141656481524e-05, "loss": 0.9227, "step": 29512 }, { "epoch": 1.78, "learning_rate": 4.02426327463673e-05, "loss": 0.6881, "step": 29514 }, { "epoch": 1.78, "learning_rate": 4.024112383625308e-05, "loss": 0.5637, "step": 29516 }, { "epoch": 1.78, "learning_rate": 4.023961492613885e-05, "loss": 0.7912, "step": 29518 }, { "epoch": 1.78, "learning_rate": 4.023810601602463e-05, "loss": 0.7615, "step": 29520 }, { "epoch": 1.78, "learning_rate": 4.02365971059104e-05, "loss": 0.636, "step": 29522 }, { "epoch": 1.78, "learning_rate": 4.0235088195796176e-05, "loss": 0.9468, "step": 29524 }, { "epoch": 1.78, "learning_rate": 4.0233579285681956e-05, "loss": 0.6501, "step": 29526 }, { "epoch": 1.78, "learning_rate": 4.023207037556773e-05, "loss": 0.7194, "step": 29528 }, { "epoch": 1.78, "learning_rate": 4.02305614654535e-05, "loss": 0.6972, "step": 29530 }, { "epoch": 1.78, "learning_rate": 4.022905255533928e-05, "loss": 0.7619, "step": 29532 }, { "epoch": 1.78, "learning_rate": 4.0227543645225055e-05, "loss": 0.6285, "step": 29534 }, { "epoch": 1.78, "learning_rate": 4.022603473511083e-05, "loss": 0.9648, "step": 29536 }, { "epoch": 1.78, "learning_rate": 4.022452582499661e-05, "loss": 0.631, "step": 29538 }, { "epoch": 1.78, "learning_rate": 4.022301691488238e-05, "loss": 0.7146, "step": 29540 }, { "epoch": 1.78, "learning_rate": 4.022150800476816e-05, "loss": 0.5287, "step": 29542 }, { "epoch": 1.78, "learning_rate": 4.0219999094653935e-05, "loss": 0.8142, "step": 29544 }, { "epoch": 1.78, "learning_rate": 4.021849018453971e-05, "loss": 1.0113, "step": 29546 }, { "epoch": 1.78, "learning_rate": 4.021698127442549e-05, "loss": 0.7846, "step": 29548 }, { "epoch": 1.78, "learning_rate": 4.021547236431126e-05, "loss": 1.0644, "step": 29550 }, { "epoch": 1.78, "learning_rate": 4.0213963454197034e-05, "loss": 0.7972, "step": 29552 }, { "epoch": 1.78, "learning_rate": 4.0212454544082814e-05, "loss": 0.8483, "step": 29554 }, { "epoch": 1.78, "learning_rate": 4.021094563396859e-05, "loss": 0.8713, "step": 29556 }, { "epoch": 1.78, "learning_rate": 4.020943672385436e-05, "loss": 0.728, "step": 29558 }, { "epoch": 1.78, "learning_rate": 4.020792781374014e-05, "loss": 0.8408, "step": 29560 }, { "epoch": 1.78, "learning_rate": 4.020641890362591e-05, "loss": 0.9447, "step": 29562 }, { "epoch": 1.78, "learning_rate": 4.0204909993511686e-05, "loss": 0.8478, "step": 29564 }, { "epoch": 1.78, "learning_rate": 4.0203401083397466e-05, "loss": 0.5802, "step": 29566 }, { "epoch": 1.78, "learning_rate": 4.020189217328324e-05, "loss": 0.7747, "step": 29568 }, { "epoch": 1.78, "learning_rate": 4.020038326316902e-05, "loss": 0.7671, "step": 29570 }, { "epoch": 1.78, "learning_rate": 4.0198874353054785e-05, "loss": 0.8014, "step": 29572 }, { "epoch": 1.79, "learning_rate": 4.0197365442940565e-05, "loss": 0.8132, "step": 29574 }, { "epoch": 1.79, "learning_rate": 4.0195856532826345e-05, "loss": 0.7923, "step": 29576 }, { "epoch": 1.79, "learning_rate": 4.019434762271212e-05, "loss": 0.6865, "step": 29578 }, { "epoch": 1.79, "learning_rate": 4.019283871259789e-05, "loss": 0.9648, "step": 29580 }, { "epoch": 1.79, "learning_rate": 4.019132980248367e-05, "loss": 0.837, "step": 29582 }, { "epoch": 1.79, "learning_rate": 4.0189820892369445e-05, "loss": 0.7605, "step": 29584 }, { "epoch": 1.79, "learning_rate": 4.018831198225522e-05, "loss": 0.696, "step": 29586 }, { "epoch": 1.79, "learning_rate": 4.018680307214099e-05, "loss": 0.7135, "step": 29588 }, { "epoch": 1.79, "learning_rate": 4.018529416202677e-05, "loss": 0.7739, "step": 29590 }, { "epoch": 1.79, "learning_rate": 4.018378525191255e-05, "loss": 0.7328, "step": 29592 }, { "epoch": 1.79, "learning_rate": 4.018227634179832e-05, "loss": 0.6604, "step": 29594 }, { "epoch": 1.79, "learning_rate": 4.01807674316841e-05, "loss": 0.8409, "step": 29596 }, { "epoch": 1.79, "learning_rate": 4.017925852156988e-05, "loss": 0.7411, "step": 29598 }, { "epoch": 1.79, "learning_rate": 4.017774961145564e-05, "loss": 0.8402, "step": 29600 }, { "epoch": 1.79, "learning_rate": 4.017624070134142e-05, "loss": 0.6027, "step": 29602 }, { "epoch": 1.79, "learning_rate": 4.0174731791227196e-05, "loss": 0.8137, "step": 29604 }, { "epoch": 1.79, "learning_rate": 4.0173222881112976e-05, "loss": 0.935, "step": 29606 }, { "epoch": 1.79, "learning_rate": 4.017171397099875e-05, "loss": 0.748, "step": 29608 }, { "epoch": 1.79, "learning_rate": 4.017020506088452e-05, "loss": 0.9314, "step": 29610 }, { "epoch": 1.79, "learning_rate": 4.01686961507703e-05, "loss": 0.7028, "step": 29612 }, { "epoch": 1.79, "learning_rate": 4.0167187240656075e-05, "loss": 0.6542, "step": 29614 }, { "epoch": 1.79, "learning_rate": 4.016567833054185e-05, "loss": 0.6465, "step": 29616 }, { "epoch": 1.79, "learning_rate": 4.016416942042763e-05, "loss": 1.022, "step": 29618 }, { "epoch": 1.79, "learning_rate": 4.01626605103134e-05, "loss": 0.7782, "step": 29620 }, { "epoch": 1.79, "learning_rate": 4.0161151600199175e-05, "loss": 0.9511, "step": 29622 }, { "epoch": 1.79, "learning_rate": 4.0159642690084955e-05, "loss": 0.6144, "step": 29624 }, { "epoch": 1.79, "learning_rate": 4.015813377997073e-05, "loss": 0.7167, "step": 29626 }, { "epoch": 1.79, "learning_rate": 4.015662486985651e-05, "loss": 0.8238, "step": 29628 }, { "epoch": 1.79, "learning_rate": 4.015511595974228e-05, "loss": 1.1727, "step": 29630 }, { "epoch": 1.79, "learning_rate": 4.0153607049628054e-05, "loss": 0.8357, "step": 29632 }, { "epoch": 1.79, "learning_rate": 4.0152098139513834e-05, "loss": 0.7863, "step": 29634 }, { "epoch": 1.79, "learning_rate": 4.015058922939961e-05, "loss": 0.9562, "step": 29636 }, { "epoch": 1.79, "learning_rate": 4.014908031928538e-05, "loss": 0.9589, "step": 29638 }, { "epoch": 1.79, "learning_rate": 4.014757140917116e-05, "loss": 0.966, "step": 29640 }, { "epoch": 1.79, "learning_rate": 4.014606249905693e-05, "loss": 0.8133, "step": 29642 }, { "epoch": 1.79, "learning_rate": 4.0144553588942706e-05, "loss": 0.7757, "step": 29644 }, { "epoch": 1.79, "learning_rate": 4.0143044678828486e-05, "loss": 0.8651, "step": 29646 }, { "epoch": 1.79, "learning_rate": 4.014153576871426e-05, "loss": 0.6957, "step": 29648 }, { "epoch": 1.79, "learning_rate": 4.014002685860003e-05, "loss": 1.0697, "step": 29650 }, { "epoch": 1.79, "learning_rate": 4.013851794848581e-05, "loss": 0.9842, "step": 29652 }, { "epoch": 1.79, "learning_rate": 4.0137009038371585e-05, "loss": 0.7702, "step": 29654 }, { "epoch": 1.79, "learning_rate": 4.0135500128257365e-05, "loss": 0.813, "step": 29656 }, { "epoch": 1.79, "learning_rate": 4.013399121814314e-05, "loss": 0.8769, "step": 29658 }, { "epoch": 1.79, "learning_rate": 4.013248230802891e-05, "loss": 0.7467, "step": 29660 }, { "epoch": 1.79, "learning_rate": 4.013097339791469e-05, "loss": 0.7447, "step": 29662 }, { "epoch": 1.79, "learning_rate": 4.0129464487800464e-05, "loss": 0.8746, "step": 29664 }, { "epoch": 1.79, "learning_rate": 4.012795557768624e-05, "loss": 0.8286, "step": 29666 }, { "epoch": 1.79, "learning_rate": 4.012644666757202e-05, "loss": 0.8246, "step": 29668 }, { "epoch": 1.79, "learning_rate": 4.012493775745779e-05, "loss": 0.54, "step": 29670 }, { "epoch": 1.79, "learning_rate": 4.0123428847343564e-05, "loss": 0.4655, "step": 29672 }, { "epoch": 1.79, "learning_rate": 4.0121919937229344e-05, "loss": 0.8001, "step": 29674 }, { "epoch": 1.79, "learning_rate": 4.012041102711512e-05, "loss": 0.7624, "step": 29676 }, { "epoch": 1.79, "learning_rate": 4.011890211700089e-05, "loss": 0.8269, "step": 29678 }, { "epoch": 1.79, "learning_rate": 4.011739320688667e-05, "loss": 0.9742, "step": 29680 }, { "epoch": 1.79, "learning_rate": 4.011588429677244e-05, "loss": 0.8824, "step": 29682 }, { "epoch": 1.79, "learning_rate": 4.011437538665822e-05, "loss": 0.661, "step": 29684 }, { "epoch": 1.79, "learning_rate": 4.011286647654399e-05, "loss": 0.786, "step": 29686 }, { "epoch": 1.79, "learning_rate": 4.011135756642977e-05, "loss": 0.6534, "step": 29688 }, { "epoch": 1.79, "learning_rate": 4.010984865631555e-05, "loss": 0.6002, "step": 29690 }, { "epoch": 1.79, "learning_rate": 4.010833974620132e-05, "loss": 0.7558, "step": 29692 }, { "epoch": 1.79, "learning_rate": 4.0106830836087095e-05, "loss": 0.8585, "step": 29694 }, { "epoch": 1.79, "learning_rate": 4.0105321925972875e-05, "loss": 0.7548, "step": 29696 }, { "epoch": 1.79, "learning_rate": 4.010381301585865e-05, "loss": 1.275, "step": 29698 }, { "epoch": 1.79, "learning_rate": 4.010230410574442e-05, "loss": 0.7013, "step": 29700 }, { "epoch": 1.79, "learning_rate": 4.0100795195630195e-05, "loss": 0.9668, "step": 29702 }, { "epoch": 1.79, "learning_rate": 4.0099286285515974e-05, "loss": 0.8302, "step": 29704 }, { "epoch": 1.79, "learning_rate": 4.0097777375401754e-05, "loss": 0.828, "step": 29706 }, { "epoch": 1.79, "learning_rate": 4.009626846528752e-05, "loss": 0.6715, "step": 29708 }, { "epoch": 1.79, "learning_rate": 4.00947595551733e-05, "loss": 0.8628, "step": 29710 }, { "epoch": 1.79, "learning_rate": 4.009325064505908e-05, "loss": 0.8029, "step": 29712 }, { "epoch": 1.79, "learning_rate": 4.009174173494485e-05, "loss": 0.7902, "step": 29714 }, { "epoch": 1.79, "learning_rate": 4.009023282483063e-05, "loss": 0.703, "step": 29716 }, { "epoch": 1.79, "learning_rate": 4.00887239147164e-05, "loss": 0.899, "step": 29718 }, { "epoch": 1.79, "learning_rate": 4.008721500460218e-05, "loss": 0.6077, "step": 29720 }, { "epoch": 1.79, "learning_rate": 4.008570609448795e-05, "loss": 0.7951, "step": 29722 }, { "epoch": 1.79, "learning_rate": 4.0084197184373726e-05, "loss": 0.9425, "step": 29724 }, { "epoch": 1.79, "learning_rate": 4.0082688274259506e-05, "loss": 0.8831, "step": 29726 }, { "epoch": 1.79, "learning_rate": 4.008117936414528e-05, "loss": 0.5989, "step": 29728 }, { "epoch": 1.79, "learning_rate": 4.007967045403105e-05, "loss": 0.5589, "step": 29730 }, { "epoch": 1.79, "learning_rate": 4.007816154391683e-05, "loss": 0.6257, "step": 29732 }, { "epoch": 1.79, "learning_rate": 4.0076652633802605e-05, "loss": 0.7473, "step": 29734 }, { "epoch": 1.79, "learning_rate": 4.007514372368838e-05, "loss": 0.645, "step": 29736 }, { "epoch": 1.79, "learning_rate": 4.007363481357416e-05, "loss": 0.9009, "step": 29738 }, { "epoch": 1.8, "learning_rate": 4.007212590345993e-05, "loss": 0.881, "step": 29740 }, { "epoch": 1.8, "learning_rate": 4.007061699334571e-05, "loss": 0.9311, "step": 29742 }, { "epoch": 1.8, "learning_rate": 4.0069108083231484e-05, "loss": 0.8284, "step": 29744 }, { "epoch": 1.8, "learning_rate": 4.006759917311726e-05, "loss": 0.7919, "step": 29746 }, { "epoch": 1.8, "learning_rate": 4.006609026300304e-05, "loss": 0.8909, "step": 29748 }, { "epoch": 1.8, "learning_rate": 4.006458135288881e-05, "loss": 0.8019, "step": 29750 }, { "epoch": 1.8, "learning_rate": 4.0063072442774584e-05, "loss": 0.7216, "step": 29752 }, { "epoch": 1.8, "learning_rate": 4.0061563532660364e-05, "loss": 0.8798, "step": 29754 }, { "epoch": 1.8, "learning_rate": 4.006005462254614e-05, "loss": 0.7479, "step": 29756 }, { "epoch": 1.8, "learning_rate": 4.005854571243191e-05, "loss": 0.9299, "step": 29758 }, { "epoch": 1.8, "learning_rate": 4.005703680231769e-05, "loss": 0.9455, "step": 29760 }, { "epoch": 1.8, "learning_rate": 4.005552789220346e-05, "loss": 0.9628, "step": 29762 }, { "epoch": 1.8, "learning_rate": 4.0054018982089236e-05, "loss": 0.6057, "step": 29764 }, { "epoch": 1.8, "learning_rate": 4.0052510071975016e-05, "loss": 0.6517, "step": 29766 }, { "epoch": 1.8, "learning_rate": 4.005100116186079e-05, "loss": 0.7088, "step": 29768 }, { "epoch": 1.8, "learning_rate": 4.004949225174657e-05, "loss": 0.8874, "step": 29770 }, { "epoch": 1.8, "learning_rate": 4.004798334163234e-05, "loss": 0.681, "step": 29772 }, { "epoch": 1.8, "learning_rate": 4.0046474431518115e-05, "loss": 0.8512, "step": 29774 }, { "epoch": 1.8, "learning_rate": 4.0044965521403895e-05, "loss": 1.173, "step": 29776 }, { "epoch": 1.8, "learning_rate": 4.004345661128966e-05, "loss": 0.6178, "step": 29778 }, { "epoch": 1.8, "learning_rate": 4.004194770117544e-05, "loss": 0.7971, "step": 29780 }, { "epoch": 1.8, "learning_rate": 4.004043879106122e-05, "loss": 0.7148, "step": 29782 }, { "epoch": 1.8, "learning_rate": 4.0038929880946994e-05, "loss": 0.8474, "step": 29784 }, { "epoch": 1.8, "learning_rate": 4.003742097083277e-05, "loss": 0.6285, "step": 29786 }, { "epoch": 1.8, "learning_rate": 4.003591206071855e-05, "loss": 0.8464, "step": 29788 }, { "epoch": 1.8, "learning_rate": 4.003440315060432e-05, "loss": 0.6881, "step": 29790 }, { "epoch": 1.8, "learning_rate": 4.00328942404901e-05, "loss": 0.9151, "step": 29792 }, { "epoch": 1.8, "learning_rate": 4.0031385330375873e-05, "loss": 0.6792, "step": 29794 }, { "epoch": 1.8, "learning_rate": 4.0029876420261647e-05, "loss": 1.5132, "step": 29796 }, { "epoch": 1.8, "learning_rate": 4.0028367510147427e-05, "loss": 0.8815, "step": 29798 }, { "epoch": 1.8, "learning_rate": 4.002685860003319e-05, "loss": 0.7168, "step": 29800 }, { "epoch": 1.8, "learning_rate": 4.002534968991897e-05, "loss": 0.8947, "step": 29802 }, { "epoch": 1.8, "learning_rate": 4.002384077980475e-05, "loss": 0.7234, "step": 29804 }, { "epoch": 1.8, "learning_rate": 4.0022331869690526e-05, "loss": 0.6777, "step": 29806 }, { "epoch": 1.8, "learning_rate": 4.00208229595763e-05, "loss": 0.6584, "step": 29808 }, { "epoch": 1.8, "learning_rate": 4.001931404946208e-05, "loss": 0.8478, "step": 29810 }, { "epoch": 1.8, "learning_rate": 4.001780513934785e-05, "loss": 0.7812, "step": 29812 }, { "epoch": 1.8, "learning_rate": 4.0016296229233625e-05, "loss": 0.872, "step": 29814 }, { "epoch": 1.8, "learning_rate": 4.00147873191194e-05, "loss": 0.9634, "step": 29816 }, { "epoch": 1.8, "learning_rate": 4.001327840900518e-05, "loss": 0.8583, "step": 29818 }, { "epoch": 1.8, "learning_rate": 4.001176949889096e-05, "loss": 0.5642, "step": 29820 }, { "epoch": 1.8, "learning_rate": 4.0010260588776724e-05, "loss": 0.7196, "step": 29822 }, { "epoch": 1.8, "learning_rate": 4.0008751678662504e-05, "loss": 0.8393, "step": 29824 }, { "epoch": 1.8, "learning_rate": 4.0007242768548284e-05, "loss": 0.8942, "step": 29826 }, { "epoch": 1.8, "learning_rate": 4.000573385843405e-05, "loss": 0.9481, "step": 29828 }, { "epoch": 1.8, "learning_rate": 4.000422494831983e-05, "loss": 0.8827, "step": 29830 }, { "epoch": 1.8, "learning_rate": 4.0002716038205604e-05, "loss": 0.7576, "step": 29832 }, { "epoch": 1.8, "learning_rate": 4.0001207128091383e-05, "loss": 0.8617, "step": 29834 }, { "epoch": 1.8, "learning_rate": 3.9999698217977157e-05, "loss": 0.7741, "step": 29836 }, { "epoch": 1.8, "learning_rate": 3.999818930786293e-05, "loss": 1.0552, "step": 29838 }, { "epoch": 1.8, "learning_rate": 3.999668039774871e-05, "loss": 0.7042, "step": 29840 }, { "epoch": 1.8, "learning_rate": 3.999517148763449e-05, "loss": 0.6239, "step": 29842 }, { "epoch": 1.8, "learning_rate": 3.9993662577520256e-05, "loss": 0.6406, "step": 29844 }, { "epoch": 1.8, "learning_rate": 3.9992153667406036e-05, "loss": 1.087, "step": 29846 }, { "epoch": 1.8, "learning_rate": 3.999064475729181e-05, "loss": 0.8999, "step": 29848 }, { "epoch": 1.8, "learning_rate": 3.998913584717758e-05, "loss": 0.8243, "step": 29850 }, { "epoch": 1.8, "learning_rate": 3.998762693706336e-05, "loss": 0.7975, "step": 29852 }, { "epoch": 1.8, "learning_rate": 3.9986118026949135e-05, "loss": 1.1227, "step": 29854 }, { "epoch": 1.8, "learning_rate": 3.9984609116834915e-05, "loss": 0.7056, "step": 29856 }, { "epoch": 1.8, "learning_rate": 3.998310020672069e-05, "loss": 1.0031, "step": 29858 }, { "epoch": 1.8, "learning_rate": 3.998159129660646e-05, "loss": 0.9146, "step": 29860 }, { "epoch": 1.8, "learning_rate": 3.998008238649224e-05, "loss": 0.7506, "step": 29862 }, { "epoch": 1.8, "learning_rate": 3.9978573476378014e-05, "loss": 0.5823, "step": 29864 }, { "epoch": 1.8, "learning_rate": 3.997706456626379e-05, "loss": 0.6079, "step": 29866 }, { "epoch": 1.8, "learning_rate": 3.997555565614957e-05, "loss": 0.9087, "step": 29868 }, { "epoch": 1.8, "learning_rate": 3.997404674603534e-05, "loss": 0.6717, "step": 29870 }, { "epoch": 1.8, "learning_rate": 3.9972537835921113e-05, "loss": 0.8946, "step": 29872 }, { "epoch": 1.8, "learning_rate": 3.997102892580689e-05, "loss": 0.6513, "step": 29874 }, { "epoch": 1.8, "learning_rate": 3.9969520015692666e-05, "loss": 0.5165, "step": 29876 }, { "epoch": 1.8, "learning_rate": 3.996801110557844e-05, "loss": 0.8745, "step": 29878 }, { "epoch": 1.8, "learning_rate": 3.996650219546422e-05, "loss": 0.8023, "step": 29880 }, { "epoch": 1.8, "learning_rate": 3.996499328534999e-05, "loss": 0.7888, "step": 29882 }, { "epoch": 1.8, "learning_rate": 3.996348437523577e-05, "loss": 1.1225, "step": 29884 }, { "epoch": 1.8, "learning_rate": 3.9961975465121546e-05, "loss": 0.8125, "step": 29886 }, { "epoch": 1.8, "learning_rate": 3.996046655500732e-05, "loss": 0.7336, "step": 29888 }, { "epoch": 1.8, "learning_rate": 3.99589576448931e-05, "loss": 0.7213, "step": 29890 }, { "epoch": 1.8, "learning_rate": 3.9957448734778865e-05, "loss": 1.0392, "step": 29892 }, { "epoch": 1.8, "learning_rate": 3.9955939824664645e-05, "loss": 0.6793, "step": 29894 }, { "epoch": 1.8, "learning_rate": 3.9954430914550425e-05, "loss": 0.6343, "step": 29896 }, { "epoch": 1.8, "learning_rate": 3.99529220044362e-05, "loss": 0.7378, "step": 29898 }, { "epoch": 1.8, "learning_rate": 3.995141309432197e-05, "loss": 0.7758, "step": 29900 }, { "epoch": 1.8, "learning_rate": 3.994990418420775e-05, "loss": 0.8359, "step": 29902 }, { "epoch": 1.8, "learning_rate": 3.9948395274093524e-05, "loss": 1.034, "step": 29904 }, { "epoch": 1.81, "learning_rate": 3.9946886363979304e-05, "loss": 1.1366, "step": 29906 }, { "epoch": 1.81, "learning_rate": 3.994537745386508e-05, "loss": 0.8159, "step": 29908 }, { "epoch": 1.81, "learning_rate": 3.994386854375085e-05, "loss": 0.8756, "step": 29910 }, { "epoch": 1.81, "learning_rate": 3.994235963363663e-05, "loss": 0.7118, "step": 29912 }, { "epoch": 1.81, "learning_rate": 3.9940850723522397e-05, "loss": 1.0413, "step": 29914 }, { "epoch": 1.81, "learning_rate": 3.9939341813408176e-05, "loss": 0.7831, "step": 29916 }, { "epoch": 1.81, "learning_rate": 3.9937832903293956e-05, "loss": 1.3763, "step": 29918 }, { "epoch": 1.81, "learning_rate": 3.993632399317973e-05, "loss": 0.8691, "step": 29920 }, { "epoch": 1.81, "learning_rate": 3.99348150830655e-05, "loss": 0.7538, "step": 29922 }, { "epoch": 1.81, "learning_rate": 3.993330617295128e-05, "loss": 0.9486, "step": 29924 }, { "epoch": 1.81, "learning_rate": 3.9931797262837056e-05, "loss": 0.7295, "step": 29926 }, { "epoch": 1.81, "learning_rate": 3.993028835272283e-05, "loss": 0.8348, "step": 29928 }, { "epoch": 1.81, "learning_rate": 3.99287794426086e-05, "loss": 0.9206, "step": 29930 }, { "epoch": 1.81, "learning_rate": 3.992727053249438e-05, "loss": 0.8761, "step": 29932 }, { "epoch": 1.81, "learning_rate": 3.992576162238016e-05, "loss": 0.6709, "step": 29934 }, { "epoch": 1.81, "learning_rate": 3.992425271226593e-05, "loss": 0.7617, "step": 29936 }, { "epoch": 1.81, "learning_rate": 3.992274380215171e-05, "loss": 0.8096, "step": 29938 }, { "epoch": 1.81, "learning_rate": 3.992123489203749e-05, "loss": 1.0115, "step": 29940 }, { "epoch": 1.81, "learning_rate": 3.9919725981923254e-05, "loss": 0.8156, "step": 29942 }, { "epoch": 1.81, "learning_rate": 3.9918217071809034e-05, "loss": 0.5616, "step": 29944 }, { "epoch": 1.81, "learning_rate": 3.991670816169481e-05, "loss": 0.778, "step": 29946 }, { "epoch": 1.81, "learning_rate": 3.991519925158059e-05, "loss": 0.9423, "step": 29948 }, { "epoch": 1.81, "learning_rate": 3.991369034146636e-05, "loss": 0.839, "step": 29950 }, { "epoch": 1.81, "learning_rate": 3.991218143135213e-05, "loss": 0.8854, "step": 29952 }, { "epoch": 1.81, "learning_rate": 3.991067252123791e-05, "loss": 0.727, "step": 29954 }, { "epoch": 1.81, "learning_rate": 3.990916361112369e-05, "loss": 0.8651, "step": 29956 }, { "epoch": 1.81, "learning_rate": 3.990765470100946e-05, "loss": 0.8837, "step": 29958 }, { "epoch": 1.81, "learning_rate": 3.990614579089524e-05, "loss": 0.7851, "step": 29960 }, { "epoch": 1.81, "learning_rate": 3.990463688078101e-05, "loss": 0.901, "step": 29962 }, { "epoch": 1.81, "learning_rate": 3.9903127970666786e-05, "loss": 0.7025, "step": 29964 }, { "epoch": 1.81, "learning_rate": 3.9901619060552566e-05, "loss": 0.7921, "step": 29966 }, { "epoch": 1.81, "learning_rate": 3.990011015043834e-05, "loss": 0.9163, "step": 29968 }, { "epoch": 1.81, "learning_rate": 3.989860124032412e-05, "loss": 0.9046, "step": 29970 }, { "epoch": 1.81, "learning_rate": 3.989709233020989e-05, "loss": 0.7711, "step": 29972 }, { "epoch": 1.81, "learning_rate": 3.9895583420095665e-05, "loss": 0.6485, "step": 29974 }, { "epoch": 1.81, "learning_rate": 3.9894074509981445e-05, "loss": 0.6844, "step": 29976 }, { "epoch": 1.81, "learning_rate": 3.989256559986722e-05, "loss": 0.8894, "step": 29978 }, { "epoch": 1.81, "learning_rate": 3.989105668975299e-05, "loss": 0.7148, "step": 29980 }, { "epoch": 1.81, "learning_rate": 3.988954777963877e-05, "loss": 0.6763, "step": 29982 }, { "epoch": 1.81, "learning_rate": 3.9888038869524544e-05, "loss": 0.8777, "step": 29984 }, { "epoch": 1.81, "learning_rate": 3.988652995941032e-05, "loss": 0.7253, "step": 29986 }, { "epoch": 1.81, "learning_rate": 3.98850210492961e-05, "loss": 0.8187, "step": 29988 }, { "epoch": 1.81, "learning_rate": 3.988351213918187e-05, "loss": 0.8895, "step": 29990 }, { "epoch": 1.81, "learning_rate": 3.988200322906764e-05, "loss": 0.9477, "step": 29992 }, { "epoch": 1.81, "learning_rate": 3.988049431895342e-05, "loss": 0.7381, "step": 29994 }, { "epoch": 1.81, "learning_rate": 3.9878985408839196e-05, "loss": 0.9018, "step": 29996 }, { "epoch": 1.81, "learning_rate": 3.9877476498724976e-05, "loss": 1.0134, "step": 29998 }, { "epoch": 1.81, "learning_rate": 3.987596758861075e-05, "loss": 0.7996, "step": 30000 }, { "epoch": 1.81, "learning_rate": 3.987445867849652e-05, "loss": 0.7876, "step": 30002 }, { "epoch": 1.81, "learning_rate": 3.98729497683823e-05, "loss": 0.7914, "step": 30004 }, { "epoch": 1.81, "learning_rate": 3.9871440858268075e-05, "loss": 0.944, "step": 30006 }, { "epoch": 1.81, "learning_rate": 3.986993194815385e-05, "loss": 0.864, "step": 30008 }, { "epoch": 1.81, "learning_rate": 3.986842303803963e-05, "loss": 1.1838, "step": 30010 }, { "epoch": 1.81, "learning_rate": 3.98669141279254e-05, "loss": 0.9205, "step": 30012 }, { "epoch": 1.81, "learning_rate": 3.9865405217811175e-05, "loss": 0.4999, "step": 30014 }, { "epoch": 1.81, "learning_rate": 3.9863896307696955e-05, "loss": 0.9624, "step": 30016 }, { "epoch": 1.81, "learning_rate": 3.986238739758273e-05, "loss": 0.6487, "step": 30018 }, { "epoch": 1.81, "learning_rate": 3.986087848746851e-05, "loss": 0.7366, "step": 30020 }, { "epoch": 1.81, "learning_rate": 3.985936957735428e-05, "loss": 1.1664, "step": 30022 }, { "epoch": 1.81, "learning_rate": 3.9857860667240054e-05, "loss": 0.7464, "step": 30024 }, { "epoch": 1.81, "learning_rate": 3.9856351757125834e-05, "loss": 0.8006, "step": 30026 }, { "epoch": 1.81, "learning_rate": 3.98548428470116e-05, "loss": 1.1848, "step": 30028 }, { "epoch": 1.81, "learning_rate": 3.985333393689738e-05, "loss": 0.6932, "step": 30030 }, { "epoch": 1.81, "learning_rate": 3.985182502678316e-05, "loss": 0.7829, "step": 30032 }, { "epoch": 1.81, "learning_rate": 3.985031611666893e-05, "loss": 0.7761, "step": 30034 }, { "epoch": 1.81, "learning_rate": 3.9848807206554706e-05, "loss": 0.7958, "step": 30036 }, { "epoch": 1.81, "learning_rate": 3.9847298296440486e-05, "loss": 0.9801, "step": 30038 }, { "epoch": 1.81, "learning_rate": 3.984578938632626e-05, "loss": 0.8139, "step": 30040 }, { "epoch": 1.81, "learning_rate": 3.984428047621203e-05, "loss": 0.6092, "step": 30042 }, { "epoch": 1.81, "learning_rate": 3.9842771566097806e-05, "loss": 0.6195, "step": 30044 }, { "epoch": 1.81, "learning_rate": 3.9841262655983585e-05, "loss": 0.8607, "step": 30046 }, { "epoch": 1.81, "learning_rate": 3.9839753745869365e-05, "loss": 0.5861, "step": 30048 }, { "epoch": 1.81, "learning_rate": 3.983824483575513e-05, "loss": 0.8405, "step": 30050 }, { "epoch": 1.81, "learning_rate": 3.983673592564091e-05, "loss": 0.9013, "step": 30052 }, { "epoch": 1.81, "learning_rate": 3.983522701552669e-05, "loss": 0.7726, "step": 30054 }, { "epoch": 1.81, "learning_rate": 3.983371810541246e-05, "loss": 0.6595, "step": 30056 }, { "epoch": 1.81, "learning_rate": 3.983220919529824e-05, "loss": 0.5791, "step": 30058 }, { "epoch": 1.81, "learning_rate": 3.983070028518401e-05, "loss": 0.7856, "step": 30060 }, { "epoch": 1.81, "learning_rate": 3.982919137506979e-05, "loss": 0.6232, "step": 30062 }, { "epoch": 1.81, "learning_rate": 3.9827682464955564e-05, "loss": 0.5157, "step": 30064 }, { "epoch": 1.81, "learning_rate": 3.982617355484134e-05, "loss": 0.7063, "step": 30066 }, { "epoch": 1.81, "learning_rate": 3.982466464472712e-05, "loss": 0.837, "step": 30068 }, { "epoch": 1.81, "learning_rate": 3.98231557346129e-05, "loss": 0.765, "step": 30070 }, { "epoch": 1.82, "learning_rate": 3.982164682449866e-05, "loss": 0.6635, "step": 30072 }, { "epoch": 1.82, "learning_rate": 3.982013791438444e-05, "loss": 0.4937, "step": 30074 }, { "epoch": 1.82, "learning_rate": 3.9818629004270216e-05, "loss": 0.658, "step": 30076 }, { "epoch": 1.82, "learning_rate": 3.981712009415599e-05, "loss": 0.674, "step": 30078 }, { "epoch": 1.82, "learning_rate": 3.981561118404177e-05, "loss": 1.0306, "step": 30080 }, { "epoch": 1.82, "learning_rate": 3.981410227392754e-05, "loss": 0.6347, "step": 30082 }, { "epoch": 1.82, "learning_rate": 3.981259336381332e-05, "loss": 0.6089, "step": 30084 }, { "epoch": 1.82, "learning_rate": 3.9811084453699095e-05, "loss": 0.664, "step": 30086 }, { "epoch": 1.82, "learning_rate": 3.980957554358487e-05, "loss": 0.8001, "step": 30088 }, { "epoch": 1.82, "learning_rate": 3.980806663347065e-05, "loss": 0.8312, "step": 30090 }, { "epoch": 1.82, "learning_rate": 3.980655772335642e-05, "loss": 0.8567, "step": 30092 }, { "epoch": 1.82, "learning_rate": 3.9805048813242195e-05, "loss": 0.7192, "step": 30094 }, { "epoch": 1.82, "learning_rate": 3.9803539903127975e-05, "loss": 0.6305, "step": 30096 }, { "epoch": 1.82, "learning_rate": 3.980203099301375e-05, "loss": 0.6559, "step": 30098 }, { "epoch": 1.82, "learning_rate": 3.980052208289952e-05, "loss": 0.6341, "step": 30100 }, { "epoch": 1.82, "learning_rate": 3.97990131727853e-05, "loss": 0.7724, "step": 30102 }, { "epoch": 1.82, "learning_rate": 3.9797504262671074e-05, "loss": 0.6688, "step": 30104 }, { "epoch": 1.82, "learning_rate": 3.979599535255685e-05, "loss": 0.8672, "step": 30106 }, { "epoch": 1.82, "learning_rate": 3.979448644244263e-05, "loss": 0.8683, "step": 30108 }, { "epoch": 1.82, "learning_rate": 3.97929775323284e-05, "loss": 0.7927, "step": 30110 }, { "epoch": 1.82, "learning_rate": 3.979146862221418e-05, "loss": 0.7719, "step": 30112 }, { "epoch": 1.82, "learning_rate": 3.978995971209995e-05, "loss": 1.276, "step": 30114 }, { "epoch": 1.82, "learning_rate": 3.9788450801985726e-05, "loss": 0.9583, "step": 30116 }, { "epoch": 1.82, "learning_rate": 3.9786941891871506e-05, "loss": 1.2683, "step": 30118 }, { "epoch": 1.82, "learning_rate": 3.978543298175728e-05, "loss": 0.8349, "step": 30120 }, { "epoch": 1.82, "learning_rate": 3.978392407164305e-05, "loss": 1.1007, "step": 30122 }, { "epoch": 1.82, "learning_rate": 3.978241516152883e-05, "loss": 0.7883, "step": 30124 }, { "epoch": 1.82, "learning_rate": 3.9780906251414605e-05, "loss": 0.4197, "step": 30126 }, { "epoch": 1.82, "learning_rate": 3.977939734130038e-05, "loss": 0.7286, "step": 30128 }, { "epoch": 1.82, "learning_rate": 3.977788843118616e-05, "loss": 0.7657, "step": 30130 }, { "epoch": 1.82, "learning_rate": 3.977637952107193e-05, "loss": 0.8675, "step": 30132 }, { "epoch": 1.82, "learning_rate": 3.977487061095771e-05, "loss": 1.003, "step": 30134 }, { "epoch": 1.82, "learning_rate": 3.9773361700843485e-05, "loss": 0.9126, "step": 30136 }, { "epoch": 1.82, "learning_rate": 3.977185279072926e-05, "loss": 0.6352, "step": 30138 }, { "epoch": 1.82, "learning_rate": 3.977034388061504e-05, "loss": 1.0564, "step": 30140 }, { "epoch": 1.82, "learning_rate": 3.9768834970500804e-05, "loss": 0.7393, "step": 30142 }, { "epoch": 1.82, "learning_rate": 3.9767326060386584e-05, "loss": 0.6152, "step": 30144 }, { "epoch": 1.82, "learning_rate": 3.9765817150272364e-05, "loss": 0.7764, "step": 30146 }, { "epoch": 1.82, "learning_rate": 3.976430824015814e-05, "loss": 0.8659, "step": 30148 }, { "epoch": 1.82, "learning_rate": 3.976279933004391e-05, "loss": 0.7759, "step": 30150 }, { "epoch": 1.82, "learning_rate": 3.976129041992969e-05, "loss": 0.9261, "step": 30152 }, { "epoch": 1.82, "learning_rate": 3.975978150981546e-05, "loss": 0.7037, "step": 30154 }, { "epoch": 1.82, "learning_rate": 3.9758272599701236e-05, "loss": 0.9194, "step": 30156 }, { "epoch": 1.82, "learning_rate": 3.975676368958701e-05, "loss": 0.9055, "step": 30158 }, { "epoch": 1.82, "learning_rate": 3.975525477947279e-05, "loss": 1.0059, "step": 30160 }, { "epoch": 1.82, "learning_rate": 3.975374586935857e-05, "loss": 0.7816, "step": 30162 }, { "epoch": 1.82, "learning_rate": 3.9752236959244335e-05, "loss": 0.8243, "step": 30164 }, { "epoch": 1.82, "learning_rate": 3.9750728049130115e-05, "loss": 0.9703, "step": 30166 }, { "epoch": 1.82, "learning_rate": 3.9749219139015895e-05, "loss": 0.9335, "step": 30168 }, { "epoch": 1.82, "learning_rate": 3.974771022890167e-05, "loss": 0.7488, "step": 30170 }, { "epoch": 1.82, "learning_rate": 3.974620131878744e-05, "loss": 0.8424, "step": 30172 }, { "epoch": 1.82, "learning_rate": 3.9744692408673215e-05, "loss": 0.7451, "step": 30174 }, { "epoch": 1.82, "learning_rate": 3.9743183498558994e-05, "loss": 0.9736, "step": 30176 }, { "epoch": 1.82, "learning_rate": 3.974167458844477e-05, "loss": 0.6653, "step": 30178 }, { "epoch": 1.82, "learning_rate": 3.974016567833054e-05, "loss": 0.6377, "step": 30180 }, { "epoch": 1.82, "learning_rate": 3.973865676821632e-05, "loss": 0.8643, "step": 30182 }, { "epoch": 1.82, "learning_rate": 3.97371478581021e-05, "loss": 0.7764, "step": 30184 }, { "epoch": 1.82, "learning_rate": 3.973563894798787e-05, "loss": 0.7184, "step": 30186 }, { "epoch": 1.82, "learning_rate": 3.973413003787365e-05, "loss": 0.7571, "step": 30188 }, { "epoch": 1.82, "learning_rate": 3.973262112775942e-05, "loss": 0.5864, "step": 30190 }, { "epoch": 1.82, "learning_rate": 3.973111221764519e-05, "loss": 0.6507, "step": 30192 }, { "epoch": 1.82, "learning_rate": 3.972960330753097e-05, "loss": 0.5928, "step": 30194 }, { "epoch": 1.82, "learning_rate": 3.9728094397416746e-05, "loss": 1.0561, "step": 30196 }, { "epoch": 1.82, "learning_rate": 3.9726585487302526e-05, "loss": 0.8169, "step": 30198 }, { "epoch": 1.82, "learning_rate": 3.97250765771883e-05, "loss": 0.6734, "step": 30200 }, { "epoch": 1.82, "learning_rate": 3.972356766707407e-05, "loss": 0.6539, "step": 30202 }, { "epoch": 1.82, "learning_rate": 3.972205875695985e-05, "loss": 0.7423, "step": 30204 }, { "epoch": 1.82, "learning_rate": 3.9720549846845625e-05, "loss": 0.9874, "step": 30206 }, { "epoch": 1.82, "learning_rate": 3.97190409367314e-05, "loss": 0.6537, "step": 30208 }, { "epoch": 1.82, "learning_rate": 3.971753202661718e-05, "loss": 0.9354, "step": 30210 }, { "epoch": 1.82, "learning_rate": 3.971602311650295e-05, "loss": 0.6822, "step": 30212 }, { "epoch": 1.82, "learning_rate": 3.9714514206388724e-05, "loss": 0.9327, "step": 30214 }, { "epoch": 1.82, "learning_rate": 3.9713005296274504e-05, "loss": 0.741, "step": 30216 }, { "epoch": 1.82, "learning_rate": 3.971149638616028e-05, "loss": 0.6717, "step": 30218 }, { "epoch": 1.82, "learning_rate": 3.970998747604605e-05, "loss": 0.9505, "step": 30220 }, { "epoch": 1.82, "learning_rate": 3.970847856593183e-05, "loss": 0.7876, "step": 30222 }, { "epoch": 1.82, "learning_rate": 3.9706969655817604e-05, "loss": 0.7665, "step": 30224 }, { "epoch": 1.82, "learning_rate": 3.9705460745703384e-05, "loss": 0.9051, "step": 30226 }, { "epoch": 1.82, "learning_rate": 3.970395183558916e-05, "loss": 1.0801, "step": 30228 }, { "epoch": 1.82, "learning_rate": 3.970244292547493e-05, "loss": 0.7341, "step": 30230 }, { "epoch": 1.82, "learning_rate": 3.970093401536071e-05, "loss": 0.96, "step": 30232 }, { "epoch": 1.82, "learning_rate": 3.969942510524648e-05, "loss": 0.9056, "step": 30234 }, { "epoch": 1.82, "learning_rate": 3.9697916195132256e-05, "loss": 0.8755, "step": 30236 }, { "epoch": 1.83, "learning_rate": 3.9696407285018036e-05, "loss": 0.8363, "step": 30238 }, { "epoch": 1.83, "learning_rate": 3.969489837490381e-05, "loss": 1.0962, "step": 30240 }, { "epoch": 1.83, "learning_rate": 3.969338946478958e-05, "loss": 0.6478, "step": 30242 }, { "epoch": 1.83, "learning_rate": 3.969188055467536e-05, "loss": 0.7496, "step": 30244 }, { "epoch": 1.83, "learning_rate": 3.9690371644561135e-05, "loss": 0.9082, "step": 30246 }, { "epoch": 1.83, "learning_rate": 3.9688862734446915e-05, "loss": 0.8868, "step": 30248 }, { "epoch": 1.83, "learning_rate": 3.968735382433268e-05, "loss": 0.6711, "step": 30250 }, { "epoch": 1.83, "learning_rate": 3.968584491421846e-05, "loss": 0.8985, "step": 30252 }, { "epoch": 1.83, "learning_rate": 3.968433600410424e-05, "loss": 0.8734, "step": 30254 }, { "epoch": 1.83, "learning_rate": 3.968282709399001e-05, "loss": 0.6574, "step": 30256 }, { "epoch": 1.83, "learning_rate": 3.968131818387579e-05, "loss": 0.6594, "step": 30258 }, { "epoch": 1.83, "learning_rate": 3.967980927376157e-05, "loss": 0.7733, "step": 30260 }, { "epoch": 1.83, "learning_rate": 3.967830036364734e-05, "loss": 0.8515, "step": 30262 }, { "epoch": 1.83, "learning_rate": 3.9676791453533114e-05, "loss": 1.0072, "step": 30264 }, { "epoch": 1.83, "learning_rate": 3.9675282543418894e-05, "loss": 0.9636, "step": 30266 }, { "epoch": 1.83, "learning_rate": 3.967377363330467e-05, "loss": 0.9727, "step": 30268 }, { "epoch": 1.83, "learning_rate": 3.967226472319044e-05, "loss": 0.9138, "step": 30270 }, { "epoch": 1.83, "learning_rate": 3.967075581307621e-05, "loss": 0.7203, "step": 30272 }, { "epoch": 1.83, "learning_rate": 3.966924690296199e-05, "loss": 0.6503, "step": 30274 }, { "epoch": 1.83, "learning_rate": 3.966773799284777e-05, "loss": 0.7324, "step": 30276 }, { "epoch": 1.83, "learning_rate": 3.966622908273354e-05, "loss": 0.8223, "step": 30278 }, { "epoch": 1.83, "learning_rate": 3.966472017261932e-05, "loss": 0.7323, "step": 30280 }, { "epoch": 1.83, "learning_rate": 3.96632112625051e-05, "loss": 1.0152, "step": 30282 }, { "epoch": 1.83, "learning_rate": 3.966170235239087e-05, "loss": 0.705, "step": 30284 }, { "epoch": 1.83, "learning_rate": 3.9660193442276645e-05, "loss": 0.6926, "step": 30286 }, { "epoch": 1.83, "learning_rate": 3.965868453216242e-05, "loss": 0.8487, "step": 30288 }, { "epoch": 1.83, "learning_rate": 3.96571756220482e-05, "loss": 0.6436, "step": 30290 }, { "epoch": 1.83, "learning_rate": 3.965566671193397e-05, "loss": 0.6886, "step": 30292 }, { "epoch": 1.83, "learning_rate": 3.9654157801819744e-05, "loss": 0.5319, "step": 30294 }, { "epoch": 1.83, "learning_rate": 3.9652648891705524e-05, "loss": 0.9397, "step": 30296 }, { "epoch": 1.83, "learning_rate": 3.9651139981591304e-05, "loss": 0.8675, "step": 30298 }, { "epoch": 1.83, "learning_rate": 3.964963107147707e-05, "loss": 0.8465, "step": 30300 }, { "epoch": 1.83, "learning_rate": 3.964812216136285e-05, "loss": 0.9358, "step": 30302 }, { "epoch": 1.83, "learning_rate": 3.9646613251248624e-05, "loss": 0.8385, "step": 30304 }, { "epoch": 1.83, "learning_rate": 3.96451043411344e-05, "loss": 0.8491, "step": 30306 }, { "epoch": 1.83, "learning_rate": 3.9643595431020177e-05, "loss": 1.0349, "step": 30308 }, { "epoch": 1.83, "learning_rate": 3.964208652090595e-05, "loss": 0.9642, "step": 30310 }, { "epoch": 1.83, "learning_rate": 3.964057761079173e-05, "loss": 0.9345, "step": 30312 }, { "epoch": 1.83, "learning_rate": 3.96390687006775e-05, "loss": 0.8418, "step": 30314 }, { "epoch": 1.83, "learning_rate": 3.9637559790563276e-05, "loss": 0.9425, "step": 30316 }, { "epoch": 1.83, "learning_rate": 3.9636050880449056e-05, "loss": 0.8387, "step": 30318 }, { "epoch": 1.83, "learning_rate": 3.963454197033483e-05, "loss": 0.731, "step": 30320 }, { "epoch": 1.83, "learning_rate": 3.96330330602206e-05, "loss": 0.5838, "step": 30322 }, { "epoch": 1.83, "learning_rate": 3.963152415010638e-05, "loss": 0.8518, "step": 30324 }, { "epoch": 1.83, "learning_rate": 3.9630015239992155e-05, "loss": 0.5674, "step": 30326 }, { "epoch": 1.83, "learning_rate": 3.962850632987793e-05, "loss": 1.2947, "step": 30328 }, { "epoch": 1.83, "learning_rate": 3.962699741976371e-05, "loss": 0.9737, "step": 30330 }, { "epoch": 1.83, "learning_rate": 3.962548850964948e-05, "loss": 0.7078, "step": 30332 }, { "epoch": 1.83, "learning_rate": 3.962397959953526e-05, "loss": 0.8401, "step": 30334 }, { "epoch": 1.83, "learning_rate": 3.9622470689421034e-05, "loss": 1.0584, "step": 30336 }, { "epoch": 1.83, "learning_rate": 3.962096177930681e-05, "loss": 0.8567, "step": 30338 }, { "epoch": 1.83, "learning_rate": 3.961945286919259e-05, "loss": 0.8198, "step": 30340 }, { "epoch": 1.83, "learning_rate": 3.961794395907836e-05, "loss": 0.6531, "step": 30342 }, { "epoch": 1.83, "learning_rate": 3.9616435048964134e-05, "loss": 0.7419, "step": 30344 }, { "epoch": 1.83, "learning_rate": 3.9614926138849913e-05, "loss": 0.6217, "step": 30346 }, { "epoch": 1.83, "learning_rate": 3.9613417228735687e-05, "loss": 0.6759, "step": 30348 }, { "epoch": 1.83, "learning_rate": 3.961190831862146e-05, "loss": 0.7136, "step": 30350 }, { "epoch": 1.83, "learning_rate": 3.961039940850724e-05, "loss": 0.8204, "step": 30352 }, { "epoch": 1.83, "learning_rate": 3.960889049839301e-05, "loss": 0.7855, "step": 30354 }, { "epoch": 1.83, "learning_rate": 3.9607381588278786e-05, "loss": 0.9814, "step": 30356 }, { "epoch": 1.83, "learning_rate": 3.9605872678164566e-05, "loss": 0.8749, "step": 30358 }, { "epoch": 1.83, "learning_rate": 3.960436376805034e-05, "loss": 0.9315, "step": 30360 }, { "epoch": 1.83, "learning_rate": 3.960285485793612e-05, "loss": 0.7748, "step": 30362 }, { "epoch": 1.83, "learning_rate": 3.9601345947821885e-05, "loss": 0.9345, "step": 30364 }, { "epoch": 1.83, "learning_rate": 3.9599837037707665e-05, "loss": 1.0608, "step": 30366 }, { "epoch": 1.83, "learning_rate": 3.9598328127593445e-05, "loss": 0.6567, "step": 30368 }, { "epoch": 1.83, "learning_rate": 3.959681921747921e-05, "loss": 0.7116, "step": 30370 }, { "epoch": 1.83, "learning_rate": 3.959531030736499e-05, "loss": 0.554, "step": 30372 }, { "epoch": 1.83, "learning_rate": 3.959380139725077e-05, "loss": 0.7563, "step": 30374 }, { "epoch": 1.83, "learning_rate": 3.9592292487136544e-05, "loss": 0.5734, "step": 30376 }, { "epoch": 1.83, "learning_rate": 3.959078357702232e-05, "loss": 0.8459, "step": 30378 }, { "epoch": 1.83, "learning_rate": 3.95892746669081e-05, "loss": 0.7637, "step": 30380 }, { "epoch": 1.83, "learning_rate": 3.958776575679387e-05, "loss": 0.9033, "step": 30382 }, { "epoch": 1.83, "learning_rate": 3.958625684667965e-05, "loss": 1.175, "step": 30384 }, { "epoch": 1.83, "learning_rate": 3.9584747936565417e-05, "loss": 0.7908, "step": 30386 }, { "epoch": 1.83, "learning_rate": 3.9583239026451196e-05, "loss": 0.8314, "step": 30388 }, { "epoch": 1.83, "learning_rate": 3.9581730116336976e-05, "loss": 0.5791, "step": 30390 }, { "epoch": 1.83, "learning_rate": 3.958022120622274e-05, "loss": 0.7206, "step": 30392 }, { "epoch": 1.83, "learning_rate": 3.957871229610852e-05, "loss": 0.7198, "step": 30394 }, { "epoch": 1.83, "learning_rate": 3.95772033859943e-05, "loss": 0.6281, "step": 30396 }, { "epoch": 1.83, "learning_rate": 3.9575694475880076e-05, "loss": 0.6598, "step": 30398 }, { "epoch": 1.83, "learning_rate": 3.957418556576585e-05, "loss": 1.0508, "step": 30400 }, { "epoch": 1.83, "learning_rate": 3.957267665565162e-05, "loss": 0.8677, "step": 30402 }, { "epoch": 1.84, "learning_rate": 3.95711677455374e-05, "loss": 0.6953, "step": 30404 }, { "epoch": 1.84, "learning_rate": 3.9569658835423175e-05, "loss": 0.6995, "step": 30406 }, { "epoch": 1.84, "learning_rate": 3.956814992530895e-05, "loss": 0.7186, "step": 30408 }, { "epoch": 1.84, "learning_rate": 3.956664101519473e-05, "loss": 0.7957, "step": 30410 }, { "epoch": 1.84, "learning_rate": 3.956513210508051e-05, "loss": 0.9739, "step": 30412 }, { "epoch": 1.84, "learning_rate": 3.9563623194966274e-05, "loss": 0.7815, "step": 30414 }, { "epoch": 1.84, "learning_rate": 3.9562114284852054e-05, "loss": 0.8236, "step": 30416 }, { "epoch": 1.84, "learning_rate": 3.956060537473783e-05, "loss": 0.783, "step": 30418 }, { "epoch": 1.84, "learning_rate": 3.95590964646236e-05, "loss": 0.554, "step": 30420 }, { "epoch": 1.84, "learning_rate": 3.955758755450938e-05, "loss": 0.8846, "step": 30422 }, { "epoch": 1.84, "learning_rate": 3.9556078644395153e-05, "loss": 0.3885, "step": 30424 }, { "epoch": 1.84, "learning_rate": 3.955456973428093e-05, "loss": 0.7886, "step": 30426 }, { "epoch": 1.84, "learning_rate": 3.9553060824166706e-05, "loss": 0.804, "step": 30428 }, { "epoch": 1.84, "learning_rate": 3.955155191405248e-05, "loss": 0.9333, "step": 30430 }, { "epoch": 1.84, "learning_rate": 3.955004300393826e-05, "loss": 0.6372, "step": 30432 }, { "epoch": 1.84, "learning_rate": 3.954853409382403e-05, "loss": 1.0634, "step": 30434 }, { "epoch": 1.84, "learning_rate": 3.9547025183709806e-05, "loss": 0.6765, "step": 30436 }, { "epoch": 1.84, "learning_rate": 3.9545516273595586e-05, "loss": 0.8598, "step": 30438 }, { "epoch": 1.84, "learning_rate": 3.954400736348136e-05, "loss": 0.897, "step": 30440 }, { "epoch": 1.84, "learning_rate": 3.954249845336713e-05, "loss": 0.7847, "step": 30442 }, { "epoch": 1.84, "learning_rate": 3.954098954325291e-05, "loss": 0.7523, "step": 30444 }, { "epoch": 1.84, "learning_rate": 3.9539480633138685e-05, "loss": 0.8976, "step": 30446 }, { "epoch": 1.84, "learning_rate": 3.9537971723024465e-05, "loss": 0.884, "step": 30448 }, { "epoch": 1.84, "learning_rate": 3.953646281291024e-05, "loss": 0.7421, "step": 30450 }, { "epoch": 1.84, "learning_rate": 3.953495390279601e-05, "loss": 0.7518, "step": 30452 }, { "epoch": 1.84, "learning_rate": 3.953344499268179e-05, "loss": 0.8612, "step": 30454 }, { "epoch": 1.84, "learning_rate": 3.9531936082567564e-05, "loss": 0.8656, "step": 30456 }, { "epoch": 1.84, "learning_rate": 3.953042717245334e-05, "loss": 0.6983, "step": 30458 }, { "epoch": 1.84, "learning_rate": 3.952891826233912e-05, "loss": 1.0113, "step": 30460 }, { "epoch": 1.84, "learning_rate": 3.952740935222489e-05, "loss": 0.9277, "step": 30462 }, { "epoch": 1.84, "learning_rate": 3.952590044211066e-05, "loss": 0.5475, "step": 30464 }, { "epoch": 1.84, "learning_rate": 3.952439153199644e-05, "loss": 0.759, "step": 30466 }, { "epoch": 1.84, "learning_rate": 3.9522882621882216e-05, "loss": 1.105, "step": 30468 }, { "epoch": 1.84, "learning_rate": 3.952137371176799e-05, "loss": 0.6789, "step": 30470 }, { "epoch": 1.84, "learning_rate": 3.951986480165377e-05, "loss": 0.857, "step": 30472 }, { "epoch": 1.84, "learning_rate": 3.951835589153954e-05, "loss": 0.7552, "step": 30474 }, { "epoch": 1.84, "learning_rate": 3.951684698142532e-05, "loss": 0.8404, "step": 30476 }, { "epoch": 1.84, "learning_rate": 3.951533807131109e-05, "loss": 0.9576, "step": 30478 }, { "epoch": 1.84, "learning_rate": 3.951382916119687e-05, "loss": 0.8899, "step": 30480 }, { "epoch": 1.84, "learning_rate": 3.951232025108265e-05, "loss": 0.7794, "step": 30482 }, { "epoch": 1.84, "learning_rate": 3.9510811340968415e-05, "loss": 1.0443, "step": 30484 }, { "epoch": 1.84, "learning_rate": 3.9509302430854195e-05, "loss": 0.7582, "step": 30486 }, { "epoch": 1.84, "learning_rate": 3.9507793520739975e-05, "loss": 0.8973, "step": 30488 }, { "epoch": 1.84, "learning_rate": 3.950628461062575e-05, "loss": 0.8325, "step": 30490 }, { "epoch": 1.84, "learning_rate": 3.950477570051152e-05, "loss": 0.7836, "step": 30492 }, { "epoch": 1.84, "learning_rate": 3.95032667903973e-05, "loss": 0.7243, "step": 30494 }, { "epoch": 1.84, "learning_rate": 3.9501757880283074e-05, "loss": 0.8414, "step": 30496 }, { "epoch": 1.84, "learning_rate": 3.9500248970168854e-05, "loss": 0.6828, "step": 30498 }, { "epoch": 1.84, "learning_rate": 3.949874006005462e-05, "loss": 0.6883, "step": 30500 }, { "epoch": 1.84, "learning_rate": 3.94972311499404e-05, "loss": 1.2062, "step": 30502 }, { "epoch": 1.84, "learning_rate": 3.949572223982618e-05, "loss": 0.8182, "step": 30504 }, { "epoch": 1.84, "learning_rate": 3.9494213329711946e-05, "loss": 0.6228, "step": 30506 }, { "epoch": 1.84, "learning_rate": 3.9492704419597726e-05, "loss": 0.9665, "step": 30508 }, { "epoch": 1.84, "learning_rate": 3.9491195509483506e-05, "loss": 0.5976, "step": 30510 }, { "epoch": 1.84, "learning_rate": 3.948968659936928e-05, "loss": 0.5533, "step": 30512 }, { "epoch": 1.84, "learning_rate": 3.948817768925505e-05, "loss": 0.854, "step": 30514 }, { "epoch": 1.84, "learning_rate": 3.9486668779140826e-05, "loss": 0.9183, "step": 30516 }, { "epoch": 1.84, "learning_rate": 3.9485159869026605e-05, "loss": 1.1295, "step": 30518 }, { "epoch": 1.84, "learning_rate": 3.948365095891238e-05, "loss": 0.7277, "step": 30520 }, { "epoch": 1.84, "learning_rate": 3.948214204879815e-05, "loss": 0.8835, "step": 30522 }, { "epoch": 1.84, "learning_rate": 3.948063313868393e-05, "loss": 0.6303, "step": 30524 }, { "epoch": 1.84, "learning_rate": 3.947912422856971e-05, "loss": 0.8573, "step": 30526 }, { "epoch": 1.84, "learning_rate": 3.947761531845548e-05, "loss": 0.7937, "step": 30528 }, { "epoch": 1.84, "learning_rate": 3.947610640834126e-05, "loss": 0.8375, "step": 30530 }, { "epoch": 1.84, "learning_rate": 3.947459749822703e-05, "loss": 0.7563, "step": 30532 }, { "epoch": 1.84, "learning_rate": 3.9473088588112804e-05, "loss": 0.7851, "step": 30534 }, { "epoch": 1.84, "learning_rate": 3.9471579677998584e-05, "loss": 0.8138, "step": 30536 }, { "epoch": 1.84, "learning_rate": 3.947007076788436e-05, "loss": 0.8648, "step": 30538 }, { "epoch": 1.84, "learning_rate": 3.946856185777014e-05, "loss": 0.7757, "step": 30540 }, { "epoch": 1.84, "learning_rate": 3.946705294765591e-05, "loss": 0.7637, "step": 30542 }, { "epoch": 1.84, "learning_rate": 3.946554403754168e-05, "loss": 0.605, "step": 30544 }, { "epoch": 1.84, "learning_rate": 3.946403512742746e-05, "loss": 0.6975, "step": 30546 }, { "epoch": 1.84, "learning_rate": 3.9462526217313236e-05, "loss": 0.6104, "step": 30548 }, { "epoch": 1.84, "learning_rate": 3.946101730719901e-05, "loss": 0.7059, "step": 30550 }, { "epoch": 1.84, "learning_rate": 3.945950839708479e-05, "loss": 0.8251, "step": 30552 }, { "epoch": 1.84, "learning_rate": 3.945799948697056e-05, "loss": 0.892, "step": 30554 }, { "epoch": 1.84, "learning_rate": 3.9456490576856336e-05, "loss": 1.2617, "step": 30556 }, { "epoch": 1.84, "learning_rate": 3.9454981666742115e-05, "loss": 0.8048, "step": 30558 }, { "epoch": 1.84, "learning_rate": 3.945347275662789e-05, "loss": 0.8496, "step": 30560 }, { "epoch": 1.84, "learning_rate": 3.945196384651367e-05, "loss": 0.9698, "step": 30562 }, { "epoch": 1.84, "learning_rate": 3.945045493639944e-05, "loss": 0.5882, "step": 30564 }, { "epoch": 1.84, "learning_rate": 3.9448946026285215e-05, "loss": 0.6739, "step": 30566 }, { "epoch": 1.85, "learning_rate": 3.9447437116170995e-05, "loss": 0.7123, "step": 30568 }, { "epoch": 1.85, "learning_rate": 3.944592820605677e-05, "loss": 0.9023, "step": 30570 }, { "epoch": 1.85, "learning_rate": 3.944441929594254e-05, "loss": 0.6753, "step": 30572 }, { "epoch": 1.85, "learning_rate": 3.944291038582832e-05, "loss": 0.6267, "step": 30574 }, { "epoch": 1.85, "learning_rate": 3.9441401475714094e-05, "loss": 0.8844, "step": 30576 }, { "epoch": 1.85, "learning_rate": 3.943989256559987e-05, "loss": 0.8016, "step": 30578 }, { "epoch": 1.85, "learning_rate": 3.943838365548565e-05, "loss": 0.9347, "step": 30580 }, { "epoch": 1.85, "learning_rate": 3.943687474537142e-05, "loss": 0.5804, "step": 30582 }, { "epoch": 1.85, "learning_rate": 3.943536583525719e-05, "loss": 0.7553, "step": 30584 }, { "epoch": 1.85, "learning_rate": 3.943385692514297e-05, "loss": 1.0231, "step": 30586 }, { "epoch": 1.85, "learning_rate": 3.9432348015028746e-05, "loss": 0.7577, "step": 30588 }, { "epoch": 1.85, "learning_rate": 3.9430839104914526e-05, "loss": 0.7952, "step": 30590 }, { "epoch": 1.85, "learning_rate": 3.942933019480029e-05, "loss": 0.6678, "step": 30592 }, { "epoch": 1.85, "learning_rate": 3.942782128468607e-05, "loss": 0.5212, "step": 30594 }, { "epoch": 1.85, "learning_rate": 3.942631237457185e-05, "loss": 0.5793, "step": 30596 }, { "epoch": 1.85, "learning_rate": 3.942480346445762e-05, "loss": 1.1309, "step": 30598 }, { "epoch": 1.85, "learning_rate": 3.94232945543434e-05, "loss": 0.9734, "step": 30600 }, { "epoch": 1.85, "learning_rate": 3.942178564422918e-05, "loss": 0.5858, "step": 30602 }, { "epoch": 1.85, "learning_rate": 3.942027673411495e-05, "loss": 0.7033, "step": 30604 }, { "epoch": 1.85, "learning_rate": 3.9418767824000725e-05, "loss": 0.6078, "step": 30606 }, { "epoch": 1.85, "learning_rate": 3.94172589138865e-05, "loss": 0.6449, "step": 30608 }, { "epoch": 1.85, "learning_rate": 3.941575000377228e-05, "loss": 0.7278, "step": 30610 }, { "epoch": 1.85, "learning_rate": 3.941424109365806e-05, "loss": 0.5445, "step": 30612 }, { "epoch": 1.85, "learning_rate": 3.9412732183543824e-05, "loss": 0.6902, "step": 30614 }, { "epoch": 1.85, "learning_rate": 3.9411223273429604e-05, "loss": 0.6734, "step": 30616 }, { "epoch": 1.85, "learning_rate": 3.9409714363315384e-05, "loss": 0.6312, "step": 30618 }, { "epoch": 1.85, "learning_rate": 3.940820545320115e-05, "loss": 0.9782, "step": 30620 }, { "epoch": 1.85, "learning_rate": 3.940669654308693e-05, "loss": 0.5067, "step": 30622 }, { "epoch": 1.85, "learning_rate": 3.940518763297271e-05, "loss": 0.7194, "step": 30624 }, { "epoch": 1.85, "learning_rate": 3.940367872285848e-05, "loss": 0.8946, "step": 30626 }, { "epoch": 1.85, "learning_rate": 3.9402169812744256e-05, "loss": 0.674, "step": 30628 }, { "epoch": 1.85, "learning_rate": 3.940066090263003e-05, "loss": 0.6296, "step": 30630 }, { "epoch": 1.85, "learning_rate": 3.939915199251581e-05, "loss": 0.6606, "step": 30632 }, { "epoch": 1.85, "learning_rate": 3.939764308240158e-05, "loss": 0.8897, "step": 30634 }, { "epoch": 1.85, "learning_rate": 3.9396134172287355e-05, "loss": 0.6793, "step": 30636 }, { "epoch": 1.85, "learning_rate": 3.9394625262173135e-05, "loss": 0.8657, "step": 30638 }, { "epoch": 1.85, "learning_rate": 3.9393116352058915e-05, "loss": 0.6796, "step": 30640 }, { "epoch": 1.85, "learning_rate": 3.939160744194468e-05, "loss": 0.9557, "step": 30642 }, { "epoch": 1.85, "learning_rate": 3.939009853183046e-05, "loss": 1.1426, "step": 30644 }, { "epoch": 1.85, "learning_rate": 3.9388589621716235e-05, "loss": 0.7112, "step": 30646 }, { "epoch": 1.85, "learning_rate": 3.938708071160201e-05, "loss": 0.7654, "step": 30648 }, { "epoch": 1.85, "learning_rate": 3.938557180148779e-05, "loss": 0.637, "step": 30650 }, { "epoch": 1.85, "learning_rate": 3.938406289137356e-05, "loss": 0.7764, "step": 30652 }, { "epoch": 1.85, "learning_rate": 3.938255398125934e-05, "loss": 0.7603, "step": 30654 }, { "epoch": 1.85, "learning_rate": 3.9381045071145114e-05, "loss": 0.9655, "step": 30656 }, { "epoch": 1.85, "learning_rate": 3.937953616103089e-05, "loss": 0.7848, "step": 30658 }, { "epoch": 1.85, "learning_rate": 3.937802725091667e-05, "loss": 0.6262, "step": 30660 }, { "epoch": 1.85, "learning_rate": 3.937651834080244e-05, "loss": 0.5568, "step": 30662 }, { "epoch": 1.85, "learning_rate": 3.937500943068821e-05, "loss": 0.8605, "step": 30664 }, { "epoch": 1.85, "learning_rate": 3.937350052057399e-05, "loss": 0.6063, "step": 30666 }, { "epoch": 1.85, "learning_rate": 3.9371991610459766e-05, "loss": 0.7235, "step": 30668 }, { "epoch": 1.85, "learning_rate": 3.937048270034554e-05, "loss": 0.9654, "step": 30670 }, { "epoch": 1.85, "learning_rate": 3.936897379023132e-05, "loss": 0.7039, "step": 30672 }, { "epoch": 1.85, "learning_rate": 3.936746488011709e-05, "loss": 0.7023, "step": 30674 }, { "epoch": 1.85, "learning_rate": 3.936595597000287e-05, "loss": 0.85, "step": 30676 }, { "epoch": 1.85, "learning_rate": 3.9364447059888645e-05, "loss": 0.7112, "step": 30678 }, { "epoch": 1.85, "learning_rate": 3.936293814977442e-05, "loss": 0.8756, "step": 30680 }, { "epoch": 1.85, "learning_rate": 3.93614292396602e-05, "loss": 0.9402, "step": 30682 }, { "epoch": 1.85, "learning_rate": 3.935992032954597e-05, "loss": 0.6621, "step": 30684 }, { "epoch": 1.85, "learning_rate": 3.9358411419431745e-05, "loss": 0.7327, "step": 30686 }, { "epoch": 1.85, "learning_rate": 3.9356902509317524e-05, "loss": 1.1423, "step": 30688 }, { "epoch": 1.85, "learning_rate": 3.93553935992033e-05, "loss": 0.8528, "step": 30690 }, { "epoch": 1.85, "learning_rate": 3.935388468908907e-05, "loss": 0.7936, "step": 30692 }, { "epoch": 1.85, "learning_rate": 3.935237577897485e-05, "loss": 0.5657, "step": 30694 }, { "epoch": 1.85, "learning_rate": 3.9350866868860624e-05, "loss": 0.4992, "step": 30696 }, { "epoch": 1.85, "learning_rate": 3.93493579587464e-05, "loss": 0.6543, "step": 30698 }, { "epoch": 1.85, "learning_rate": 3.934784904863218e-05, "loss": 0.6242, "step": 30700 }, { "epoch": 1.85, "learning_rate": 3.934634013851795e-05, "loss": 0.589, "step": 30702 }, { "epoch": 1.85, "learning_rate": 3.934483122840373e-05, "loss": 1.0085, "step": 30704 }, { "epoch": 1.85, "learning_rate": 3.9343322318289496e-05, "loss": 0.8042, "step": 30706 }, { "epoch": 1.85, "learning_rate": 3.9341813408175276e-05, "loss": 0.9912, "step": 30708 }, { "epoch": 1.85, "learning_rate": 3.9340304498061056e-05, "loss": 0.9095, "step": 30710 }, { "epoch": 1.85, "learning_rate": 3.933879558794683e-05, "loss": 1.0138, "step": 30712 }, { "epoch": 1.85, "learning_rate": 3.93372866778326e-05, "loss": 0.8139, "step": 30714 }, { "epoch": 1.85, "learning_rate": 3.933577776771838e-05, "loss": 0.8143, "step": 30716 }, { "epoch": 1.85, "learning_rate": 3.9334268857604155e-05, "loss": 0.7375, "step": 30718 }, { "epoch": 1.85, "learning_rate": 3.933275994748993e-05, "loss": 0.9982, "step": 30720 }, { "epoch": 1.85, "learning_rate": 3.93312510373757e-05, "loss": 0.6434, "step": 30722 }, { "epoch": 1.85, "learning_rate": 3.932974212726148e-05, "loss": 0.8121, "step": 30724 }, { "epoch": 1.85, "learning_rate": 3.932823321714726e-05, "loss": 0.5205, "step": 30726 }, { "epoch": 1.85, "learning_rate": 3.932672430703303e-05, "loss": 0.6836, "step": 30728 }, { "epoch": 1.85, "learning_rate": 3.932521539691881e-05, "loss": 0.7863, "step": 30730 }, { "epoch": 1.85, "learning_rate": 3.932370648680459e-05, "loss": 0.7348, "step": 30732 }, { "epoch": 1.86, "learning_rate": 3.9322197576690354e-05, "loss": 1.0271, "step": 30734 }, { "epoch": 1.86, "learning_rate": 3.9320688666576134e-05, "loss": 0.8863, "step": 30736 }, { "epoch": 1.86, "learning_rate": 3.9319179756461914e-05, "loss": 0.7635, "step": 30738 }, { "epoch": 1.86, "learning_rate": 3.931767084634769e-05, "loss": 0.901, "step": 30740 }, { "epoch": 1.86, "learning_rate": 3.931616193623346e-05, "loss": 0.6858, "step": 30742 }, { "epoch": 1.86, "learning_rate": 3.931465302611923e-05, "loss": 0.7619, "step": 30744 }, { "epoch": 1.86, "learning_rate": 3.931314411600501e-05, "loss": 0.6027, "step": 30746 }, { "epoch": 1.86, "learning_rate": 3.9311635205890786e-05, "loss": 0.9648, "step": 30748 }, { "epoch": 1.86, "learning_rate": 3.931012629577656e-05, "loss": 0.7717, "step": 30750 }, { "epoch": 1.86, "learning_rate": 3.930861738566234e-05, "loss": 0.8574, "step": 30752 }, { "epoch": 1.86, "learning_rate": 3.930710847554812e-05, "loss": 0.7021, "step": 30754 }, { "epoch": 1.86, "learning_rate": 3.9305599565433885e-05, "loss": 0.9928, "step": 30756 }, { "epoch": 1.86, "learning_rate": 3.9304090655319665e-05, "loss": 0.8123, "step": 30758 }, { "epoch": 1.86, "learning_rate": 3.930258174520544e-05, "loss": 0.6353, "step": 30760 }, { "epoch": 1.86, "learning_rate": 3.930107283509121e-05, "loss": 0.7524, "step": 30762 }, { "epoch": 1.86, "learning_rate": 3.929956392497699e-05, "loss": 0.5974, "step": 30764 }, { "epoch": 1.86, "learning_rate": 3.9298055014862764e-05, "loss": 0.6858, "step": 30766 }, { "epoch": 1.86, "learning_rate": 3.9296546104748544e-05, "loss": 0.8034, "step": 30768 }, { "epoch": 1.86, "learning_rate": 3.929503719463432e-05, "loss": 0.4757, "step": 30770 }, { "epoch": 1.86, "learning_rate": 3.929352828452009e-05, "loss": 0.7651, "step": 30772 }, { "epoch": 1.86, "learning_rate": 3.929201937440587e-05, "loss": 0.6392, "step": 30774 }, { "epoch": 1.86, "learning_rate": 3.9290510464291644e-05, "loss": 0.6556, "step": 30776 }, { "epoch": 1.86, "learning_rate": 3.928900155417742e-05, "loss": 1.0567, "step": 30778 }, { "epoch": 1.86, "learning_rate": 3.92874926440632e-05, "loss": 0.8929, "step": 30780 }, { "epoch": 1.86, "learning_rate": 3.928598373394897e-05, "loss": 0.7435, "step": 30782 }, { "epoch": 1.86, "learning_rate": 3.928447482383474e-05, "loss": 0.6975, "step": 30784 }, { "epoch": 1.86, "learning_rate": 3.928296591372052e-05, "loss": 0.6074, "step": 30786 }, { "epoch": 1.86, "learning_rate": 3.9281457003606296e-05, "loss": 0.5689, "step": 30788 }, { "epoch": 1.86, "learning_rate": 3.9279948093492076e-05, "loss": 0.6822, "step": 30790 }, { "epoch": 1.86, "learning_rate": 3.927843918337785e-05, "loss": 0.6548, "step": 30792 }, { "epoch": 1.86, "learning_rate": 3.927693027326362e-05, "loss": 0.8223, "step": 30794 }, { "epoch": 1.86, "learning_rate": 3.92754213631494e-05, "loss": 0.5378, "step": 30796 }, { "epoch": 1.86, "learning_rate": 3.9273912453035175e-05, "loss": 0.8228, "step": 30798 }, { "epoch": 1.86, "learning_rate": 3.927240354292095e-05, "loss": 0.7331, "step": 30800 }, { "epoch": 1.86, "learning_rate": 3.927089463280673e-05, "loss": 0.8327, "step": 30802 }, { "epoch": 1.86, "learning_rate": 3.92693857226925e-05, "loss": 0.7393, "step": 30804 }, { "epoch": 1.86, "learning_rate": 3.9267876812578274e-05, "loss": 0.5973, "step": 30806 }, { "epoch": 1.86, "learning_rate": 3.9266367902464054e-05, "loss": 0.4568, "step": 30808 }, { "epoch": 1.86, "learning_rate": 3.926485899234983e-05, "loss": 0.7714, "step": 30810 }, { "epoch": 1.86, "learning_rate": 3.92633500822356e-05, "loss": 0.9167, "step": 30812 }, { "epoch": 1.86, "learning_rate": 3.926184117212138e-05, "loss": 0.5079, "step": 30814 }, { "epoch": 1.86, "learning_rate": 3.9260332262007154e-05, "loss": 0.5902, "step": 30816 }, { "epoch": 1.86, "learning_rate": 3.9258823351892933e-05, "loss": 0.6227, "step": 30818 }, { "epoch": 1.86, "learning_rate": 3.92573144417787e-05, "loss": 0.5703, "step": 30820 }, { "epoch": 1.86, "learning_rate": 3.925580553166448e-05, "loss": 0.7312, "step": 30822 }, { "epoch": 1.86, "learning_rate": 3.925429662155026e-05, "loss": 0.7048, "step": 30824 }, { "epoch": 1.86, "learning_rate": 3.925278771143603e-05, "loss": 0.7427, "step": 30826 }, { "epoch": 1.86, "learning_rate": 3.9251278801321806e-05, "loss": 0.6108, "step": 30828 }, { "epoch": 1.86, "learning_rate": 3.9249769891207586e-05, "loss": 0.8176, "step": 30830 }, { "epoch": 1.86, "learning_rate": 3.924826098109336e-05, "loss": 0.8553, "step": 30832 }, { "epoch": 1.86, "learning_rate": 3.924675207097913e-05, "loss": 0.6192, "step": 30834 }, { "epoch": 1.86, "learning_rate": 3.9245243160864905e-05, "loss": 0.6461, "step": 30836 }, { "epoch": 1.86, "learning_rate": 3.9243734250750685e-05, "loss": 0.7618, "step": 30838 }, { "epoch": 1.86, "learning_rate": 3.9242225340636465e-05, "loss": 0.7707, "step": 30840 }, { "epoch": 1.86, "learning_rate": 3.924071643052223e-05, "loss": 0.8542, "step": 30842 }, { "epoch": 1.86, "learning_rate": 3.923920752040801e-05, "loss": 0.4946, "step": 30844 }, { "epoch": 1.86, "learning_rate": 3.923769861029379e-05, "loss": 0.7346, "step": 30846 }, { "epoch": 1.86, "learning_rate": 3.923618970017956e-05, "loss": 0.8033, "step": 30848 }, { "epoch": 1.86, "learning_rate": 3.923468079006534e-05, "loss": 0.7551, "step": 30850 }, { "epoch": 1.86, "learning_rate": 3.923317187995112e-05, "loss": 0.7862, "step": 30852 }, { "epoch": 1.86, "learning_rate": 3.923166296983689e-05, "loss": 0.6913, "step": 30854 }, { "epoch": 1.86, "learning_rate": 3.9230154059722663e-05, "loss": 0.5343, "step": 30856 }, { "epoch": 1.86, "learning_rate": 3.922864514960844e-05, "loss": 0.5992, "step": 30858 }, { "epoch": 1.86, "learning_rate": 3.9227136239494217e-05, "loss": 0.9549, "step": 30860 }, { "epoch": 1.86, "learning_rate": 3.922562732937999e-05, "loss": 0.9894, "step": 30862 }, { "epoch": 1.86, "learning_rate": 3.922411841926576e-05, "loss": 0.7312, "step": 30864 }, { "epoch": 1.86, "learning_rate": 3.922260950915154e-05, "loss": 0.6226, "step": 30866 }, { "epoch": 1.86, "learning_rate": 3.922110059903732e-05, "loss": 0.9553, "step": 30868 }, { "epoch": 1.86, "learning_rate": 3.921959168892309e-05, "loss": 0.5885, "step": 30870 }, { "epoch": 1.86, "learning_rate": 3.921808277880887e-05, "loss": 0.6336, "step": 30872 }, { "epoch": 1.86, "learning_rate": 3.921657386869464e-05, "loss": 0.8872, "step": 30874 }, { "epoch": 1.86, "learning_rate": 3.921506495858042e-05, "loss": 0.6619, "step": 30876 }, { "epoch": 1.86, "learning_rate": 3.9213556048466195e-05, "loss": 0.7768, "step": 30878 }, { "epoch": 1.86, "learning_rate": 3.921204713835197e-05, "loss": 0.6669, "step": 30880 }, { "epoch": 1.86, "learning_rate": 3.921053822823775e-05, "loss": 1.0477, "step": 30882 }, { "epoch": 1.86, "learning_rate": 3.920902931812352e-05, "loss": 0.6006, "step": 30884 }, { "epoch": 1.86, "learning_rate": 3.9207520408009294e-05, "loss": 0.6399, "step": 30886 }, { "epoch": 1.86, "learning_rate": 3.9206011497895074e-05, "loss": 0.7214, "step": 30888 }, { "epoch": 1.86, "learning_rate": 3.920450258778085e-05, "loss": 0.8414, "step": 30890 }, { "epoch": 1.86, "learning_rate": 3.920299367766662e-05, "loss": 0.7153, "step": 30892 }, { "epoch": 1.86, "learning_rate": 3.92014847675524e-05, "loss": 0.7777, "step": 30894 }, { "epoch": 1.86, "learning_rate": 3.9199975857438173e-05, "loss": 0.6158, "step": 30896 }, { "epoch": 1.86, "learning_rate": 3.9198466947323947e-05, "loss": 1.1198, "step": 30898 }, { "epoch": 1.87, "learning_rate": 3.9196958037209726e-05, "loss": 1.0158, "step": 30900 }, { "epoch": 1.87, "learning_rate": 3.91954491270955e-05, "loss": 0.7915, "step": 30902 }, { "epoch": 1.87, "learning_rate": 3.919394021698128e-05, "loss": 0.6723, "step": 30904 }, { "epoch": 1.87, "learning_rate": 3.919243130686705e-05, "loss": 0.9165, "step": 30906 }, { "epoch": 1.87, "learning_rate": 3.9190922396752826e-05, "loss": 0.6004, "step": 30908 }, { "epoch": 1.87, "learning_rate": 3.9189413486638606e-05, "loss": 0.7735, "step": 30910 }, { "epoch": 1.87, "learning_rate": 3.918790457652438e-05, "loss": 0.8759, "step": 30912 }, { "epoch": 1.87, "learning_rate": 3.918639566641015e-05, "loss": 0.7085, "step": 30914 }, { "epoch": 1.87, "learning_rate": 3.918488675629593e-05, "loss": 0.4974, "step": 30916 }, { "epoch": 1.87, "learning_rate": 3.9183377846181705e-05, "loss": 0.6128, "step": 30918 }, { "epoch": 1.87, "learning_rate": 3.918186893606748e-05, "loss": 1.0752, "step": 30920 }, { "epoch": 1.87, "learning_rate": 3.918036002595326e-05, "loss": 1.1337, "step": 30922 }, { "epoch": 1.87, "learning_rate": 3.917885111583903e-05, "loss": 0.8825, "step": 30924 }, { "epoch": 1.87, "learning_rate": 3.917734220572481e-05, "loss": 0.9036, "step": 30926 }, { "epoch": 1.87, "learning_rate": 3.9175833295610584e-05, "loss": 0.5608, "step": 30928 }, { "epoch": 1.87, "learning_rate": 3.917432438549636e-05, "loss": 0.7225, "step": 30930 }, { "epoch": 1.87, "learning_rate": 3.917281547538214e-05, "loss": 0.9486, "step": 30932 }, { "epoch": 1.87, "learning_rate": 3.9171306565267903e-05, "loss": 0.6486, "step": 30934 }, { "epoch": 1.87, "learning_rate": 3.916979765515368e-05, "loss": 0.9277, "step": 30936 }, { "epoch": 1.87, "learning_rate": 3.916828874503946e-05, "loss": 0.8193, "step": 30938 }, { "epoch": 1.87, "learning_rate": 3.9166779834925236e-05, "loss": 0.7196, "step": 30940 }, { "epoch": 1.87, "learning_rate": 3.916527092481101e-05, "loss": 0.9729, "step": 30942 }, { "epoch": 1.87, "learning_rate": 3.916376201469679e-05, "loss": 0.5943, "step": 30944 }, { "epoch": 1.87, "learning_rate": 3.916225310458256e-05, "loss": 0.779, "step": 30946 }, { "epoch": 1.87, "learning_rate": 3.9160744194468336e-05, "loss": 0.7409, "step": 30948 }, { "epoch": 1.87, "learning_rate": 3.915923528435411e-05, "loss": 0.6455, "step": 30950 }, { "epoch": 1.87, "learning_rate": 3.915772637423989e-05, "loss": 0.7693, "step": 30952 }, { "epoch": 1.87, "learning_rate": 3.915621746412567e-05, "loss": 1.0236, "step": 30954 }, { "epoch": 1.87, "learning_rate": 3.9154708554011435e-05, "loss": 0.8769, "step": 30956 }, { "epoch": 1.87, "learning_rate": 3.9153199643897215e-05, "loss": 0.5446, "step": 30958 }, { "epoch": 1.87, "learning_rate": 3.9151690733782995e-05, "loss": 0.5258, "step": 30960 }, { "epoch": 1.87, "learning_rate": 3.915018182366876e-05, "loss": 0.9402, "step": 30962 }, { "epoch": 1.87, "learning_rate": 3.914867291355454e-05, "loss": 0.7634, "step": 30964 }, { "epoch": 1.87, "learning_rate": 3.914716400344032e-05, "loss": 0.6866, "step": 30966 }, { "epoch": 1.87, "learning_rate": 3.9145655093326094e-05, "loss": 0.677, "step": 30968 }, { "epoch": 1.87, "learning_rate": 3.914414618321187e-05, "loss": 0.735, "step": 30970 }, { "epoch": 1.87, "learning_rate": 3.914263727309764e-05, "loss": 0.9443, "step": 30972 }, { "epoch": 1.87, "learning_rate": 3.914112836298342e-05, "loss": 1.0881, "step": 30974 }, { "epoch": 1.87, "learning_rate": 3.913961945286919e-05, "loss": 0.5201, "step": 30976 }, { "epoch": 1.87, "learning_rate": 3.9138110542754966e-05, "loss": 0.7224, "step": 30978 }, { "epoch": 1.87, "learning_rate": 3.9136601632640746e-05, "loss": 0.6722, "step": 30980 }, { "epoch": 1.87, "learning_rate": 3.9135092722526526e-05, "loss": 0.785, "step": 30982 }, { "epoch": 1.87, "learning_rate": 3.913358381241229e-05, "loss": 0.6865, "step": 30984 }, { "epoch": 1.87, "learning_rate": 3.913207490229807e-05, "loss": 0.8224, "step": 30986 }, { "epoch": 1.87, "learning_rate": 3.9130565992183846e-05, "loss": 0.9409, "step": 30988 }, { "epoch": 1.87, "learning_rate": 3.9129057082069626e-05, "loss": 0.7979, "step": 30990 }, { "epoch": 1.87, "learning_rate": 3.91275481719554e-05, "loss": 0.7758, "step": 30992 }, { "epoch": 1.87, "learning_rate": 3.912603926184117e-05, "loss": 1.1056, "step": 30994 }, { "epoch": 1.87, "learning_rate": 3.912453035172695e-05, "loss": 0.6742, "step": 30996 }, { "epoch": 1.87, "learning_rate": 3.9123021441612725e-05, "loss": 0.8387, "step": 30998 }, { "epoch": 1.87, "learning_rate": 3.91215125314985e-05, "loss": 0.6161, "step": 31000 }, { "epoch": 1.87, "learning_rate": 3.912000362138428e-05, "loss": 0.9744, "step": 31002 }, { "epoch": 1.87, "learning_rate": 3.911849471127005e-05, "loss": 0.5776, "step": 31004 }, { "epoch": 1.87, "learning_rate": 3.9116985801155824e-05, "loss": 0.9733, "step": 31006 }, { "epoch": 1.87, "learning_rate": 3.9115476891041604e-05, "loss": 0.8344, "step": 31008 }, { "epoch": 1.87, "learning_rate": 3.911396798092738e-05, "loss": 0.6048, "step": 31010 }, { "epoch": 1.87, "learning_rate": 3.911245907081315e-05, "loss": 0.9827, "step": 31012 }, { "epoch": 1.87, "learning_rate": 3.911095016069893e-05, "loss": 0.8145, "step": 31014 }, { "epoch": 1.87, "learning_rate": 3.91094412505847e-05, "loss": 0.7538, "step": 31016 }, { "epoch": 1.87, "learning_rate": 3.910793234047048e-05, "loss": 0.8475, "step": 31018 }, { "epoch": 1.87, "learning_rate": 3.9106423430356256e-05, "loss": 0.7883, "step": 31020 }, { "epoch": 1.87, "learning_rate": 3.910491452024203e-05, "loss": 0.7495, "step": 31022 }, { "epoch": 1.87, "learning_rate": 3.910340561012781e-05, "loss": 0.8468, "step": 31024 }, { "epoch": 1.87, "learning_rate": 3.910189670001358e-05, "loss": 0.6217, "step": 31026 }, { "epoch": 1.87, "learning_rate": 3.9100387789899356e-05, "loss": 0.7224, "step": 31028 }, { "epoch": 1.87, "learning_rate": 3.9098878879785135e-05, "loss": 0.764, "step": 31030 }, { "epoch": 1.87, "learning_rate": 3.909736996967091e-05, "loss": 0.6883, "step": 31032 }, { "epoch": 1.87, "learning_rate": 3.909586105955668e-05, "loss": 1.0216, "step": 31034 }, { "epoch": 1.87, "learning_rate": 3.909435214944246e-05, "loss": 0.9086, "step": 31036 }, { "epoch": 1.87, "learning_rate": 3.9092843239328235e-05, "loss": 0.979, "step": 31038 }, { "epoch": 1.87, "learning_rate": 3.9091334329214015e-05, "loss": 0.6671, "step": 31040 }, { "epoch": 1.87, "learning_rate": 3.908982541909979e-05, "loss": 0.9551, "step": 31042 }, { "epoch": 1.87, "learning_rate": 3.908831650898556e-05, "loss": 0.7784, "step": 31044 }, { "epoch": 1.87, "learning_rate": 3.908680759887134e-05, "loss": 0.7283, "step": 31046 }, { "epoch": 1.87, "learning_rate": 3.908529868875711e-05, "loss": 0.7612, "step": 31048 }, { "epoch": 1.87, "learning_rate": 3.908378977864289e-05, "loss": 0.6437, "step": 31050 }, { "epoch": 1.87, "learning_rate": 3.908228086852867e-05, "loss": 0.8907, "step": 31052 }, { "epoch": 1.87, "learning_rate": 3.908077195841444e-05, "loss": 0.7212, "step": 31054 }, { "epoch": 1.87, "learning_rate": 3.907926304830021e-05, "loss": 0.7173, "step": 31056 }, { "epoch": 1.87, "learning_rate": 3.907775413818599e-05, "loss": 0.8129, "step": 31058 }, { "epoch": 1.87, "learning_rate": 3.9076245228071766e-05, "loss": 0.7137, "step": 31060 }, { "epoch": 1.87, "learning_rate": 3.907473631795754e-05, "loss": 0.61, "step": 31062 }, { "epoch": 1.87, "learning_rate": 3.907322740784331e-05, "loss": 0.7859, "step": 31064 }, { "epoch": 1.88, "learning_rate": 3.907171849772909e-05, "loss": 0.7507, "step": 31066 }, { "epoch": 1.88, "learning_rate": 3.907020958761487e-05, "loss": 0.6794, "step": 31068 }, { "epoch": 1.88, "learning_rate": 3.906870067750064e-05, "loss": 0.6995, "step": 31070 }, { "epoch": 1.88, "learning_rate": 3.906719176738642e-05, "loss": 0.7296, "step": 31072 }, { "epoch": 1.88, "learning_rate": 3.90656828572722e-05, "loss": 0.6872, "step": 31074 }, { "epoch": 1.88, "learning_rate": 3.9064173947157965e-05, "loss": 0.8539, "step": 31076 }, { "epoch": 1.88, "learning_rate": 3.9062665037043745e-05, "loss": 0.6805, "step": 31078 }, { "epoch": 1.88, "learning_rate": 3.906115612692952e-05, "loss": 0.7561, "step": 31080 }, { "epoch": 1.88, "learning_rate": 3.90596472168153e-05, "loss": 0.567, "step": 31082 }, { "epoch": 1.88, "learning_rate": 3.905813830670107e-05, "loss": 0.867, "step": 31084 }, { "epoch": 1.88, "learning_rate": 3.9056629396586844e-05, "loss": 0.5458, "step": 31086 }, { "epoch": 1.88, "learning_rate": 3.9055120486472624e-05, "loss": 0.7186, "step": 31088 }, { "epoch": 1.88, "learning_rate": 3.9053611576358404e-05, "loss": 0.8341, "step": 31090 }, { "epoch": 1.88, "learning_rate": 3.905210266624417e-05, "loss": 0.7996, "step": 31092 }, { "epoch": 1.88, "learning_rate": 3.905059375612995e-05, "loss": 0.7688, "step": 31094 }, { "epoch": 1.88, "learning_rate": 3.904908484601573e-05, "loss": 0.9101, "step": 31096 }, { "epoch": 1.88, "learning_rate": 3.9047575935901496e-05, "loss": 0.7177, "step": 31098 }, { "epoch": 1.88, "learning_rate": 3.9046067025787276e-05, "loss": 1.0124, "step": 31100 }, { "epoch": 1.88, "learning_rate": 3.904455811567305e-05, "loss": 0.9326, "step": 31102 }, { "epoch": 1.88, "learning_rate": 3.904304920555883e-05, "loss": 0.7956, "step": 31104 }, { "epoch": 1.88, "learning_rate": 3.90415402954446e-05, "loss": 1.0552, "step": 31106 }, { "epoch": 1.88, "learning_rate": 3.9040031385330375e-05, "loss": 1.0657, "step": 31108 }, { "epoch": 1.88, "learning_rate": 3.9038522475216155e-05, "loss": 0.9031, "step": 31110 }, { "epoch": 1.88, "learning_rate": 3.903701356510193e-05, "loss": 0.7276, "step": 31112 }, { "epoch": 1.88, "learning_rate": 3.90355046549877e-05, "loss": 0.7106, "step": 31114 }, { "epoch": 1.88, "learning_rate": 3.903399574487348e-05, "loss": 0.6518, "step": 31116 }, { "epoch": 1.88, "learning_rate": 3.9032486834759255e-05, "loss": 0.7998, "step": 31118 }, { "epoch": 1.88, "learning_rate": 3.903097792464503e-05, "loss": 0.8179, "step": 31120 }, { "epoch": 1.88, "learning_rate": 3.902946901453081e-05, "loss": 0.6011, "step": 31122 }, { "epoch": 1.88, "learning_rate": 3.902796010441658e-05, "loss": 1.0088, "step": 31124 }, { "epoch": 1.88, "learning_rate": 3.9026451194302354e-05, "loss": 0.9128, "step": 31126 }, { "epoch": 1.88, "learning_rate": 3.9024942284188134e-05, "loss": 1.1266, "step": 31128 }, { "epoch": 1.88, "learning_rate": 3.902343337407391e-05, "loss": 0.6456, "step": 31130 }, { "epoch": 1.88, "learning_rate": 3.902192446395969e-05, "loss": 0.7206, "step": 31132 }, { "epoch": 1.88, "learning_rate": 3.902041555384546e-05, "loss": 1.1185, "step": 31134 }, { "epoch": 1.88, "learning_rate": 3.901890664373123e-05, "loss": 0.6335, "step": 31136 }, { "epoch": 1.88, "learning_rate": 3.901739773361701e-05, "loss": 0.9152, "step": 31138 }, { "epoch": 1.88, "learning_rate": 3.9015888823502786e-05, "loss": 0.8265, "step": 31140 }, { "epoch": 1.88, "learning_rate": 3.901437991338856e-05, "loss": 0.5367, "step": 31142 }, { "epoch": 1.88, "learning_rate": 3.901287100327434e-05, "loss": 0.6272, "step": 31144 }, { "epoch": 1.88, "learning_rate": 3.901136209316011e-05, "loss": 0.6676, "step": 31146 }, { "epoch": 1.88, "learning_rate": 3.9009853183045885e-05, "loss": 0.8926, "step": 31148 }, { "epoch": 1.88, "learning_rate": 3.9008344272931665e-05, "loss": 0.8841, "step": 31150 }, { "epoch": 1.88, "learning_rate": 3.900683536281744e-05, "loss": 0.7436, "step": 31152 }, { "epoch": 1.88, "learning_rate": 3.900532645270322e-05, "loss": 0.8818, "step": 31154 }, { "epoch": 1.88, "learning_rate": 3.900381754258899e-05, "loss": 0.5693, "step": 31156 }, { "epoch": 1.88, "learning_rate": 3.9002308632474765e-05, "loss": 0.9016, "step": 31158 }, { "epoch": 1.88, "learning_rate": 3.9000799722360544e-05, "loss": 0.5686, "step": 31160 }, { "epoch": 1.88, "learning_rate": 3.899929081224631e-05, "loss": 1.0639, "step": 31162 }, { "epoch": 1.88, "learning_rate": 3.899778190213209e-05, "loss": 0.7134, "step": 31164 }, { "epoch": 1.88, "learning_rate": 3.899627299201787e-05, "loss": 0.974, "step": 31166 }, { "epoch": 1.88, "learning_rate": 3.8994764081903644e-05, "loss": 0.7555, "step": 31168 }, { "epoch": 1.88, "learning_rate": 3.899325517178942e-05, "loss": 0.7375, "step": 31170 }, { "epoch": 1.88, "learning_rate": 3.89917462616752e-05, "loss": 0.9141, "step": 31172 }, { "epoch": 1.88, "learning_rate": 3.899023735156097e-05, "loss": 0.6077, "step": 31174 }, { "epoch": 1.88, "learning_rate": 3.898872844144674e-05, "loss": 1.1764, "step": 31176 }, { "epoch": 1.88, "learning_rate": 3.8987219531332516e-05, "loss": 0.8287, "step": 31178 }, { "epoch": 1.88, "learning_rate": 3.8985710621218296e-05, "loss": 0.6652, "step": 31180 }, { "epoch": 1.88, "learning_rate": 3.8984201711104076e-05, "loss": 0.9134, "step": 31182 }, { "epoch": 1.88, "learning_rate": 3.898269280098984e-05, "loss": 0.8268, "step": 31184 }, { "epoch": 1.88, "learning_rate": 3.898118389087562e-05, "loss": 0.7445, "step": 31186 }, { "epoch": 1.88, "learning_rate": 3.89796749807614e-05, "loss": 0.7008, "step": 31188 }, { "epoch": 1.88, "learning_rate": 3.897816607064717e-05, "loss": 0.7813, "step": 31190 }, { "epoch": 1.88, "learning_rate": 3.897665716053295e-05, "loss": 1.0457, "step": 31192 }, { "epoch": 1.88, "learning_rate": 3.897514825041872e-05, "loss": 0.6069, "step": 31194 }, { "epoch": 1.88, "learning_rate": 3.89736393403045e-05, "loss": 0.8419, "step": 31196 }, { "epoch": 1.88, "learning_rate": 3.8972130430190275e-05, "loss": 0.5861, "step": 31198 }, { "epoch": 1.88, "learning_rate": 3.897062152007605e-05, "loss": 0.7571, "step": 31200 }, { "epoch": 1.88, "learning_rate": 3.896911260996183e-05, "loss": 0.7883, "step": 31202 }, { "epoch": 1.88, "learning_rate": 3.896760369984761e-05, "loss": 0.8429, "step": 31204 }, { "epoch": 1.88, "learning_rate": 3.8966094789733374e-05, "loss": 0.7791, "step": 31206 }, { "epoch": 1.88, "learning_rate": 3.8964585879619154e-05, "loss": 0.642, "step": 31208 }, { "epoch": 1.88, "learning_rate": 3.8963076969504934e-05, "loss": 0.7143, "step": 31210 }, { "epoch": 1.88, "learning_rate": 3.89615680593907e-05, "loss": 0.8415, "step": 31212 }, { "epoch": 1.88, "learning_rate": 3.896005914927648e-05, "loss": 0.994, "step": 31214 }, { "epoch": 1.88, "learning_rate": 3.895855023916225e-05, "loss": 0.7212, "step": 31216 }, { "epoch": 1.88, "learning_rate": 3.895704132904803e-05, "loss": 0.938, "step": 31218 }, { "epoch": 1.88, "learning_rate": 3.8955532418933806e-05, "loss": 0.7269, "step": 31220 }, { "epoch": 1.88, "learning_rate": 3.895402350881958e-05, "loss": 0.8601, "step": 31222 }, { "epoch": 1.88, "learning_rate": 3.895251459870536e-05, "loss": 0.784, "step": 31224 }, { "epoch": 1.88, "learning_rate": 3.895100568859113e-05, "loss": 0.6488, "step": 31226 }, { "epoch": 1.88, "learning_rate": 3.8949496778476905e-05, "loss": 0.6673, "step": 31228 }, { "epoch": 1.88, "learning_rate": 3.8947987868362685e-05, "loss": 0.6439, "step": 31230 }, { "epoch": 1.89, "learning_rate": 3.894647895824846e-05, "loss": 0.804, "step": 31232 }, { "epoch": 1.89, "learning_rate": 3.894497004813423e-05, "loss": 0.6429, "step": 31234 }, { "epoch": 1.89, "learning_rate": 3.894346113802001e-05, "loss": 0.8363, "step": 31236 }, { "epoch": 1.89, "learning_rate": 3.8941952227905784e-05, "loss": 0.9979, "step": 31238 }, { "epoch": 1.89, "learning_rate": 3.894044331779156e-05, "loss": 0.6522, "step": 31240 }, { "epoch": 1.89, "learning_rate": 3.893893440767734e-05, "loss": 0.769, "step": 31242 }, { "epoch": 1.89, "learning_rate": 3.893742549756311e-05, "loss": 0.9392, "step": 31244 }, { "epoch": 1.89, "learning_rate": 3.893591658744889e-05, "loss": 0.612, "step": 31246 }, { "epoch": 1.89, "learning_rate": 3.8934407677334664e-05, "loss": 0.8465, "step": 31248 }, { "epoch": 1.89, "learning_rate": 3.893289876722044e-05, "loss": 0.8037, "step": 31250 }, { "epoch": 1.89, "learning_rate": 3.893138985710622e-05, "loss": 0.5506, "step": 31252 }, { "epoch": 1.89, "learning_rate": 3.892988094699199e-05, "loss": 0.6287, "step": 31254 }, { "epoch": 1.89, "learning_rate": 3.892837203687776e-05, "loss": 1.1586, "step": 31256 }, { "epoch": 1.89, "learning_rate": 3.892686312676354e-05, "loss": 1.0957, "step": 31258 }, { "epoch": 1.89, "learning_rate": 3.8925354216649316e-05, "loss": 0.6356, "step": 31260 }, { "epoch": 1.89, "learning_rate": 3.892384530653509e-05, "loss": 0.7728, "step": 31262 }, { "epoch": 1.89, "learning_rate": 3.892233639642087e-05, "loss": 0.5033, "step": 31264 }, { "epoch": 1.89, "learning_rate": 3.892082748630664e-05, "loss": 0.6633, "step": 31266 }, { "epoch": 1.89, "learning_rate": 3.891931857619242e-05, "loss": 0.9319, "step": 31268 }, { "epoch": 1.89, "learning_rate": 3.8917809666078195e-05, "loss": 0.59, "step": 31270 }, { "epoch": 1.89, "learning_rate": 3.891630075596397e-05, "loss": 0.9898, "step": 31272 }, { "epoch": 1.89, "learning_rate": 3.891479184584975e-05, "loss": 0.9398, "step": 31274 }, { "epoch": 1.89, "learning_rate": 3.8913282935735515e-05, "loss": 0.7918, "step": 31276 }, { "epoch": 1.89, "learning_rate": 3.8911774025621294e-05, "loss": 1.0249, "step": 31278 }, { "epoch": 1.89, "learning_rate": 3.8910265115507074e-05, "loss": 0.7929, "step": 31280 }, { "epoch": 1.89, "learning_rate": 3.890875620539285e-05, "loss": 0.7396, "step": 31282 }, { "epoch": 1.89, "learning_rate": 3.890724729527862e-05, "loss": 0.8176, "step": 31284 }, { "epoch": 1.89, "learning_rate": 3.89057383851644e-05, "loss": 0.948, "step": 31286 }, { "epoch": 1.89, "learning_rate": 3.8904229475050174e-05, "loss": 0.882, "step": 31288 }, { "epoch": 1.89, "learning_rate": 3.890272056493595e-05, "loss": 0.6147, "step": 31290 }, { "epoch": 1.89, "learning_rate": 3.890121165482172e-05, "loss": 0.7611, "step": 31292 }, { "epoch": 1.89, "learning_rate": 3.88997027447075e-05, "loss": 0.9057, "step": 31294 }, { "epoch": 1.89, "learning_rate": 3.889819383459328e-05, "loss": 0.7921, "step": 31296 }, { "epoch": 1.89, "learning_rate": 3.8896684924479046e-05, "loss": 0.4532, "step": 31298 }, { "epoch": 1.89, "learning_rate": 3.8895176014364826e-05, "loss": 0.7183, "step": 31300 }, { "epoch": 1.89, "learning_rate": 3.8893667104250606e-05, "loss": 0.6933, "step": 31302 }, { "epoch": 1.89, "learning_rate": 3.889215819413637e-05, "loss": 0.5115, "step": 31304 }, { "epoch": 1.89, "learning_rate": 3.889064928402215e-05, "loss": 0.6506, "step": 31306 }, { "epoch": 1.89, "learning_rate": 3.8889140373907925e-05, "loss": 0.7768, "step": 31308 }, { "epoch": 1.89, "learning_rate": 3.8887631463793705e-05, "loss": 0.7857, "step": 31310 }, { "epoch": 1.89, "learning_rate": 3.888612255367948e-05, "loss": 0.5434, "step": 31312 }, { "epoch": 1.89, "learning_rate": 3.888461364356525e-05, "loss": 0.8254, "step": 31314 }, { "epoch": 1.89, "learning_rate": 3.888310473345103e-05, "loss": 0.6536, "step": 31316 }, { "epoch": 1.89, "learning_rate": 3.888159582333681e-05, "loss": 0.7607, "step": 31318 }, { "epoch": 1.89, "learning_rate": 3.888008691322258e-05, "loss": 0.7617, "step": 31320 }, { "epoch": 1.89, "learning_rate": 3.887857800310836e-05, "loss": 0.7485, "step": 31322 }, { "epoch": 1.89, "learning_rate": 3.887706909299414e-05, "loss": 0.6977, "step": 31324 }, { "epoch": 1.89, "learning_rate": 3.8875560182879904e-05, "loss": 0.8736, "step": 31326 }, { "epoch": 1.89, "learning_rate": 3.8874051272765684e-05, "loss": 0.6649, "step": 31328 }, { "epoch": 1.89, "learning_rate": 3.887254236265146e-05, "loss": 0.6633, "step": 31330 }, { "epoch": 1.89, "learning_rate": 3.8871033452537237e-05, "loss": 0.8782, "step": 31332 }, { "epoch": 1.89, "learning_rate": 3.886952454242301e-05, "loss": 0.5461, "step": 31334 }, { "epoch": 1.89, "learning_rate": 3.886801563230878e-05, "loss": 1.2231, "step": 31336 }, { "epoch": 1.89, "learning_rate": 3.886650672219456e-05, "loss": 0.8736, "step": 31338 }, { "epoch": 1.89, "learning_rate": 3.8864997812080336e-05, "loss": 0.768, "step": 31340 }, { "epoch": 1.89, "learning_rate": 3.886348890196611e-05, "loss": 0.6525, "step": 31342 }, { "epoch": 1.89, "learning_rate": 3.886197999185189e-05, "loss": 1.0838, "step": 31344 }, { "epoch": 1.89, "learning_rate": 3.886047108173766e-05, "loss": 0.7853, "step": 31346 }, { "epoch": 1.89, "learning_rate": 3.8858962171623435e-05, "loss": 1.0055, "step": 31348 }, { "epoch": 1.89, "learning_rate": 3.8857453261509215e-05, "loss": 0.7715, "step": 31350 }, { "epoch": 1.89, "learning_rate": 3.885594435139499e-05, "loss": 0.9135, "step": 31352 }, { "epoch": 1.89, "learning_rate": 3.885443544128076e-05, "loss": 1.0051, "step": 31354 }, { "epoch": 1.89, "learning_rate": 3.885292653116654e-05, "loss": 0.5083, "step": 31356 }, { "epoch": 1.89, "learning_rate": 3.8851417621052314e-05, "loss": 0.6978, "step": 31358 }, { "epoch": 1.89, "learning_rate": 3.8849908710938094e-05, "loss": 0.7925, "step": 31360 }, { "epoch": 1.89, "learning_rate": 3.884839980082387e-05, "loss": 0.689, "step": 31362 }, { "epoch": 1.89, "learning_rate": 3.884689089070964e-05, "loss": 0.6561, "step": 31364 }, { "epoch": 1.89, "learning_rate": 3.884538198059542e-05, "loss": 0.6304, "step": 31366 }, { "epoch": 1.89, "learning_rate": 3.8843873070481193e-05, "loss": 0.8489, "step": 31368 }, { "epoch": 1.89, "learning_rate": 3.8842364160366967e-05, "loss": 0.8128, "step": 31370 }, { "epoch": 1.89, "learning_rate": 3.8840855250252747e-05, "loss": 0.6728, "step": 31372 }, { "epoch": 1.89, "learning_rate": 3.883934634013852e-05, "loss": 0.6957, "step": 31374 }, { "epoch": 1.89, "learning_rate": 3.883783743002429e-05, "loss": 0.8094, "step": 31376 }, { "epoch": 1.89, "learning_rate": 3.883632851991007e-05, "loss": 0.6845, "step": 31378 }, { "epoch": 1.89, "learning_rate": 3.8834819609795846e-05, "loss": 0.6296, "step": 31380 }, { "epoch": 1.89, "learning_rate": 3.8833310699681626e-05, "loss": 0.5557, "step": 31382 }, { "epoch": 1.89, "learning_rate": 3.88318017895674e-05, "loss": 0.6, "step": 31384 }, { "epoch": 1.89, "learning_rate": 3.883029287945317e-05, "loss": 0.7974, "step": 31386 }, { "epoch": 1.89, "learning_rate": 3.882878396933895e-05, "loss": 0.653, "step": 31388 }, { "epoch": 1.89, "learning_rate": 3.882727505922472e-05, "loss": 0.8801, "step": 31390 }, { "epoch": 1.89, "learning_rate": 3.88257661491105e-05, "loss": 0.7063, "step": 31392 }, { "epoch": 1.89, "learning_rate": 3.882425723899628e-05, "loss": 0.9409, "step": 31394 }, { "epoch": 1.89, "learning_rate": 3.882274832888205e-05, "loss": 0.9162, "step": 31396 }, { "epoch": 1.9, "learning_rate": 3.8821239418767824e-05, "loss": 0.6486, "step": 31398 }, { "epoch": 1.9, "learning_rate": 3.8819730508653604e-05, "loss": 0.7716, "step": 31400 }, { "epoch": 1.9, "learning_rate": 3.881822159853938e-05, "loss": 0.899, "step": 31402 }, { "epoch": 1.9, "learning_rate": 3.881671268842515e-05, "loss": 0.7207, "step": 31404 }, { "epoch": 1.9, "learning_rate": 3.8815203778310924e-05, "loss": 0.7979, "step": 31406 }, { "epoch": 1.9, "learning_rate": 3.8813694868196703e-05, "loss": 0.5627, "step": 31408 }, { "epoch": 1.9, "learning_rate": 3.881218595808248e-05, "loss": 0.9556, "step": 31410 }, { "epoch": 1.9, "learning_rate": 3.881067704796825e-05, "loss": 0.5532, "step": 31412 }, { "epoch": 1.9, "learning_rate": 3.880916813785403e-05, "loss": 0.8615, "step": 31414 }, { "epoch": 1.9, "learning_rate": 3.880765922773981e-05, "loss": 0.7689, "step": 31416 }, { "epoch": 1.9, "learning_rate": 3.880615031762558e-05, "loss": 0.5508, "step": 31418 }, { "epoch": 1.9, "learning_rate": 3.8804641407511356e-05, "loss": 0.7764, "step": 31420 }, { "epoch": 1.9, "learning_rate": 3.880313249739713e-05, "loss": 0.4321, "step": 31422 }, { "epoch": 1.9, "learning_rate": 3.880162358728291e-05, "loss": 0.9299, "step": 31424 }, { "epoch": 1.9, "learning_rate": 3.880011467716868e-05, "loss": 0.7309, "step": 31426 }, { "epoch": 1.9, "learning_rate": 3.8798605767054455e-05, "loss": 0.8702, "step": 31428 }, { "epoch": 1.9, "learning_rate": 3.8797096856940235e-05, "loss": 0.9145, "step": 31430 }, { "epoch": 1.9, "learning_rate": 3.8795587946826015e-05, "loss": 0.8568, "step": 31432 }, { "epoch": 1.9, "learning_rate": 3.879407903671178e-05, "loss": 0.8784, "step": 31434 }, { "epoch": 1.9, "learning_rate": 3.879257012659756e-05, "loss": 0.7744, "step": 31436 }, { "epoch": 1.9, "learning_rate": 3.8791061216483334e-05, "loss": 0.9587, "step": 31438 }, { "epoch": 1.9, "learning_rate": 3.878955230636911e-05, "loss": 0.9478, "step": 31440 }, { "epoch": 1.9, "learning_rate": 3.878804339625489e-05, "loss": 0.5905, "step": 31442 }, { "epoch": 1.9, "learning_rate": 3.878653448614066e-05, "loss": 0.723, "step": 31444 }, { "epoch": 1.9, "learning_rate": 3.878502557602644e-05, "loss": 0.5372, "step": 31446 }, { "epoch": 1.9, "learning_rate": 3.878351666591221e-05, "loss": 0.7548, "step": 31448 }, { "epoch": 1.9, "learning_rate": 3.8782007755797986e-05, "loss": 0.7384, "step": 31450 }, { "epoch": 1.9, "learning_rate": 3.8780498845683766e-05, "loss": 0.8489, "step": 31452 }, { "epoch": 1.9, "learning_rate": 3.877898993556954e-05, "loss": 0.7586, "step": 31454 }, { "epoch": 1.9, "learning_rate": 3.877748102545531e-05, "loss": 1.0241, "step": 31456 }, { "epoch": 1.9, "learning_rate": 3.877597211534109e-05, "loss": 0.6768, "step": 31458 }, { "epoch": 1.9, "learning_rate": 3.8774463205226866e-05, "loss": 0.7802, "step": 31460 }, { "epoch": 1.9, "learning_rate": 3.877295429511264e-05, "loss": 0.7701, "step": 31462 }, { "epoch": 1.9, "learning_rate": 3.877144538499842e-05, "loss": 1.1505, "step": 31464 }, { "epoch": 1.9, "learning_rate": 3.876993647488419e-05, "loss": 0.8319, "step": 31466 }, { "epoch": 1.9, "learning_rate": 3.8768427564769965e-05, "loss": 0.5897, "step": 31468 }, { "epoch": 1.9, "learning_rate": 3.8766918654655745e-05, "loss": 1.0664, "step": 31470 }, { "epoch": 1.9, "learning_rate": 3.876540974454152e-05, "loss": 0.5984, "step": 31472 }, { "epoch": 1.9, "learning_rate": 3.87639008344273e-05, "loss": 0.7386, "step": 31474 }, { "epoch": 1.9, "learning_rate": 3.876239192431307e-05, "loss": 0.8686, "step": 31476 }, { "epoch": 1.9, "learning_rate": 3.8760883014198844e-05, "loss": 0.7003, "step": 31478 }, { "epoch": 1.9, "learning_rate": 3.8759374104084624e-05, "loss": 0.69, "step": 31480 }, { "epoch": 1.9, "learning_rate": 3.87578651939704e-05, "loss": 0.7561, "step": 31482 }, { "epoch": 1.9, "learning_rate": 3.875635628385617e-05, "loss": 0.4911, "step": 31484 }, { "epoch": 1.9, "learning_rate": 3.875484737374195e-05, "loss": 0.8916, "step": 31486 }, { "epoch": 1.9, "learning_rate": 3.875333846362772e-05, "loss": 0.8676, "step": 31488 }, { "epoch": 1.9, "learning_rate": 3.8751829553513496e-05, "loss": 0.9022, "step": 31490 }, { "epoch": 1.9, "learning_rate": 3.8750320643399276e-05, "loss": 0.7823, "step": 31492 }, { "epoch": 1.9, "learning_rate": 3.874881173328505e-05, "loss": 0.8715, "step": 31494 }, { "epoch": 1.9, "learning_rate": 3.874730282317083e-05, "loss": 0.8464, "step": 31496 }, { "epoch": 1.9, "learning_rate": 3.87457939130566e-05, "loss": 0.8931, "step": 31498 }, { "epoch": 1.9, "learning_rate": 3.8744285002942376e-05, "loss": 0.9271, "step": 31500 }, { "epoch": 1.9, "learning_rate": 3.8742776092828156e-05, "loss": 0.8275, "step": 31502 }, { "epoch": 1.9, "learning_rate": 3.874126718271392e-05, "loss": 0.8564, "step": 31504 }, { "epoch": 1.9, "learning_rate": 3.87397582725997e-05, "loss": 0.5704, "step": 31506 }, { "epoch": 1.9, "learning_rate": 3.873824936248548e-05, "loss": 0.7262, "step": 31508 }, { "epoch": 1.9, "learning_rate": 3.8736740452371255e-05, "loss": 0.9176, "step": 31510 }, { "epoch": 1.9, "learning_rate": 3.873523154225703e-05, "loss": 0.7238, "step": 31512 }, { "epoch": 1.9, "learning_rate": 3.873372263214281e-05, "loss": 0.6323, "step": 31514 }, { "epoch": 1.9, "learning_rate": 3.873221372202858e-05, "loss": 0.7909, "step": 31516 }, { "epoch": 1.9, "learning_rate": 3.8730704811914354e-05, "loss": 0.8829, "step": 31518 }, { "epoch": 1.9, "learning_rate": 3.872919590180013e-05, "loss": 0.6783, "step": 31520 }, { "epoch": 1.9, "learning_rate": 3.872768699168591e-05, "loss": 0.6248, "step": 31522 }, { "epoch": 1.9, "learning_rate": 3.872617808157169e-05, "loss": 0.5621, "step": 31524 }, { "epoch": 1.9, "learning_rate": 3.872466917145745e-05, "loss": 0.5326, "step": 31526 }, { "epoch": 1.9, "learning_rate": 3.872316026134323e-05, "loss": 0.9084, "step": 31528 }, { "epoch": 1.9, "learning_rate": 3.872165135122901e-05, "loss": 0.9082, "step": 31530 }, { "epoch": 1.9, "learning_rate": 3.8720142441114786e-05, "loss": 0.9354, "step": 31532 }, { "epoch": 1.9, "learning_rate": 3.871863353100056e-05, "loss": 0.7819, "step": 31534 }, { "epoch": 1.9, "learning_rate": 3.871712462088633e-05, "loss": 0.5874, "step": 31536 }, { "epoch": 1.9, "learning_rate": 3.871561571077211e-05, "loss": 0.7792, "step": 31538 }, { "epoch": 1.9, "learning_rate": 3.8714106800657886e-05, "loss": 0.6868, "step": 31540 }, { "epoch": 1.9, "learning_rate": 3.871259789054366e-05, "loss": 0.7328, "step": 31542 }, { "epoch": 1.9, "learning_rate": 3.871108898042944e-05, "loss": 0.93, "step": 31544 }, { "epoch": 1.9, "learning_rate": 3.870958007031522e-05, "loss": 0.8358, "step": 31546 }, { "epoch": 1.9, "learning_rate": 3.8708071160200985e-05, "loss": 0.8741, "step": 31548 }, { "epoch": 1.9, "learning_rate": 3.8706562250086765e-05, "loss": 0.8233, "step": 31550 }, { "epoch": 1.9, "learning_rate": 3.870505333997254e-05, "loss": 1.0396, "step": 31552 }, { "epoch": 1.9, "learning_rate": 3.870354442985831e-05, "loss": 0.7036, "step": 31554 }, { "epoch": 1.9, "learning_rate": 3.870203551974409e-05, "loss": 0.8619, "step": 31556 }, { "epoch": 1.9, "learning_rate": 3.8700526609629864e-05, "loss": 0.8096, "step": 31558 }, { "epoch": 1.9, "learning_rate": 3.8699017699515644e-05, "loss": 1.0358, "step": 31560 }, { "epoch": 1.9, "learning_rate": 3.869750878940142e-05, "loss": 0.7982, "step": 31562 }, { "epoch": 1.91, "learning_rate": 3.869599987928719e-05, "loss": 0.8983, "step": 31564 }, { "epoch": 1.91, "learning_rate": 3.869449096917297e-05, "loss": 0.8592, "step": 31566 }, { "epoch": 1.91, "learning_rate": 3.869298205905874e-05, "loss": 0.626, "step": 31568 }, { "epoch": 1.91, "learning_rate": 3.8691473148944516e-05, "loss": 0.7008, "step": 31570 }, { "epoch": 1.91, "learning_rate": 3.8689964238830296e-05, "loss": 0.5277, "step": 31572 }, { "epoch": 1.91, "learning_rate": 3.868845532871607e-05, "loss": 0.6032, "step": 31574 }, { "epoch": 1.91, "learning_rate": 3.868694641860184e-05, "loss": 0.6066, "step": 31576 }, { "epoch": 1.91, "learning_rate": 3.868543750848762e-05, "loss": 0.7394, "step": 31578 }, { "epoch": 1.91, "learning_rate": 3.8683928598373396e-05, "loss": 1.1877, "step": 31580 }, { "epoch": 1.91, "learning_rate": 3.8682419688259175e-05, "loss": 0.7011, "step": 31582 }, { "epoch": 1.91, "learning_rate": 3.868091077814495e-05, "loss": 0.8168, "step": 31584 }, { "epoch": 1.91, "learning_rate": 3.867940186803072e-05, "loss": 0.8203, "step": 31586 }, { "epoch": 1.91, "learning_rate": 3.86778929579165e-05, "loss": 0.8357, "step": 31588 }, { "epoch": 1.91, "learning_rate": 3.8676384047802275e-05, "loss": 0.7649, "step": 31590 }, { "epoch": 1.91, "learning_rate": 3.867487513768805e-05, "loss": 0.9697, "step": 31592 }, { "epoch": 1.91, "learning_rate": 3.867336622757383e-05, "loss": 0.889, "step": 31594 }, { "epoch": 1.91, "learning_rate": 3.86718573174596e-05, "loss": 0.9153, "step": 31596 }, { "epoch": 1.91, "learning_rate": 3.8670348407345374e-05, "loss": 0.7324, "step": 31598 }, { "epoch": 1.91, "learning_rate": 3.8668839497231154e-05, "loss": 0.9364, "step": 31600 }, { "epoch": 1.91, "learning_rate": 3.866733058711693e-05, "loss": 0.8956, "step": 31602 }, { "epoch": 1.91, "learning_rate": 3.86658216770027e-05, "loss": 0.8142, "step": 31604 }, { "epoch": 1.91, "learning_rate": 3.866431276688848e-05, "loss": 0.7147, "step": 31606 }, { "epoch": 1.91, "learning_rate": 3.866280385677425e-05, "loss": 0.9551, "step": 31608 }, { "epoch": 1.91, "learning_rate": 3.866129494666003e-05, "loss": 0.8729, "step": 31610 }, { "epoch": 1.91, "learning_rate": 3.8659786036545806e-05, "loss": 0.7374, "step": 31612 }, { "epoch": 1.91, "learning_rate": 3.865827712643158e-05, "loss": 0.6437, "step": 31614 }, { "epoch": 1.91, "learning_rate": 3.865676821631736e-05, "loss": 0.7336, "step": 31616 }, { "epoch": 1.91, "learning_rate": 3.8655259306203126e-05, "loss": 0.999, "step": 31618 }, { "epoch": 1.91, "learning_rate": 3.8653750396088905e-05, "loss": 0.5964, "step": 31620 }, { "epoch": 1.91, "learning_rate": 3.8652241485974685e-05, "loss": 0.5689, "step": 31622 }, { "epoch": 1.91, "learning_rate": 3.865073257586046e-05, "loss": 0.7282, "step": 31624 }, { "epoch": 1.91, "learning_rate": 3.864922366574623e-05, "loss": 0.8333, "step": 31626 }, { "epoch": 1.91, "learning_rate": 3.864771475563201e-05, "loss": 0.9128, "step": 31628 }, { "epoch": 1.91, "learning_rate": 3.8646205845517785e-05, "loss": 0.7739, "step": 31630 }, { "epoch": 1.91, "learning_rate": 3.8644696935403565e-05, "loss": 0.4602, "step": 31632 }, { "epoch": 1.91, "learning_rate": 3.864318802528933e-05, "loss": 0.8253, "step": 31634 }, { "epoch": 1.91, "learning_rate": 3.864167911517511e-05, "loss": 0.6628, "step": 31636 }, { "epoch": 1.91, "learning_rate": 3.864017020506089e-05, "loss": 0.905, "step": 31638 }, { "epoch": 1.91, "learning_rate": 3.863866129494666e-05, "loss": 1.0175, "step": 31640 }, { "epoch": 1.91, "learning_rate": 3.863715238483244e-05, "loss": 1.2821, "step": 31642 }, { "epoch": 1.91, "learning_rate": 3.863564347471822e-05, "loss": 0.803, "step": 31644 }, { "epoch": 1.91, "learning_rate": 3.863413456460399e-05, "loss": 1.0125, "step": 31646 }, { "epoch": 1.91, "learning_rate": 3.863262565448976e-05, "loss": 0.8411, "step": 31648 }, { "epoch": 1.91, "learning_rate": 3.8631116744375536e-05, "loss": 1.1166, "step": 31650 }, { "epoch": 1.91, "learning_rate": 3.8629607834261316e-05, "loss": 0.6211, "step": 31652 }, { "epoch": 1.91, "learning_rate": 3.862809892414709e-05, "loss": 0.6522, "step": 31654 }, { "epoch": 1.91, "learning_rate": 3.862659001403286e-05, "loss": 0.6819, "step": 31656 }, { "epoch": 1.91, "learning_rate": 3.862508110391864e-05, "loss": 0.7523, "step": 31658 }, { "epoch": 1.91, "learning_rate": 3.862357219380442e-05, "loss": 0.7025, "step": 31660 }, { "epoch": 1.91, "learning_rate": 3.862206328369019e-05, "loss": 0.5996, "step": 31662 }, { "epoch": 1.91, "learning_rate": 3.862055437357597e-05, "loss": 0.914, "step": 31664 }, { "epoch": 1.91, "learning_rate": 3.861904546346174e-05, "loss": 0.7261, "step": 31666 }, { "epoch": 1.91, "learning_rate": 3.8617536553347515e-05, "loss": 0.9226, "step": 31668 }, { "epoch": 1.91, "learning_rate": 3.8616027643233295e-05, "loss": 0.7254, "step": 31670 }, { "epoch": 1.91, "learning_rate": 3.861451873311907e-05, "loss": 0.546, "step": 31672 }, { "epoch": 1.91, "learning_rate": 3.861300982300485e-05, "loss": 0.8624, "step": 31674 }, { "epoch": 1.91, "learning_rate": 3.861150091289062e-05, "loss": 0.9686, "step": 31676 }, { "epoch": 1.91, "learning_rate": 3.8609992002776394e-05, "loss": 0.5654, "step": 31678 }, { "epoch": 1.91, "learning_rate": 3.8608483092662174e-05, "loss": 0.7626, "step": 31680 }, { "epoch": 1.91, "learning_rate": 3.860697418254795e-05, "loss": 0.8929, "step": 31682 }, { "epoch": 1.91, "learning_rate": 3.860546527243372e-05, "loss": 0.7913, "step": 31684 }, { "epoch": 1.91, "learning_rate": 3.86039563623195e-05, "loss": 0.8024, "step": 31686 }, { "epoch": 1.91, "learning_rate": 3.860244745220527e-05, "loss": 0.9347, "step": 31688 }, { "epoch": 1.91, "learning_rate": 3.8600938542091046e-05, "loss": 0.6852, "step": 31690 }, { "epoch": 1.91, "learning_rate": 3.8599429631976826e-05, "loss": 0.6169, "step": 31692 }, { "epoch": 1.91, "learning_rate": 3.85979207218626e-05, "loss": 0.6637, "step": 31694 }, { "epoch": 1.91, "learning_rate": 3.859641181174838e-05, "loss": 0.6683, "step": 31696 }, { "epoch": 1.91, "learning_rate": 3.859490290163415e-05, "loss": 0.6611, "step": 31698 }, { "epoch": 1.91, "learning_rate": 3.8593393991519925e-05, "loss": 0.7673, "step": 31700 }, { "epoch": 1.91, "learning_rate": 3.8591885081405705e-05, "loss": 0.6201, "step": 31702 }, { "epoch": 1.91, "learning_rate": 3.859037617129148e-05, "loss": 0.8105, "step": 31704 }, { "epoch": 1.91, "learning_rate": 3.858886726117725e-05, "loss": 0.8718, "step": 31706 }, { "epoch": 1.91, "learning_rate": 3.858735835106303e-05, "loss": 0.6229, "step": 31708 }, { "epoch": 1.91, "learning_rate": 3.8585849440948805e-05, "loss": 0.6837, "step": 31710 }, { "epoch": 1.91, "learning_rate": 3.858434053083458e-05, "loss": 0.8785, "step": 31712 }, { "epoch": 1.91, "learning_rate": 3.858283162072036e-05, "loss": 0.9089, "step": 31714 }, { "epoch": 1.91, "learning_rate": 3.858132271060613e-05, "loss": 0.7944, "step": 31716 }, { "epoch": 1.91, "learning_rate": 3.8579813800491904e-05, "loss": 0.7342, "step": 31718 }, { "epoch": 1.91, "learning_rate": 3.8578304890377684e-05, "loss": 0.6952, "step": 31720 }, { "epoch": 1.91, "learning_rate": 3.857679598026346e-05, "loss": 0.6692, "step": 31722 }, { "epoch": 1.91, "learning_rate": 3.857528707014924e-05, "loss": 0.7481, "step": 31724 }, { "epoch": 1.91, "learning_rate": 3.857377816003501e-05, "loss": 0.7662, "step": 31726 }, { "epoch": 1.92, "learning_rate": 3.857226924992078e-05, "loss": 0.7433, "step": 31728 }, { "epoch": 1.92, "learning_rate": 3.857076033980656e-05, "loss": 0.8405, "step": 31730 }, { "epoch": 1.92, "learning_rate": 3.856925142969233e-05, "loss": 0.906, "step": 31732 }, { "epoch": 1.92, "learning_rate": 3.856774251957811e-05, "loss": 0.6914, "step": 31734 }, { "epoch": 1.92, "learning_rate": 3.856623360946389e-05, "loss": 0.9584, "step": 31736 }, { "epoch": 1.92, "learning_rate": 3.856472469934966e-05, "loss": 0.9227, "step": 31738 }, { "epoch": 1.92, "learning_rate": 3.8563215789235435e-05, "loss": 0.743, "step": 31740 }, { "epoch": 1.92, "learning_rate": 3.8561706879121215e-05, "loss": 0.6336, "step": 31742 }, { "epoch": 1.92, "learning_rate": 3.856019796900699e-05, "loss": 0.9223, "step": 31744 }, { "epoch": 1.92, "learning_rate": 3.855868905889277e-05, "loss": 0.8568, "step": 31746 }, { "epoch": 1.92, "learning_rate": 3.8557180148778535e-05, "loss": 0.7222, "step": 31748 }, { "epoch": 1.92, "learning_rate": 3.8555671238664314e-05, "loss": 0.739, "step": 31750 }, { "epoch": 1.92, "learning_rate": 3.8554162328550094e-05, "loss": 0.6737, "step": 31752 }, { "epoch": 1.92, "learning_rate": 3.855265341843586e-05, "loss": 1.2001, "step": 31754 }, { "epoch": 1.92, "learning_rate": 3.855114450832164e-05, "loss": 0.8499, "step": 31756 }, { "epoch": 1.92, "learning_rate": 3.854963559820742e-05, "loss": 0.9682, "step": 31758 }, { "epoch": 1.92, "learning_rate": 3.8548126688093194e-05, "loss": 0.7662, "step": 31760 }, { "epoch": 1.92, "learning_rate": 3.854661777797897e-05, "loss": 0.7419, "step": 31762 }, { "epoch": 1.92, "learning_rate": 3.854510886786474e-05, "loss": 0.8057, "step": 31764 }, { "epoch": 1.92, "learning_rate": 3.854359995775052e-05, "loss": 1.1191, "step": 31766 }, { "epoch": 1.92, "learning_rate": 3.854209104763629e-05, "loss": 0.8764, "step": 31768 }, { "epoch": 1.92, "learning_rate": 3.8540582137522066e-05, "loss": 0.8505, "step": 31770 }, { "epoch": 1.92, "learning_rate": 3.8539073227407846e-05, "loss": 0.5735, "step": 31772 }, { "epoch": 1.92, "learning_rate": 3.8537564317293626e-05, "loss": 1.2405, "step": 31774 }, { "epoch": 1.92, "learning_rate": 3.853605540717939e-05, "loss": 0.6735, "step": 31776 }, { "epoch": 1.92, "learning_rate": 3.853454649706517e-05, "loss": 0.5949, "step": 31778 }, { "epoch": 1.92, "learning_rate": 3.8533037586950945e-05, "loss": 0.9376, "step": 31780 }, { "epoch": 1.92, "learning_rate": 3.853152867683672e-05, "loss": 0.4921, "step": 31782 }, { "epoch": 1.92, "learning_rate": 3.85300197667225e-05, "loss": 0.7965, "step": 31784 }, { "epoch": 1.92, "learning_rate": 3.852851085660827e-05, "loss": 0.5893, "step": 31786 }, { "epoch": 1.92, "learning_rate": 3.852700194649405e-05, "loss": 0.8592, "step": 31788 }, { "epoch": 1.92, "learning_rate": 3.8525493036379824e-05, "loss": 0.9226, "step": 31790 }, { "epoch": 1.92, "learning_rate": 3.85239841262656e-05, "loss": 0.7467, "step": 31792 }, { "epoch": 1.92, "learning_rate": 3.852247521615138e-05, "loss": 0.4784, "step": 31794 }, { "epoch": 1.92, "learning_rate": 3.852096630603716e-05, "loss": 0.7726, "step": 31796 }, { "epoch": 1.92, "learning_rate": 3.8519457395922924e-05, "loss": 0.5477, "step": 31798 }, { "epoch": 1.92, "learning_rate": 3.8517948485808704e-05, "loss": 0.7523, "step": 31800 }, { "epoch": 1.92, "learning_rate": 3.851643957569448e-05, "loss": 0.7598, "step": 31802 }, { "epoch": 1.92, "learning_rate": 3.851493066558025e-05, "loss": 0.992, "step": 31804 }, { "epoch": 1.92, "learning_rate": 3.851342175546603e-05, "loss": 0.5649, "step": 31806 }, { "epoch": 1.92, "learning_rate": 3.85119128453518e-05, "loss": 0.5649, "step": 31808 }, { "epoch": 1.92, "learning_rate": 3.851040393523758e-05, "loss": 0.6531, "step": 31810 }, { "epoch": 1.92, "learning_rate": 3.8508895025123356e-05, "loss": 0.4824, "step": 31812 }, { "epoch": 1.92, "learning_rate": 3.850738611500913e-05, "loss": 0.7575, "step": 31814 }, { "epoch": 1.92, "learning_rate": 3.850587720489491e-05, "loss": 0.7412, "step": 31816 }, { "epoch": 1.92, "learning_rate": 3.850436829478068e-05, "loss": 1.1262, "step": 31818 }, { "epoch": 1.92, "learning_rate": 3.8502859384666455e-05, "loss": 0.9567, "step": 31820 }, { "epoch": 1.92, "learning_rate": 3.8501350474552235e-05, "loss": 0.5497, "step": 31822 }, { "epoch": 1.92, "learning_rate": 3.849984156443801e-05, "loss": 0.6862, "step": 31824 }, { "epoch": 1.92, "learning_rate": 3.849833265432378e-05, "loss": 0.727, "step": 31826 }, { "epoch": 1.92, "learning_rate": 3.849682374420956e-05, "loss": 0.8985, "step": 31828 }, { "epoch": 1.92, "learning_rate": 3.8495314834095334e-05, "loss": 0.8712, "step": 31830 }, { "epoch": 1.92, "learning_rate": 3.849380592398111e-05, "loss": 1.2673, "step": 31832 }, { "epoch": 1.92, "learning_rate": 3.849229701386689e-05, "loss": 0.7229, "step": 31834 }, { "epoch": 1.92, "learning_rate": 3.849078810375266e-05, "loss": 0.6936, "step": 31836 }, { "epoch": 1.92, "learning_rate": 3.848927919363844e-05, "loss": 0.9799, "step": 31838 }, { "epoch": 1.92, "learning_rate": 3.8487770283524214e-05, "loss": 0.9204, "step": 31840 }, { "epoch": 1.92, "learning_rate": 3.848626137340999e-05, "loss": 0.8416, "step": 31842 }, { "epoch": 1.92, "learning_rate": 3.8484752463295767e-05, "loss": 0.9224, "step": 31844 }, { "epoch": 1.92, "learning_rate": 3.848324355318153e-05, "loss": 0.9214, "step": 31846 }, { "epoch": 1.92, "learning_rate": 3.848173464306731e-05, "loss": 0.8446, "step": 31848 }, { "epoch": 1.92, "learning_rate": 3.848022573295309e-05, "loss": 0.8142, "step": 31850 }, { "epoch": 1.92, "learning_rate": 3.8478716822838866e-05, "loss": 0.6256, "step": 31852 }, { "epoch": 1.92, "learning_rate": 3.847720791272464e-05, "loss": 0.9502, "step": 31854 }, { "epoch": 1.92, "learning_rate": 3.847569900261042e-05, "loss": 0.6038, "step": 31856 }, { "epoch": 1.92, "learning_rate": 3.847419009249619e-05, "loss": 0.5728, "step": 31858 }, { "epoch": 1.92, "learning_rate": 3.847268118238197e-05, "loss": 0.7881, "step": 31860 }, { "epoch": 1.92, "learning_rate": 3.847117227226774e-05, "loss": 0.8229, "step": 31862 }, { "epoch": 1.92, "learning_rate": 3.846966336215352e-05, "loss": 0.643, "step": 31864 }, { "epoch": 1.92, "learning_rate": 3.84681544520393e-05, "loss": 0.8082, "step": 31866 }, { "epoch": 1.92, "learning_rate": 3.8466645541925064e-05, "loss": 0.8993, "step": 31868 }, { "epoch": 1.92, "learning_rate": 3.8465136631810844e-05, "loss": 0.7088, "step": 31870 }, { "epoch": 1.92, "learning_rate": 3.8463627721696624e-05, "loss": 0.8046, "step": 31872 }, { "epoch": 1.92, "learning_rate": 3.84621188115824e-05, "loss": 0.5772, "step": 31874 }, { "epoch": 1.92, "learning_rate": 3.846060990146817e-05, "loss": 0.6877, "step": 31876 }, { "epoch": 1.92, "learning_rate": 3.8459100991353944e-05, "loss": 0.9263, "step": 31878 }, { "epoch": 1.92, "learning_rate": 3.8457592081239723e-05, "loss": 0.6269, "step": 31880 }, { "epoch": 1.92, "learning_rate": 3.8456083171125497e-05, "loss": 0.9072, "step": 31882 }, { "epoch": 1.92, "learning_rate": 3.845457426101127e-05, "loss": 0.7625, "step": 31884 }, { "epoch": 1.92, "learning_rate": 3.845306535089705e-05, "loss": 0.591, "step": 31886 }, { "epoch": 1.92, "learning_rate": 3.845155644078283e-05, "loss": 0.9965, "step": 31888 }, { "epoch": 1.92, "learning_rate": 3.8450047530668596e-05, "loss": 0.7466, "step": 31890 }, { "epoch": 1.92, "learning_rate": 3.8448538620554376e-05, "loss": 0.6159, "step": 31892 }, { "epoch": 1.93, "learning_rate": 3.844702971044015e-05, "loss": 0.8225, "step": 31894 }, { "epoch": 1.93, "learning_rate": 3.844552080032592e-05, "loss": 0.6959, "step": 31896 }, { "epoch": 1.93, "learning_rate": 3.84440118902117e-05, "loss": 0.6954, "step": 31898 }, { "epoch": 1.93, "learning_rate": 3.8442502980097475e-05, "loss": 0.7126, "step": 31900 }, { "epoch": 1.93, "learning_rate": 3.8440994069983255e-05, "loss": 1.1147, "step": 31902 }, { "epoch": 1.93, "learning_rate": 3.843948515986903e-05, "loss": 0.6987, "step": 31904 }, { "epoch": 1.93, "learning_rate": 3.84379762497548e-05, "loss": 1.0141, "step": 31906 }, { "epoch": 1.93, "learning_rate": 3.843646733964058e-05, "loss": 0.7738, "step": 31908 }, { "epoch": 1.93, "learning_rate": 3.8434958429526354e-05, "loss": 0.8765, "step": 31910 }, { "epoch": 1.93, "learning_rate": 3.843344951941213e-05, "loss": 0.6105, "step": 31912 }, { "epoch": 1.93, "learning_rate": 3.843194060929791e-05, "loss": 0.73, "step": 31914 }, { "epoch": 1.93, "learning_rate": 3.843043169918368e-05, "loss": 0.8561, "step": 31916 }, { "epoch": 1.93, "learning_rate": 3.8428922789069454e-05, "loss": 0.7048, "step": 31918 }, { "epoch": 1.93, "learning_rate": 3.8427413878955233e-05, "loss": 0.7482, "step": 31920 }, { "epoch": 1.93, "learning_rate": 3.8425904968841007e-05, "loss": 0.6913, "step": 31922 }, { "epoch": 1.93, "learning_rate": 3.8424396058726786e-05, "loss": 0.7809, "step": 31924 }, { "epoch": 1.93, "learning_rate": 3.842288714861256e-05, "loss": 0.7614, "step": 31926 }, { "epoch": 1.93, "learning_rate": 3.842137823849833e-05, "loss": 0.8647, "step": 31928 }, { "epoch": 1.93, "learning_rate": 3.841986932838411e-05, "loss": 0.7324, "step": 31930 }, { "epoch": 1.93, "learning_rate": 3.8418360418269886e-05, "loss": 0.5441, "step": 31932 }, { "epoch": 1.93, "learning_rate": 3.841685150815566e-05, "loss": 0.545, "step": 31934 }, { "epoch": 1.93, "learning_rate": 3.841534259804144e-05, "loss": 0.7339, "step": 31936 }, { "epoch": 1.93, "learning_rate": 3.841383368792721e-05, "loss": 0.946, "step": 31938 }, { "epoch": 1.93, "learning_rate": 3.8412324777812985e-05, "loss": 0.8248, "step": 31940 }, { "epoch": 1.93, "learning_rate": 3.8410815867698765e-05, "loss": 0.762, "step": 31942 }, { "epoch": 1.93, "learning_rate": 3.840930695758454e-05, "loss": 0.9641, "step": 31944 }, { "epoch": 1.93, "learning_rate": 3.840779804747031e-05, "loss": 0.8467, "step": 31946 }, { "epoch": 1.93, "learning_rate": 3.840628913735609e-05, "loss": 0.6415, "step": 31948 }, { "epoch": 1.93, "learning_rate": 3.8404780227241864e-05, "loss": 0.4312, "step": 31950 }, { "epoch": 1.93, "learning_rate": 3.8403271317127644e-05, "loss": 0.882, "step": 31952 }, { "epoch": 1.93, "learning_rate": 3.840176240701342e-05, "loss": 0.8797, "step": 31954 }, { "epoch": 1.93, "learning_rate": 3.840025349689919e-05, "loss": 0.9121, "step": 31956 }, { "epoch": 1.93, "learning_rate": 3.839874458678497e-05, "loss": 0.7215, "step": 31958 }, { "epoch": 1.93, "learning_rate": 3.839723567667074e-05, "loss": 0.6182, "step": 31960 }, { "epoch": 1.93, "learning_rate": 3.8395726766556516e-05, "loss": 0.4064, "step": 31962 }, { "epoch": 1.93, "learning_rate": 3.8394217856442296e-05, "loss": 0.6503, "step": 31964 }, { "epoch": 1.93, "learning_rate": 3.839270894632807e-05, "loss": 0.7749, "step": 31966 }, { "epoch": 1.93, "learning_rate": 3.839120003621384e-05, "loss": 0.5566, "step": 31968 }, { "epoch": 1.93, "learning_rate": 3.838969112609962e-05, "loss": 0.6907, "step": 31970 }, { "epoch": 1.93, "learning_rate": 3.8388182215985396e-05, "loss": 0.7198, "step": 31972 }, { "epoch": 1.93, "learning_rate": 3.8386673305871176e-05, "loss": 0.8131, "step": 31974 }, { "epoch": 1.93, "learning_rate": 3.838516439575694e-05, "loss": 0.9132, "step": 31976 }, { "epoch": 1.93, "learning_rate": 3.838365548564272e-05, "loss": 0.6997, "step": 31978 }, { "epoch": 1.93, "learning_rate": 3.83821465755285e-05, "loss": 0.7006, "step": 31980 }, { "epoch": 1.93, "learning_rate": 3.838063766541427e-05, "loss": 0.8499, "step": 31982 }, { "epoch": 1.93, "learning_rate": 3.837912875530005e-05, "loss": 1.075, "step": 31984 }, { "epoch": 1.93, "learning_rate": 3.837761984518583e-05, "loss": 0.8026, "step": 31986 }, { "epoch": 1.93, "learning_rate": 3.83761109350716e-05, "loss": 0.7748, "step": 31988 }, { "epoch": 1.93, "learning_rate": 3.8374602024957374e-05, "loss": 0.6177, "step": 31990 }, { "epoch": 1.93, "learning_rate": 3.837309311484315e-05, "loss": 1.094, "step": 31992 }, { "epoch": 1.93, "learning_rate": 3.837158420472893e-05, "loss": 0.6652, "step": 31994 }, { "epoch": 1.93, "learning_rate": 3.83700752946147e-05, "loss": 0.876, "step": 31996 }, { "epoch": 1.93, "learning_rate": 3.8368566384500473e-05, "loss": 0.7882, "step": 31998 }, { "epoch": 1.93, "learning_rate": 3.836705747438625e-05, "loss": 0.6721, "step": 32000 }, { "epoch": 1.93, "learning_rate": 3.836554856427203e-05, "loss": 0.8596, "step": 32002 }, { "epoch": 1.93, "learning_rate": 3.83640396541578e-05, "loss": 0.5535, "step": 32004 }, { "epoch": 1.93, "learning_rate": 3.836253074404358e-05, "loss": 0.7494, "step": 32006 }, { "epoch": 1.93, "learning_rate": 3.836102183392935e-05, "loss": 0.6401, "step": 32008 }, { "epoch": 1.93, "learning_rate": 3.8359512923815126e-05, "loss": 0.6736, "step": 32010 }, { "epoch": 1.93, "learning_rate": 3.8358004013700906e-05, "loss": 0.8099, "step": 32012 }, { "epoch": 1.93, "learning_rate": 3.835649510358668e-05, "loss": 0.8478, "step": 32014 }, { "epoch": 1.93, "learning_rate": 3.835498619347246e-05, "loss": 0.9669, "step": 32016 }, { "epoch": 1.93, "learning_rate": 3.835347728335823e-05, "loss": 0.9176, "step": 32018 }, { "epoch": 1.93, "learning_rate": 3.8351968373244005e-05, "loss": 0.7357, "step": 32020 }, { "epoch": 1.93, "learning_rate": 3.8350459463129785e-05, "loss": 0.6108, "step": 32022 }, { "epoch": 1.93, "learning_rate": 3.834895055301556e-05, "loss": 0.7181, "step": 32024 }, { "epoch": 1.93, "learning_rate": 3.834744164290133e-05, "loss": 0.7005, "step": 32026 }, { "epoch": 1.93, "learning_rate": 3.834593273278711e-05, "loss": 0.5724, "step": 32028 }, { "epoch": 1.93, "learning_rate": 3.8344423822672884e-05, "loss": 0.8031, "step": 32030 }, { "epoch": 1.93, "learning_rate": 3.834291491255866e-05, "loss": 0.653, "step": 32032 }, { "epoch": 1.93, "learning_rate": 3.834140600244444e-05, "loss": 0.7546, "step": 32034 }, { "epoch": 1.93, "learning_rate": 3.833989709233021e-05, "loss": 0.7624, "step": 32036 }, { "epoch": 1.93, "learning_rate": 3.833838818221599e-05, "loss": 0.7763, "step": 32038 }, { "epoch": 1.93, "learning_rate": 3.833687927210176e-05, "loss": 0.6007, "step": 32040 }, { "epoch": 1.93, "learning_rate": 3.8335370361987536e-05, "loss": 0.9185, "step": 32042 }, { "epoch": 1.93, "learning_rate": 3.8333861451873316e-05, "loss": 0.6878, "step": 32044 }, { "epoch": 1.93, "learning_rate": 3.833235254175909e-05, "loss": 0.917, "step": 32046 }, { "epoch": 1.93, "learning_rate": 3.833084363164486e-05, "loss": 0.7012, "step": 32048 }, { "epoch": 1.93, "learning_rate": 3.832933472153064e-05, "loss": 0.7983, "step": 32050 }, { "epoch": 1.93, "learning_rate": 3.8327825811416416e-05, "loss": 0.7804, "step": 32052 }, { "epoch": 1.93, "learning_rate": 3.832631690130219e-05, "loss": 0.7699, "step": 32054 }, { "epoch": 1.93, "learning_rate": 3.832480799118797e-05, "loss": 0.6743, "step": 32056 }, { "epoch": 1.93, "learning_rate": 3.832329908107374e-05, "loss": 0.6502, "step": 32058 }, { "epoch": 1.94, "learning_rate": 3.8321790170959515e-05, "loss": 0.6376, "step": 32060 }, { "epoch": 1.94, "learning_rate": 3.8320281260845295e-05, "loss": 0.8631, "step": 32062 }, { "epoch": 1.94, "learning_rate": 3.831877235073107e-05, "loss": 0.6893, "step": 32064 }, { "epoch": 1.94, "learning_rate": 3.831726344061685e-05, "loss": 0.7233, "step": 32066 }, { "epoch": 1.94, "learning_rate": 3.831575453050262e-05, "loss": 0.8646, "step": 32068 }, { "epoch": 1.94, "learning_rate": 3.8314245620388394e-05, "loss": 0.5625, "step": 32070 }, { "epoch": 1.94, "learning_rate": 3.8312736710274174e-05, "loss": 0.4408, "step": 32072 }, { "epoch": 1.94, "learning_rate": 3.831122780015995e-05, "loss": 0.8194, "step": 32074 }, { "epoch": 1.94, "learning_rate": 3.830971889004572e-05, "loss": 0.597, "step": 32076 }, { "epoch": 1.94, "learning_rate": 3.83082099799315e-05, "loss": 0.6252, "step": 32078 }, { "epoch": 1.94, "learning_rate": 3.830670106981727e-05, "loss": 0.6344, "step": 32080 }, { "epoch": 1.94, "learning_rate": 3.8305192159703046e-05, "loss": 0.8802, "step": 32082 }, { "epoch": 1.94, "learning_rate": 3.8303683249588826e-05, "loss": 1.0132, "step": 32084 }, { "epoch": 1.94, "learning_rate": 3.83021743394746e-05, "loss": 0.7873, "step": 32086 }, { "epoch": 1.94, "learning_rate": 3.830066542936038e-05, "loss": 0.4256, "step": 32088 }, { "epoch": 1.94, "learning_rate": 3.8299156519246146e-05, "loss": 0.7817, "step": 32090 }, { "epoch": 1.94, "learning_rate": 3.8297647609131925e-05, "loss": 0.5293, "step": 32092 }, { "epoch": 1.94, "learning_rate": 3.8296138699017705e-05, "loss": 0.9074, "step": 32094 }, { "epoch": 1.94, "learning_rate": 3.829462978890347e-05, "loss": 0.6899, "step": 32096 }, { "epoch": 1.94, "learning_rate": 3.829312087878925e-05, "loss": 0.7574, "step": 32098 }, { "epoch": 1.94, "learning_rate": 3.829161196867503e-05, "loss": 0.7662, "step": 32100 }, { "epoch": 1.94, "learning_rate": 3.8290103058560805e-05, "loss": 0.7639, "step": 32102 }, { "epoch": 1.94, "learning_rate": 3.828859414844658e-05, "loss": 0.7063, "step": 32104 }, { "epoch": 1.94, "learning_rate": 3.828708523833235e-05, "loss": 0.7538, "step": 32106 }, { "epoch": 1.94, "learning_rate": 3.828557632821813e-05, "loss": 0.5058, "step": 32108 }, { "epoch": 1.94, "learning_rate": 3.8284067418103904e-05, "loss": 0.677, "step": 32110 }, { "epoch": 1.94, "learning_rate": 3.828255850798968e-05, "loss": 0.8418, "step": 32112 }, { "epoch": 1.94, "learning_rate": 3.828104959787546e-05, "loss": 0.6637, "step": 32114 }, { "epoch": 1.94, "learning_rate": 3.827954068776124e-05, "loss": 0.7607, "step": 32116 }, { "epoch": 1.94, "learning_rate": 3.8278031777647e-05, "loss": 1.1993, "step": 32118 }, { "epoch": 1.94, "learning_rate": 3.827652286753278e-05, "loss": 0.7328, "step": 32120 }, { "epoch": 1.94, "learning_rate": 3.8275013957418556e-05, "loss": 0.7474, "step": 32122 }, { "epoch": 1.94, "learning_rate": 3.8273505047304336e-05, "loss": 0.4334, "step": 32124 }, { "epoch": 1.94, "learning_rate": 3.827199613719011e-05, "loss": 0.6487, "step": 32126 }, { "epoch": 1.94, "learning_rate": 3.827048722707588e-05, "loss": 0.6755, "step": 32128 }, { "epoch": 1.94, "learning_rate": 3.826897831696166e-05, "loss": 0.7446, "step": 32130 }, { "epoch": 1.94, "learning_rate": 3.8267469406847435e-05, "loss": 0.8574, "step": 32132 }, { "epoch": 1.94, "learning_rate": 3.826596049673321e-05, "loss": 0.5986, "step": 32134 }, { "epoch": 1.94, "learning_rate": 3.826445158661899e-05, "loss": 0.8606, "step": 32136 }, { "epoch": 1.94, "learning_rate": 3.826294267650476e-05, "loss": 1.1264, "step": 32138 }, { "epoch": 1.94, "learning_rate": 3.8261433766390535e-05, "loss": 0.8804, "step": 32140 }, { "epoch": 1.94, "learning_rate": 3.8259924856276315e-05, "loss": 0.5572, "step": 32142 }, { "epoch": 1.94, "learning_rate": 3.825841594616209e-05, "loss": 0.7517, "step": 32144 }, { "epoch": 1.94, "learning_rate": 3.825690703604786e-05, "loss": 0.7976, "step": 32146 }, { "epoch": 1.94, "learning_rate": 3.825539812593364e-05, "loss": 0.5773, "step": 32148 }, { "epoch": 1.94, "learning_rate": 3.8253889215819414e-05, "loss": 0.9227, "step": 32150 }, { "epoch": 1.94, "learning_rate": 3.8252380305705194e-05, "loss": 0.4806, "step": 32152 }, { "epoch": 1.94, "learning_rate": 3.825087139559097e-05, "loss": 0.6859, "step": 32154 }, { "epoch": 1.94, "learning_rate": 3.824936248547674e-05, "loss": 0.8839, "step": 32156 }, { "epoch": 1.94, "learning_rate": 3.824785357536252e-05, "loss": 0.9772, "step": 32158 }, { "epoch": 1.94, "learning_rate": 3.824634466524829e-05, "loss": 0.7242, "step": 32160 }, { "epoch": 1.94, "learning_rate": 3.8244835755134066e-05, "loss": 0.6795, "step": 32162 }, { "epoch": 1.94, "learning_rate": 3.8243326845019846e-05, "loss": 0.6669, "step": 32164 }, { "epoch": 1.94, "learning_rate": 3.824181793490562e-05, "loss": 0.8756, "step": 32166 }, { "epoch": 1.94, "learning_rate": 3.824030902479139e-05, "loss": 0.9928, "step": 32168 }, { "epoch": 1.94, "learning_rate": 3.823880011467717e-05, "loss": 0.7702, "step": 32170 }, { "epoch": 1.94, "learning_rate": 3.8237291204562945e-05, "loss": 0.8835, "step": 32172 }, { "epoch": 1.94, "learning_rate": 3.8235782294448725e-05, "loss": 0.7508, "step": 32174 }, { "epoch": 1.94, "learning_rate": 3.82342733843345e-05, "loss": 0.7243, "step": 32176 }, { "epoch": 1.94, "learning_rate": 3.823276447422027e-05, "loss": 0.6347, "step": 32178 }, { "epoch": 1.94, "learning_rate": 3.823125556410605e-05, "loss": 0.925, "step": 32180 }, { "epoch": 1.94, "learning_rate": 3.8229746653991825e-05, "loss": 0.6698, "step": 32182 }, { "epoch": 1.94, "learning_rate": 3.82282377438776e-05, "loss": 0.7988, "step": 32184 }, { "epoch": 1.94, "learning_rate": 3.822672883376338e-05, "loss": 0.5538, "step": 32186 }, { "epoch": 1.94, "learning_rate": 3.822521992364915e-05, "loss": 1.0798, "step": 32188 }, { "epoch": 1.94, "learning_rate": 3.8223711013534924e-05, "loss": 0.9669, "step": 32190 }, { "epoch": 1.94, "learning_rate": 3.8222202103420704e-05, "loss": 0.7932, "step": 32192 }, { "epoch": 1.94, "learning_rate": 3.822069319330648e-05, "loss": 0.6262, "step": 32194 }, { "epoch": 1.94, "learning_rate": 3.821918428319225e-05, "loss": 0.8194, "step": 32196 }, { "epoch": 1.94, "learning_rate": 3.821767537307803e-05, "loss": 0.6868, "step": 32198 }, { "epoch": 1.94, "learning_rate": 3.82161664629638e-05, "loss": 0.5396, "step": 32200 }, { "epoch": 1.94, "learning_rate": 3.821465755284958e-05, "loss": 0.8207, "step": 32202 }, { "epoch": 1.94, "learning_rate": 3.821314864273535e-05, "loss": 0.8915, "step": 32204 }, { "epoch": 1.94, "learning_rate": 3.821163973262113e-05, "loss": 0.6345, "step": 32206 }, { "epoch": 1.94, "learning_rate": 3.821013082250691e-05, "loss": 0.5401, "step": 32208 }, { "epoch": 1.94, "learning_rate": 3.8208621912392675e-05, "loss": 0.4116, "step": 32210 }, { "epoch": 1.94, "learning_rate": 3.8207113002278455e-05, "loss": 0.9323, "step": 32212 }, { "epoch": 1.94, "learning_rate": 3.8205604092164235e-05, "loss": 0.8219, "step": 32214 }, { "epoch": 1.94, "learning_rate": 3.820409518205001e-05, "loss": 0.6123, "step": 32216 }, { "epoch": 1.94, "learning_rate": 3.820258627193578e-05, "loss": 0.8201, "step": 32218 }, { "epoch": 1.94, "learning_rate": 3.8201077361821555e-05, "loss": 0.7289, "step": 32220 }, { "epoch": 1.94, "learning_rate": 3.8199568451707335e-05, "loss": 0.5132, "step": 32222 }, { "epoch": 1.94, "learning_rate": 3.819805954159311e-05, "loss": 0.6816, "step": 32224 }, { "epoch": 1.95, "learning_rate": 3.819655063147888e-05, "loss": 0.7712, "step": 32226 }, { "epoch": 1.95, "learning_rate": 3.819504172136466e-05, "loss": 0.6566, "step": 32228 }, { "epoch": 1.95, "learning_rate": 3.819353281125044e-05, "loss": 0.8057, "step": 32230 }, { "epoch": 1.95, "learning_rate": 3.819202390113621e-05, "loss": 0.785, "step": 32232 }, { "epoch": 1.95, "learning_rate": 3.819051499102199e-05, "loss": 0.7171, "step": 32234 }, { "epoch": 1.95, "learning_rate": 3.818900608090776e-05, "loss": 0.6513, "step": 32236 }, { "epoch": 1.95, "learning_rate": 3.818749717079354e-05, "loss": 0.6377, "step": 32238 }, { "epoch": 1.95, "learning_rate": 3.818598826067931e-05, "loss": 0.8655, "step": 32240 }, { "epoch": 1.95, "learning_rate": 3.8184479350565086e-05, "loss": 0.725, "step": 32242 }, { "epoch": 1.95, "learning_rate": 3.8182970440450866e-05, "loss": 0.6973, "step": 32244 }, { "epoch": 1.95, "learning_rate": 3.818146153033664e-05, "loss": 0.9818, "step": 32246 }, { "epoch": 1.95, "learning_rate": 3.817995262022241e-05, "loss": 0.7658, "step": 32248 }, { "epoch": 1.95, "learning_rate": 3.817844371010819e-05, "loss": 0.6255, "step": 32250 }, { "epoch": 1.95, "learning_rate": 3.8176934799993965e-05, "loss": 0.8366, "step": 32252 }, { "epoch": 1.95, "learning_rate": 3.817542588987974e-05, "loss": 0.9668, "step": 32254 }, { "epoch": 1.95, "learning_rate": 3.817391697976552e-05, "loss": 0.5882, "step": 32256 }, { "epoch": 1.95, "learning_rate": 3.817240806965129e-05, "loss": 0.7831, "step": 32258 }, { "epoch": 1.95, "learning_rate": 3.8170899159537065e-05, "loss": 0.7394, "step": 32260 }, { "epoch": 1.95, "learning_rate": 3.8169390249422844e-05, "loss": 0.7783, "step": 32262 }, { "epoch": 1.95, "learning_rate": 3.816788133930862e-05, "loss": 0.7599, "step": 32264 }, { "epoch": 1.95, "learning_rate": 3.81663724291944e-05, "loss": 0.6058, "step": 32266 }, { "epoch": 1.95, "learning_rate": 3.816486351908017e-05, "loss": 0.7196, "step": 32268 }, { "epoch": 1.95, "learning_rate": 3.8163354608965944e-05, "loss": 0.5351, "step": 32270 }, { "epoch": 1.95, "learning_rate": 3.8161845698851724e-05, "loss": 0.7143, "step": 32272 }, { "epoch": 1.95, "learning_rate": 3.81603367887375e-05, "loss": 0.7661, "step": 32274 }, { "epoch": 1.95, "learning_rate": 3.815882787862327e-05, "loss": 1.0953, "step": 32276 }, { "epoch": 1.95, "learning_rate": 3.815731896850905e-05, "loss": 0.8542, "step": 32278 }, { "epoch": 1.95, "learning_rate": 3.815581005839482e-05, "loss": 0.846, "step": 32280 }, { "epoch": 1.95, "learning_rate": 3.8154301148280596e-05, "loss": 0.815, "step": 32282 }, { "epoch": 1.95, "learning_rate": 3.8152792238166376e-05, "loss": 0.5912, "step": 32284 }, { "epoch": 1.95, "learning_rate": 3.815128332805215e-05, "loss": 0.7001, "step": 32286 }, { "epoch": 1.95, "learning_rate": 3.814977441793793e-05, "loss": 0.6879, "step": 32288 }, { "epoch": 1.95, "learning_rate": 3.81482655078237e-05, "loss": 0.8497, "step": 32290 }, { "epoch": 1.95, "learning_rate": 3.8146756597709475e-05, "loss": 0.5658, "step": 32292 }, { "epoch": 1.95, "learning_rate": 3.8145247687595255e-05, "loss": 0.6838, "step": 32294 }, { "epoch": 1.95, "learning_rate": 3.814373877748103e-05, "loss": 0.6903, "step": 32296 }, { "epoch": 1.95, "learning_rate": 3.81422298673668e-05, "loss": 0.6431, "step": 32298 }, { "epoch": 1.95, "learning_rate": 3.814072095725258e-05, "loss": 0.7118, "step": 32300 }, { "epoch": 1.95, "learning_rate": 3.8139212047138354e-05, "loss": 0.8379, "step": 32302 }, { "epoch": 1.95, "learning_rate": 3.813770313702413e-05, "loss": 0.8426, "step": 32304 }, { "epoch": 1.95, "learning_rate": 3.813619422690991e-05, "loss": 0.7694, "step": 32306 }, { "epoch": 1.95, "learning_rate": 3.813468531679568e-05, "loss": 0.868, "step": 32308 }, { "epoch": 1.95, "learning_rate": 3.8133176406681454e-05, "loss": 0.6332, "step": 32310 }, { "epoch": 1.95, "learning_rate": 3.8131667496567234e-05, "loss": 0.9897, "step": 32312 }, { "epoch": 1.95, "learning_rate": 3.813015858645301e-05, "loss": 0.6523, "step": 32314 }, { "epoch": 1.95, "learning_rate": 3.8128649676338787e-05, "loss": 0.9634, "step": 32316 }, { "epoch": 1.95, "learning_rate": 3.812714076622455e-05, "loss": 1.1142, "step": 32318 }, { "epoch": 1.95, "learning_rate": 3.812563185611033e-05, "loss": 0.7533, "step": 32320 }, { "epoch": 1.95, "learning_rate": 3.812412294599611e-05, "loss": 0.7882, "step": 32322 }, { "epoch": 1.95, "learning_rate": 3.812261403588188e-05, "loss": 0.8418, "step": 32324 }, { "epoch": 1.95, "learning_rate": 3.812110512576766e-05, "loss": 1.0373, "step": 32326 }, { "epoch": 1.95, "learning_rate": 3.811959621565344e-05, "loss": 0.6006, "step": 32328 }, { "epoch": 1.95, "learning_rate": 3.811808730553921e-05, "loss": 0.5851, "step": 32330 }, { "epoch": 1.95, "learning_rate": 3.8116578395424985e-05, "loss": 0.647, "step": 32332 }, { "epoch": 1.95, "learning_rate": 3.811506948531076e-05, "loss": 0.8136, "step": 32334 }, { "epoch": 1.95, "learning_rate": 3.811356057519654e-05, "loss": 0.6214, "step": 32336 }, { "epoch": 1.95, "learning_rate": 3.811205166508232e-05, "loss": 0.9095, "step": 32338 }, { "epoch": 1.95, "learning_rate": 3.8110542754968084e-05, "loss": 0.5556, "step": 32340 }, { "epoch": 1.95, "learning_rate": 3.8109033844853864e-05, "loss": 0.9081, "step": 32342 }, { "epoch": 1.95, "learning_rate": 3.8107524934739644e-05, "loss": 0.6741, "step": 32344 }, { "epoch": 1.95, "learning_rate": 3.810601602462541e-05, "loss": 0.8829, "step": 32346 }, { "epoch": 1.95, "learning_rate": 3.810450711451119e-05, "loss": 0.8192, "step": 32348 }, { "epoch": 1.95, "learning_rate": 3.8102998204396964e-05, "loss": 0.5429, "step": 32350 }, { "epoch": 1.95, "learning_rate": 3.8101489294282744e-05, "loss": 0.6569, "step": 32352 }, { "epoch": 1.95, "learning_rate": 3.809998038416852e-05, "loss": 1.0997, "step": 32354 }, { "epoch": 1.95, "learning_rate": 3.809847147405429e-05, "loss": 1.1858, "step": 32356 }, { "epoch": 1.95, "learning_rate": 3.809696256394007e-05, "loss": 0.5457, "step": 32358 }, { "epoch": 1.95, "learning_rate": 3.809545365382584e-05, "loss": 0.7994, "step": 32360 }, { "epoch": 1.95, "learning_rate": 3.8093944743711616e-05, "loss": 1.0664, "step": 32362 }, { "epoch": 1.95, "learning_rate": 3.8092435833597396e-05, "loss": 0.4461, "step": 32364 }, { "epoch": 1.95, "learning_rate": 3.809092692348317e-05, "loss": 0.8287, "step": 32366 }, { "epoch": 1.95, "learning_rate": 3.808941801336894e-05, "loss": 0.8653, "step": 32368 }, { "epoch": 1.95, "learning_rate": 3.808790910325472e-05, "loss": 1.0903, "step": 32370 }, { "epoch": 1.95, "learning_rate": 3.8086400193140495e-05, "loss": 1.0027, "step": 32372 }, { "epoch": 1.95, "learning_rate": 3.808489128302627e-05, "loss": 0.7783, "step": 32374 }, { "epoch": 1.95, "learning_rate": 3.808338237291205e-05, "loss": 0.7083, "step": 32376 }, { "epoch": 1.95, "learning_rate": 3.808187346279782e-05, "loss": 0.8654, "step": 32378 }, { "epoch": 1.95, "learning_rate": 3.80803645526836e-05, "loss": 0.7611, "step": 32380 }, { "epoch": 1.95, "learning_rate": 3.8078855642569374e-05, "loss": 0.8201, "step": 32382 }, { "epoch": 1.95, "learning_rate": 3.807734673245515e-05, "loss": 0.9389, "step": 32384 }, { "epoch": 1.95, "learning_rate": 3.807583782234093e-05, "loss": 0.5582, "step": 32386 }, { "epoch": 1.95, "learning_rate": 3.80743289122267e-05, "loss": 0.6527, "step": 32388 }, { "epoch": 1.95, "learning_rate": 3.8072820002112474e-05, "loss": 0.8616, "step": 32390 }, { "epoch": 1.96, "learning_rate": 3.8071311091998253e-05, "loss": 0.8674, "step": 32392 }, { "epoch": 1.96, "learning_rate": 3.8069802181884027e-05, "loss": 0.6628, "step": 32394 }, { "epoch": 1.96, "learning_rate": 3.80682932717698e-05, "loss": 0.8873, "step": 32396 }, { "epoch": 1.96, "learning_rate": 3.806678436165558e-05, "loss": 0.461, "step": 32398 }, { "epoch": 1.96, "learning_rate": 3.806527545154135e-05, "loss": 0.5399, "step": 32400 }, { "epoch": 1.96, "learning_rate": 3.806376654142713e-05, "loss": 0.754, "step": 32402 }, { "epoch": 1.96, "learning_rate": 3.8062257631312906e-05, "loss": 1.046, "step": 32404 }, { "epoch": 1.96, "learning_rate": 3.806074872119868e-05, "loss": 1.3037, "step": 32406 }, { "epoch": 1.96, "learning_rate": 3.805923981108446e-05, "loss": 0.6372, "step": 32408 }, { "epoch": 1.96, "learning_rate": 3.805773090097023e-05, "loss": 0.6019, "step": 32410 }, { "epoch": 1.96, "learning_rate": 3.8056221990856005e-05, "loss": 0.7796, "step": 32412 }, { "epoch": 1.96, "learning_rate": 3.8054713080741785e-05, "loss": 0.7969, "step": 32414 }, { "epoch": 1.96, "learning_rate": 3.805320417062756e-05, "loss": 0.9909, "step": 32416 }, { "epoch": 1.96, "learning_rate": 3.805169526051333e-05, "loss": 0.8346, "step": 32418 }, { "epoch": 1.96, "learning_rate": 3.805018635039911e-05, "loss": 0.8723, "step": 32420 }, { "epoch": 1.96, "learning_rate": 3.8048677440284884e-05, "loss": 0.7523, "step": 32422 }, { "epoch": 1.96, "learning_rate": 3.804716853017066e-05, "loss": 0.6388, "step": 32424 }, { "epoch": 1.96, "learning_rate": 3.804565962005644e-05, "loss": 0.8149, "step": 32426 }, { "epoch": 1.96, "learning_rate": 3.804415070994221e-05, "loss": 0.6549, "step": 32428 }, { "epoch": 1.96, "learning_rate": 3.804264179982799e-05, "loss": 0.7597, "step": 32430 }, { "epoch": 1.96, "learning_rate": 3.804113288971376e-05, "loss": 0.5722, "step": 32432 }, { "epoch": 1.96, "learning_rate": 3.8039623979599537e-05, "loss": 0.8303, "step": 32434 }, { "epoch": 1.96, "learning_rate": 3.8038115069485316e-05, "loss": 0.7397, "step": 32436 }, { "epoch": 1.96, "learning_rate": 3.803660615937108e-05, "loss": 0.7534, "step": 32438 }, { "epoch": 1.96, "learning_rate": 3.803509724925686e-05, "loss": 1.0149, "step": 32440 }, { "epoch": 1.96, "learning_rate": 3.803358833914264e-05, "loss": 0.7735, "step": 32442 }, { "epoch": 1.96, "learning_rate": 3.8032079429028416e-05, "loss": 0.9051, "step": 32444 }, { "epoch": 1.96, "learning_rate": 3.803057051891419e-05, "loss": 0.8233, "step": 32446 }, { "epoch": 1.96, "learning_rate": 3.802906160879996e-05, "loss": 0.6716, "step": 32448 }, { "epoch": 1.96, "learning_rate": 3.802755269868574e-05, "loss": 0.6566, "step": 32450 }, { "epoch": 1.96, "learning_rate": 3.802604378857152e-05, "loss": 0.8053, "step": 32452 }, { "epoch": 1.96, "learning_rate": 3.802453487845729e-05, "loss": 0.6806, "step": 32454 }, { "epoch": 1.96, "learning_rate": 3.802302596834307e-05, "loss": 1.0879, "step": 32456 }, { "epoch": 1.96, "learning_rate": 3.802151705822885e-05, "loss": 0.6469, "step": 32458 }, { "epoch": 1.96, "learning_rate": 3.8020008148114614e-05, "loss": 0.5425, "step": 32460 }, { "epoch": 1.96, "learning_rate": 3.8018499238000394e-05, "loss": 0.5557, "step": 32462 }, { "epoch": 1.96, "learning_rate": 3.801699032788617e-05, "loss": 0.713, "step": 32464 }, { "epoch": 1.96, "learning_rate": 3.801548141777195e-05, "loss": 1.0747, "step": 32466 }, { "epoch": 1.96, "learning_rate": 3.801397250765772e-05, "loss": 0.844, "step": 32468 }, { "epoch": 1.96, "learning_rate": 3.8012463597543493e-05, "loss": 0.6948, "step": 32470 }, { "epoch": 1.96, "learning_rate": 3.801095468742927e-05, "loss": 0.552, "step": 32472 }, { "epoch": 1.96, "learning_rate": 3.8009445777315046e-05, "loss": 1.0293, "step": 32474 }, { "epoch": 1.96, "learning_rate": 3.800793686720082e-05, "loss": 0.9916, "step": 32476 }, { "epoch": 1.96, "learning_rate": 3.80064279570866e-05, "loss": 0.7209, "step": 32478 }, { "epoch": 1.96, "learning_rate": 3.800491904697237e-05, "loss": 0.7502, "step": 32480 }, { "epoch": 1.96, "learning_rate": 3.8003410136858146e-05, "loss": 0.8308, "step": 32482 }, { "epoch": 1.96, "learning_rate": 3.8001901226743926e-05, "loss": 0.6959, "step": 32484 }, { "epoch": 1.96, "learning_rate": 3.80003923166297e-05, "loss": 0.7513, "step": 32486 }, { "epoch": 1.96, "learning_rate": 3.799888340651547e-05, "loss": 0.7416, "step": 32488 }, { "epoch": 1.96, "learning_rate": 3.799737449640125e-05, "loss": 0.7803, "step": 32490 }, { "epoch": 1.96, "learning_rate": 3.7995865586287025e-05, "loss": 0.8563, "step": 32492 }, { "epoch": 1.96, "learning_rate": 3.7994356676172805e-05, "loss": 0.8034, "step": 32494 }, { "epoch": 1.96, "learning_rate": 3.799284776605858e-05, "loss": 0.6668, "step": 32496 }, { "epoch": 1.96, "learning_rate": 3.799133885594435e-05, "loss": 0.8124, "step": 32498 }, { "epoch": 1.96, "learning_rate": 3.798982994583013e-05, "loss": 0.8207, "step": 32500 }, { "epoch": 1.96, "learning_rate": 3.7988321035715904e-05, "loss": 0.7782, "step": 32502 }, { "epoch": 1.96, "learning_rate": 3.798681212560168e-05, "loss": 1.0326, "step": 32504 }, { "epoch": 1.96, "learning_rate": 3.798530321548746e-05, "loss": 0.7692, "step": 32506 }, { "epoch": 1.96, "learning_rate": 3.798379430537323e-05, "loss": 0.6362, "step": 32508 }, { "epoch": 1.96, "learning_rate": 3.7982285395259e-05, "loss": 0.8926, "step": 32510 }, { "epoch": 1.96, "learning_rate": 3.798077648514478e-05, "loss": 0.8919, "step": 32512 }, { "epoch": 1.96, "learning_rate": 3.7979267575030556e-05, "loss": 1.0362, "step": 32514 }, { "epoch": 1.96, "learning_rate": 3.7977758664916336e-05, "loss": 0.7673, "step": 32516 }, { "epoch": 1.96, "learning_rate": 3.797624975480211e-05, "loss": 0.8744, "step": 32518 }, { "epoch": 1.96, "learning_rate": 3.797474084468788e-05, "loss": 0.8351, "step": 32520 }, { "epoch": 1.96, "learning_rate": 3.797323193457366e-05, "loss": 1.0608, "step": 32522 }, { "epoch": 1.96, "learning_rate": 3.7971723024459436e-05, "loss": 0.7494, "step": 32524 }, { "epoch": 1.96, "learning_rate": 3.797021411434521e-05, "loss": 0.7459, "step": 32526 }, { "epoch": 1.96, "learning_rate": 3.796870520423099e-05, "loss": 0.6507, "step": 32528 }, { "epoch": 1.96, "learning_rate": 3.796719629411676e-05, "loss": 0.7573, "step": 32530 }, { "epoch": 1.96, "learning_rate": 3.7965687384002535e-05, "loss": 0.6743, "step": 32532 }, { "epoch": 1.96, "learning_rate": 3.7964178473888315e-05, "loss": 0.7678, "step": 32534 }, { "epoch": 1.96, "learning_rate": 3.796266956377409e-05, "loss": 0.7629, "step": 32536 }, { "epoch": 1.96, "learning_rate": 3.796116065365986e-05, "loss": 0.8231, "step": 32538 }, { "epoch": 1.96, "learning_rate": 3.795965174354564e-05, "loss": 0.6905, "step": 32540 }, { "epoch": 1.96, "learning_rate": 3.7958142833431414e-05, "loss": 0.7373, "step": 32542 }, { "epoch": 1.96, "learning_rate": 3.7956633923317194e-05, "loss": 0.7944, "step": 32544 }, { "epoch": 1.96, "learning_rate": 3.795512501320296e-05, "loss": 0.7207, "step": 32546 }, { "epoch": 1.96, "learning_rate": 3.795361610308874e-05, "loss": 0.5038, "step": 32548 }, { "epoch": 1.96, "learning_rate": 3.795210719297452e-05, "loss": 0.7323, "step": 32550 }, { "epoch": 1.96, "learning_rate": 3.7950598282860286e-05, "loss": 0.5194, "step": 32552 }, { "epoch": 1.96, "learning_rate": 3.7949089372746066e-05, "loss": 0.8028, "step": 32554 }, { "epoch": 1.96, "learning_rate": 3.7947580462631846e-05, "loss": 0.771, "step": 32556 }, { "epoch": 1.97, "learning_rate": 3.794607155251762e-05, "loss": 0.7364, "step": 32558 }, { "epoch": 1.97, "learning_rate": 3.794456264240339e-05, "loss": 0.8558, "step": 32560 }, { "epoch": 1.97, "learning_rate": 3.7943053732289166e-05, "loss": 0.8188, "step": 32562 }, { "epoch": 1.97, "learning_rate": 3.7941544822174946e-05, "loss": 0.9529, "step": 32564 }, { "epoch": 1.97, "learning_rate": 3.7940035912060725e-05, "loss": 0.7765, "step": 32566 }, { "epoch": 1.97, "learning_rate": 3.793852700194649e-05, "loss": 0.8279, "step": 32568 }, { "epoch": 1.97, "learning_rate": 3.793701809183227e-05, "loss": 0.6374, "step": 32570 }, { "epoch": 1.97, "learning_rate": 3.793550918171805e-05, "loss": 0.8279, "step": 32572 }, { "epoch": 1.97, "learning_rate": 3.793400027160382e-05, "loss": 0.7217, "step": 32574 }, { "epoch": 1.97, "learning_rate": 3.79324913614896e-05, "loss": 0.6005, "step": 32576 }, { "epoch": 1.97, "learning_rate": 3.793098245137537e-05, "loss": 0.8523, "step": 32578 }, { "epoch": 1.97, "learning_rate": 3.792947354126115e-05, "loss": 0.7026, "step": 32580 }, { "epoch": 1.97, "learning_rate": 3.7927964631146924e-05, "loss": 0.7964, "step": 32582 }, { "epoch": 1.97, "learning_rate": 3.79264557210327e-05, "loss": 0.7021, "step": 32584 }, { "epoch": 1.97, "learning_rate": 3.792494681091848e-05, "loss": 0.6203, "step": 32586 }, { "epoch": 1.97, "learning_rate": 3.792343790080425e-05, "loss": 0.9565, "step": 32588 }, { "epoch": 1.97, "learning_rate": 3.792192899069002e-05, "loss": 0.5439, "step": 32590 }, { "epoch": 1.97, "learning_rate": 3.79204200805758e-05, "loss": 0.8201, "step": 32592 }, { "epoch": 1.97, "learning_rate": 3.7918911170461576e-05, "loss": 0.708, "step": 32594 }, { "epoch": 1.97, "learning_rate": 3.791740226034735e-05, "loss": 0.6378, "step": 32596 }, { "epoch": 1.97, "learning_rate": 3.791589335023313e-05, "loss": 1.0031, "step": 32598 }, { "epoch": 1.97, "learning_rate": 3.79143844401189e-05, "loss": 0.5025, "step": 32600 }, { "epoch": 1.97, "learning_rate": 3.7912875530004676e-05, "loss": 0.6599, "step": 32602 }, { "epoch": 1.97, "learning_rate": 3.7911366619890455e-05, "loss": 0.9674, "step": 32604 }, { "epoch": 1.97, "learning_rate": 3.790985770977623e-05, "loss": 0.7536, "step": 32606 }, { "epoch": 1.97, "learning_rate": 3.790834879966201e-05, "loss": 0.4997, "step": 32608 }, { "epoch": 1.97, "learning_rate": 3.790683988954778e-05, "loss": 0.4945, "step": 32610 }, { "epoch": 1.97, "learning_rate": 3.7905330979433555e-05, "loss": 0.84, "step": 32612 }, { "epoch": 1.97, "learning_rate": 3.7903822069319335e-05, "loss": 0.6794, "step": 32614 }, { "epoch": 1.97, "learning_rate": 3.790231315920511e-05, "loss": 0.7834, "step": 32616 }, { "epoch": 1.97, "learning_rate": 3.790080424909088e-05, "loss": 0.9081, "step": 32618 }, { "epoch": 1.97, "learning_rate": 3.789929533897666e-05, "loss": 1.0855, "step": 32620 }, { "epoch": 1.97, "learning_rate": 3.7897786428862434e-05, "loss": 0.7286, "step": 32622 }, { "epoch": 1.97, "learning_rate": 3.789627751874821e-05, "loss": 0.7555, "step": 32624 }, { "epoch": 1.97, "learning_rate": 3.789476860863399e-05, "loss": 0.9772, "step": 32626 }, { "epoch": 1.97, "learning_rate": 3.789325969851976e-05, "loss": 0.7916, "step": 32628 }, { "epoch": 1.97, "learning_rate": 3.789175078840554e-05, "loss": 0.7604, "step": 32630 }, { "epoch": 1.97, "learning_rate": 3.789024187829131e-05, "loss": 1.088, "step": 32632 }, { "epoch": 1.97, "learning_rate": 3.7888732968177086e-05, "loss": 0.5332, "step": 32634 }, { "epoch": 1.97, "learning_rate": 3.7887224058062866e-05, "loss": 0.6368, "step": 32636 }, { "epoch": 1.97, "learning_rate": 3.788571514794864e-05, "loss": 0.7447, "step": 32638 }, { "epoch": 1.97, "learning_rate": 3.788420623783441e-05, "loss": 0.5519, "step": 32640 }, { "epoch": 1.97, "learning_rate": 3.788269732772019e-05, "loss": 0.6922, "step": 32642 }, { "epoch": 1.97, "learning_rate": 3.7881188417605965e-05, "loss": 0.667, "step": 32644 }, { "epoch": 1.97, "learning_rate": 3.787967950749174e-05, "loss": 1.0706, "step": 32646 }, { "epoch": 1.97, "learning_rate": 3.787817059737752e-05, "loss": 0.8096, "step": 32648 }, { "epoch": 1.97, "learning_rate": 3.787666168726329e-05, "loss": 0.5344, "step": 32650 }, { "epoch": 1.97, "learning_rate": 3.7875152777149065e-05, "loss": 0.8213, "step": 32652 }, { "epoch": 1.97, "learning_rate": 3.7873643867034845e-05, "loss": 1.1777, "step": 32654 }, { "epoch": 1.97, "learning_rate": 3.787213495692062e-05, "loss": 0.8234, "step": 32656 }, { "epoch": 1.97, "learning_rate": 3.78706260468064e-05, "loss": 0.7006, "step": 32658 }, { "epoch": 1.97, "learning_rate": 3.7869117136692164e-05, "loss": 0.8588, "step": 32660 }, { "epoch": 1.97, "learning_rate": 3.7867608226577944e-05, "loss": 0.459, "step": 32662 }, { "epoch": 1.97, "learning_rate": 3.7866099316463724e-05, "loss": 0.8164, "step": 32664 }, { "epoch": 1.97, "learning_rate": 3.78645904063495e-05, "loss": 0.766, "step": 32666 }, { "epoch": 1.97, "learning_rate": 3.786308149623527e-05, "loss": 0.8836, "step": 32668 }, { "epoch": 1.97, "learning_rate": 3.786157258612105e-05, "loss": 0.6602, "step": 32670 }, { "epoch": 1.97, "learning_rate": 3.786006367600682e-05, "loss": 0.7874, "step": 32672 }, { "epoch": 1.97, "learning_rate": 3.7858554765892596e-05, "loss": 0.7782, "step": 32674 }, { "epoch": 1.97, "learning_rate": 3.785704585577837e-05, "loss": 0.8855, "step": 32676 }, { "epoch": 1.97, "learning_rate": 3.785553694566415e-05, "loss": 0.7033, "step": 32678 }, { "epoch": 1.97, "learning_rate": 3.785402803554993e-05, "loss": 0.7505, "step": 32680 }, { "epoch": 1.97, "learning_rate": 3.7852519125435695e-05, "loss": 0.7031, "step": 32682 }, { "epoch": 1.97, "learning_rate": 3.7851010215321475e-05, "loss": 0.6884, "step": 32684 }, { "epoch": 1.97, "learning_rate": 3.7849501305207255e-05, "loss": 0.6305, "step": 32686 }, { "epoch": 1.97, "learning_rate": 3.784799239509302e-05, "loss": 0.8967, "step": 32688 }, { "epoch": 1.97, "learning_rate": 3.78464834849788e-05, "loss": 0.7082, "step": 32690 }, { "epoch": 1.97, "learning_rate": 3.7844974574864575e-05, "loss": 0.8559, "step": 32692 }, { "epoch": 1.97, "learning_rate": 3.7843465664750355e-05, "loss": 0.7499, "step": 32694 }, { "epoch": 1.97, "learning_rate": 3.784195675463613e-05, "loss": 0.9781, "step": 32696 }, { "epoch": 1.97, "learning_rate": 3.78404478445219e-05, "loss": 0.5664, "step": 32698 }, { "epoch": 1.97, "learning_rate": 3.783893893440768e-05, "loss": 0.7579, "step": 32700 }, { "epoch": 1.97, "learning_rate": 3.7837430024293454e-05, "loss": 1.1003, "step": 32702 }, { "epoch": 1.97, "learning_rate": 3.783592111417923e-05, "loss": 0.5895, "step": 32704 }, { "epoch": 1.97, "learning_rate": 3.783441220406501e-05, "loss": 0.7787, "step": 32706 }, { "epoch": 1.97, "learning_rate": 3.783290329395078e-05, "loss": 0.6495, "step": 32708 }, { "epoch": 1.97, "learning_rate": 3.783139438383655e-05, "loss": 0.9209, "step": 32710 }, { "epoch": 1.97, "learning_rate": 3.782988547372233e-05, "loss": 0.7234, "step": 32712 }, { "epoch": 1.97, "learning_rate": 3.7828376563608106e-05, "loss": 0.6893, "step": 32714 }, { "epoch": 1.97, "learning_rate": 3.7826867653493886e-05, "loss": 0.6863, "step": 32716 }, { "epoch": 1.97, "learning_rate": 3.782535874337966e-05, "loss": 0.6434, "step": 32718 }, { "epoch": 1.97, "learning_rate": 3.782384983326543e-05, "loss": 0.8355, "step": 32720 }, { "epoch": 1.98, "learning_rate": 3.782234092315121e-05, "loss": 0.6111, "step": 32722 }, { "epoch": 1.98, "learning_rate": 3.7820832013036985e-05, "loss": 0.6982, "step": 32724 }, { "epoch": 1.98, "learning_rate": 3.781932310292276e-05, "loss": 0.68, "step": 32726 }, { "epoch": 1.98, "learning_rate": 3.781781419280854e-05, "loss": 0.7247, "step": 32728 }, { "epoch": 1.98, "learning_rate": 3.781630528269431e-05, "loss": 0.7688, "step": 32730 }, { "epoch": 1.98, "learning_rate": 3.7814796372580085e-05, "loss": 0.7712, "step": 32732 }, { "epoch": 1.98, "learning_rate": 3.7813287462465864e-05, "loss": 0.6554, "step": 32734 }, { "epoch": 1.98, "learning_rate": 3.781177855235164e-05, "loss": 0.9682, "step": 32736 }, { "epoch": 1.98, "learning_rate": 3.781026964223741e-05, "loss": 0.7099, "step": 32738 }, { "epoch": 1.98, "learning_rate": 3.780876073212319e-05, "loss": 0.8371, "step": 32740 }, { "epoch": 1.98, "learning_rate": 3.7807251822008964e-05, "loss": 0.8025, "step": 32742 }, { "epoch": 1.98, "learning_rate": 3.7805742911894744e-05, "loss": 0.5748, "step": 32744 }, { "epoch": 1.98, "learning_rate": 3.780423400178052e-05, "loss": 0.7773, "step": 32746 }, { "epoch": 1.98, "learning_rate": 3.780272509166629e-05, "loss": 0.801, "step": 32748 }, { "epoch": 1.98, "learning_rate": 3.780121618155207e-05, "loss": 0.7091, "step": 32750 }, { "epoch": 1.98, "learning_rate": 3.779970727143784e-05, "loss": 0.7421, "step": 32752 }, { "epoch": 1.98, "learning_rate": 3.7798198361323616e-05, "loss": 0.6613, "step": 32754 }, { "epoch": 1.98, "learning_rate": 3.7796689451209396e-05, "loss": 0.9714, "step": 32756 }, { "epoch": 1.98, "learning_rate": 3.779518054109517e-05, "loss": 0.7485, "step": 32758 }, { "epoch": 1.98, "learning_rate": 3.779367163098094e-05, "loss": 0.7595, "step": 32760 }, { "epoch": 1.98, "learning_rate": 3.779216272086672e-05, "loss": 0.7278, "step": 32762 }, { "epoch": 1.98, "learning_rate": 3.7790653810752495e-05, "loss": 0.6945, "step": 32764 }, { "epoch": 1.98, "learning_rate": 3.778914490063827e-05, "loss": 0.8913, "step": 32766 }, { "epoch": 1.98, "learning_rate": 3.778763599052405e-05, "loss": 0.5816, "step": 32768 }, { "epoch": 1.98, "learning_rate": 3.778612708040982e-05, "loss": 0.5436, "step": 32770 }, { "epoch": 1.98, "learning_rate": 3.77846181702956e-05, "loss": 0.6812, "step": 32772 }, { "epoch": 1.98, "learning_rate": 3.778310926018137e-05, "loss": 0.6686, "step": 32774 }, { "epoch": 1.98, "learning_rate": 3.778160035006715e-05, "loss": 0.5276, "step": 32776 }, { "epoch": 1.98, "learning_rate": 3.778009143995293e-05, "loss": 0.8453, "step": 32778 }, { "epoch": 1.98, "learning_rate": 3.77785825298387e-05, "loss": 1.0025, "step": 32780 }, { "epoch": 1.98, "learning_rate": 3.7777073619724474e-05, "loss": 0.6183, "step": 32782 }, { "epoch": 1.98, "learning_rate": 3.7775564709610254e-05, "loss": 0.6412, "step": 32784 }, { "epoch": 1.98, "learning_rate": 3.777405579949603e-05, "loss": 0.6399, "step": 32786 }, { "epoch": 1.98, "learning_rate": 3.77725468893818e-05, "loss": 0.6899, "step": 32788 }, { "epoch": 1.98, "learning_rate": 3.777103797926757e-05, "loss": 0.7747, "step": 32790 }, { "epoch": 1.98, "learning_rate": 3.776952906915335e-05, "loss": 0.8692, "step": 32792 }, { "epoch": 1.98, "learning_rate": 3.776802015903913e-05, "loss": 0.7922, "step": 32794 }, { "epoch": 1.98, "learning_rate": 3.77665112489249e-05, "loss": 0.5256, "step": 32796 }, { "epoch": 1.98, "learning_rate": 3.776500233881068e-05, "loss": 0.5139, "step": 32798 }, { "epoch": 1.98, "learning_rate": 3.776349342869646e-05, "loss": 0.699, "step": 32800 }, { "epoch": 1.98, "learning_rate": 3.7761984518582225e-05, "loss": 0.656, "step": 32802 }, { "epoch": 1.98, "learning_rate": 3.7760475608468005e-05, "loss": 0.5136, "step": 32804 }, { "epoch": 1.98, "learning_rate": 3.775896669835378e-05, "loss": 0.7364, "step": 32806 }, { "epoch": 1.98, "learning_rate": 3.775745778823956e-05, "loss": 0.5497, "step": 32808 }, { "epoch": 1.98, "learning_rate": 3.775594887812533e-05, "loss": 0.903, "step": 32810 }, { "epoch": 1.98, "learning_rate": 3.7754439968011104e-05, "loss": 0.5586, "step": 32812 }, { "epoch": 1.98, "learning_rate": 3.7752931057896884e-05, "loss": 0.7055, "step": 32814 }, { "epoch": 1.98, "learning_rate": 3.775142214778266e-05, "loss": 0.6166, "step": 32816 }, { "epoch": 1.98, "learning_rate": 3.774991323766843e-05, "loss": 0.913, "step": 32818 }, { "epoch": 1.98, "learning_rate": 3.774840432755421e-05, "loss": 0.9323, "step": 32820 }, { "epoch": 1.98, "learning_rate": 3.7746895417439984e-05, "loss": 0.6317, "step": 32822 }, { "epoch": 1.98, "learning_rate": 3.774538650732576e-05, "loss": 0.8887, "step": 32824 }, { "epoch": 1.98, "learning_rate": 3.774387759721154e-05, "loss": 0.7109, "step": 32826 }, { "epoch": 1.98, "learning_rate": 3.774236868709731e-05, "loss": 0.6621, "step": 32828 }, { "epoch": 1.98, "learning_rate": 3.774085977698309e-05, "loss": 0.6354, "step": 32830 }, { "epoch": 1.98, "learning_rate": 3.773935086686886e-05, "loss": 0.8147, "step": 32832 }, { "epoch": 1.98, "learning_rate": 3.7737841956754636e-05, "loss": 0.5559, "step": 32834 }, { "epoch": 1.98, "learning_rate": 3.7736333046640416e-05, "loss": 0.9506, "step": 32836 }, { "epoch": 1.98, "learning_rate": 3.773482413652619e-05, "loss": 0.7476, "step": 32838 }, { "epoch": 1.98, "learning_rate": 3.773331522641196e-05, "loss": 0.6166, "step": 32840 }, { "epoch": 1.98, "learning_rate": 3.773180631629774e-05, "loss": 0.7057, "step": 32842 }, { "epoch": 1.98, "learning_rate": 3.7730297406183515e-05, "loss": 0.8005, "step": 32844 }, { "epoch": 1.98, "learning_rate": 3.772878849606929e-05, "loss": 0.7564, "step": 32846 }, { "epoch": 1.98, "learning_rate": 3.772727958595507e-05, "loss": 0.8168, "step": 32848 }, { "epoch": 1.98, "learning_rate": 3.772577067584084e-05, "loss": 0.7473, "step": 32850 }, { "epoch": 1.98, "learning_rate": 3.7724261765726614e-05, "loss": 0.64, "step": 32852 }, { "epoch": 1.98, "learning_rate": 3.7722752855612394e-05, "loss": 0.5664, "step": 32854 }, { "epoch": 1.98, "learning_rate": 3.772124394549817e-05, "loss": 0.4764, "step": 32856 }, { "epoch": 1.98, "learning_rate": 3.771973503538395e-05, "loss": 0.6768, "step": 32858 }, { "epoch": 1.98, "learning_rate": 3.771822612526972e-05, "loss": 0.9139, "step": 32860 }, { "epoch": 1.98, "learning_rate": 3.7716717215155494e-05, "loss": 0.936, "step": 32862 }, { "epoch": 1.98, "learning_rate": 3.7715208305041274e-05, "loss": 0.6841, "step": 32864 }, { "epoch": 1.98, "learning_rate": 3.771369939492705e-05, "loss": 0.7692, "step": 32866 }, { "epoch": 1.98, "learning_rate": 3.771219048481282e-05, "loss": 0.6949, "step": 32868 }, { "epoch": 1.98, "learning_rate": 3.77106815746986e-05, "loss": 0.9973, "step": 32870 }, { "epoch": 1.98, "learning_rate": 3.770917266458437e-05, "loss": 0.8348, "step": 32872 }, { "epoch": 1.98, "learning_rate": 3.7707663754470146e-05, "loss": 1.1426, "step": 32874 }, { "epoch": 1.98, "learning_rate": 3.7706154844355926e-05, "loss": 0.8905, "step": 32876 }, { "epoch": 1.98, "learning_rate": 3.77046459342417e-05, "loss": 0.9089, "step": 32878 }, { "epoch": 1.98, "learning_rate": 3.770313702412748e-05, "loss": 0.9902, "step": 32880 }, { "epoch": 1.98, "learning_rate": 3.770162811401325e-05, "loss": 0.6959, "step": 32882 }, { "epoch": 1.98, "learning_rate": 3.7700119203899025e-05, "loss": 0.9266, "step": 32884 }, { "epoch": 1.98, "learning_rate": 3.7698610293784805e-05, "loss": 0.7746, "step": 32886 }, { "epoch": 1.99, "learning_rate": 3.769710138367057e-05, "loss": 0.8929, "step": 32888 }, { "epoch": 1.99, "learning_rate": 3.769559247355635e-05, "loss": 0.6054, "step": 32890 }, { "epoch": 1.99, "learning_rate": 3.769408356344213e-05, "loss": 0.7872, "step": 32892 }, { "epoch": 1.99, "learning_rate": 3.7692574653327904e-05, "loss": 0.6029, "step": 32894 }, { "epoch": 1.99, "learning_rate": 3.769106574321368e-05, "loss": 0.6732, "step": 32896 }, { "epoch": 1.99, "learning_rate": 3.768955683309946e-05, "loss": 0.6111, "step": 32898 }, { "epoch": 1.99, "learning_rate": 3.768804792298523e-05, "loss": 0.463, "step": 32900 }, { "epoch": 1.99, "learning_rate": 3.7686539012871004e-05, "loss": 0.6986, "step": 32902 }, { "epoch": 1.99, "learning_rate": 3.768503010275678e-05, "loss": 0.7036, "step": 32904 }, { "epoch": 1.99, "learning_rate": 3.7683521192642557e-05, "loss": 0.8492, "step": 32906 }, { "epoch": 1.99, "learning_rate": 3.7682012282528336e-05, "loss": 0.8116, "step": 32908 }, { "epoch": 1.99, "learning_rate": 3.76805033724141e-05, "loss": 0.7456, "step": 32910 }, { "epoch": 1.99, "learning_rate": 3.767899446229988e-05, "loss": 0.6351, "step": 32912 }, { "epoch": 1.99, "learning_rate": 3.767748555218566e-05, "loss": 0.6614, "step": 32914 }, { "epoch": 1.99, "learning_rate": 3.767597664207143e-05, "loss": 0.5982, "step": 32916 }, { "epoch": 1.99, "learning_rate": 3.767446773195721e-05, "loss": 0.7222, "step": 32918 }, { "epoch": 1.99, "learning_rate": 3.767295882184298e-05, "loss": 0.7207, "step": 32920 }, { "epoch": 1.99, "learning_rate": 3.767144991172876e-05, "loss": 0.5628, "step": 32922 }, { "epoch": 1.99, "learning_rate": 3.7669941001614535e-05, "loss": 0.6086, "step": 32924 }, { "epoch": 1.99, "learning_rate": 3.766843209150031e-05, "loss": 0.491, "step": 32926 }, { "epoch": 1.99, "learning_rate": 3.766692318138609e-05, "loss": 0.87, "step": 32928 }, { "epoch": 1.99, "learning_rate": 3.766541427127186e-05, "loss": 0.6396, "step": 32930 }, { "epoch": 1.99, "learning_rate": 3.7663905361157634e-05, "loss": 1.0127, "step": 32932 }, { "epoch": 1.99, "learning_rate": 3.7662396451043414e-05, "loss": 0.5877, "step": 32934 }, { "epoch": 1.99, "learning_rate": 3.766088754092919e-05, "loss": 0.8372, "step": 32936 }, { "epoch": 1.99, "learning_rate": 3.765937863081496e-05, "loss": 0.8165, "step": 32938 }, { "epoch": 1.99, "learning_rate": 3.765786972070074e-05, "loss": 0.7785, "step": 32940 }, { "epoch": 1.99, "learning_rate": 3.7656360810586513e-05, "loss": 1.1941, "step": 32942 }, { "epoch": 1.99, "learning_rate": 3.7654851900472293e-05, "loss": 0.7218, "step": 32944 }, { "epoch": 1.99, "learning_rate": 3.7654097445415177e-05, "loss": 0.6343, "step": 32946 }, { "epoch": 1.99, "learning_rate": 3.7652588535300956e-05, "loss": 0.7558, "step": 32948 }, { "epoch": 1.99, "learning_rate": 3.765107962518673e-05, "loss": 0.8765, "step": 32950 }, { "epoch": 1.99, "learning_rate": 3.764957071507251e-05, "loss": 0.8001, "step": 32952 }, { "epoch": 1.99, "learning_rate": 3.7648061804958276e-05, "loss": 0.8853, "step": 32954 }, { "epoch": 1.99, "learning_rate": 3.7646552894844056e-05, "loss": 0.821, "step": 32956 }, { "epoch": 1.99, "learning_rate": 3.7645043984729836e-05, "loss": 0.6379, "step": 32958 }, { "epoch": 1.99, "learning_rate": 3.76435350746156e-05, "loss": 0.5796, "step": 32960 }, { "epoch": 1.99, "learning_rate": 3.764202616450138e-05, "loss": 0.9021, "step": 32962 }, { "epoch": 1.99, "learning_rate": 3.764051725438716e-05, "loss": 0.8039, "step": 32964 }, { "epoch": 1.99, "learning_rate": 3.7639008344272935e-05, "loss": 1.1155, "step": 32966 }, { "epoch": 1.99, "learning_rate": 3.763749943415871e-05, "loss": 0.592, "step": 32968 }, { "epoch": 1.99, "learning_rate": 3.763599052404448e-05, "loss": 0.7968, "step": 32970 }, { "epoch": 1.99, "learning_rate": 3.763448161393026e-05, "loss": 0.8759, "step": 32972 }, { "epoch": 1.99, "learning_rate": 3.7632972703816034e-05, "loss": 0.8196, "step": 32974 }, { "epoch": 1.99, "learning_rate": 3.763146379370181e-05, "loss": 0.8088, "step": 32976 }, { "epoch": 1.99, "learning_rate": 3.762995488358759e-05, "loss": 0.5598, "step": 32978 }, { "epoch": 1.99, "learning_rate": 3.762844597347337e-05, "loss": 0.7789, "step": 32980 }, { "epoch": 1.99, "learning_rate": 3.7626937063359133e-05, "loss": 0.7542, "step": 32982 }, { "epoch": 1.99, "learning_rate": 3.762542815324491e-05, "loss": 0.566, "step": 32984 }, { "epoch": 1.99, "learning_rate": 3.7623919243130687e-05, "loss": 0.7838, "step": 32986 }, { "epoch": 1.99, "learning_rate": 3.7622410333016466e-05, "loss": 0.8698, "step": 32988 }, { "epoch": 1.99, "learning_rate": 3.762090142290224e-05, "loss": 0.9274, "step": 32990 }, { "epoch": 1.99, "learning_rate": 3.761939251278801e-05, "loss": 0.7775, "step": 32992 }, { "epoch": 1.99, "learning_rate": 3.761788360267379e-05, "loss": 0.8195, "step": 32994 }, { "epoch": 1.99, "learning_rate": 3.7616374692559566e-05, "loss": 1.089, "step": 32996 }, { "epoch": 1.99, "learning_rate": 3.761486578244534e-05, "loss": 0.9975, "step": 32998 }, { "epoch": 1.99, "learning_rate": 3.761335687233112e-05, "loss": 0.7585, "step": 33000 }, { "epoch": 1.99, "learning_rate": 3.761184796221689e-05, "loss": 0.6907, "step": 33002 }, { "epoch": 1.99, "learning_rate": 3.7610339052102665e-05, "loss": 0.8257, "step": 33004 }, { "epoch": 1.99, "learning_rate": 3.7608830141988445e-05, "loss": 0.873, "step": 33006 }, { "epoch": 1.99, "learning_rate": 3.760732123187422e-05, "loss": 0.5053, "step": 33008 }, { "epoch": 1.99, "learning_rate": 3.760581232175999e-05, "loss": 0.6179, "step": 33010 }, { "epoch": 1.99, "learning_rate": 3.760430341164577e-05, "loss": 0.8427, "step": 33012 }, { "epoch": 1.99, "learning_rate": 3.7602794501531544e-05, "loss": 0.8266, "step": 33014 }, { "epoch": 1.99, "learning_rate": 3.7601285591417324e-05, "loss": 0.5377, "step": 33016 }, { "epoch": 1.99, "learning_rate": 3.75997766813031e-05, "loss": 0.6302, "step": 33018 }, { "epoch": 1.99, "learning_rate": 3.759826777118887e-05, "loss": 0.868, "step": 33020 }, { "epoch": 1.99, "learning_rate": 3.759675886107465e-05, "loss": 0.6432, "step": 33022 }, { "epoch": 1.99, "learning_rate": 3.759524995096042e-05, "loss": 0.5941, "step": 33024 }, { "epoch": 1.99, "learning_rate": 3.7593741040846196e-05, "loss": 0.8245, "step": 33026 }, { "epoch": 1.99, "learning_rate": 3.7592232130731976e-05, "loss": 0.6902, "step": 33028 }, { "epoch": 1.99, "learning_rate": 3.759072322061775e-05, "loss": 0.7887, "step": 33030 }, { "epoch": 1.99, "learning_rate": 3.758921431050352e-05, "loss": 0.7241, "step": 33032 }, { "epoch": 1.99, "learning_rate": 3.75877054003893e-05, "loss": 0.7307, "step": 33034 }, { "epoch": 1.99, "learning_rate": 3.7586196490275076e-05, "loss": 0.7312, "step": 33036 }, { "epoch": 1.99, "learning_rate": 3.758468758016085e-05, "loss": 0.8148, "step": 33038 }, { "epoch": 1.99, "learning_rate": 3.758317867004663e-05, "loss": 0.716, "step": 33040 }, { "epoch": 1.99, "learning_rate": 3.75816697599324e-05, "loss": 0.5519, "step": 33042 }, { "epoch": 1.99, "learning_rate": 3.758016084981818e-05, "loss": 0.9221, "step": 33044 }, { "epoch": 1.99, "learning_rate": 3.7578651939703955e-05, "loss": 0.6612, "step": 33046 }, { "epoch": 1.99, "learning_rate": 3.757714302958973e-05, "loss": 0.7703, "step": 33048 }, { "epoch": 1.99, "learning_rate": 3.757563411947551e-05, "loss": 0.7726, "step": 33050 }, { "epoch": 1.99, "learning_rate": 3.757412520936128e-05, "loss": 0.7213, "step": 33052 }, { "epoch": 2.0, "learning_rate": 3.7572616299247054e-05, "loss": 0.7323, "step": 33054 }, { "epoch": 2.0, "learning_rate": 3.7571107389132834e-05, "loss": 0.7168, "step": 33056 }, { "epoch": 2.0, "learning_rate": 3.756959847901861e-05, "loss": 0.8739, "step": 33058 }, { "epoch": 2.0, "learning_rate": 3.75688440239615e-05, "loss": 0.703, "step": 33060 }, { "epoch": 2.0, "learning_rate": 3.756733511384727e-05, "loss": 0.4959, "step": 33062 }, { "epoch": 2.0, "learning_rate": 3.756582620373304e-05, "loss": 0.7708, "step": 33064 }, { "epoch": 2.0, "learning_rate": 3.756431729361882e-05, "loss": 0.6642, "step": 33066 }, { "epoch": 2.0, "learning_rate": 3.7562808383504596e-05, "loss": 0.6075, "step": 33068 }, { "epoch": 2.0, "learning_rate": 3.756129947339037e-05, "loss": 0.7553, "step": 33070 }, { "epoch": 2.0, "learning_rate": 3.755979056327615e-05, "loss": 0.7661, "step": 33072 }, { "epoch": 2.0, "learning_rate": 3.755828165316192e-05, "loss": 0.8012, "step": 33074 }, { "epoch": 2.0, "learning_rate": 3.7556772743047696e-05, "loss": 0.6217, "step": 33076 }, { "epoch": 2.0, "learning_rate": 3.7555263832933476e-05, "loss": 0.4824, "step": 33078 }, { "epoch": 2.0, "learning_rate": 3.755375492281925e-05, "loss": 0.8913, "step": 33080 }, { "epoch": 2.0, "learning_rate": 3.755224601270502e-05, "loss": 0.8002, "step": 33082 }, { "epoch": 2.0, "learning_rate": 3.75507371025908e-05, "loss": 1.0613, "step": 33084 }, { "epoch": 2.0, "learning_rate": 3.7549228192476575e-05, "loss": 0.7886, "step": 33086 }, { "epoch": 2.0, "learning_rate": 3.7547719282362355e-05, "loss": 1.0882, "step": 33088 }, { "epoch": 2.0, "learning_rate": 3.754621037224813e-05, "loss": 0.6823, "step": 33090 }, { "epoch": 2.0, "learning_rate": 3.75447014621339e-05, "loss": 0.6616, "step": 33092 }, { "epoch": 2.0, "learning_rate": 3.754319255201968e-05, "loss": 0.4721, "step": 33094 }, { "epoch": 2.0, "learning_rate": 3.7541683641905454e-05, "loss": 0.7978, "step": 33096 }, { "epoch": 2.0, "learning_rate": 3.754017473179123e-05, "loss": 0.7514, "step": 33098 }, { "epoch": 2.0, "learning_rate": 3.753866582167701e-05, "loss": 1.2807, "step": 33100 }, { "epoch": 2.0, "learning_rate": 3.753715691156278e-05, "loss": 0.5993, "step": 33102 }, { "epoch": 2.0, "learning_rate": 3.753564800144855e-05, "loss": 0.7437, "step": 33104 }, { "epoch": 2.0, "learning_rate": 3.753413909133433e-05, "loss": 0.541, "step": 33106 }, { "epoch": 2.0, "learning_rate": 3.7532630181220106e-05, "loss": 0.4849, "step": 33108 }, { "epoch": 2.0, "learning_rate": 3.7531121271105886e-05, "loss": 0.6618, "step": 33110 }, { "epoch": 2.0, "learning_rate": 3.752961236099166e-05, "loss": 0.4773, "step": 33112 }, { "epoch": 2.0, "learning_rate": 3.752810345087743e-05, "loss": 0.893, "step": 33114 }, { "epoch": 2.0, "learning_rate": 3.752659454076321e-05, "loss": 0.7675, "step": 33116 }, { "epoch": 2.0, "learning_rate": 3.752508563064898e-05, "loss": 0.3954, "step": 33118 }, { "epoch": 2.0, "learning_rate": 3.752357672053476e-05, "loss": 0.8423, "step": 33120 }, { "epoch": 2.0, "learning_rate": 3.752206781042054e-05, "loss": 0.9176, "step": 33122 }, { "epoch": 2.0, "learning_rate": 3.752055890030631e-05, "loss": 0.819, "step": 33124 }, { "epoch": 2.0, "learning_rate": 3.7519049990192085e-05, "loss": 0.9532, "step": 33126 }, { "epoch": 2.0, "learning_rate": 3.7517541080077865e-05, "loss": 0.7837, "step": 33128 }, { "epoch": 2.0, "learning_rate": 3.751603216996364e-05, "loss": 0.7815, "step": 33130 }, { "epoch": 2.0, "learning_rate": 3.751452325984941e-05, "loss": 0.9953, "step": 33132 }, { "epoch": 2.0, "learning_rate": 3.7513014349735184e-05, "loss": 0.6597, "step": 33134 }, { "epoch": 2.0, "eval_cer": 0.11752051855467642, "eval_loss": 0.7549927830696106, "eval_runtime": 9268.6028, "eval_samples_per_second": 1.787, "eval_steps_per_second": 0.223, "step": 33134 }, { "epoch": 2.0, "learning_rate": 3.7511505439620964e-05, "loss": 0.5867, "step": 33136 }, { "epoch": 2.0, "learning_rate": 3.7509996529506744e-05, "loss": 0.6831, "step": 33138 }, { "epoch": 2.0, "learning_rate": 3.750848761939251e-05, "loss": 0.6572, "step": 33140 }, { "epoch": 2.0, "learning_rate": 3.750697870927829e-05, "loss": 0.7063, "step": 33142 }, { "epoch": 2.0, "learning_rate": 3.750546979916407e-05, "loss": 0.7889, "step": 33144 }, { "epoch": 2.0, "learning_rate": 3.7503960889049836e-05, "loss": 0.6846, "step": 33146 }, { "epoch": 2.0, "learning_rate": 3.7502451978935616e-05, "loss": 0.7962, "step": 33148 }, { "epoch": 2.0, "learning_rate": 3.750094306882139e-05, "loss": 0.5805, "step": 33150 }, { "epoch": 2.0, "learning_rate": 3.749943415870717e-05, "loss": 0.7082, "step": 33152 }, { "epoch": 2.0, "learning_rate": 3.749792524859294e-05, "loss": 0.4978, "step": 33154 }, { "epoch": 2.0, "learning_rate": 3.7496416338478716e-05, "loss": 0.7086, "step": 33156 }, { "epoch": 2.0, "learning_rate": 3.7494907428364495e-05, "loss": 0.7125, "step": 33158 }, { "epoch": 2.0, "learning_rate": 3.7493398518250275e-05, "loss": 0.4987, "step": 33160 }, { "epoch": 2.0, "learning_rate": 3.749188960813604e-05, "loss": 0.703, "step": 33162 }, { "epoch": 2.0, "learning_rate": 3.749038069802182e-05, "loss": 0.5951, "step": 33164 }, { "epoch": 2.0, "learning_rate": 3.7488871787907595e-05, "loss": 0.7113, "step": 33166 }, { "epoch": 2.0, "learning_rate": 3.748736287779337e-05, "loss": 0.7076, "step": 33168 }, { "epoch": 2.0, "learning_rate": 3.748585396767915e-05, "loss": 0.5025, "step": 33170 }, { "epoch": 2.0, "learning_rate": 3.748434505756492e-05, "loss": 0.3515, "step": 33172 }, { "epoch": 2.0, "learning_rate": 3.74828361474507e-05, "loss": 0.5569, "step": 33174 }, { "epoch": 2.0, "learning_rate": 3.7481327237336474e-05, "loss": 0.6583, "step": 33176 }, { "epoch": 2.0, "learning_rate": 3.747981832722225e-05, "loss": 0.7374, "step": 33178 }, { "epoch": 2.0, "learning_rate": 3.747830941710803e-05, "loss": 0.4621, "step": 33180 }, { "epoch": 2.0, "learning_rate": 3.74768005069938e-05, "loss": 0.8699, "step": 33182 }, { "epoch": 2.0, "learning_rate": 3.747529159687957e-05, "loss": 0.4724, "step": 33184 }, { "epoch": 2.0, "learning_rate": 3.747378268676535e-05, "loss": 0.457, "step": 33186 }, { "epoch": 2.0, "learning_rate": 3.7472273776651126e-05, "loss": 0.4631, "step": 33188 }, { "epoch": 2.0, "learning_rate": 3.74707648665369e-05, "loss": 0.5386, "step": 33190 }, { "epoch": 2.0, "learning_rate": 3.746925595642268e-05, "loss": 0.6403, "step": 33192 }, { "epoch": 2.0, "learning_rate": 3.746774704630845e-05, "loss": 0.7931, "step": 33194 }, { "epoch": 2.0, "learning_rate": 3.7466238136194225e-05, "loss": 0.6745, "step": 33196 }, { "epoch": 2.0, "learning_rate": 3.7464729226080005e-05, "loss": 0.3986, "step": 33198 }, { "epoch": 2.0, "learning_rate": 3.746322031596578e-05, "loss": 0.6079, "step": 33200 }, { "epoch": 2.0, "learning_rate": 3.746171140585156e-05, "loss": 0.8134, "step": 33202 }, { "epoch": 2.0, "learning_rate": 3.746020249573733e-05, "loss": 0.4342, "step": 33204 }, { "epoch": 2.0, "learning_rate": 3.7458693585623105e-05, "loss": 0.7227, "step": 33206 }, { "epoch": 2.0, "learning_rate": 3.7457184675508885e-05, "loss": 0.4739, "step": 33208 }, { "epoch": 2.0, "learning_rate": 3.745567576539466e-05, "loss": 0.7297, "step": 33210 }, { "epoch": 2.0, "learning_rate": 3.745416685528043e-05, "loss": 0.6119, "step": 33212 }, { "epoch": 2.0, "learning_rate": 3.745265794516621e-05, "loss": 0.577, "step": 33214 }, { "epoch": 2.0, "learning_rate": 3.7451149035051984e-05, "loss": 0.709, "step": 33216 }, { "epoch": 2.0, "learning_rate": 3.744964012493776e-05, "loss": 0.8127, "step": 33218 }, { "epoch": 2.01, "learning_rate": 3.744813121482354e-05, "loss": 0.5751, "step": 33220 }, { "epoch": 2.01, "learning_rate": 3.744662230470931e-05, "loss": 0.4372, "step": 33222 }, { "epoch": 2.01, "learning_rate": 3.744511339459509e-05, "loss": 0.5734, "step": 33224 }, { "epoch": 2.01, "learning_rate": 3.744360448448086e-05, "loss": 0.6934, "step": 33226 }, { "epoch": 2.01, "learning_rate": 3.7442095574366636e-05, "loss": 0.8103, "step": 33228 }, { "epoch": 2.01, "learning_rate": 3.7440586664252416e-05, "loss": 0.626, "step": 33230 }, { "epoch": 2.01, "learning_rate": 3.743907775413818e-05, "loss": 0.679, "step": 33232 }, { "epoch": 2.01, "learning_rate": 3.743756884402396e-05, "loss": 0.47, "step": 33234 }, { "epoch": 2.01, "learning_rate": 3.743605993390974e-05, "loss": 0.3966, "step": 33236 }, { "epoch": 2.01, "learning_rate": 3.7434551023795515e-05, "loss": 0.6961, "step": 33238 }, { "epoch": 2.01, "learning_rate": 3.743304211368129e-05, "loss": 0.4998, "step": 33240 }, { "epoch": 2.01, "learning_rate": 3.743153320356707e-05, "loss": 0.7425, "step": 33242 }, { "epoch": 2.01, "learning_rate": 3.743002429345284e-05, "loss": 0.7038, "step": 33244 }, { "epoch": 2.01, "learning_rate": 3.7428515383338615e-05, "loss": 0.5452, "step": 33246 }, { "epoch": 2.01, "learning_rate": 3.742700647322439e-05, "loss": 0.5982, "step": 33248 }, { "epoch": 2.01, "learning_rate": 3.742549756311017e-05, "loss": 0.6406, "step": 33250 }, { "epoch": 2.01, "learning_rate": 3.742398865299595e-05, "loss": 0.6961, "step": 33252 }, { "epoch": 2.01, "learning_rate": 3.7422479742881714e-05, "loss": 0.4419, "step": 33254 }, { "epoch": 2.01, "learning_rate": 3.7420970832767494e-05, "loss": 0.4157, "step": 33256 }, { "epoch": 2.01, "learning_rate": 3.7419461922653274e-05, "loss": 0.8763, "step": 33258 }, { "epoch": 2.01, "learning_rate": 3.741795301253905e-05, "loss": 0.5205, "step": 33260 }, { "epoch": 2.01, "learning_rate": 3.741644410242482e-05, "loss": 0.6717, "step": 33262 }, { "epoch": 2.01, "learning_rate": 3.741493519231059e-05, "loss": 0.5464, "step": 33264 }, { "epoch": 2.01, "learning_rate": 3.741342628219637e-05, "loss": 0.4906, "step": 33266 }, { "epoch": 2.01, "learning_rate": 3.7411917372082146e-05, "loss": 0.7826, "step": 33268 }, { "epoch": 2.01, "learning_rate": 3.741040846196792e-05, "loss": 0.6133, "step": 33270 }, { "epoch": 2.01, "learning_rate": 3.74088995518537e-05, "loss": 0.7917, "step": 33272 }, { "epoch": 2.01, "learning_rate": 3.740739064173948e-05, "loss": 0.694, "step": 33274 }, { "epoch": 2.01, "learning_rate": 3.7405881731625245e-05, "loss": 0.5147, "step": 33276 }, { "epoch": 2.01, "learning_rate": 3.7404372821511025e-05, "loss": 0.4889, "step": 33278 }, { "epoch": 2.01, "learning_rate": 3.74028639113968e-05, "loss": 0.505, "step": 33280 }, { "epoch": 2.01, "learning_rate": 3.740135500128257e-05, "loss": 0.7422, "step": 33282 }, { "epoch": 2.01, "learning_rate": 3.739984609116835e-05, "loss": 0.4621, "step": 33284 }, { "epoch": 2.01, "learning_rate": 3.7398337181054125e-05, "loss": 0.6892, "step": 33286 }, { "epoch": 2.01, "learning_rate": 3.7396828270939904e-05, "loss": 1.0642, "step": 33288 }, { "epoch": 2.01, "learning_rate": 3.739531936082568e-05, "loss": 0.5093, "step": 33290 }, { "epoch": 2.01, "learning_rate": 3.739381045071145e-05, "loss": 0.5447, "step": 33292 }, { "epoch": 2.01, "learning_rate": 3.739230154059723e-05, "loss": 0.5443, "step": 33294 }, { "epoch": 2.01, "learning_rate": 3.7390792630483004e-05, "loss": 0.7634, "step": 33296 }, { "epoch": 2.01, "learning_rate": 3.738928372036878e-05, "loss": 0.6517, "step": 33298 }, { "epoch": 2.01, "learning_rate": 3.738777481025456e-05, "loss": 0.5934, "step": 33300 }, { "epoch": 2.01, "learning_rate": 3.738626590014033e-05, "loss": 0.5475, "step": 33302 }, { "epoch": 2.01, "learning_rate": 3.73847569900261e-05, "loss": 0.729, "step": 33304 }, { "epoch": 2.01, "learning_rate": 3.738324807991188e-05, "loss": 0.6513, "step": 33306 }, { "epoch": 2.01, "learning_rate": 3.7381739169797656e-05, "loss": 0.5518, "step": 33308 }, { "epoch": 2.01, "learning_rate": 3.738023025968343e-05, "loss": 0.8182, "step": 33310 }, { "epoch": 2.01, "learning_rate": 3.737872134956921e-05, "loss": 0.4568, "step": 33312 }, { "epoch": 2.01, "learning_rate": 3.737721243945498e-05, "loss": 0.3186, "step": 33314 }, { "epoch": 2.01, "learning_rate": 3.737570352934076e-05, "loss": 0.8259, "step": 33316 }, { "epoch": 2.01, "learning_rate": 3.7374194619226535e-05, "loss": 0.7461, "step": 33318 }, { "epoch": 2.01, "learning_rate": 3.737268570911231e-05, "loss": 0.7134, "step": 33320 }, { "epoch": 2.01, "learning_rate": 3.737117679899809e-05, "loss": 0.8315, "step": 33322 }, { "epoch": 2.01, "learning_rate": 3.736966788888386e-05, "loss": 0.4946, "step": 33324 }, { "epoch": 2.01, "learning_rate": 3.7368158978769634e-05, "loss": 0.6155, "step": 33326 }, { "epoch": 2.01, "learning_rate": 3.7366650068655414e-05, "loss": 0.4808, "step": 33328 }, { "epoch": 2.01, "learning_rate": 3.736514115854119e-05, "loss": 0.5793, "step": 33330 }, { "epoch": 2.01, "learning_rate": 3.736363224842696e-05, "loss": 0.8201, "step": 33332 }, { "epoch": 2.01, "learning_rate": 3.736212333831274e-05, "loss": 0.8299, "step": 33334 }, { "epoch": 2.01, "learning_rate": 3.7360614428198514e-05, "loss": 0.5042, "step": 33336 }, { "epoch": 2.01, "learning_rate": 3.7359105518084294e-05, "loss": 0.7249, "step": 33338 }, { "epoch": 2.01, "learning_rate": 3.735759660797007e-05, "loss": 0.6453, "step": 33340 }, { "epoch": 2.01, "learning_rate": 3.735608769785584e-05, "loss": 0.5555, "step": 33342 }, { "epoch": 2.01, "learning_rate": 3.735457878774162e-05, "loss": 0.9636, "step": 33344 }, { "epoch": 2.01, "learning_rate": 3.7353069877627386e-05, "loss": 0.6517, "step": 33346 }, { "epoch": 2.01, "learning_rate": 3.7351560967513166e-05, "loss": 0.414, "step": 33348 }, { "epoch": 2.01, "learning_rate": 3.7350052057398946e-05, "loss": 0.4671, "step": 33350 }, { "epoch": 2.01, "learning_rate": 3.734854314728472e-05, "loss": 0.4417, "step": 33352 }, { "epoch": 2.01, "learning_rate": 3.734703423717049e-05, "loss": 0.5791, "step": 33354 }, { "epoch": 2.01, "learning_rate": 3.734552532705627e-05, "loss": 0.6385, "step": 33356 }, { "epoch": 2.01, "learning_rate": 3.7344016416942045e-05, "loss": 0.6825, "step": 33358 }, { "epoch": 2.01, "learning_rate": 3.734250750682782e-05, "loss": 0.6265, "step": 33360 }, { "epoch": 2.01, "learning_rate": 3.734099859671359e-05, "loss": 0.6331, "step": 33362 }, { "epoch": 2.01, "learning_rate": 3.733948968659937e-05, "loss": 0.5219, "step": 33364 }, { "epoch": 2.01, "learning_rate": 3.733798077648515e-05, "loss": 0.5812, "step": 33366 }, { "epoch": 2.01, "learning_rate": 3.733647186637092e-05, "loss": 0.5968, "step": 33368 }, { "epoch": 2.01, "learning_rate": 3.73349629562567e-05, "loss": 0.6831, "step": 33370 }, { "epoch": 2.01, "learning_rate": 3.733345404614248e-05, "loss": 0.4977, "step": 33372 }, { "epoch": 2.01, "learning_rate": 3.733194513602825e-05, "loss": 0.5594, "step": 33374 }, { "epoch": 2.01, "learning_rate": 3.7330436225914024e-05, "loss": 0.5961, "step": 33376 }, { "epoch": 2.01, "learning_rate": 3.73289273157998e-05, "loss": 0.6017, "step": 33378 }, { "epoch": 2.01, "learning_rate": 3.732741840568558e-05, "loss": 0.58, "step": 33380 }, { "epoch": 2.01, "learning_rate": 3.732590949557135e-05, "loss": 0.5691, "step": 33382 }, { "epoch": 2.01, "learning_rate": 3.732440058545712e-05, "loss": 0.6101, "step": 33384 }, { "epoch": 2.02, "learning_rate": 3.73228916753429e-05, "loss": 0.4668, "step": 33386 }, { "epoch": 2.02, "learning_rate": 3.732138276522868e-05, "loss": 0.8017, "step": 33388 }, { "epoch": 2.02, "learning_rate": 3.731987385511445e-05, "loss": 0.4971, "step": 33390 }, { "epoch": 2.02, "learning_rate": 3.731836494500023e-05, "loss": 0.6596, "step": 33392 }, { "epoch": 2.02, "learning_rate": 3.7316856034886e-05, "loss": 0.5066, "step": 33394 }, { "epoch": 2.02, "learning_rate": 3.7315347124771775e-05, "loss": 0.7925, "step": 33396 }, { "epoch": 2.02, "learning_rate": 3.7313838214657555e-05, "loss": 0.7853, "step": 33398 }, { "epoch": 2.02, "learning_rate": 3.731232930454333e-05, "loss": 0.5351, "step": 33400 }, { "epoch": 2.02, "learning_rate": 3.731082039442911e-05, "loss": 0.6918, "step": 33402 }, { "epoch": 2.02, "learning_rate": 3.730931148431488e-05, "loss": 0.6248, "step": 33404 }, { "epoch": 2.02, "learning_rate": 3.7307802574200654e-05, "loss": 0.7132, "step": 33406 }, { "epoch": 2.02, "learning_rate": 3.7306293664086434e-05, "loss": 0.5889, "step": 33408 }, { "epoch": 2.02, "learning_rate": 3.730478475397221e-05, "loss": 0.4622, "step": 33410 }, { "epoch": 2.02, "learning_rate": 3.730327584385798e-05, "loss": 0.6096, "step": 33412 }, { "epoch": 2.02, "learning_rate": 3.730176693374376e-05, "loss": 0.5911, "step": 33414 }, { "epoch": 2.02, "learning_rate": 3.7300258023629534e-05, "loss": 0.652, "step": 33416 }, { "epoch": 2.02, "learning_rate": 3.729874911351531e-05, "loss": 0.5633, "step": 33418 }, { "epoch": 2.02, "learning_rate": 3.7297240203401087e-05, "loss": 0.7635, "step": 33420 }, { "epoch": 2.02, "learning_rate": 3.729573129328686e-05, "loss": 0.5287, "step": 33422 }, { "epoch": 2.02, "learning_rate": 3.729422238317264e-05, "loss": 0.6655, "step": 33424 }, { "epoch": 2.02, "learning_rate": 3.729271347305841e-05, "loss": 0.581, "step": 33426 }, { "epoch": 2.02, "learning_rate": 3.7291204562944186e-05, "loss": 0.5591, "step": 33428 }, { "epoch": 2.02, "learning_rate": 3.7289695652829966e-05, "loss": 0.668, "step": 33430 }, { "epoch": 2.02, "learning_rate": 3.728818674271574e-05, "loss": 0.4453, "step": 33432 }, { "epoch": 2.02, "learning_rate": 3.728667783260151e-05, "loss": 0.3231, "step": 33434 }, { "epoch": 2.02, "learning_rate": 3.728516892248729e-05, "loss": 0.5156, "step": 33436 }, { "epoch": 2.02, "learning_rate": 3.7283660012373065e-05, "loss": 0.7725, "step": 33438 }, { "epoch": 2.02, "learning_rate": 3.728215110225884e-05, "loss": 0.5223, "step": 33440 }, { "epoch": 2.02, "learning_rate": 3.728064219214462e-05, "loss": 0.5554, "step": 33442 }, { "epoch": 2.02, "learning_rate": 3.727913328203039e-05, "loss": 0.4538, "step": 33444 }, { "epoch": 2.02, "learning_rate": 3.7277624371916164e-05, "loss": 0.6592, "step": 33446 }, { "epoch": 2.02, "learning_rate": 3.7276115461801944e-05, "loss": 0.8038, "step": 33448 }, { "epoch": 2.02, "learning_rate": 3.727460655168772e-05, "loss": 0.6513, "step": 33450 }, { "epoch": 2.02, "learning_rate": 3.72730976415735e-05, "loss": 0.593, "step": 33452 }, { "epoch": 2.02, "learning_rate": 3.727158873145927e-05, "loss": 0.6839, "step": 33454 }, { "epoch": 2.02, "learning_rate": 3.7270079821345043e-05, "loss": 0.703, "step": 33456 }, { "epoch": 2.02, "learning_rate": 3.726857091123082e-05, "loss": 0.8592, "step": 33458 }, { "epoch": 2.02, "learning_rate": 3.726706200111659e-05, "loss": 0.6715, "step": 33460 }, { "epoch": 2.02, "learning_rate": 3.726555309100237e-05, "loss": 0.6411, "step": 33462 }, { "epoch": 2.02, "learning_rate": 3.726404418088815e-05, "loss": 0.6109, "step": 33464 }, { "epoch": 2.02, "learning_rate": 3.726253527077392e-05, "loss": 0.7599, "step": 33466 }, { "epoch": 2.02, "learning_rate": 3.7261026360659696e-05, "loss": 0.6089, "step": 33468 }, { "epoch": 2.02, "learning_rate": 3.7259517450545476e-05, "loss": 0.485, "step": 33470 }, { "epoch": 2.02, "learning_rate": 3.725800854043125e-05, "loss": 0.9038, "step": 33472 }, { "epoch": 2.02, "learning_rate": 3.725649963031702e-05, "loss": 0.7288, "step": 33474 }, { "epoch": 2.02, "learning_rate": 3.7254990720202795e-05, "loss": 0.6483, "step": 33476 }, { "epoch": 2.02, "learning_rate": 3.7253481810088575e-05, "loss": 0.6184, "step": 33478 }, { "epoch": 2.02, "learning_rate": 3.7251972899974355e-05, "loss": 0.6469, "step": 33480 }, { "epoch": 2.02, "learning_rate": 3.725046398986012e-05, "loss": 0.6674, "step": 33482 }, { "epoch": 2.02, "learning_rate": 3.72489550797459e-05, "loss": 0.6848, "step": 33484 }, { "epoch": 2.02, "learning_rate": 3.724744616963168e-05, "loss": 0.5553, "step": 33486 }, { "epoch": 2.02, "learning_rate": 3.7245937259517454e-05, "loss": 0.6131, "step": 33488 }, { "epoch": 2.02, "learning_rate": 3.724442834940323e-05, "loss": 0.5717, "step": 33490 }, { "epoch": 2.02, "learning_rate": 3.7242919439289e-05, "loss": 0.371, "step": 33492 }, { "epoch": 2.02, "learning_rate": 3.724141052917478e-05, "loss": 0.5663, "step": 33494 }, { "epoch": 2.02, "learning_rate": 3.7239901619060553e-05, "loss": 0.7104, "step": 33496 }, { "epoch": 2.02, "learning_rate": 3.7238392708946327e-05, "loss": 0.4667, "step": 33498 }, { "epoch": 2.02, "learning_rate": 3.7236883798832106e-05, "loss": 0.5184, "step": 33500 }, { "epoch": 2.02, "learning_rate": 3.7235374888717886e-05, "loss": 0.4276, "step": 33502 }, { "epoch": 2.02, "learning_rate": 3.723386597860365e-05, "loss": 0.8263, "step": 33504 }, { "epoch": 2.02, "learning_rate": 3.723235706848943e-05, "loss": 0.5735, "step": 33506 }, { "epoch": 2.02, "learning_rate": 3.7230848158375206e-05, "loss": 0.3981, "step": 33508 }, { "epoch": 2.02, "learning_rate": 3.722933924826098e-05, "loss": 0.6081, "step": 33510 }, { "epoch": 2.02, "learning_rate": 3.722783033814676e-05, "loss": 0.8165, "step": 33512 }, { "epoch": 2.02, "learning_rate": 3.722632142803253e-05, "loss": 0.6859, "step": 33514 }, { "epoch": 2.02, "learning_rate": 3.722481251791831e-05, "loss": 0.4978, "step": 33516 }, { "epoch": 2.02, "learning_rate": 3.7223303607804085e-05, "loss": 0.7816, "step": 33518 }, { "epoch": 2.02, "learning_rate": 3.722179469768986e-05, "loss": 0.5747, "step": 33520 }, { "epoch": 2.02, "learning_rate": 3.722028578757564e-05, "loss": 0.737, "step": 33522 }, { "epoch": 2.02, "learning_rate": 3.721877687746141e-05, "loss": 0.4385, "step": 33524 }, { "epoch": 2.02, "learning_rate": 3.7217267967347184e-05, "loss": 0.8806, "step": 33526 }, { "epoch": 2.02, "learning_rate": 3.7215759057232964e-05, "loss": 0.5862, "step": 33528 }, { "epoch": 2.02, "learning_rate": 3.721425014711874e-05, "loss": 0.8497, "step": 33530 }, { "epoch": 2.02, "learning_rate": 3.721274123700451e-05, "loss": 0.6609, "step": 33532 }, { "epoch": 2.02, "learning_rate": 3.721123232689029e-05, "loss": 0.8101, "step": 33534 }, { "epoch": 2.02, "learning_rate": 3.720972341677606e-05, "loss": 0.5743, "step": 33536 }, { "epoch": 2.02, "learning_rate": 3.720821450666184e-05, "loss": 0.5633, "step": 33538 }, { "epoch": 2.02, "learning_rate": 3.7206705596547616e-05, "loss": 0.5486, "step": 33540 }, { "epoch": 2.02, "learning_rate": 3.720519668643339e-05, "loss": 0.7034, "step": 33542 }, { "epoch": 2.02, "learning_rate": 3.720368777631917e-05, "loss": 0.5188, "step": 33544 }, { "epoch": 2.02, "learning_rate": 3.720217886620494e-05, "loss": 0.621, "step": 33546 }, { "epoch": 2.02, "learning_rate": 3.7200669956090716e-05, "loss": 0.9574, "step": 33548 }, { "epoch": 2.02, "learning_rate": 3.7199161045976496e-05, "loss": 0.5482, "step": 33550 }, { "epoch": 2.03, "learning_rate": 3.719765213586227e-05, "loss": 0.9008, "step": 33552 }, { "epoch": 2.03, "learning_rate": 3.719614322574804e-05, "loss": 0.5824, "step": 33554 }, { "epoch": 2.03, "learning_rate": 3.719463431563382e-05, "loss": 0.5478, "step": 33556 }, { "epoch": 2.03, "learning_rate": 3.7193125405519595e-05, "loss": 0.673, "step": 33558 }, { "epoch": 2.03, "learning_rate": 3.719161649540537e-05, "loss": 0.613, "step": 33560 }, { "epoch": 2.03, "learning_rate": 3.719010758529115e-05, "loss": 0.6413, "step": 33562 }, { "epoch": 2.03, "learning_rate": 3.718859867517692e-05, "loss": 0.6258, "step": 33564 }, { "epoch": 2.03, "learning_rate": 3.71870897650627e-05, "loss": 0.874, "step": 33566 }, { "epoch": 2.03, "learning_rate": 3.7185580854948474e-05, "loss": 0.6166, "step": 33568 }, { "epoch": 2.03, "learning_rate": 3.718407194483425e-05, "loss": 0.5308, "step": 33570 }, { "epoch": 2.03, "learning_rate": 3.718256303472003e-05, "loss": 0.7688, "step": 33572 }, { "epoch": 2.03, "learning_rate": 3.7181054124605793e-05, "loss": 1.0028, "step": 33574 }, { "epoch": 2.03, "learning_rate": 3.717954521449157e-05, "loss": 0.6841, "step": 33576 }, { "epoch": 2.03, "learning_rate": 3.717803630437735e-05, "loss": 0.7412, "step": 33578 }, { "epoch": 2.03, "learning_rate": 3.7176527394263126e-05, "loss": 0.674, "step": 33580 }, { "epoch": 2.03, "learning_rate": 3.71750184841489e-05, "loss": 0.5136, "step": 33582 }, { "epoch": 2.03, "learning_rate": 3.717350957403468e-05, "loss": 0.6278, "step": 33584 }, { "epoch": 2.03, "learning_rate": 3.717200066392045e-05, "loss": 0.692, "step": 33586 }, { "epoch": 2.03, "learning_rate": 3.717049175380623e-05, "loss": 0.7588, "step": 33588 }, { "epoch": 2.03, "learning_rate": 3.7168982843692e-05, "loss": 0.7509, "step": 33590 }, { "epoch": 2.03, "learning_rate": 3.716747393357778e-05, "loss": 0.634, "step": 33592 }, { "epoch": 2.03, "learning_rate": 3.716596502346356e-05, "loss": 0.4316, "step": 33594 }, { "epoch": 2.03, "learning_rate": 3.7164456113349325e-05, "loss": 0.6568, "step": 33596 }, { "epoch": 2.03, "learning_rate": 3.7162947203235105e-05, "loss": 0.606, "step": 33598 }, { "epoch": 2.03, "learning_rate": 3.7161438293120885e-05, "loss": 0.754, "step": 33600 }, { "epoch": 2.03, "learning_rate": 3.715992938300666e-05, "loss": 0.6012, "step": 33602 }, { "epoch": 2.03, "learning_rate": 3.715842047289243e-05, "loss": 0.5325, "step": 33604 }, { "epoch": 2.03, "learning_rate": 3.7156911562778204e-05, "loss": 0.6247, "step": 33606 }, { "epoch": 2.03, "learning_rate": 3.7155402652663984e-05, "loss": 0.7808, "step": 33608 }, { "epoch": 2.03, "learning_rate": 3.715389374254976e-05, "loss": 0.5363, "step": 33610 }, { "epoch": 2.03, "learning_rate": 3.715238483243553e-05, "loss": 0.8398, "step": 33612 }, { "epoch": 2.03, "learning_rate": 3.715087592232131e-05, "loss": 0.4134, "step": 33614 }, { "epoch": 2.03, "learning_rate": 3.714936701220709e-05, "loss": 0.6274, "step": 33616 }, { "epoch": 2.03, "learning_rate": 3.7147858102092856e-05, "loss": 0.5245, "step": 33618 }, { "epoch": 2.03, "learning_rate": 3.7146349191978636e-05, "loss": 0.7347, "step": 33620 }, { "epoch": 2.03, "learning_rate": 3.714484028186441e-05, "loss": 0.5331, "step": 33622 }, { "epoch": 2.03, "learning_rate": 3.714333137175018e-05, "loss": 0.3454, "step": 33624 }, { "epoch": 2.03, "learning_rate": 3.714182246163596e-05, "loss": 0.6683, "step": 33626 }, { "epoch": 2.03, "learning_rate": 3.7140313551521736e-05, "loss": 0.5203, "step": 33628 }, { "epoch": 2.03, "learning_rate": 3.7138804641407515e-05, "loss": 0.6568, "step": 33630 }, { "epoch": 2.03, "learning_rate": 3.713729573129329e-05, "loss": 0.655, "step": 33632 }, { "epoch": 2.03, "learning_rate": 3.713578682117906e-05, "loss": 0.5184, "step": 33634 }, { "epoch": 2.03, "learning_rate": 3.713427791106484e-05, "loss": 0.6246, "step": 33636 }, { "epoch": 2.03, "learning_rate": 3.7132769000950615e-05, "loss": 0.5624, "step": 33638 }, { "epoch": 2.03, "learning_rate": 3.713126009083639e-05, "loss": 0.5995, "step": 33640 }, { "epoch": 2.03, "learning_rate": 3.712975118072217e-05, "loss": 0.6856, "step": 33642 }, { "epoch": 2.03, "learning_rate": 3.712824227060794e-05, "loss": 0.3942, "step": 33644 }, { "epoch": 2.03, "learning_rate": 3.7126733360493714e-05, "loss": 0.6058, "step": 33646 }, { "epoch": 2.03, "learning_rate": 3.7125224450379494e-05, "loss": 0.6995, "step": 33648 }, { "epoch": 2.03, "learning_rate": 3.712371554026527e-05, "loss": 0.5028, "step": 33650 }, { "epoch": 2.03, "learning_rate": 3.712220663015105e-05, "loss": 0.6735, "step": 33652 }, { "epoch": 2.03, "learning_rate": 3.712069772003682e-05, "loss": 0.5995, "step": 33654 }, { "epoch": 2.03, "learning_rate": 3.711918880992259e-05, "loss": 0.6067, "step": 33656 }, { "epoch": 2.03, "learning_rate": 3.711767989980837e-05, "loss": 0.5803, "step": 33658 }, { "epoch": 2.03, "learning_rate": 3.7116170989694146e-05, "loss": 0.763, "step": 33660 }, { "epoch": 2.03, "learning_rate": 3.711466207957992e-05, "loss": 0.4911, "step": 33662 }, { "epoch": 2.03, "learning_rate": 3.71131531694657e-05, "loss": 0.6456, "step": 33664 }, { "epoch": 2.03, "learning_rate": 3.711164425935147e-05, "loss": 0.6239, "step": 33666 }, { "epoch": 2.03, "learning_rate": 3.7110135349237245e-05, "loss": 0.5013, "step": 33668 }, { "epoch": 2.03, "learning_rate": 3.7108626439123025e-05, "loss": 0.437, "step": 33670 }, { "epoch": 2.03, "learning_rate": 3.71071175290088e-05, "loss": 0.5206, "step": 33672 }, { "epoch": 2.03, "learning_rate": 3.710560861889457e-05, "loss": 0.5811, "step": 33674 }, { "epoch": 2.03, "learning_rate": 3.710409970878035e-05, "loss": 0.6837, "step": 33676 }, { "epoch": 2.03, "learning_rate": 3.7102590798666125e-05, "loss": 0.5812, "step": 33678 }, { "epoch": 2.03, "learning_rate": 3.7101081888551905e-05, "loss": 0.594, "step": 33680 }, { "epoch": 2.03, "learning_rate": 3.709957297843768e-05, "loss": 0.7346, "step": 33682 }, { "epoch": 2.03, "learning_rate": 3.709806406832345e-05, "loss": 0.6358, "step": 33684 }, { "epoch": 2.03, "learning_rate": 3.709655515820923e-05, "loss": 0.6173, "step": 33686 }, { "epoch": 2.03, "learning_rate": 3.7095046248095e-05, "loss": 0.594, "step": 33688 }, { "epoch": 2.03, "learning_rate": 3.709353733798078e-05, "loss": 0.8001, "step": 33690 }, { "epoch": 2.03, "learning_rate": 3.709202842786656e-05, "loss": 0.6513, "step": 33692 }, { "epoch": 2.03, "learning_rate": 3.709051951775233e-05, "loss": 0.6983, "step": 33694 }, { "epoch": 2.03, "learning_rate": 3.70890106076381e-05, "loss": 0.6748, "step": 33696 }, { "epoch": 2.03, "learning_rate": 3.708750169752388e-05, "loss": 0.8487, "step": 33698 }, { "epoch": 2.03, "learning_rate": 3.7085992787409656e-05, "loss": 0.8891, "step": 33700 }, { "epoch": 2.03, "learning_rate": 3.7084483877295436e-05, "loss": 0.5742, "step": 33702 }, { "epoch": 2.03, "learning_rate": 3.70829749671812e-05, "loss": 0.5284, "step": 33704 }, { "epoch": 2.03, "learning_rate": 3.708146605706698e-05, "loss": 0.5663, "step": 33706 }, { "epoch": 2.03, "learning_rate": 3.707995714695276e-05, "loss": 0.5719, "step": 33708 }, { "epoch": 2.03, "learning_rate": 3.707844823683853e-05, "loss": 0.7317, "step": 33710 }, { "epoch": 2.03, "learning_rate": 3.707693932672431e-05, "loss": 0.5768, "step": 33712 }, { "epoch": 2.03, "learning_rate": 3.707543041661009e-05, "loss": 0.8217, "step": 33714 }, { "epoch": 2.04, "learning_rate": 3.707392150649586e-05, "loss": 0.7254, "step": 33716 }, { "epoch": 2.04, "learning_rate": 3.7072412596381635e-05, "loss": 0.5886, "step": 33718 }, { "epoch": 2.04, "learning_rate": 3.707090368626741e-05, "loss": 0.6779, "step": 33720 }, { "epoch": 2.04, "learning_rate": 3.706939477615319e-05, "loss": 0.7948, "step": 33722 }, { "epoch": 2.04, "learning_rate": 3.706788586603896e-05, "loss": 0.4594, "step": 33724 }, { "epoch": 2.04, "learning_rate": 3.7066376955924734e-05, "loss": 0.8627, "step": 33726 }, { "epoch": 2.04, "learning_rate": 3.7064868045810514e-05, "loss": 0.5596, "step": 33728 }, { "epoch": 2.04, "learning_rate": 3.7063359135696294e-05, "loss": 0.8149, "step": 33730 }, { "epoch": 2.04, "learning_rate": 3.706185022558206e-05, "loss": 0.6482, "step": 33732 }, { "epoch": 2.04, "learning_rate": 3.706034131546784e-05, "loss": 0.4736, "step": 33734 }, { "epoch": 2.04, "learning_rate": 3.705883240535361e-05, "loss": 0.7269, "step": 33736 }, { "epoch": 2.04, "learning_rate": 3.7057323495239386e-05, "loss": 0.8931, "step": 33738 }, { "epoch": 2.04, "learning_rate": 3.7055814585125166e-05, "loss": 0.5342, "step": 33740 }, { "epoch": 2.04, "learning_rate": 3.705430567501094e-05, "loss": 0.5491, "step": 33742 }, { "epoch": 2.04, "learning_rate": 3.705279676489672e-05, "loss": 0.5547, "step": 33744 }, { "epoch": 2.04, "learning_rate": 3.705128785478249e-05, "loss": 0.6383, "step": 33746 }, { "epoch": 2.04, "learning_rate": 3.7049778944668265e-05, "loss": 0.3788, "step": 33748 }, { "epoch": 2.04, "learning_rate": 3.7048270034554045e-05, "loss": 0.5809, "step": 33750 }, { "epoch": 2.04, "learning_rate": 3.704676112443982e-05, "loss": 0.4215, "step": 33752 }, { "epoch": 2.04, "learning_rate": 3.704525221432559e-05, "loss": 0.819, "step": 33754 }, { "epoch": 2.04, "learning_rate": 3.704374330421137e-05, "loss": 0.3995, "step": 33756 }, { "epoch": 2.04, "learning_rate": 3.7042234394097145e-05, "loss": 0.7404, "step": 33758 }, { "epoch": 2.04, "learning_rate": 3.704072548398292e-05, "loss": 0.664, "step": 33760 }, { "epoch": 2.04, "learning_rate": 3.70392165738687e-05, "loss": 0.6057, "step": 33762 }, { "epoch": 2.04, "learning_rate": 3.703770766375447e-05, "loss": 0.8373, "step": 33764 }, { "epoch": 2.04, "learning_rate": 3.703619875364025e-05, "loss": 0.6139, "step": 33766 }, { "epoch": 2.04, "learning_rate": 3.7034689843526024e-05, "loss": 0.4566, "step": 33768 }, { "epoch": 2.04, "learning_rate": 3.70331809334118e-05, "loss": 0.4843, "step": 33770 }, { "epoch": 2.04, "learning_rate": 3.703167202329758e-05, "loss": 0.4839, "step": 33772 }, { "epoch": 2.04, "learning_rate": 3.703016311318335e-05, "loss": 0.7371, "step": 33774 }, { "epoch": 2.04, "learning_rate": 3.702865420306912e-05, "loss": 0.4878, "step": 33776 }, { "epoch": 2.04, "learning_rate": 3.70271452929549e-05, "loss": 0.5849, "step": 33778 }, { "epoch": 2.04, "learning_rate": 3.7025636382840676e-05, "loss": 0.3845, "step": 33780 }, { "epoch": 2.04, "learning_rate": 3.702412747272645e-05, "loss": 0.6929, "step": 33782 }, { "epoch": 2.04, "learning_rate": 3.702261856261223e-05, "loss": 0.5254, "step": 33784 }, { "epoch": 2.04, "learning_rate": 3.7021109652498e-05, "loss": 0.5979, "step": 33786 }, { "epoch": 2.04, "learning_rate": 3.7019600742383775e-05, "loss": 0.6735, "step": 33788 }, { "epoch": 2.04, "learning_rate": 3.7018091832269555e-05, "loss": 0.6749, "step": 33790 }, { "epoch": 2.04, "learning_rate": 3.701658292215533e-05, "loss": 0.7101, "step": 33792 }, { "epoch": 2.04, "learning_rate": 3.701507401204111e-05, "loss": 0.7114, "step": 33794 }, { "epoch": 2.04, "learning_rate": 3.701356510192688e-05, "loss": 0.6454, "step": 33796 }, { "epoch": 2.04, "learning_rate": 3.7012056191812655e-05, "loss": 0.765, "step": 33798 }, { "epoch": 2.04, "learning_rate": 3.7010547281698434e-05, "loss": 0.5342, "step": 33800 }, { "epoch": 2.04, "learning_rate": 3.70090383715842e-05, "loss": 0.7145, "step": 33802 }, { "epoch": 2.04, "learning_rate": 3.700752946146998e-05, "loss": 0.6618, "step": 33804 }, { "epoch": 2.04, "learning_rate": 3.700602055135576e-05, "loss": 0.5595, "step": 33806 }, { "epoch": 2.04, "learning_rate": 3.7004511641241534e-05, "loss": 0.7689, "step": 33808 }, { "epoch": 2.04, "learning_rate": 3.700300273112731e-05, "loss": 0.5766, "step": 33810 }, { "epoch": 2.04, "learning_rate": 3.700149382101309e-05, "loss": 0.6604, "step": 33812 }, { "epoch": 2.04, "learning_rate": 3.699998491089886e-05, "loss": 0.5633, "step": 33814 }, { "epoch": 2.04, "learning_rate": 3.699847600078464e-05, "loss": 0.5742, "step": 33816 }, { "epoch": 2.04, "learning_rate": 3.6996967090670406e-05, "loss": 0.6992, "step": 33818 }, { "epoch": 2.04, "learning_rate": 3.6995458180556186e-05, "loss": 0.8802, "step": 33820 }, { "epoch": 2.04, "learning_rate": 3.6993949270441966e-05, "loss": 0.574, "step": 33822 }, { "epoch": 2.04, "learning_rate": 3.699244036032773e-05, "loss": 1.0306, "step": 33824 }, { "epoch": 2.04, "learning_rate": 3.699093145021351e-05, "loss": 0.4505, "step": 33826 }, { "epoch": 2.04, "learning_rate": 3.698942254009929e-05, "loss": 0.4467, "step": 33828 }, { "epoch": 2.04, "learning_rate": 3.6987913629985065e-05, "loss": 0.8027, "step": 33830 }, { "epoch": 2.04, "learning_rate": 3.698640471987084e-05, "loss": 0.6854, "step": 33832 }, { "epoch": 2.04, "learning_rate": 3.698489580975661e-05, "loss": 0.4373, "step": 33834 }, { "epoch": 2.04, "learning_rate": 3.698338689964239e-05, "loss": 0.4559, "step": 33836 }, { "epoch": 2.04, "learning_rate": 3.6981877989528164e-05, "loss": 0.4538, "step": 33838 }, { "epoch": 2.04, "learning_rate": 3.698036907941394e-05, "loss": 0.6272, "step": 33840 }, { "epoch": 2.04, "learning_rate": 3.697886016929972e-05, "loss": 0.7358, "step": 33842 }, { "epoch": 2.04, "learning_rate": 3.69773512591855e-05, "loss": 0.8026, "step": 33844 }, { "epoch": 2.04, "learning_rate": 3.6975842349071264e-05, "loss": 0.5637, "step": 33846 }, { "epoch": 2.04, "learning_rate": 3.6974333438957044e-05, "loss": 0.5597, "step": 33848 }, { "epoch": 2.04, "learning_rate": 3.697282452884282e-05, "loss": 0.4913, "step": 33850 }, { "epoch": 2.04, "learning_rate": 3.697131561872859e-05, "loss": 0.6579, "step": 33852 }, { "epoch": 2.04, "learning_rate": 3.696980670861437e-05, "loss": 0.5472, "step": 33854 }, { "epoch": 2.04, "learning_rate": 3.696829779850014e-05, "loss": 0.6896, "step": 33856 }, { "epoch": 2.04, "learning_rate": 3.696678888838592e-05, "loss": 0.6642, "step": 33858 }, { "epoch": 2.04, "learning_rate": 3.6965279978271696e-05, "loss": 0.751, "step": 33860 }, { "epoch": 2.04, "learning_rate": 3.696377106815747e-05, "loss": 0.6889, "step": 33862 }, { "epoch": 2.04, "learning_rate": 3.696226215804325e-05, "loss": 0.5425, "step": 33864 }, { "epoch": 2.04, "learning_rate": 3.696075324792902e-05, "loss": 0.6867, "step": 33866 }, { "epoch": 2.04, "learning_rate": 3.6959244337814795e-05, "loss": 0.5231, "step": 33868 }, { "epoch": 2.04, "learning_rate": 3.6957735427700575e-05, "loss": 0.7121, "step": 33870 }, { "epoch": 2.04, "learning_rate": 3.695622651758635e-05, "loss": 0.6238, "step": 33872 }, { "epoch": 2.04, "learning_rate": 3.695471760747212e-05, "loss": 0.7354, "step": 33874 }, { "epoch": 2.04, "learning_rate": 3.69532086973579e-05, "loss": 0.4182, "step": 33876 }, { "epoch": 2.04, "learning_rate": 3.6951699787243674e-05, "loss": 0.6532, "step": 33878 }, { "epoch": 2.04, "learning_rate": 3.6950190877129454e-05, "loss": 0.5952, "step": 33880 }, { "epoch": 2.05, "learning_rate": 3.694868196701523e-05, "loss": 0.5579, "step": 33882 }, { "epoch": 2.05, "learning_rate": 3.6947173056901e-05, "loss": 0.7829, "step": 33884 }, { "epoch": 2.05, "learning_rate": 3.694566414678678e-05, "loss": 0.6536, "step": 33886 }, { "epoch": 2.05, "learning_rate": 3.6944155236672554e-05, "loss": 0.7616, "step": 33888 }, { "epoch": 2.05, "learning_rate": 3.694264632655833e-05, "loss": 0.6646, "step": 33890 }, { "epoch": 2.05, "learning_rate": 3.6941137416444107e-05, "loss": 0.743, "step": 33892 }, { "epoch": 2.05, "learning_rate": 3.693962850632988e-05, "loss": 0.643, "step": 33894 }, { "epoch": 2.05, "learning_rate": 3.693811959621565e-05, "loss": 0.5499, "step": 33896 }, { "epoch": 2.05, "learning_rate": 3.693661068610143e-05, "loss": 0.6946, "step": 33898 }, { "epoch": 2.05, "learning_rate": 3.6935101775987206e-05, "loss": 0.773, "step": 33900 }, { "epoch": 2.05, "learning_rate": 3.693359286587298e-05, "loss": 0.5946, "step": 33902 }, { "epoch": 2.05, "learning_rate": 3.693208395575876e-05, "loss": 0.5597, "step": 33904 }, { "epoch": 2.05, "learning_rate": 3.693057504564453e-05, "loss": 0.5868, "step": 33906 }, { "epoch": 2.05, "learning_rate": 3.692906613553031e-05, "loss": 1.2357, "step": 33908 }, { "epoch": 2.05, "learning_rate": 3.6927557225416085e-05, "loss": 0.5999, "step": 33910 }, { "epoch": 2.05, "learning_rate": 3.692604831530186e-05, "loss": 0.5231, "step": 33912 }, { "epoch": 2.05, "learning_rate": 3.692453940518764e-05, "loss": 0.825, "step": 33914 }, { "epoch": 2.05, "learning_rate": 3.692303049507341e-05, "loss": 0.604, "step": 33916 }, { "epoch": 2.05, "learning_rate": 3.6921521584959184e-05, "loss": 0.6822, "step": 33918 }, { "epoch": 2.05, "learning_rate": 3.6920012674844964e-05, "loss": 0.6159, "step": 33920 }, { "epoch": 2.05, "learning_rate": 3.691850376473074e-05, "loss": 0.6466, "step": 33922 }, { "epoch": 2.05, "learning_rate": 3.691699485461651e-05, "loss": 0.5342, "step": 33924 }, { "epoch": 2.05, "learning_rate": 3.691548594450229e-05, "loss": 0.4591, "step": 33926 }, { "epoch": 2.05, "learning_rate": 3.6913977034388064e-05, "loss": 0.5373, "step": 33928 }, { "epoch": 2.05, "learning_rate": 3.6912468124273843e-05, "loss": 0.6515, "step": 33930 }, { "epoch": 2.05, "learning_rate": 3.691095921415961e-05, "loss": 0.467, "step": 33932 }, { "epoch": 2.05, "learning_rate": 3.690945030404539e-05, "loss": 0.5894, "step": 33934 }, { "epoch": 2.05, "learning_rate": 3.690794139393117e-05, "loss": 0.3286, "step": 33936 }, { "epoch": 2.05, "learning_rate": 3.6906432483816936e-05, "loss": 0.6477, "step": 33938 }, { "epoch": 2.05, "learning_rate": 3.6904923573702716e-05, "loss": 0.6434, "step": 33940 }, { "epoch": 2.05, "learning_rate": 3.6903414663588496e-05, "loss": 0.4167, "step": 33942 }, { "epoch": 2.05, "learning_rate": 3.690190575347427e-05, "loss": 0.3866, "step": 33944 }, { "epoch": 2.05, "learning_rate": 3.690039684336004e-05, "loss": 0.5739, "step": 33946 }, { "epoch": 2.05, "learning_rate": 3.6898887933245815e-05, "loss": 0.6856, "step": 33948 }, { "epoch": 2.05, "learning_rate": 3.6897379023131595e-05, "loss": 0.9318, "step": 33950 }, { "epoch": 2.05, "learning_rate": 3.689587011301737e-05, "loss": 0.6412, "step": 33952 }, { "epoch": 2.05, "learning_rate": 3.689436120290314e-05, "loss": 0.5207, "step": 33954 }, { "epoch": 2.05, "learning_rate": 3.689285229278892e-05, "loss": 0.7762, "step": 33956 }, { "epoch": 2.05, "learning_rate": 3.68913433826747e-05, "loss": 0.5209, "step": 33958 }, { "epoch": 2.05, "learning_rate": 3.688983447256047e-05, "loss": 0.453, "step": 33960 }, { "epoch": 2.05, "learning_rate": 3.688832556244625e-05, "loss": 0.6903, "step": 33962 }, { "epoch": 2.05, "learning_rate": 3.688681665233202e-05, "loss": 0.6294, "step": 33964 }, { "epoch": 2.05, "learning_rate": 3.68853077422178e-05, "loss": 0.7351, "step": 33966 }, { "epoch": 2.05, "learning_rate": 3.6883798832103573e-05, "loss": 0.6851, "step": 33968 }, { "epoch": 2.05, "learning_rate": 3.6882289921989347e-05, "loss": 0.587, "step": 33970 }, { "epoch": 2.05, "learning_rate": 3.6880781011875126e-05, "loss": 0.6296, "step": 33972 }, { "epoch": 2.05, "learning_rate": 3.68792721017609e-05, "loss": 0.4783, "step": 33974 }, { "epoch": 2.05, "learning_rate": 3.687776319164667e-05, "loss": 0.9522, "step": 33976 }, { "epoch": 2.05, "learning_rate": 3.687625428153245e-05, "loss": 0.7167, "step": 33978 }, { "epoch": 2.05, "learning_rate": 3.6874745371418226e-05, "loss": 0.775, "step": 33980 }, { "epoch": 2.05, "learning_rate": 3.6873236461304e-05, "loss": 0.491, "step": 33982 }, { "epoch": 2.05, "learning_rate": 3.687172755118978e-05, "loss": 0.5208, "step": 33984 }, { "epoch": 2.05, "learning_rate": 3.687021864107555e-05, "loss": 0.6256, "step": 33986 }, { "epoch": 2.05, "learning_rate": 3.6868709730961325e-05, "loss": 0.7181, "step": 33988 }, { "epoch": 2.05, "learning_rate": 3.6867200820847105e-05, "loss": 0.3888, "step": 33990 }, { "epoch": 2.05, "learning_rate": 3.686569191073288e-05, "loss": 0.5031, "step": 33992 }, { "epoch": 2.05, "learning_rate": 3.686418300061866e-05, "loss": 0.5665, "step": 33994 }, { "epoch": 2.05, "learning_rate": 3.686267409050443e-05, "loss": 0.7365, "step": 33996 }, { "epoch": 2.05, "learning_rate": 3.6861165180390204e-05, "loss": 0.6896, "step": 33998 }, { "epoch": 2.05, "learning_rate": 3.6859656270275984e-05, "loss": 0.8189, "step": 34000 }, { "epoch": 2.05, "learning_rate": 3.685814736016176e-05, "loss": 0.5495, "step": 34002 }, { "epoch": 2.05, "learning_rate": 3.685663845004753e-05, "loss": 0.6006, "step": 34004 }, { "epoch": 2.05, "learning_rate": 3.685512953993331e-05, "loss": 0.5791, "step": 34006 }, { "epoch": 2.05, "learning_rate": 3.6853620629819083e-05, "loss": 0.4541, "step": 34008 }, { "epoch": 2.05, "learning_rate": 3.6852111719704857e-05, "loss": 0.637, "step": 34010 }, { "epoch": 2.05, "learning_rate": 3.6850602809590636e-05, "loss": 0.5754, "step": 34012 }, { "epoch": 2.05, "learning_rate": 3.684909389947641e-05, "loss": 0.5678, "step": 34014 }, { "epoch": 2.05, "learning_rate": 3.684758498936218e-05, "loss": 0.579, "step": 34016 }, { "epoch": 2.05, "learning_rate": 3.684607607924796e-05, "loss": 0.5706, "step": 34018 }, { "epoch": 2.05, "learning_rate": 3.6844567169133736e-05, "loss": 0.7679, "step": 34020 }, { "epoch": 2.05, "learning_rate": 3.6843058259019516e-05, "loss": 0.6539, "step": 34022 }, { "epoch": 2.05, "learning_rate": 3.684154934890529e-05, "loss": 0.8377, "step": 34024 }, { "epoch": 2.05, "learning_rate": 3.684004043879106e-05, "loss": 0.6761, "step": 34026 }, { "epoch": 2.05, "learning_rate": 3.683853152867684e-05, "loss": 0.6201, "step": 34028 }, { "epoch": 2.05, "learning_rate": 3.6837022618562615e-05, "loss": 0.5145, "step": 34030 }, { "epoch": 2.05, "learning_rate": 3.683551370844839e-05, "loss": 0.6515, "step": 34032 }, { "epoch": 2.05, "learning_rate": 3.683400479833417e-05, "loss": 0.6874, "step": 34034 }, { "epoch": 2.05, "learning_rate": 3.683249588821994e-05, "loss": 0.8638, "step": 34036 }, { "epoch": 2.05, "learning_rate": 3.6830986978105714e-05, "loss": 0.6711, "step": 34038 }, { "epoch": 2.05, "learning_rate": 3.6829478067991494e-05, "loss": 0.8777, "step": 34040 }, { "epoch": 2.05, "learning_rate": 3.682796915787727e-05, "loss": 0.6749, "step": 34042 }, { "epoch": 2.05, "learning_rate": 3.682646024776305e-05, "loss": 0.7702, "step": 34044 }, { "epoch": 2.05, "learning_rate": 3.6824951337648813e-05, "loss": 0.5284, "step": 34046 }, { "epoch": 2.06, "learning_rate": 3.682344242753459e-05, "loss": 0.5536, "step": 34048 }, { "epoch": 2.06, "learning_rate": 3.682193351742037e-05, "loss": 0.6826, "step": 34050 }, { "epoch": 2.06, "learning_rate": 3.682042460730614e-05, "loss": 0.6863, "step": 34052 }, { "epoch": 2.06, "learning_rate": 3.681891569719192e-05, "loss": 0.8018, "step": 34054 }, { "epoch": 2.06, "learning_rate": 3.68174067870777e-05, "loss": 0.761, "step": 34056 }, { "epoch": 2.06, "learning_rate": 3.681589787696347e-05, "loss": 0.7171, "step": 34058 }, { "epoch": 2.06, "learning_rate": 3.6814388966849246e-05, "loss": 0.6426, "step": 34060 }, { "epoch": 2.06, "learning_rate": 3.681288005673502e-05, "loss": 0.4672, "step": 34062 }, { "epoch": 2.06, "learning_rate": 3.68113711466208e-05, "loss": 0.5555, "step": 34064 }, { "epoch": 2.06, "learning_rate": 3.680986223650657e-05, "loss": 0.5476, "step": 34066 }, { "epoch": 2.06, "learning_rate": 3.6808353326392345e-05, "loss": 0.8643, "step": 34068 }, { "epoch": 2.06, "learning_rate": 3.6806844416278125e-05, "loss": 0.6122, "step": 34070 }, { "epoch": 2.06, "learning_rate": 3.6805335506163905e-05, "loss": 0.7293, "step": 34072 }, { "epoch": 2.06, "learning_rate": 3.680382659604967e-05, "loss": 0.6503, "step": 34074 }, { "epoch": 2.06, "learning_rate": 3.680231768593545e-05, "loss": 0.5798, "step": 34076 }, { "epoch": 2.06, "learning_rate": 3.6800808775821224e-05, "loss": 0.5156, "step": 34078 }, { "epoch": 2.06, "learning_rate": 3.6799299865707004e-05, "loss": 0.6922, "step": 34080 }, { "epoch": 2.06, "learning_rate": 3.679779095559278e-05, "loss": 0.7994, "step": 34082 }, { "epoch": 2.06, "learning_rate": 3.679628204547855e-05, "loss": 0.6477, "step": 34084 }, { "epoch": 2.06, "learning_rate": 3.679477313536433e-05, "loss": 0.5333, "step": 34086 }, { "epoch": 2.06, "learning_rate": 3.67932642252501e-05, "loss": 0.5571, "step": 34088 }, { "epoch": 2.06, "learning_rate": 3.6791755315135876e-05, "loss": 0.9063, "step": 34090 }, { "epoch": 2.06, "learning_rate": 3.6790246405021656e-05, "loss": 0.5036, "step": 34092 }, { "epoch": 2.06, "learning_rate": 3.678873749490743e-05, "loss": 0.6012, "step": 34094 }, { "epoch": 2.06, "learning_rate": 3.67872285847932e-05, "loss": 0.6054, "step": 34096 }, { "epoch": 2.06, "learning_rate": 3.678571967467898e-05, "loss": 0.5229, "step": 34098 }, { "epoch": 2.06, "learning_rate": 3.6784210764564756e-05, "loss": 0.5268, "step": 34100 }, { "epoch": 2.06, "learning_rate": 3.678270185445053e-05, "loss": 0.7835, "step": 34102 }, { "epoch": 2.06, "learning_rate": 3.678119294433631e-05, "loss": 0.8773, "step": 34104 }, { "epoch": 2.06, "learning_rate": 3.677968403422208e-05, "loss": 0.6183, "step": 34106 }, { "epoch": 2.06, "learning_rate": 3.677817512410786e-05, "loss": 0.6559, "step": 34108 }, { "epoch": 2.06, "learning_rate": 3.6776666213993635e-05, "loss": 0.4924, "step": 34110 }, { "epoch": 2.06, "learning_rate": 3.677515730387941e-05, "loss": 0.6648, "step": 34112 }, { "epoch": 2.06, "learning_rate": 3.677364839376519e-05, "loss": 0.4742, "step": 34114 }, { "epoch": 2.06, "learning_rate": 3.677213948365096e-05, "loss": 0.5595, "step": 34116 }, { "epoch": 2.06, "learning_rate": 3.6770630573536734e-05, "loss": 0.5482, "step": 34118 }, { "epoch": 2.06, "learning_rate": 3.6769121663422514e-05, "loss": 0.5415, "step": 34120 }, { "epoch": 2.06, "learning_rate": 3.676761275330829e-05, "loss": 0.5693, "step": 34122 }, { "epoch": 2.06, "learning_rate": 3.676610384319406e-05, "loss": 0.6228, "step": 34124 }, { "epoch": 2.06, "learning_rate": 3.676459493307984e-05, "loss": 0.6281, "step": 34126 }, { "epoch": 2.06, "learning_rate": 3.676308602296561e-05, "loss": 0.6131, "step": 34128 }, { "epoch": 2.06, "learning_rate": 3.676157711285139e-05, "loss": 0.735, "step": 34130 }, { "epoch": 2.06, "learning_rate": 3.6760068202737166e-05, "loss": 0.8454, "step": 34132 }, { "epoch": 2.06, "learning_rate": 3.675855929262294e-05, "loss": 0.5658, "step": 34134 }, { "epoch": 2.06, "learning_rate": 3.675705038250872e-05, "loss": 0.6931, "step": 34136 }, { "epoch": 2.06, "learning_rate": 3.675554147239449e-05, "loss": 0.558, "step": 34138 }, { "epoch": 2.06, "learning_rate": 3.6754032562280266e-05, "loss": 0.8912, "step": 34140 }, { "epoch": 2.06, "learning_rate": 3.6752523652166045e-05, "loss": 0.7134, "step": 34142 }, { "epoch": 2.06, "learning_rate": 3.675101474205182e-05, "loss": 0.9017, "step": 34144 }, { "epoch": 2.06, "learning_rate": 3.674950583193759e-05, "loss": 0.5251, "step": 34146 }, { "epoch": 2.06, "learning_rate": 3.674799692182337e-05, "loss": 0.6, "step": 34148 }, { "epoch": 2.06, "learning_rate": 3.6746488011709145e-05, "loss": 0.6155, "step": 34150 }, { "epoch": 2.06, "learning_rate": 3.674497910159492e-05, "loss": 0.5117, "step": 34152 }, { "epoch": 2.06, "learning_rate": 3.67434701914807e-05, "loss": 0.7636, "step": 34154 }, { "epoch": 2.06, "learning_rate": 3.674196128136647e-05, "loss": 1.1006, "step": 34156 }, { "epoch": 2.06, "learning_rate": 3.674045237125225e-05, "loss": 0.613, "step": 34158 }, { "epoch": 2.06, "learning_rate": 3.673894346113802e-05, "loss": 0.5741, "step": 34160 }, { "epoch": 2.06, "learning_rate": 3.67374345510238e-05, "loss": 0.4695, "step": 34162 }, { "epoch": 2.06, "learning_rate": 3.673592564090958e-05, "loss": 0.8212, "step": 34164 }, { "epoch": 2.06, "learning_rate": 3.673441673079534e-05, "loss": 0.7145, "step": 34166 }, { "epoch": 2.06, "learning_rate": 3.673290782068112e-05, "loss": 0.6394, "step": 34168 }, { "epoch": 2.06, "learning_rate": 3.67313989105669e-05, "loss": 0.448, "step": 34170 }, { "epoch": 2.06, "learning_rate": 3.6729890000452676e-05, "loss": 0.5024, "step": 34172 }, { "epoch": 2.06, "learning_rate": 3.672838109033845e-05, "loss": 0.7372, "step": 34174 }, { "epoch": 2.06, "learning_rate": 3.672687218022422e-05, "loss": 0.7597, "step": 34176 }, { "epoch": 2.06, "learning_rate": 3.672536327011e-05, "loss": 0.4826, "step": 34178 }, { "epoch": 2.06, "learning_rate": 3.6723854359995775e-05, "loss": 0.6685, "step": 34180 }, { "epoch": 2.06, "learning_rate": 3.672234544988155e-05, "loss": 0.5899, "step": 34182 }, { "epoch": 2.06, "learning_rate": 3.672083653976733e-05, "loss": 0.5286, "step": 34184 }, { "epoch": 2.06, "learning_rate": 3.671932762965311e-05, "loss": 0.6931, "step": 34186 }, { "epoch": 2.06, "learning_rate": 3.6717818719538875e-05, "loss": 0.772, "step": 34188 }, { "epoch": 2.06, "learning_rate": 3.6716309809424655e-05, "loss": 0.8788, "step": 34190 }, { "epoch": 2.06, "learning_rate": 3.671480089931043e-05, "loss": 0.6569, "step": 34192 }, { "epoch": 2.06, "learning_rate": 3.671329198919621e-05, "loss": 0.4798, "step": 34194 }, { "epoch": 2.06, "learning_rate": 3.671178307908198e-05, "loss": 0.5162, "step": 34196 }, { "epoch": 2.06, "learning_rate": 3.6710274168967754e-05, "loss": 0.4972, "step": 34198 }, { "epoch": 2.06, "learning_rate": 3.6708765258853534e-05, "loss": 0.8104, "step": 34200 }, { "epoch": 2.06, "learning_rate": 3.670725634873931e-05, "loss": 0.5158, "step": 34202 }, { "epoch": 2.06, "learning_rate": 3.670574743862508e-05, "loss": 0.7784, "step": 34204 }, { "epoch": 2.06, "learning_rate": 3.670423852851086e-05, "loss": 0.5186, "step": 34206 }, { "epoch": 2.06, "learning_rate": 3.670272961839663e-05, "loss": 0.6651, "step": 34208 }, { "epoch": 2.06, "learning_rate": 3.6701220708282406e-05, "loss": 0.5171, "step": 34210 }, { "epoch": 2.06, "learning_rate": 3.6699711798168186e-05, "loss": 0.4752, "step": 34212 }, { "epoch": 2.07, "learning_rate": 3.669820288805396e-05, "loss": 0.5401, "step": 34214 }, { "epoch": 2.07, "learning_rate": 3.669669397793973e-05, "loss": 0.7811, "step": 34216 }, { "epoch": 2.07, "learning_rate": 3.669518506782551e-05, "loss": 0.5297, "step": 34218 }, { "epoch": 2.07, "learning_rate": 3.6693676157711285e-05, "loss": 0.866, "step": 34220 }, { "epoch": 2.07, "learning_rate": 3.6692167247597065e-05, "loss": 0.6894, "step": 34222 }, { "epoch": 2.07, "learning_rate": 3.669065833748284e-05, "loss": 0.4137, "step": 34224 }, { "epoch": 2.07, "learning_rate": 3.668914942736861e-05, "loss": 0.6065, "step": 34226 }, { "epoch": 2.07, "learning_rate": 3.668764051725439e-05, "loss": 0.4614, "step": 34228 }, { "epoch": 2.07, "learning_rate": 3.6686131607140165e-05, "loss": 0.4733, "step": 34230 }, { "epoch": 2.07, "learning_rate": 3.668462269702594e-05, "loss": 0.6488, "step": 34232 }, { "epoch": 2.07, "learning_rate": 3.668311378691172e-05, "loss": 0.7742, "step": 34234 }, { "epoch": 2.07, "learning_rate": 3.668160487679749e-05, "loss": 0.572, "step": 34236 }, { "epoch": 2.07, "learning_rate": 3.6680095966683264e-05, "loss": 0.7538, "step": 34238 }, { "epoch": 2.07, "learning_rate": 3.6678587056569044e-05, "loss": 0.7199, "step": 34240 }, { "epoch": 2.07, "learning_rate": 3.667707814645482e-05, "loss": 0.5332, "step": 34242 }, { "epoch": 2.07, "learning_rate": 3.66755692363406e-05, "loss": 0.5368, "step": 34244 }, { "epoch": 2.07, "learning_rate": 3.667406032622637e-05, "loss": 0.5834, "step": 34246 }, { "epoch": 2.07, "learning_rate": 3.667255141611214e-05, "loss": 0.75, "step": 34248 }, { "epoch": 2.07, "learning_rate": 3.667104250599792e-05, "loss": 0.5604, "step": 34250 }, { "epoch": 2.07, "learning_rate": 3.6669533595883696e-05, "loss": 0.7168, "step": 34252 }, { "epoch": 2.07, "learning_rate": 3.666802468576947e-05, "loss": 0.8169, "step": 34254 }, { "epoch": 2.07, "learning_rate": 3.666651577565525e-05, "loss": 0.6153, "step": 34256 }, { "epoch": 2.07, "learning_rate": 3.666500686554102e-05, "loss": 0.7024, "step": 34258 }, { "epoch": 2.07, "learning_rate": 3.6663497955426795e-05, "loss": 0.7675, "step": 34260 }, { "epoch": 2.07, "learning_rate": 3.6661989045312575e-05, "loss": 0.6825, "step": 34262 }, { "epoch": 2.07, "learning_rate": 3.666048013519835e-05, "loss": 0.7958, "step": 34264 }, { "epoch": 2.07, "learning_rate": 3.665897122508412e-05, "loss": 0.591, "step": 34266 }, { "epoch": 2.07, "learning_rate": 3.66574623149699e-05, "loss": 0.7574, "step": 34268 }, { "epoch": 2.07, "learning_rate": 3.6655953404855675e-05, "loss": 0.4178, "step": 34270 }, { "epoch": 2.07, "learning_rate": 3.6654444494741454e-05, "loss": 0.5883, "step": 34272 }, { "epoch": 2.07, "learning_rate": 3.665293558462722e-05, "loss": 0.927, "step": 34274 }, { "epoch": 2.07, "learning_rate": 3.6651426674513e-05, "loss": 0.6267, "step": 34276 }, { "epoch": 2.07, "learning_rate": 3.664991776439878e-05, "loss": 0.588, "step": 34278 }, { "epoch": 2.07, "learning_rate": 3.664840885428455e-05, "loss": 0.5328, "step": 34280 }, { "epoch": 2.07, "learning_rate": 3.664689994417033e-05, "loss": 0.6075, "step": 34282 }, { "epoch": 2.07, "learning_rate": 3.664539103405611e-05, "loss": 0.7603, "step": 34284 }, { "epoch": 2.07, "learning_rate": 3.664388212394188e-05, "loss": 0.6928, "step": 34286 }, { "epoch": 2.07, "learning_rate": 3.664237321382765e-05, "loss": 0.8573, "step": 34288 }, { "epoch": 2.07, "learning_rate": 3.6640864303713426e-05, "loss": 0.5641, "step": 34290 }, { "epoch": 2.07, "learning_rate": 3.6639355393599206e-05, "loss": 0.5465, "step": 34292 }, { "epoch": 2.07, "learning_rate": 3.6637846483484986e-05, "loss": 0.4739, "step": 34294 }, { "epoch": 2.07, "learning_rate": 3.663633757337075e-05, "loss": 0.8598, "step": 34296 }, { "epoch": 2.07, "learning_rate": 3.663482866325653e-05, "loss": 0.7047, "step": 34298 }, { "epoch": 2.07, "learning_rate": 3.663331975314231e-05, "loss": 0.6037, "step": 34300 }, { "epoch": 2.07, "learning_rate": 3.663181084302808e-05, "loss": 0.8411, "step": 34302 }, { "epoch": 2.07, "learning_rate": 3.663030193291386e-05, "loss": 0.4216, "step": 34304 }, { "epoch": 2.07, "learning_rate": 3.662879302279963e-05, "loss": 0.792, "step": 34306 }, { "epoch": 2.07, "learning_rate": 3.662728411268541e-05, "loss": 0.5526, "step": 34308 }, { "epoch": 2.07, "learning_rate": 3.6625775202571184e-05, "loss": 0.5249, "step": 34310 }, { "epoch": 2.07, "learning_rate": 3.662426629245696e-05, "loss": 0.726, "step": 34312 }, { "epoch": 2.07, "learning_rate": 3.662275738234274e-05, "loss": 0.5352, "step": 34314 }, { "epoch": 2.07, "learning_rate": 3.662124847222851e-05, "loss": 0.5745, "step": 34316 }, { "epoch": 2.07, "learning_rate": 3.6619739562114284e-05, "loss": 0.6006, "step": 34318 }, { "epoch": 2.07, "learning_rate": 3.6618230652000064e-05, "loss": 0.5067, "step": 34320 }, { "epoch": 2.07, "learning_rate": 3.661672174188584e-05, "loss": 0.5173, "step": 34322 }, { "epoch": 2.07, "learning_rate": 3.661521283177161e-05, "loss": 0.5235, "step": 34324 }, { "epoch": 2.07, "learning_rate": 3.661370392165739e-05, "loss": 0.6228, "step": 34326 }, { "epoch": 2.07, "learning_rate": 3.661219501154316e-05, "loss": 0.3787, "step": 34328 }, { "epoch": 2.07, "learning_rate": 3.6610686101428936e-05, "loss": 0.6801, "step": 34330 }, { "epoch": 2.07, "learning_rate": 3.6609177191314716e-05, "loss": 0.5455, "step": 34332 }, { "epoch": 2.07, "learning_rate": 3.660766828120049e-05, "loss": 0.6305, "step": 34334 }, { "epoch": 2.07, "learning_rate": 3.660615937108627e-05, "loss": 0.5203, "step": 34336 }, { "epoch": 2.07, "learning_rate": 3.660465046097204e-05, "loss": 0.4288, "step": 34338 }, { "epoch": 2.07, "learning_rate": 3.6603141550857815e-05, "loss": 0.6164, "step": 34340 }, { "epoch": 2.07, "learning_rate": 3.6601632640743595e-05, "loss": 0.5619, "step": 34342 }, { "epoch": 2.07, "learning_rate": 3.660012373062937e-05, "loss": 0.8013, "step": 34344 }, { "epoch": 2.07, "learning_rate": 3.659861482051514e-05, "loss": 0.5817, "step": 34346 }, { "epoch": 2.07, "learning_rate": 3.659710591040092e-05, "loss": 0.655, "step": 34348 }, { "epoch": 2.07, "learning_rate": 3.6595597000286694e-05, "loss": 0.4861, "step": 34350 }, { "epoch": 2.07, "learning_rate": 3.659408809017247e-05, "loss": 0.6716, "step": 34352 }, { "epoch": 2.07, "learning_rate": 3.659257918005825e-05, "loss": 0.6, "step": 34354 }, { "epoch": 2.07, "learning_rate": 3.659107026994402e-05, "loss": 0.5039, "step": 34356 }, { "epoch": 2.07, "learning_rate": 3.65895613598298e-05, "loss": 0.8387, "step": 34358 }, { "epoch": 2.07, "learning_rate": 3.6588052449715574e-05, "loss": 0.5682, "step": 34360 }, { "epoch": 2.07, "learning_rate": 3.658654353960135e-05, "loss": 0.4079, "step": 34362 }, { "epoch": 2.07, "learning_rate": 3.658503462948713e-05, "loss": 0.8446, "step": 34364 }, { "epoch": 2.07, "learning_rate": 3.65835257193729e-05, "loss": 0.4415, "step": 34366 }, { "epoch": 2.07, "learning_rate": 3.658201680925867e-05, "loss": 0.4388, "step": 34368 }, { "epoch": 2.07, "learning_rate": 3.658050789914445e-05, "loss": 0.5469, "step": 34370 }, { "epoch": 2.07, "learning_rate": 3.6578998989030226e-05, "loss": 0.6485, "step": 34372 }, { "epoch": 2.07, "learning_rate": 3.6577490078916e-05, "loss": 0.8618, "step": 34374 }, { "epoch": 2.07, "learning_rate": 3.657598116880178e-05, "loss": 0.4794, "step": 34376 }, { "epoch": 2.07, "learning_rate": 3.657447225868755e-05, "loss": 0.5706, "step": 34378 }, { "epoch": 2.08, "learning_rate": 3.6572963348573325e-05, "loss": 0.4146, "step": 34380 }, { "epoch": 2.08, "learning_rate": 3.6571454438459105e-05, "loss": 0.6924, "step": 34382 }, { "epoch": 2.08, "learning_rate": 3.656994552834488e-05, "loss": 0.475, "step": 34384 }, { "epoch": 2.08, "learning_rate": 3.656843661823066e-05, "loss": 0.8337, "step": 34386 }, { "epoch": 2.08, "learning_rate": 3.6566927708116424e-05, "loss": 0.6508, "step": 34388 }, { "epoch": 2.08, "learning_rate": 3.6565418798002204e-05, "loss": 0.5686, "step": 34390 }, { "epoch": 2.08, "learning_rate": 3.6563909887887984e-05, "loss": 0.5712, "step": 34392 }, { "epoch": 2.08, "learning_rate": 3.656240097777375e-05, "loss": 0.5436, "step": 34394 }, { "epoch": 2.08, "learning_rate": 3.656089206765953e-05, "loss": 0.4988, "step": 34396 }, { "epoch": 2.08, "learning_rate": 3.655938315754531e-05, "loss": 0.6347, "step": 34398 }, { "epoch": 2.08, "learning_rate": 3.6557874247431084e-05, "loss": 0.6839, "step": 34400 }, { "epoch": 2.08, "learning_rate": 3.655636533731686e-05, "loss": 0.587, "step": 34402 }, { "epoch": 2.08, "learning_rate": 3.655485642720263e-05, "loss": 0.6827, "step": 34404 }, { "epoch": 2.08, "learning_rate": 3.655334751708841e-05, "loss": 0.5469, "step": 34406 }, { "epoch": 2.08, "learning_rate": 3.655183860697419e-05, "loss": 0.5909, "step": 34408 }, { "epoch": 2.08, "learning_rate": 3.6550329696859956e-05, "loss": 0.6735, "step": 34410 }, { "epoch": 2.08, "learning_rate": 3.6548820786745736e-05, "loss": 0.5593, "step": 34412 }, { "epoch": 2.08, "learning_rate": 3.6547311876631516e-05, "loss": 0.43, "step": 34414 }, { "epoch": 2.08, "learning_rate": 3.654580296651728e-05, "loss": 0.5261, "step": 34416 }, { "epoch": 2.08, "learning_rate": 3.654429405640306e-05, "loss": 0.8663, "step": 34418 }, { "epoch": 2.08, "learning_rate": 3.6542785146288835e-05, "loss": 0.5333, "step": 34420 }, { "epoch": 2.08, "learning_rate": 3.6541276236174615e-05, "loss": 0.7775, "step": 34422 }, { "epoch": 2.08, "learning_rate": 3.653976732606039e-05, "loss": 0.4739, "step": 34424 }, { "epoch": 2.08, "learning_rate": 3.653825841594616e-05, "loss": 0.7268, "step": 34426 }, { "epoch": 2.08, "learning_rate": 3.653674950583194e-05, "loss": 0.5804, "step": 34428 }, { "epoch": 2.08, "learning_rate": 3.6535240595717714e-05, "loss": 0.5322, "step": 34430 }, { "epoch": 2.08, "learning_rate": 3.653373168560349e-05, "loss": 0.5984, "step": 34432 }, { "epoch": 2.08, "learning_rate": 3.653222277548927e-05, "loss": 0.475, "step": 34434 }, { "epoch": 2.08, "learning_rate": 3.653071386537504e-05, "loss": 0.8712, "step": 34436 }, { "epoch": 2.08, "learning_rate": 3.6529204955260814e-05, "loss": 0.4716, "step": 34438 }, { "epoch": 2.08, "learning_rate": 3.6527696045146594e-05, "loss": 0.5348, "step": 34440 }, { "epoch": 2.08, "learning_rate": 3.652618713503237e-05, "loss": 0.6539, "step": 34442 }, { "epoch": 2.08, "learning_rate": 3.652467822491814e-05, "loss": 0.7928, "step": 34444 }, { "epoch": 2.08, "learning_rate": 3.652316931480392e-05, "loss": 0.6317, "step": 34446 }, { "epoch": 2.08, "learning_rate": 3.652166040468969e-05, "loss": 0.8605, "step": 34448 }, { "epoch": 2.08, "learning_rate": 3.652015149457547e-05, "loss": 0.4214, "step": 34450 }, { "epoch": 2.08, "learning_rate": 3.6518642584461246e-05, "loss": 0.6404, "step": 34452 }, { "epoch": 2.08, "learning_rate": 3.651713367434702e-05, "loss": 0.4854, "step": 34454 }, { "epoch": 2.08, "learning_rate": 3.65156247642328e-05, "loss": 0.6344, "step": 34456 }, { "epoch": 2.08, "learning_rate": 3.651411585411857e-05, "loss": 0.4756, "step": 34458 }, { "epoch": 2.08, "learning_rate": 3.6512606944004345e-05, "loss": 0.3535, "step": 34460 }, { "epoch": 2.08, "learning_rate": 3.6511098033890125e-05, "loss": 0.6081, "step": 34462 }, { "epoch": 2.08, "learning_rate": 3.65095891237759e-05, "loss": 0.6064, "step": 34464 }, { "epoch": 2.08, "learning_rate": 3.650808021366167e-05, "loss": 0.5858, "step": 34466 }, { "epoch": 2.08, "learning_rate": 3.650657130354745e-05, "loss": 0.6956, "step": 34468 }, { "epoch": 2.08, "learning_rate": 3.6505062393433224e-05, "loss": 0.8101, "step": 34470 }, { "epoch": 2.08, "learning_rate": 3.6503553483319004e-05, "loss": 0.7836, "step": 34472 }, { "epoch": 2.08, "learning_rate": 3.650204457320478e-05, "loss": 0.5486, "step": 34474 }, { "epoch": 2.08, "learning_rate": 3.650053566309055e-05, "loss": 0.5767, "step": 34476 }, { "epoch": 2.08, "learning_rate": 3.649902675297633e-05, "loss": 0.7331, "step": 34478 }, { "epoch": 2.08, "learning_rate": 3.6497517842862103e-05, "loss": 0.7816, "step": 34480 }, { "epoch": 2.08, "learning_rate": 3.6496008932747877e-05, "loss": 0.8138, "step": 34482 }, { "epoch": 2.08, "learning_rate": 3.6494500022633656e-05, "loss": 0.5372, "step": 34484 }, { "epoch": 2.08, "learning_rate": 3.649299111251943e-05, "loss": 0.6479, "step": 34486 }, { "epoch": 2.08, "learning_rate": 3.64914822024052e-05, "loss": 0.6709, "step": 34488 }, { "epoch": 2.08, "learning_rate": 3.648997329229098e-05, "loss": 0.6559, "step": 34490 }, { "epoch": 2.08, "learning_rate": 3.6488464382176756e-05, "loss": 0.6121, "step": 34492 }, { "epoch": 2.08, "learning_rate": 3.648695547206253e-05, "loss": 0.7797, "step": 34494 }, { "epoch": 2.08, "learning_rate": 3.648544656194831e-05, "loss": 0.5305, "step": 34496 }, { "epoch": 2.08, "learning_rate": 3.648393765183408e-05, "loss": 0.5866, "step": 34498 }, { "epoch": 2.08, "learning_rate": 3.648242874171986e-05, "loss": 0.7974, "step": 34500 }, { "epoch": 2.08, "learning_rate": 3.648091983160563e-05, "loss": 0.3636, "step": 34502 }, { "epoch": 2.08, "learning_rate": 3.647941092149141e-05, "loss": 0.6818, "step": 34504 }, { "epoch": 2.08, "learning_rate": 3.647790201137719e-05, "loss": 0.6184, "step": 34506 }, { "epoch": 2.08, "learning_rate": 3.647639310126296e-05, "loss": 0.5254, "step": 34508 }, { "epoch": 2.08, "learning_rate": 3.6474884191148734e-05, "loss": 0.4914, "step": 34510 }, { "epoch": 2.08, "learning_rate": 3.6473375281034514e-05, "loss": 0.5938, "step": 34512 }, { "epoch": 2.08, "learning_rate": 3.647186637092029e-05, "loss": 0.539, "step": 34514 }, { "epoch": 2.08, "learning_rate": 3.647035746080606e-05, "loss": 0.6511, "step": 34516 }, { "epoch": 2.08, "learning_rate": 3.6468848550691833e-05, "loss": 0.7241, "step": 34518 }, { "epoch": 2.08, "learning_rate": 3.6467339640577613e-05, "loss": 0.7099, "step": 34520 }, { "epoch": 2.08, "learning_rate": 3.646583073046339e-05, "loss": 0.6023, "step": 34522 }, { "epoch": 2.08, "learning_rate": 3.646432182034916e-05, "loss": 0.4542, "step": 34524 }, { "epoch": 2.08, "learning_rate": 3.646281291023494e-05, "loss": 0.7061, "step": 34526 }, { "epoch": 2.08, "learning_rate": 3.646130400012072e-05, "loss": 0.7601, "step": 34528 }, { "epoch": 2.08, "learning_rate": 3.6459795090006486e-05, "loss": 0.4761, "step": 34530 }, { "epoch": 2.08, "learning_rate": 3.6458286179892266e-05, "loss": 0.688, "step": 34532 }, { "epoch": 2.08, "learning_rate": 3.645677726977804e-05, "loss": 0.4891, "step": 34534 }, { "epoch": 2.08, "learning_rate": 3.645526835966382e-05, "loss": 0.394, "step": 34536 }, { "epoch": 2.08, "learning_rate": 3.645375944954959e-05, "loss": 0.7847, "step": 34538 }, { "epoch": 2.08, "learning_rate": 3.6452250539435365e-05, "loss": 0.6154, "step": 34540 }, { "epoch": 2.08, "learning_rate": 3.6450741629321145e-05, "loss": 0.6481, "step": 34542 }, { "epoch": 2.08, "learning_rate": 3.644923271920692e-05, "loss": 0.7634, "step": 34544 }, { "epoch": 2.09, "learning_rate": 3.644772380909269e-05, "loss": 0.6247, "step": 34546 }, { "epoch": 2.09, "learning_rate": 3.644621489897847e-05, "loss": 0.5476, "step": 34548 }, { "epoch": 2.09, "learning_rate": 3.6444705988864244e-05, "loss": 0.6941, "step": 34550 }, { "epoch": 2.09, "learning_rate": 3.644319707875002e-05, "loss": 0.5118, "step": 34552 }, { "epoch": 2.09, "learning_rate": 3.64416881686358e-05, "loss": 0.3041, "step": 34554 }, { "epoch": 2.09, "learning_rate": 3.644017925852157e-05, "loss": 0.397, "step": 34556 }, { "epoch": 2.09, "learning_rate": 3.6438670348407343e-05, "loss": 0.4628, "step": 34558 }, { "epoch": 2.09, "learning_rate": 3.643716143829312e-05, "loss": 0.705, "step": 34560 }, { "epoch": 2.09, "learning_rate": 3.6435652528178896e-05, "loss": 0.5358, "step": 34562 }, { "epoch": 2.09, "learning_rate": 3.6434143618064676e-05, "loss": 0.6505, "step": 34564 }, { "epoch": 2.09, "learning_rate": 3.643263470795045e-05, "loss": 0.4293, "step": 34566 }, { "epoch": 2.09, "learning_rate": 3.643112579783622e-05, "loss": 0.663, "step": 34568 }, { "epoch": 2.09, "learning_rate": 3.6429616887722e-05, "loss": 0.7152, "step": 34570 }, { "epoch": 2.09, "learning_rate": 3.6428107977607776e-05, "loss": 0.717, "step": 34572 }, { "epoch": 2.09, "learning_rate": 3.642659906749355e-05, "loss": 0.6645, "step": 34574 }, { "epoch": 2.09, "learning_rate": 3.642509015737933e-05, "loss": 0.4048, "step": 34576 }, { "epoch": 2.09, "learning_rate": 3.64235812472651e-05, "loss": 0.4077, "step": 34578 }, { "epoch": 2.09, "learning_rate": 3.6422072337150875e-05, "loss": 0.5684, "step": 34580 }, { "epoch": 2.09, "learning_rate": 3.6420563427036655e-05, "loss": 0.6441, "step": 34582 }, { "epoch": 2.09, "learning_rate": 3.641905451692243e-05, "loss": 0.7413, "step": 34584 }, { "epoch": 2.09, "learning_rate": 3.641754560680821e-05, "loss": 0.6672, "step": 34586 }, { "epoch": 2.09, "learning_rate": 3.641603669669398e-05, "loss": 0.6532, "step": 34588 }, { "epoch": 2.09, "learning_rate": 3.6414527786579754e-05, "loss": 0.7618, "step": 34590 }, { "epoch": 2.09, "learning_rate": 3.6413018876465534e-05, "loss": 0.8649, "step": 34592 }, { "epoch": 2.09, "learning_rate": 3.64115099663513e-05, "loss": 0.7052, "step": 34594 }, { "epoch": 2.09, "learning_rate": 3.641000105623708e-05, "loss": 0.6537, "step": 34596 }, { "epoch": 2.09, "learning_rate": 3.640849214612286e-05, "loss": 0.6364, "step": 34598 }, { "epoch": 2.09, "learning_rate": 3.640698323600863e-05, "loss": 0.5054, "step": 34600 }, { "epoch": 2.09, "learning_rate": 3.6405474325894406e-05, "loss": 0.6382, "step": 34602 }, { "epoch": 2.09, "learning_rate": 3.6403965415780186e-05, "loss": 0.7716, "step": 34604 }, { "epoch": 2.09, "learning_rate": 3.640245650566596e-05, "loss": 0.5414, "step": 34606 }, { "epoch": 2.09, "learning_rate": 3.640094759555173e-05, "loss": 0.629, "step": 34608 }, { "epoch": 2.09, "learning_rate": 3.639943868543751e-05, "loss": 0.5275, "step": 34610 }, { "epoch": 2.09, "learning_rate": 3.6397929775323286e-05, "loss": 0.6536, "step": 34612 }, { "epoch": 2.09, "learning_rate": 3.6396420865209065e-05, "loss": 0.6491, "step": 34614 }, { "epoch": 2.09, "learning_rate": 3.639491195509483e-05, "loss": 0.5394, "step": 34616 }, { "epoch": 2.09, "learning_rate": 3.639340304498061e-05, "loss": 0.7049, "step": 34618 }, { "epoch": 2.09, "learning_rate": 3.639189413486639e-05, "loss": 0.7644, "step": 34620 }, { "epoch": 2.09, "learning_rate": 3.6390385224752165e-05, "loss": 0.6136, "step": 34622 }, { "epoch": 2.09, "learning_rate": 3.638963076969505e-05, "loss": 0.7533, "step": 34624 }, { "epoch": 2.09, "learning_rate": 3.638812185958083e-05, "loss": 0.7062, "step": 34626 }, { "epoch": 2.09, "learning_rate": 3.63866129494666e-05, "loss": 0.6096, "step": 34628 }, { "epoch": 2.09, "learning_rate": 3.638510403935238e-05, "loss": 0.7786, "step": 34630 }, { "epoch": 2.09, "learning_rate": 3.6383595129238154e-05, "loss": 0.4987, "step": 34632 }, { "epoch": 2.09, "learning_rate": 3.638208621912393e-05, "loss": 0.8159, "step": 34634 }, { "epoch": 2.09, "learning_rate": 3.638057730900971e-05, "loss": 0.6739, "step": 34636 }, { "epoch": 2.09, "learning_rate": 3.637906839889548e-05, "loss": 0.6979, "step": 34638 }, { "epoch": 2.09, "learning_rate": 3.637755948878125e-05, "loss": 0.4974, "step": 34640 }, { "epoch": 2.09, "learning_rate": 3.637605057866703e-05, "loss": 0.497, "step": 34642 }, { "epoch": 2.09, "learning_rate": 3.6374541668552806e-05, "loss": 0.4816, "step": 34644 }, { "epoch": 2.09, "learning_rate": 3.637303275843858e-05, "loss": 0.7951, "step": 34646 }, { "epoch": 2.09, "learning_rate": 3.637152384832436e-05, "loss": 0.6617, "step": 34648 }, { "epoch": 2.09, "learning_rate": 3.637001493821013e-05, "loss": 0.6109, "step": 34650 }, { "epoch": 2.09, "learning_rate": 3.6368506028095906e-05, "loss": 0.4846, "step": 34652 }, { "epoch": 2.09, "learning_rate": 3.6366997117981685e-05, "loss": 0.6714, "step": 34654 }, { "epoch": 2.09, "learning_rate": 3.636548820786746e-05, "loss": 0.5597, "step": 34656 }, { "epoch": 2.09, "learning_rate": 3.636397929775324e-05, "loss": 0.749, "step": 34658 }, { "epoch": 2.09, "learning_rate": 3.636247038763901e-05, "loss": 0.6158, "step": 34660 }, { "epoch": 2.09, "learning_rate": 3.6360961477524785e-05, "loss": 0.5497, "step": 34662 }, { "epoch": 2.09, "learning_rate": 3.6359452567410565e-05, "loss": 0.5966, "step": 34664 }, { "epoch": 2.09, "learning_rate": 3.635794365729633e-05, "loss": 0.4638, "step": 34666 }, { "epoch": 2.09, "learning_rate": 3.635643474718211e-05, "loss": 0.4952, "step": 34668 }, { "epoch": 2.09, "learning_rate": 3.635492583706789e-05, "loss": 0.6229, "step": 34670 }, { "epoch": 2.09, "learning_rate": 3.6353416926953664e-05, "loss": 0.6639, "step": 34672 }, { "epoch": 2.09, "learning_rate": 3.635190801683944e-05, "loss": 0.6434, "step": 34674 }, { "epoch": 2.09, "learning_rate": 3.635039910672522e-05, "loss": 0.6541, "step": 34676 }, { "epoch": 2.09, "learning_rate": 3.634889019661099e-05, "loss": 0.4442, "step": 34678 }, { "epoch": 2.09, "learning_rate": 3.634738128649677e-05, "loss": 0.5272, "step": 34680 }, { "epoch": 2.09, "learning_rate": 3.6345872376382536e-05, "loss": 0.5013, "step": 34682 }, { "epoch": 2.09, "learning_rate": 3.6344363466268316e-05, "loss": 0.6003, "step": 34684 }, { "epoch": 2.09, "learning_rate": 3.6342854556154096e-05, "loss": 0.6777, "step": 34686 }, { "epoch": 2.09, "learning_rate": 3.634134564603986e-05, "loss": 0.6844, "step": 34688 }, { "epoch": 2.09, "learning_rate": 3.633983673592564e-05, "loss": 0.5953, "step": 34690 }, { "epoch": 2.09, "learning_rate": 3.633832782581142e-05, "loss": 0.7097, "step": 34692 }, { "epoch": 2.09, "learning_rate": 3.6336818915697195e-05, "loss": 0.6548, "step": 34694 }, { "epoch": 2.09, "learning_rate": 3.633531000558297e-05, "loss": 0.5824, "step": 34696 }, { "epoch": 2.09, "learning_rate": 3.633380109546874e-05, "loss": 0.7132, "step": 34698 }, { "epoch": 2.09, "learning_rate": 3.633229218535452e-05, "loss": 0.6157, "step": 34700 }, { "epoch": 2.09, "learning_rate": 3.6330783275240295e-05, "loss": 0.57, "step": 34702 }, { "epoch": 2.09, "learning_rate": 3.632927436512607e-05, "loss": 0.4745, "step": 34704 }, { "epoch": 2.09, "learning_rate": 3.632776545501185e-05, "loss": 0.6493, "step": 34706 }, { "epoch": 2.09, "learning_rate": 3.632625654489763e-05, "loss": 0.7076, "step": 34708 }, { "epoch": 2.1, "learning_rate": 3.6324747634783394e-05, "loss": 0.4902, "step": 34710 }, { "epoch": 2.1, "learning_rate": 3.6323238724669174e-05, "loss": 0.6001, "step": 34712 }, { "epoch": 2.1, "learning_rate": 3.632172981455495e-05, "loss": 0.6671, "step": 34714 }, { "epoch": 2.1, "learning_rate": 3.632022090444072e-05, "loss": 0.5306, "step": 34716 }, { "epoch": 2.1, "learning_rate": 3.63187119943265e-05, "loss": 0.6565, "step": 34718 }, { "epoch": 2.1, "learning_rate": 3.631720308421227e-05, "loss": 0.5243, "step": 34720 }, { "epoch": 2.1, "learning_rate": 3.631569417409805e-05, "loss": 0.6281, "step": 34722 }, { "epoch": 2.1, "learning_rate": 3.6314185263983826e-05, "loss": 0.5854, "step": 34724 }, { "epoch": 2.1, "learning_rate": 3.63126763538696e-05, "loss": 0.6582, "step": 34726 }, { "epoch": 2.1, "learning_rate": 3.631116744375538e-05, "loss": 0.7699, "step": 34728 }, { "epoch": 2.1, "learning_rate": 3.630965853364115e-05, "loss": 0.7675, "step": 34730 }, { "epoch": 2.1, "learning_rate": 3.6308149623526925e-05, "loss": 0.6147, "step": 34732 }, { "epoch": 2.1, "learning_rate": 3.6306640713412705e-05, "loss": 0.9044, "step": 34734 }, { "epoch": 2.1, "learning_rate": 3.630513180329848e-05, "loss": 0.554, "step": 34736 }, { "epoch": 2.1, "learning_rate": 3.630362289318425e-05, "loss": 0.6409, "step": 34738 }, { "epoch": 2.1, "learning_rate": 3.630211398307003e-05, "loss": 0.8289, "step": 34740 }, { "epoch": 2.1, "learning_rate": 3.6300605072955805e-05, "loss": 0.6067, "step": 34742 }, { "epoch": 2.1, "learning_rate": 3.6299096162841585e-05, "loss": 0.4568, "step": 34744 }, { "epoch": 2.1, "learning_rate": 3.629758725272736e-05, "loss": 0.8491, "step": 34746 }, { "epoch": 2.1, "learning_rate": 3.629607834261313e-05, "loss": 0.7424, "step": 34748 }, { "epoch": 2.1, "learning_rate": 3.629456943249891e-05, "loss": 0.8303, "step": 34750 }, { "epoch": 2.1, "learning_rate": 3.6293060522384684e-05, "loss": 0.6337, "step": 34752 }, { "epoch": 2.1, "learning_rate": 3.629155161227046e-05, "loss": 0.5209, "step": 34754 }, { "epoch": 2.1, "learning_rate": 3.629004270215624e-05, "loss": 0.5061, "step": 34756 }, { "epoch": 2.1, "learning_rate": 3.628853379204201e-05, "loss": 0.6431, "step": 34758 }, { "epoch": 2.1, "learning_rate": 3.628702488192778e-05, "loss": 0.6658, "step": 34760 }, { "epoch": 2.1, "learning_rate": 3.628551597181356e-05, "loss": 0.4191, "step": 34762 }, { "epoch": 2.1, "learning_rate": 3.6284007061699336e-05, "loss": 0.8998, "step": 34764 }, { "epoch": 2.1, "learning_rate": 3.628249815158511e-05, "loss": 0.6068, "step": 34766 }, { "epoch": 2.1, "learning_rate": 3.628098924147089e-05, "loss": 0.6885, "step": 34768 }, { "epoch": 2.1, "learning_rate": 3.627948033135666e-05, "loss": 0.6795, "step": 34770 }, { "epoch": 2.1, "learning_rate": 3.627797142124244e-05, "loss": 0.7544, "step": 34772 }, { "epoch": 2.1, "learning_rate": 3.627646251112821e-05, "loss": 0.5523, "step": 34774 }, { "epoch": 2.1, "learning_rate": 3.627495360101399e-05, "loss": 0.6612, "step": 34776 }, { "epoch": 2.1, "learning_rate": 3.627344469089977e-05, "loss": 0.6095, "step": 34778 }, { "epoch": 2.1, "learning_rate": 3.627193578078554e-05, "loss": 0.7638, "step": 34780 }, { "epoch": 2.1, "learning_rate": 3.6270426870671315e-05, "loss": 0.5719, "step": 34782 }, { "epoch": 2.1, "learning_rate": 3.6268917960557094e-05, "loss": 0.9143, "step": 34784 }, { "epoch": 2.1, "learning_rate": 3.626740905044287e-05, "loss": 0.5187, "step": 34786 }, { "epoch": 2.1, "learning_rate": 3.626590014032864e-05, "loss": 0.5539, "step": 34788 }, { "epoch": 2.1, "learning_rate": 3.626439123021442e-05, "loss": 0.6556, "step": 34790 }, { "epoch": 2.1, "learning_rate": 3.6262882320100194e-05, "loss": 0.6259, "step": 34792 }, { "epoch": 2.1, "learning_rate": 3.6261373409985974e-05, "loss": 0.5645, "step": 34794 }, { "epoch": 2.1, "learning_rate": 3.625986449987174e-05, "loss": 0.6213, "step": 34796 }, { "epoch": 2.1, "learning_rate": 3.625835558975752e-05, "loss": 0.5637, "step": 34798 }, { "epoch": 2.1, "learning_rate": 3.62568466796433e-05, "loss": 0.8157, "step": 34800 }, { "epoch": 2.1, "learning_rate": 3.6255337769529066e-05, "loss": 0.4719, "step": 34802 }, { "epoch": 2.1, "learning_rate": 3.6253828859414846e-05, "loss": 0.744, "step": 34804 }, { "epoch": 2.1, "learning_rate": 3.6252319949300626e-05, "loss": 0.4992, "step": 34806 }, { "epoch": 2.1, "learning_rate": 3.62508110391864e-05, "loss": 0.6352, "step": 34808 }, { "epoch": 2.1, "learning_rate": 3.624930212907217e-05, "loss": 0.6566, "step": 34810 }, { "epoch": 2.1, "learning_rate": 3.6247793218957945e-05, "loss": 0.8876, "step": 34812 }, { "epoch": 2.1, "learning_rate": 3.6246284308843725e-05, "loss": 0.385, "step": 34814 }, { "epoch": 2.1, "learning_rate": 3.62447753987295e-05, "loss": 0.8355, "step": 34816 }, { "epoch": 2.1, "learning_rate": 3.624326648861527e-05, "loss": 0.7702, "step": 34818 }, { "epoch": 2.1, "learning_rate": 3.624175757850105e-05, "loss": 0.9946, "step": 34820 }, { "epoch": 2.1, "learning_rate": 3.624024866838683e-05, "loss": 0.9214, "step": 34822 }, { "epoch": 2.1, "learning_rate": 3.62387397582726e-05, "loss": 0.8369, "step": 34824 }, { "epoch": 2.1, "learning_rate": 3.623723084815838e-05, "loss": 0.5678, "step": 34826 }, { "epoch": 2.1, "learning_rate": 3.623572193804415e-05, "loss": 0.7581, "step": 34828 }, { "epoch": 2.1, "learning_rate": 3.6234213027929924e-05, "loss": 0.7013, "step": 34830 }, { "epoch": 2.1, "learning_rate": 3.6232704117815704e-05, "loss": 0.6546, "step": 34832 }, { "epoch": 2.1, "learning_rate": 3.623119520770148e-05, "loss": 0.7671, "step": 34834 }, { "epoch": 2.1, "learning_rate": 3.622968629758726e-05, "loss": 0.722, "step": 34836 }, { "epoch": 2.1, "learning_rate": 3.622817738747303e-05, "loss": 0.7359, "step": 34838 }, { "epoch": 2.1, "learning_rate": 3.62266684773588e-05, "loss": 0.5246, "step": 34840 }, { "epoch": 2.1, "learning_rate": 3.622515956724458e-05, "loss": 0.5483, "step": 34842 }, { "epoch": 2.1, "learning_rate": 3.6223650657130356e-05, "loss": 0.4503, "step": 34844 }, { "epoch": 2.1, "learning_rate": 3.622214174701613e-05, "loss": 0.6541, "step": 34846 }, { "epoch": 2.1, "learning_rate": 3.622063283690191e-05, "loss": 0.841, "step": 34848 }, { "epoch": 2.1, "learning_rate": 3.621912392678768e-05, "loss": 0.5506, "step": 34850 }, { "epoch": 2.1, "learning_rate": 3.6217615016673455e-05, "loss": 0.6825, "step": 34852 }, { "epoch": 2.1, "learning_rate": 3.6216106106559235e-05, "loss": 0.6292, "step": 34854 }, { "epoch": 2.1, "learning_rate": 3.621459719644501e-05, "loss": 0.4567, "step": 34856 }, { "epoch": 2.1, "learning_rate": 3.621308828633079e-05, "loss": 0.6669, "step": 34858 }, { "epoch": 2.1, "learning_rate": 3.621157937621656e-05, "loss": 0.7442, "step": 34860 }, { "epoch": 2.1, "learning_rate": 3.6210070466102334e-05, "loss": 0.6423, "step": 34862 }, { "epoch": 2.1, "learning_rate": 3.6208561555988114e-05, "loss": 0.6831, "step": 34864 }, { "epoch": 2.1, "learning_rate": 3.620705264587389e-05, "loss": 0.7068, "step": 34866 }, { "epoch": 2.1, "learning_rate": 3.620554373575966e-05, "loss": 0.6209, "step": 34868 }, { "epoch": 2.1, "learning_rate": 3.620403482564544e-05, "loss": 0.7995, "step": 34870 }, { "epoch": 2.1, "learning_rate": 3.6202525915531214e-05, "loss": 0.5703, "step": 34872 }, { "epoch": 2.1, "learning_rate": 3.620101700541699e-05, "loss": 0.845, "step": 34874 }, { "epoch": 2.11, "learning_rate": 3.619950809530277e-05, "loss": 0.4776, "step": 34876 }, { "epoch": 2.11, "learning_rate": 3.619799918518854e-05, "loss": 0.4998, "step": 34878 }, { "epoch": 2.11, "learning_rate": 3.619649027507431e-05, "loss": 0.5824, "step": 34880 }, { "epoch": 2.11, "learning_rate": 3.619498136496009e-05, "loss": 0.7559, "step": 34882 }, { "epoch": 2.11, "learning_rate": 3.6193472454845866e-05, "loss": 0.6754, "step": 34884 }, { "epoch": 2.11, "learning_rate": 3.6191963544731646e-05, "loss": 0.762, "step": 34886 }, { "epoch": 2.11, "learning_rate": 3.619045463461741e-05, "loss": 0.5854, "step": 34888 }, { "epoch": 2.11, "learning_rate": 3.618894572450319e-05, "loss": 0.613, "step": 34890 }, { "epoch": 2.11, "learning_rate": 3.618743681438897e-05, "loss": 0.9553, "step": 34892 }, { "epoch": 2.11, "learning_rate": 3.6185927904274745e-05, "loss": 0.7819, "step": 34894 }, { "epoch": 2.11, "learning_rate": 3.618441899416052e-05, "loss": 0.8065, "step": 34896 }, { "epoch": 2.11, "learning_rate": 3.61829100840463e-05, "loss": 0.5267, "step": 34898 }, { "epoch": 2.11, "learning_rate": 3.618140117393207e-05, "loss": 0.8741, "step": 34900 }, { "epoch": 2.11, "learning_rate": 3.6179892263817844e-05, "loss": 0.6832, "step": 34902 }, { "epoch": 2.11, "learning_rate": 3.6178383353703624e-05, "loss": 0.4955, "step": 34904 }, { "epoch": 2.11, "learning_rate": 3.61768744435894e-05, "loss": 0.4436, "step": 34906 }, { "epoch": 2.11, "learning_rate": 3.617536553347518e-05, "loss": 0.6346, "step": 34908 }, { "epoch": 2.11, "learning_rate": 3.6173856623360944e-05, "loss": 0.5483, "step": 34910 }, { "epoch": 2.11, "learning_rate": 3.6172347713246724e-05, "loss": 0.8015, "step": 34912 }, { "epoch": 2.11, "learning_rate": 3.6170838803132504e-05, "loss": 0.555, "step": 34914 }, { "epoch": 2.11, "learning_rate": 3.616932989301827e-05, "loss": 0.6844, "step": 34916 }, { "epoch": 2.11, "learning_rate": 3.616782098290405e-05, "loss": 0.8535, "step": 34918 }, { "epoch": 2.11, "learning_rate": 3.616631207278983e-05, "loss": 0.6791, "step": 34920 }, { "epoch": 2.11, "learning_rate": 3.61648031626756e-05, "loss": 0.7233, "step": 34922 }, { "epoch": 2.11, "learning_rate": 3.6163294252561376e-05, "loss": 0.5542, "step": 34924 }, { "epoch": 2.11, "learning_rate": 3.616178534244715e-05, "loss": 0.5133, "step": 34926 }, { "epoch": 2.11, "learning_rate": 3.616027643233293e-05, "loss": 0.6248, "step": 34928 }, { "epoch": 2.11, "learning_rate": 3.61587675222187e-05, "loss": 0.5399, "step": 34930 }, { "epoch": 2.11, "learning_rate": 3.6157258612104475e-05, "loss": 0.4383, "step": 34932 }, { "epoch": 2.11, "learning_rate": 3.6155749701990255e-05, "loss": 0.8629, "step": 34934 }, { "epoch": 2.11, "learning_rate": 3.6154240791876035e-05, "loss": 0.6164, "step": 34936 }, { "epoch": 2.11, "learning_rate": 3.61527318817618e-05, "loss": 0.8751, "step": 34938 }, { "epoch": 2.11, "learning_rate": 3.615122297164758e-05, "loss": 0.6496, "step": 34940 }, { "epoch": 2.11, "learning_rate": 3.6149714061533354e-05, "loss": 0.5988, "step": 34942 }, { "epoch": 2.11, "learning_rate": 3.6148205151419134e-05, "loss": 0.6681, "step": 34944 }, { "epoch": 2.11, "learning_rate": 3.614669624130491e-05, "loss": 0.6698, "step": 34946 }, { "epoch": 2.11, "learning_rate": 3.614518733119068e-05, "loss": 0.5456, "step": 34948 }, { "epoch": 2.11, "learning_rate": 3.614367842107646e-05, "loss": 0.5375, "step": 34950 }, { "epoch": 2.11, "learning_rate": 3.6142169510962234e-05, "loss": 0.6059, "step": 34952 }, { "epoch": 2.11, "learning_rate": 3.614066060084801e-05, "loss": 0.5887, "step": 34954 }, { "epoch": 2.11, "learning_rate": 3.6139151690733787e-05, "loss": 0.4128, "step": 34956 }, { "epoch": 2.11, "learning_rate": 3.613764278061956e-05, "loss": 0.6149, "step": 34958 }, { "epoch": 2.11, "learning_rate": 3.613613387050533e-05, "loss": 0.5467, "step": 34960 }, { "epoch": 2.11, "learning_rate": 3.613462496039111e-05, "loss": 0.6746, "step": 34962 }, { "epoch": 2.11, "learning_rate": 3.6133116050276886e-05, "loss": 0.6417, "step": 34964 }, { "epoch": 2.11, "learning_rate": 3.613160714016266e-05, "loss": 0.8175, "step": 34966 }, { "epoch": 2.11, "learning_rate": 3.613009823004844e-05, "loss": 0.7478, "step": 34968 }, { "epoch": 2.11, "learning_rate": 3.612858931993421e-05, "loss": 0.547, "step": 34970 }, { "epoch": 2.11, "learning_rate": 3.612708040981999e-05, "loss": 0.6752, "step": 34972 }, { "epoch": 2.11, "learning_rate": 3.6125571499705765e-05, "loss": 0.7566, "step": 34974 }, { "epoch": 2.11, "learning_rate": 3.612406258959154e-05, "loss": 0.5879, "step": 34976 }, { "epoch": 2.11, "learning_rate": 3.612255367947732e-05, "loss": 0.6612, "step": 34978 }, { "epoch": 2.11, "learning_rate": 3.612104476936309e-05, "loss": 0.7004, "step": 34980 }, { "epoch": 2.11, "learning_rate": 3.6119535859248864e-05, "loss": 0.5658, "step": 34982 }, { "epoch": 2.11, "learning_rate": 3.6118026949134644e-05, "loss": 0.4331, "step": 34984 }, { "epoch": 2.11, "learning_rate": 3.611651803902042e-05, "loss": 0.5826, "step": 34986 }, { "epoch": 2.11, "learning_rate": 3.611500912890619e-05, "loss": 0.7646, "step": 34988 }, { "epoch": 2.11, "learning_rate": 3.611350021879197e-05, "loss": 0.4877, "step": 34990 }, { "epoch": 2.11, "learning_rate": 3.6111991308677743e-05, "loss": 0.5393, "step": 34992 }, { "epoch": 2.11, "learning_rate": 3.611048239856352e-05, "loss": 0.6872, "step": 34994 }, { "epoch": 2.11, "learning_rate": 3.6108973488449297e-05, "loss": 0.8358, "step": 34996 }, { "epoch": 2.11, "learning_rate": 3.610746457833507e-05, "loss": 0.4836, "step": 34998 }, { "epoch": 2.11, "learning_rate": 3.610595566822085e-05, "loss": 0.6379, "step": 35000 }, { "epoch": 2.11, "learning_rate": 3.6104446758106616e-05, "loss": 0.6398, "step": 35002 }, { "epoch": 2.11, "learning_rate": 3.6102937847992396e-05, "loss": 0.4927, "step": 35004 }, { "epoch": 2.11, "learning_rate": 3.6101428937878176e-05, "loss": 0.6055, "step": 35006 }, { "epoch": 2.11, "learning_rate": 3.609992002776395e-05, "loss": 0.6255, "step": 35008 }, { "epoch": 2.11, "learning_rate": 3.609841111764972e-05, "loss": 0.55, "step": 35010 }, { "epoch": 2.11, "learning_rate": 3.60969022075355e-05, "loss": 0.703, "step": 35012 }, { "epoch": 2.11, "learning_rate": 3.6095393297421275e-05, "loss": 0.4315, "step": 35014 }, { "epoch": 2.11, "learning_rate": 3.609388438730705e-05, "loss": 0.6587, "step": 35016 }, { "epoch": 2.11, "learning_rate": 3.609237547719283e-05, "loss": 0.6421, "step": 35018 }, { "epoch": 2.11, "learning_rate": 3.60908665670786e-05, "loss": 0.5916, "step": 35020 }, { "epoch": 2.11, "learning_rate": 3.608935765696438e-05, "loss": 0.5741, "step": 35022 }, { "epoch": 2.11, "learning_rate": 3.608784874685015e-05, "loss": 0.6917, "step": 35024 }, { "epoch": 2.11, "learning_rate": 3.608633983673593e-05, "loss": 0.3505, "step": 35026 }, { "epoch": 2.11, "learning_rate": 3.608483092662171e-05, "loss": 0.7114, "step": 35028 }, { "epoch": 2.11, "learning_rate": 3.6083322016507474e-05, "loss": 0.7483, "step": 35030 }, { "epoch": 2.11, "learning_rate": 3.6081813106393253e-05, "loss": 0.485, "step": 35032 }, { "epoch": 2.11, "learning_rate": 3.608030419627903e-05, "loss": 0.7782, "step": 35034 }, { "epoch": 2.11, "learning_rate": 3.6078795286164806e-05, "loss": 0.6525, "step": 35036 }, { "epoch": 2.11, "learning_rate": 3.607728637605058e-05, "loss": 0.6122, "step": 35038 }, { "epoch": 2.11, "learning_rate": 3.607577746593635e-05, "loss": 0.6267, "step": 35040 }, { "epoch": 2.12, "learning_rate": 3.607426855582213e-05, "loss": 0.443, "step": 35042 }, { "epoch": 2.12, "learning_rate": 3.6072759645707906e-05, "loss": 0.4978, "step": 35044 }, { "epoch": 2.12, "learning_rate": 3.607125073559368e-05, "loss": 0.726, "step": 35046 }, { "epoch": 2.12, "learning_rate": 3.606974182547946e-05, "loss": 0.4751, "step": 35048 }, { "epoch": 2.12, "learning_rate": 3.606823291536524e-05, "loss": 0.7848, "step": 35050 }, { "epoch": 2.12, "learning_rate": 3.6066724005251005e-05, "loss": 0.7492, "step": 35052 }, { "epoch": 2.12, "learning_rate": 3.6065215095136785e-05, "loss": 0.9767, "step": 35054 }, { "epoch": 2.12, "learning_rate": 3.606370618502256e-05, "loss": 0.598, "step": 35056 }, { "epoch": 2.12, "learning_rate": 3.606219727490834e-05, "loss": 0.8726, "step": 35058 }, { "epoch": 2.12, "learning_rate": 3.606068836479411e-05, "loss": 0.5781, "step": 35060 }, { "epoch": 2.12, "learning_rate": 3.6059179454679884e-05, "loss": 0.5582, "step": 35062 }, { "epoch": 2.12, "learning_rate": 3.6057670544565664e-05, "loss": 0.4736, "step": 35064 }, { "epoch": 2.12, "learning_rate": 3.605616163445144e-05, "loss": 0.4507, "step": 35066 }, { "epoch": 2.12, "learning_rate": 3.605465272433721e-05, "loss": 0.6082, "step": 35068 }, { "epoch": 2.12, "learning_rate": 3.605314381422299e-05, "loss": 0.5447, "step": 35070 }, { "epoch": 2.12, "learning_rate": 3.605163490410876e-05, "loss": 0.4702, "step": 35072 }, { "epoch": 2.12, "learning_rate": 3.6050125993994537e-05, "loss": 0.6025, "step": 35074 }, { "epoch": 2.12, "learning_rate": 3.6048617083880316e-05, "loss": 0.6775, "step": 35076 }, { "epoch": 2.12, "learning_rate": 3.604710817376609e-05, "loss": 0.7688, "step": 35078 }, { "epoch": 2.12, "learning_rate": 3.604559926365186e-05, "loss": 0.7013, "step": 35080 }, { "epoch": 2.12, "learning_rate": 3.604409035353764e-05, "loss": 0.6981, "step": 35082 }, { "epoch": 2.12, "learning_rate": 3.6042581443423416e-05, "loss": 0.7737, "step": 35084 }, { "epoch": 2.12, "learning_rate": 3.6041072533309196e-05, "loss": 0.7077, "step": 35086 }, { "epoch": 2.12, "learning_rate": 3.603956362319497e-05, "loss": 0.9178, "step": 35088 }, { "epoch": 2.12, "learning_rate": 3.603805471308074e-05, "loss": 0.7804, "step": 35090 }, { "epoch": 2.12, "learning_rate": 3.603654580296652e-05, "loss": 0.4172, "step": 35092 }, { "epoch": 2.12, "learning_rate": 3.6035036892852295e-05, "loss": 0.5306, "step": 35094 }, { "epoch": 2.12, "learning_rate": 3.603352798273807e-05, "loss": 0.5701, "step": 35096 }, { "epoch": 2.12, "learning_rate": 3.603201907262385e-05, "loss": 0.5813, "step": 35098 }, { "epoch": 2.12, "learning_rate": 3.603051016250962e-05, "loss": 0.5562, "step": 35100 }, { "epoch": 2.12, "learning_rate": 3.6029001252395394e-05, "loss": 0.5172, "step": 35102 }, { "epoch": 2.12, "learning_rate": 3.6027492342281174e-05, "loss": 0.5306, "step": 35104 }, { "epoch": 2.12, "learning_rate": 3.602598343216695e-05, "loss": 0.5539, "step": 35106 }, { "epoch": 2.12, "learning_rate": 3.602447452205273e-05, "loss": 0.5943, "step": 35108 }, { "epoch": 2.12, "learning_rate": 3.60229656119385e-05, "loss": 0.691, "step": 35110 }, { "epoch": 2.12, "learning_rate": 3.602145670182427e-05, "loss": 0.5806, "step": 35112 }, { "epoch": 2.12, "learning_rate": 3.601994779171005e-05, "loss": 0.641, "step": 35114 }, { "epoch": 2.12, "learning_rate": 3.601843888159582e-05, "loss": 0.6409, "step": 35116 }, { "epoch": 2.12, "learning_rate": 3.60169299714816e-05, "loss": 0.603, "step": 35118 }, { "epoch": 2.12, "learning_rate": 3.601542106136738e-05, "loss": 0.4372, "step": 35120 }, { "epoch": 2.12, "learning_rate": 3.601391215125315e-05, "loss": 0.442, "step": 35122 }, { "epoch": 2.12, "learning_rate": 3.6012403241138926e-05, "loss": 0.449, "step": 35124 }, { "epoch": 2.12, "learning_rate": 3.6010894331024706e-05, "loss": 0.5461, "step": 35126 }, { "epoch": 2.12, "learning_rate": 3.600938542091048e-05, "loss": 0.6902, "step": 35128 }, { "epoch": 2.12, "learning_rate": 3.600787651079625e-05, "loss": 0.6157, "step": 35130 }, { "epoch": 2.12, "learning_rate": 3.600636760068203e-05, "loss": 0.8274, "step": 35132 }, { "epoch": 2.12, "learning_rate": 3.6004858690567805e-05, "loss": 0.7411, "step": 35134 }, { "epoch": 2.12, "learning_rate": 3.6003349780453585e-05, "loss": 0.5528, "step": 35136 }, { "epoch": 2.12, "learning_rate": 3.600184087033935e-05, "loss": 0.7377, "step": 35138 }, { "epoch": 2.12, "learning_rate": 3.600033196022513e-05, "loss": 0.7819, "step": 35140 }, { "epoch": 2.12, "learning_rate": 3.599882305011091e-05, "loss": 0.5509, "step": 35142 }, { "epoch": 2.12, "learning_rate": 3.599731413999668e-05, "loss": 0.684, "step": 35144 }, { "epoch": 2.12, "learning_rate": 3.599580522988246e-05, "loss": 0.6179, "step": 35146 }, { "epoch": 2.12, "learning_rate": 3.599429631976824e-05, "loss": 0.7242, "step": 35148 }, { "epoch": 2.12, "learning_rate": 3.599278740965401e-05, "loss": 0.3611, "step": 35150 }, { "epoch": 2.12, "learning_rate": 3.599127849953978e-05, "loss": 0.6826, "step": 35152 }, { "epoch": 2.12, "learning_rate": 3.5989769589425556e-05, "loss": 0.6003, "step": 35154 }, { "epoch": 2.12, "learning_rate": 3.5988260679311336e-05, "loss": 0.471, "step": 35156 }, { "epoch": 2.12, "learning_rate": 3.598675176919711e-05, "loss": 0.7379, "step": 35158 }, { "epoch": 2.12, "learning_rate": 3.598524285908288e-05, "loss": 0.6846, "step": 35160 }, { "epoch": 2.12, "learning_rate": 3.598373394896866e-05, "loss": 0.8142, "step": 35162 }, { "epoch": 2.12, "learning_rate": 3.598222503885444e-05, "loss": 0.5982, "step": 35164 }, { "epoch": 2.12, "learning_rate": 3.598071612874021e-05, "loss": 0.728, "step": 35166 }, { "epoch": 2.12, "learning_rate": 3.597920721862599e-05, "loss": 0.6134, "step": 35168 }, { "epoch": 2.12, "learning_rate": 3.597769830851176e-05, "loss": 0.6409, "step": 35170 }, { "epoch": 2.12, "learning_rate": 3.597618939839754e-05, "loss": 0.7419, "step": 35172 }, { "epoch": 2.12, "learning_rate": 3.5974680488283315e-05, "loss": 0.8071, "step": 35174 }, { "epoch": 2.12, "learning_rate": 3.597317157816909e-05, "loss": 0.6774, "step": 35176 }, { "epoch": 2.12, "learning_rate": 3.597166266805487e-05, "loss": 0.5637, "step": 35178 }, { "epoch": 2.12, "learning_rate": 3.597015375794064e-05, "loss": 0.5185, "step": 35180 }, { "epoch": 2.12, "learning_rate": 3.5968644847826414e-05, "loss": 0.6135, "step": 35182 }, { "epoch": 2.12, "learning_rate": 3.5967135937712194e-05, "loss": 0.664, "step": 35184 }, { "epoch": 2.12, "learning_rate": 3.596562702759797e-05, "loss": 0.6748, "step": 35186 }, { "epoch": 2.12, "learning_rate": 3.596411811748374e-05, "loss": 0.8186, "step": 35188 }, { "epoch": 2.12, "learning_rate": 3.596260920736952e-05, "loss": 0.5266, "step": 35190 }, { "epoch": 2.12, "learning_rate": 3.596110029725529e-05, "loss": 0.5937, "step": 35192 }, { "epoch": 2.12, "learning_rate": 3.5959591387141066e-05, "loss": 0.6224, "step": 35194 }, { "epoch": 2.12, "learning_rate": 3.5958082477026846e-05, "loss": 0.5473, "step": 35196 }, { "epoch": 2.12, "learning_rate": 3.595657356691262e-05, "loss": 0.6292, "step": 35198 }, { "epoch": 2.12, "learning_rate": 3.59550646567984e-05, "loss": 0.6469, "step": 35200 }, { "epoch": 2.12, "learning_rate": 3.595355574668417e-05, "loss": 0.7795, "step": 35202 }, { "epoch": 2.12, "learning_rate": 3.5952046836569946e-05, "loss": 0.4163, "step": 35204 }, { "epoch": 2.12, "learning_rate": 3.5950537926455725e-05, "loss": 0.6061, "step": 35206 }, { "epoch": 2.13, "learning_rate": 3.59490290163415e-05, "loss": 0.5672, "step": 35208 }, { "epoch": 2.13, "learning_rate": 3.594752010622727e-05, "loss": 0.4164, "step": 35210 }, { "epoch": 2.13, "learning_rate": 3.594601119611305e-05, "loss": 0.7122, "step": 35212 }, { "epoch": 2.13, "learning_rate": 3.5944502285998825e-05, "loss": 0.6194, "step": 35214 }, { "epoch": 2.13, "learning_rate": 3.59429933758846e-05, "loss": 0.6223, "step": 35216 }, { "epoch": 2.13, "learning_rate": 3.594148446577038e-05, "loss": 0.8358, "step": 35218 }, { "epoch": 2.13, "learning_rate": 3.593997555565615e-05, "loss": 0.5204, "step": 35220 }, { "epoch": 2.13, "learning_rate": 3.593846664554193e-05, "loss": 0.6393, "step": 35222 }, { "epoch": 2.13, "learning_rate": 3.5936957735427704e-05, "loss": 0.5768, "step": 35224 }, { "epoch": 2.13, "learning_rate": 3.593544882531348e-05, "loss": 0.6381, "step": 35226 }, { "epoch": 2.13, "learning_rate": 3.593393991519926e-05, "loss": 0.412, "step": 35228 }, { "epoch": 2.13, "learning_rate": 3.593243100508502e-05, "loss": 0.6808, "step": 35230 }, { "epoch": 2.13, "learning_rate": 3.59309220949708e-05, "loss": 0.7637, "step": 35232 }, { "epoch": 2.13, "learning_rate": 3.592941318485658e-05, "loss": 0.4808, "step": 35234 }, { "epoch": 2.13, "learning_rate": 3.5927904274742356e-05, "loss": 0.6234, "step": 35236 }, { "epoch": 2.13, "learning_rate": 3.592639536462813e-05, "loss": 0.7169, "step": 35238 }, { "epoch": 2.13, "learning_rate": 3.592488645451391e-05, "loss": 0.9369, "step": 35240 }, { "epoch": 2.13, "learning_rate": 3.592337754439968e-05, "loss": 0.5304, "step": 35242 }, { "epoch": 2.13, "learning_rate": 3.5921868634285455e-05, "loss": 0.7027, "step": 35244 }, { "epoch": 2.13, "learning_rate": 3.592035972417123e-05, "loss": 0.5712, "step": 35246 }, { "epoch": 2.13, "learning_rate": 3.591885081405701e-05, "loss": 0.6576, "step": 35248 }, { "epoch": 2.13, "learning_rate": 3.591734190394279e-05, "loss": 0.5449, "step": 35250 }, { "epoch": 2.13, "learning_rate": 3.5915832993828555e-05, "loss": 0.5778, "step": 35252 }, { "epoch": 2.13, "learning_rate": 3.5914324083714335e-05, "loss": 0.5917, "step": 35254 }, { "epoch": 2.13, "learning_rate": 3.5912815173600115e-05, "loss": 0.7967, "step": 35256 }, { "epoch": 2.13, "learning_rate": 3.591130626348588e-05, "loss": 0.5601, "step": 35258 }, { "epoch": 2.13, "learning_rate": 3.590979735337166e-05, "loss": 0.6413, "step": 35260 }, { "epoch": 2.13, "learning_rate": 3.590828844325744e-05, "loss": 0.5703, "step": 35262 }, { "epoch": 2.13, "learning_rate": 3.5906779533143214e-05, "loss": 0.5619, "step": 35264 }, { "epoch": 2.13, "learning_rate": 3.590527062302899e-05, "loss": 0.4484, "step": 35266 }, { "epoch": 2.13, "learning_rate": 3.590376171291476e-05, "loss": 0.616, "step": 35268 }, { "epoch": 2.13, "learning_rate": 3.590225280280054e-05, "loss": 0.6469, "step": 35270 }, { "epoch": 2.13, "learning_rate": 3.590074389268632e-05, "loss": 0.6217, "step": 35272 }, { "epoch": 2.13, "learning_rate": 3.5899234982572086e-05, "loss": 0.7931, "step": 35274 }, { "epoch": 2.13, "learning_rate": 3.5897726072457866e-05, "loss": 0.3179, "step": 35276 }, { "epoch": 2.13, "learning_rate": 3.5896217162343646e-05, "loss": 0.6301, "step": 35278 }, { "epoch": 2.13, "learning_rate": 3.589470825222941e-05, "loss": 0.3048, "step": 35280 }, { "epoch": 2.13, "learning_rate": 3.589319934211519e-05, "loss": 0.5985, "step": 35282 }, { "epoch": 2.13, "learning_rate": 3.5891690432000965e-05, "loss": 0.6967, "step": 35284 }, { "epoch": 2.13, "learning_rate": 3.5890181521886745e-05, "loss": 0.4098, "step": 35286 }, { "epoch": 2.13, "learning_rate": 3.588867261177252e-05, "loss": 0.606, "step": 35288 }, { "epoch": 2.13, "learning_rate": 3.588716370165829e-05, "loss": 0.5992, "step": 35290 }, { "epoch": 2.13, "learning_rate": 3.588565479154407e-05, "loss": 0.498, "step": 35292 }, { "epoch": 2.13, "learning_rate": 3.5884145881429845e-05, "loss": 0.6317, "step": 35294 }, { "epoch": 2.13, "learning_rate": 3.588263697131562e-05, "loss": 0.6294, "step": 35296 }, { "epoch": 2.13, "learning_rate": 3.58811280612014e-05, "loss": 0.8339, "step": 35298 }, { "epoch": 2.13, "learning_rate": 3.587961915108717e-05, "loss": 0.8178, "step": 35300 }, { "epoch": 2.13, "learning_rate": 3.5878110240972944e-05, "loss": 0.5716, "step": 35302 }, { "epoch": 2.13, "learning_rate": 3.5876601330858724e-05, "loss": 0.4741, "step": 35304 }, { "epoch": 2.13, "learning_rate": 3.58750924207445e-05, "loss": 0.7358, "step": 35306 }, { "epoch": 2.13, "learning_rate": 3.587358351063027e-05, "loss": 0.5355, "step": 35308 }, { "epoch": 2.13, "learning_rate": 3.587207460051605e-05, "loss": 0.884, "step": 35310 }, { "epoch": 2.13, "learning_rate": 3.587056569040182e-05, "loss": 0.6878, "step": 35312 }, { "epoch": 2.13, "learning_rate": 3.58690567802876e-05, "loss": 0.6006, "step": 35314 }, { "epoch": 2.13, "learning_rate": 3.5867547870173376e-05, "loss": 0.6041, "step": 35316 }, { "epoch": 2.13, "learning_rate": 3.586603896005915e-05, "loss": 0.491, "step": 35318 }, { "epoch": 2.13, "learning_rate": 3.586453004994493e-05, "loss": 0.4122, "step": 35320 }, { "epoch": 2.13, "learning_rate": 3.58630211398307e-05, "loss": 0.399, "step": 35322 }, { "epoch": 2.13, "learning_rate": 3.5861512229716475e-05, "loss": 0.7813, "step": 35324 }, { "epoch": 2.13, "learning_rate": 3.5860003319602255e-05, "loss": 0.6592, "step": 35326 }, { "epoch": 2.13, "learning_rate": 3.585849440948803e-05, "loss": 0.4525, "step": 35328 }, { "epoch": 2.13, "learning_rate": 3.58569854993738e-05, "loss": 0.7169, "step": 35330 }, { "epoch": 2.13, "learning_rate": 3.585547658925958e-05, "loss": 0.521, "step": 35332 }, { "epoch": 2.13, "learning_rate": 3.5853967679145355e-05, "loss": 0.5408, "step": 35334 }, { "epoch": 2.13, "learning_rate": 3.5852458769031134e-05, "loss": 0.8543, "step": 35336 }, { "epoch": 2.13, "learning_rate": 3.585094985891691e-05, "loss": 0.66, "step": 35338 }, { "epoch": 2.13, "learning_rate": 3.584944094880268e-05, "loss": 0.6363, "step": 35340 }, { "epoch": 2.13, "learning_rate": 3.584793203868846e-05, "loss": 0.7065, "step": 35342 }, { "epoch": 2.13, "learning_rate": 3.584642312857423e-05, "loss": 0.6763, "step": 35344 }, { "epoch": 2.13, "learning_rate": 3.584491421846001e-05, "loss": 0.8212, "step": 35346 }, { "epoch": 2.13, "learning_rate": 3.584340530834579e-05, "loss": 0.5014, "step": 35348 }, { "epoch": 2.13, "learning_rate": 3.584189639823156e-05, "loss": 0.4685, "step": 35350 }, { "epoch": 2.13, "learning_rate": 3.584038748811733e-05, "loss": 0.6776, "step": 35352 }, { "epoch": 2.13, "learning_rate": 3.583887857800311e-05, "loss": 0.5173, "step": 35354 }, { "epoch": 2.13, "learning_rate": 3.5837369667888886e-05, "loss": 0.6094, "step": 35356 }, { "epoch": 2.13, "learning_rate": 3.583586075777466e-05, "loss": 0.6349, "step": 35358 }, { "epoch": 2.13, "learning_rate": 3.583435184766043e-05, "loss": 0.5128, "step": 35360 }, { "epoch": 2.13, "learning_rate": 3.583284293754621e-05, "loss": 0.5596, "step": 35362 }, { "epoch": 2.13, "learning_rate": 3.583133402743199e-05, "loss": 0.9376, "step": 35364 }, { "epoch": 2.13, "learning_rate": 3.582982511731776e-05, "loss": 0.7342, "step": 35366 }, { "epoch": 2.13, "learning_rate": 3.582831620720354e-05, "loss": 0.8292, "step": 35368 }, { "epoch": 2.13, "learning_rate": 3.582680729708932e-05, "loss": 0.5444, "step": 35370 }, { "epoch": 2.13, "learning_rate": 3.5825298386975085e-05, "loss": 0.4193, "step": 35372 }, { "epoch": 2.14, "learning_rate": 3.5823789476860864e-05, "loss": 1.1364, "step": 35374 }, { "epoch": 2.14, "learning_rate": 3.5822280566746644e-05, "loss": 0.4814, "step": 35376 }, { "epoch": 2.14, "learning_rate": 3.582077165663242e-05, "loss": 0.6804, "step": 35378 }, { "epoch": 2.14, "learning_rate": 3.581926274651819e-05, "loss": 0.3673, "step": 35380 }, { "epoch": 2.14, "learning_rate": 3.5817753836403964e-05, "loss": 0.6524, "step": 35382 }, { "epoch": 2.14, "learning_rate": 3.5816244926289744e-05, "loss": 0.5931, "step": 35384 }, { "epoch": 2.14, "learning_rate": 3.5814736016175524e-05, "loss": 0.629, "step": 35386 }, { "epoch": 2.14, "learning_rate": 3.581322710606129e-05, "loss": 0.6697, "step": 35388 }, { "epoch": 2.14, "learning_rate": 3.581171819594707e-05, "loss": 0.6257, "step": 35390 }, { "epoch": 2.14, "learning_rate": 3.581020928583285e-05, "loss": 0.5459, "step": 35392 }, { "epoch": 2.14, "learning_rate": 3.5808700375718616e-05, "loss": 0.9895, "step": 35394 }, { "epoch": 2.14, "learning_rate": 3.5807191465604396e-05, "loss": 0.8007, "step": 35396 }, { "epoch": 2.14, "learning_rate": 3.580568255549017e-05, "loss": 0.588, "step": 35398 }, { "epoch": 2.14, "learning_rate": 3.580417364537595e-05, "loss": 0.7416, "step": 35400 }, { "epoch": 2.14, "learning_rate": 3.580266473526172e-05, "loss": 0.7998, "step": 35402 }, { "epoch": 2.14, "learning_rate": 3.5801155825147495e-05, "loss": 0.619, "step": 35404 }, { "epoch": 2.14, "learning_rate": 3.5799646915033275e-05, "loss": 0.4806, "step": 35406 }, { "epoch": 2.14, "learning_rate": 3.579813800491905e-05, "loss": 0.6525, "step": 35408 }, { "epoch": 2.14, "learning_rate": 3.579662909480482e-05, "loss": 0.5893, "step": 35410 }, { "epoch": 2.14, "learning_rate": 3.57951201846906e-05, "loss": 0.6037, "step": 35412 }, { "epoch": 2.14, "learning_rate": 3.5793611274576374e-05, "loss": 0.7342, "step": 35414 }, { "epoch": 2.14, "learning_rate": 3.579210236446215e-05, "loss": 0.6346, "step": 35416 }, { "epoch": 2.14, "learning_rate": 3.579059345434793e-05, "loss": 0.6822, "step": 35418 }, { "epoch": 2.14, "learning_rate": 3.57890845442337e-05, "loss": 0.4787, "step": 35420 }, { "epoch": 2.14, "learning_rate": 3.5787575634119474e-05, "loss": 0.908, "step": 35422 }, { "epoch": 2.14, "learning_rate": 3.5786066724005254e-05, "loss": 0.6054, "step": 35424 }, { "epoch": 2.14, "learning_rate": 3.578455781389103e-05, "loss": 0.6438, "step": 35426 }, { "epoch": 2.14, "learning_rate": 3.578304890377681e-05, "loss": 0.7265, "step": 35428 }, { "epoch": 2.14, "learning_rate": 3.578153999366258e-05, "loss": 0.4109, "step": 35430 }, { "epoch": 2.14, "learning_rate": 3.578003108354835e-05, "loss": 0.8771, "step": 35432 }, { "epoch": 2.14, "learning_rate": 3.577852217343413e-05, "loss": 0.4313, "step": 35434 }, { "epoch": 2.14, "learning_rate": 3.5777013263319906e-05, "loss": 0.7383, "step": 35436 }, { "epoch": 2.14, "learning_rate": 3.577550435320568e-05, "loss": 0.6092, "step": 35438 }, { "epoch": 2.14, "learning_rate": 3.577399544309146e-05, "loss": 0.4061, "step": 35440 }, { "epoch": 2.14, "learning_rate": 3.577248653297723e-05, "loss": 0.5185, "step": 35442 }, { "epoch": 2.14, "learning_rate": 3.5770977622863005e-05, "loss": 0.731, "step": 35444 }, { "epoch": 2.14, "learning_rate": 3.5769468712748785e-05, "loss": 0.6079, "step": 35446 }, { "epoch": 2.14, "learning_rate": 3.576795980263456e-05, "loss": 0.5382, "step": 35448 }, { "epoch": 2.14, "learning_rate": 3.576645089252034e-05, "loss": 0.6805, "step": 35450 }, { "epoch": 2.14, "learning_rate": 3.576494198240611e-05, "loss": 0.6947, "step": 35452 }, { "epoch": 2.14, "learning_rate": 3.5763433072291884e-05, "loss": 0.8061, "step": 35454 }, { "epoch": 2.14, "learning_rate": 3.5761924162177664e-05, "loss": 0.6657, "step": 35456 }, { "epoch": 2.14, "learning_rate": 3.576041525206343e-05, "loss": 0.626, "step": 35458 }, { "epoch": 2.14, "learning_rate": 3.575890634194921e-05, "loss": 0.6047, "step": 35460 }, { "epoch": 2.14, "learning_rate": 3.575739743183499e-05, "loss": 0.8291, "step": 35462 }, { "epoch": 2.14, "learning_rate": 3.5755888521720764e-05, "loss": 0.7515, "step": 35464 }, { "epoch": 2.14, "learning_rate": 3.575437961160654e-05, "loss": 0.4503, "step": 35466 }, { "epoch": 2.14, "learning_rate": 3.5752870701492317e-05, "loss": 0.7279, "step": 35468 }, { "epoch": 2.14, "learning_rate": 3.575136179137809e-05, "loss": 0.5323, "step": 35470 }, { "epoch": 2.14, "learning_rate": 3.574985288126386e-05, "loss": 0.7588, "step": 35472 }, { "epoch": 2.14, "learning_rate": 3.5748343971149636e-05, "loss": 0.6571, "step": 35474 }, { "epoch": 2.14, "learning_rate": 3.5746835061035416e-05, "loss": 0.3886, "step": 35476 }, { "epoch": 2.14, "learning_rate": 3.5745326150921196e-05, "loss": 0.7417, "step": 35478 }, { "epoch": 2.14, "learning_rate": 3.574381724080696e-05, "loss": 0.5444, "step": 35480 }, { "epoch": 2.14, "learning_rate": 3.574230833069274e-05, "loss": 0.5118, "step": 35482 }, { "epoch": 2.14, "learning_rate": 3.574079942057852e-05, "loss": 0.7055, "step": 35484 }, { "epoch": 2.14, "learning_rate": 3.5739290510464295e-05, "loss": 0.858, "step": 35486 }, { "epoch": 2.14, "learning_rate": 3.573778160035007e-05, "loss": 0.5773, "step": 35488 }, { "epoch": 2.14, "learning_rate": 3.573627269023585e-05, "loss": 0.6657, "step": 35490 }, { "epoch": 2.14, "learning_rate": 3.573476378012162e-05, "loss": 0.4976, "step": 35492 }, { "epoch": 2.14, "learning_rate": 3.5733254870007394e-05, "loss": 0.8687, "step": 35494 }, { "epoch": 2.14, "learning_rate": 3.573174595989317e-05, "loss": 0.3231, "step": 35496 }, { "epoch": 2.14, "learning_rate": 3.573023704977895e-05, "loss": 0.4808, "step": 35498 }, { "epoch": 2.14, "learning_rate": 3.572872813966473e-05, "loss": 0.432, "step": 35500 }, { "epoch": 2.14, "learning_rate": 3.5727219229550494e-05, "loss": 0.83, "step": 35502 }, { "epoch": 2.14, "learning_rate": 3.5725710319436273e-05, "loss": 0.6679, "step": 35504 }, { "epoch": 2.14, "learning_rate": 3.572420140932205e-05, "loss": 0.6317, "step": 35506 }, { "epoch": 2.14, "learning_rate": 3.572269249920782e-05, "loss": 0.6342, "step": 35508 }, { "epoch": 2.14, "learning_rate": 3.57211835890936e-05, "loss": 0.6772, "step": 35510 }, { "epoch": 2.14, "learning_rate": 3.571967467897937e-05, "loss": 0.3683, "step": 35512 }, { "epoch": 2.14, "learning_rate": 3.571816576886515e-05, "loss": 0.6004, "step": 35514 }, { "epoch": 2.14, "learning_rate": 3.5716656858750926e-05, "loss": 0.7426, "step": 35516 }, { "epoch": 2.14, "learning_rate": 3.57151479486367e-05, "loss": 0.5651, "step": 35518 }, { "epoch": 2.14, "learning_rate": 3.571363903852248e-05, "loss": 0.5139, "step": 35520 }, { "epoch": 2.14, "learning_rate": 3.571213012840825e-05, "loss": 0.6218, "step": 35522 }, { "epoch": 2.14, "learning_rate": 3.5710621218294025e-05, "loss": 0.6832, "step": 35524 }, { "epoch": 2.14, "learning_rate": 3.5709112308179805e-05, "loss": 0.5811, "step": 35526 }, { "epoch": 2.14, "learning_rate": 3.570760339806558e-05, "loss": 0.4648, "step": 35528 }, { "epoch": 2.14, "learning_rate": 3.570609448795135e-05, "loss": 0.476, "step": 35530 }, { "epoch": 2.14, "learning_rate": 3.570458557783713e-05, "loss": 0.758, "step": 35532 }, { "epoch": 2.14, "learning_rate": 3.5703076667722904e-05, "loss": 0.7291, "step": 35534 }, { "epoch": 2.14, "learning_rate": 3.570156775760868e-05, "loss": 0.65, "step": 35536 }, { "epoch": 2.14, "learning_rate": 3.570005884749446e-05, "loss": 0.4598, "step": 35538 }, { "epoch": 2.15, "learning_rate": 3.569854993738023e-05, "loss": 0.6825, "step": 35540 }, { "epoch": 2.15, "learning_rate": 3.569704102726601e-05, "loss": 0.6983, "step": 35542 }, { "epoch": 2.15, "learning_rate": 3.5695532117151783e-05, "loss": 0.4915, "step": 35544 }, { "epoch": 2.15, "learning_rate": 3.5694023207037557e-05, "loss": 0.5806, "step": 35546 }, { "epoch": 2.15, "learning_rate": 3.5692514296923336e-05, "loss": 0.6476, "step": 35548 }, { "epoch": 2.15, "learning_rate": 3.569100538680911e-05, "loss": 0.466, "step": 35550 }, { "epoch": 2.15, "learning_rate": 3.568949647669488e-05, "loss": 0.536, "step": 35552 }, { "epoch": 2.15, "learning_rate": 3.568798756658066e-05, "loss": 0.5165, "step": 35554 }, { "epoch": 2.15, "learning_rate": 3.5686478656466436e-05, "loss": 0.7869, "step": 35556 }, { "epoch": 2.15, "learning_rate": 3.568496974635221e-05, "loss": 0.7445, "step": 35558 }, { "epoch": 2.15, "learning_rate": 3.568346083623799e-05, "loss": 0.669, "step": 35560 }, { "epoch": 2.15, "learning_rate": 3.568195192612376e-05, "loss": 0.5353, "step": 35562 }, { "epoch": 2.15, "learning_rate": 3.568044301600954e-05, "loss": 0.7009, "step": 35564 }, { "epoch": 2.15, "learning_rate": 3.5678934105895315e-05, "loss": 0.4302, "step": 35566 }, { "epoch": 2.15, "learning_rate": 3.567742519578109e-05, "loss": 0.5944, "step": 35568 }, { "epoch": 2.15, "learning_rate": 3.567591628566687e-05, "loss": 0.6852, "step": 35570 }, { "epoch": 2.15, "learning_rate": 3.5674407375552634e-05, "loss": 1.0456, "step": 35572 }, { "epoch": 2.15, "learning_rate": 3.5672898465438414e-05, "loss": 0.6763, "step": 35574 }, { "epoch": 2.15, "learning_rate": 3.5671389555324194e-05, "loss": 0.56, "step": 35576 }, { "epoch": 2.15, "learning_rate": 3.566988064520997e-05, "loss": 0.6221, "step": 35578 }, { "epoch": 2.15, "learning_rate": 3.566837173509574e-05, "loss": 0.6486, "step": 35580 }, { "epoch": 2.15, "learning_rate": 3.566686282498152e-05, "loss": 0.645, "step": 35582 }, { "epoch": 2.15, "learning_rate": 3.566535391486729e-05, "loss": 0.4885, "step": 35584 }, { "epoch": 2.15, "learning_rate": 3.5663845004753066e-05, "loss": 0.4215, "step": 35586 }, { "epoch": 2.15, "learning_rate": 3.566233609463884e-05, "loss": 0.6407, "step": 35588 }, { "epoch": 2.15, "learning_rate": 3.566082718452462e-05, "loss": 0.5502, "step": 35590 }, { "epoch": 2.15, "learning_rate": 3.56593182744104e-05, "loss": 0.6836, "step": 35592 }, { "epoch": 2.15, "learning_rate": 3.5657809364296166e-05, "loss": 0.7627, "step": 35594 }, { "epoch": 2.15, "learning_rate": 3.5656300454181946e-05, "loss": 0.4519, "step": 35596 }, { "epoch": 2.15, "learning_rate": 3.5654791544067726e-05, "loss": 0.4272, "step": 35598 }, { "epoch": 2.15, "learning_rate": 3.56532826339535e-05, "loss": 0.4507, "step": 35600 }, { "epoch": 2.15, "learning_rate": 3.565177372383927e-05, "loss": 0.6945, "step": 35602 }, { "epoch": 2.15, "learning_rate": 3.565026481372505e-05, "loss": 0.6066, "step": 35604 }, { "epoch": 2.15, "learning_rate": 3.5648755903610825e-05, "loss": 0.6212, "step": 35606 }, { "epoch": 2.15, "learning_rate": 3.56472469934966e-05, "loss": 0.638, "step": 35608 }, { "epoch": 2.15, "learning_rate": 3.564573808338237e-05, "loss": 0.9102, "step": 35610 }, { "epoch": 2.15, "learning_rate": 3.564422917326815e-05, "loss": 0.4385, "step": 35612 }, { "epoch": 2.15, "learning_rate": 3.564272026315393e-05, "loss": 0.5268, "step": 35614 }, { "epoch": 2.15, "learning_rate": 3.56412113530397e-05, "loss": 0.6359, "step": 35616 }, { "epoch": 2.15, "learning_rate": 3.563970244292548e-05, "loss": 0.6044, "step": 35618 }, { "epoch": 2.15, "learning_rate": 3.563819353281126e-05, "loss": 0.5989, "step": 35620 }, { "epoch": 2.15, "learning_rate": 3.5636684622697023e-05, "loss": 0.3727, "step": 35622 }, { "epoch": 2.15, "learning_rate": 3.56351757125828e-05, "loss": 0.5515, "step": 35624 }, { "epoch": 2.15, "learning_rate": 3.5633666802468576e-05, "loss": 0.5666, "step": 35626 }, { "epoch": 2.15, "learning_rate": 3.5632157892354356e-05, "loss": 0.5769, "step": 35628 }, { "epoch": 2.15, "learning_rate": 3.563064898224013e-05, "loss": 0.6399, "step": 35630 }, { "epoch": 2.15, "learning_rate": 3.56291400721259e-05, "loss": 0.6428, "step": 35632 }, { "epoch": 2.15, "learning_rate": 3.562763116201168e-05, "loss": 0.996, "step": 35634 }, { "epoch": 2.15, "learning_rate": 3.5626122251897456e-05, "loss": 0.4831, "step": 35636 }, { "epoch": 2.15, "learning_rate": 3.562461334178323e-05, "loss": 0.7388, "step": 35638 }, { "epoch": 2.15, "learning_rate": 3.562310443166901e-05, "loss": 0.5851, "step": 35640 }, { "epoch": 2.15, "learning_rate": 3.562159552155478e-05, "loss": 0.6409, "step": 35642 }, { "epoch": 2.15, "learning_rate": 3.5620086611440555e-05, "loss": 0.5632, "step": 35644 }, { "epoch": 2.15, "learning_rate": 3.5618577701326335e-05, "loss": 0.6072, "step": 35646 }, { "epoch": 2.15, "learning_rate": 3.561706879121211e-05, "loss": 0.466, "step": 35648 }, { "epoch": 2.15, "learning_rate": 3.561555988109789e-05, "loss": 0.4396, "step": 35650 }, { "epoch": 2.15, "learning_rate": 3.561405097098366e-05, "loss": 0.7572, "step": 35652 }, { "epoch": 2.15, "learning_rate": 3.5612542060869434e-05, "loss": 0.5052, "step": 35654 }, { "epoch": 2.15, "learning_rate": 3.5611033150755214e-05, "loss": 0.5463, "step": 35656 }, { "epoch": 2.15, "learning_rate": 3.560952424064099e-05, "loss": 0.4059, "step": 35658 }, { "epoch": 2.15, "learning_rate": 3.560801533052676e-05, "loss": 0.5027, "step": 35660 }, { "epoch": 2.15, "learning_rate": 3.560650642041254e-05, "loss": 0.5732, "step": 35662 }, { "epoch": 2.15, "learning_rate": 3.560499751029831e-05, "loss": 0.6007, "step": 35664 }, { "epoch": 2.15, "learning_rate": 3.5603488600184086e-05, "loss": 0.4368, "step": 35666 }, { "epoch": 2.15, "learning_rate": 3.5601979690069866e-05, "loss": 0.6908, "step": 35668 }, { "epoch": 2.15, "learning_rate": 3.560047077995564e-05, "loss": 0.7466, "step": 35670 }, { "epoch": 2.15, "learning_rate": 3.559896186984141e-05, "loss": 0.5738, "step": 35672 }, { "epoch": 2.15, "learning_rate": 3.559745295972719e-05, "loss": 0.692, "step": 35674 }, { "epoch": 2.15, "learning_rate": 3.5595944049612966e-05, "loss": 0.608, "step": 35676 }, { "epoch": 2.15, "learning_rate": 3.5594435139498745e-05, "loss": 0.5733, "step": 35678 }, { "epoch": 2.15, "learning_rate": 3.559292622938452e-05, "loss": 0.4517, "step": 35680 }, { "epoch": 2.15, "learning_rate": 3.559141731927029e-05, "loss": 0.567, "step": 35682 }, { "epoch": 2.15, "learning_rate": 3.558990840915607e-05, "loss": 0.7247, "step": 35684 }, { "epoch": 2.15, "learning_rate": 3.558839949904184e-05, "loss": 0.5033, "step": 35686 }, { "epoch": 2.15, "learning_rate": 3.558689058892762e-05, "loss": 0.6073, "step": 35688 }, { "epoch": 2.15, "learning_rate": 3.55853816788134e-05, "loss": 0.2902, "step": 35690 }, { "epoch": 2.15, "learning_rate": 3.558387276869917e-05, "loss": 0.3809, "step": 35692 }, { "epoch": 2.15, "learning_rate": 3.5582363858584944e-05, "loss": 0.726, "step": 35694 }, { "epoch": 2.15, "learning_rate": 3.5580854948470724e-05, "loss": 0.375, "step": 35696 }, { "epoch": 2.15, "learning_rate": 3.55793460383565e-05, "loss": 0.5668, "step": 35698 }, { "epoch": 2.15, "learning_rate": 3.557783712824227e-05, "loss": 0.7405, "step": 35700 }, { "epoch": 2.15, "learning_rate": 3.557632821812804e-05, "loss": 0.6452, "step": 35702 }, { "epoch": 2.15, "learning_rate": 3.557481930801382e-05, "loss": 0.7734, "step": 35704 }, { "epoch": 2.16, "learning_rate": 3.55733103978996e-05, "loss": 0.8123, "step": 35706 }, { "epoch": 2.16, "learning_rate": 3.557180148778537e-05, "loss": 0.8213, "step": 35708 }, { "epoch": 2.16, "learning_rate": 3.557029257767115e-05, "loss": 0.6553, "step": 35710 }, { "epoch": 2.16, "learning_rate": 3.556878366755693e-05, "loss": 0.5435, "step": 35712 }, { "epoch": 2.16, "learning_rate": 3.55672747574427e-05, "loss": 0.7189, "step": 35714 }, { "epoch": 2.16, "learning_rate": 3.5565765847328476e-05, "loss": 0.5485, "step": 35716 }, { "epoch": 2.16, "learning_rate": 3.556425693721425e-05, "loss": 0.6822, "step": 35718 }, { "epoch": 2.16, "learning_rate": 3.556274802710003e-05, "loss": 0.6187, "step": 35720 }, { "epoch": 2.16, "learning_rate": 3.55612391169858e-05, "loss": 0.4071, "step": 35722 }, { "epoch": 2.16, "learning_rate": 3.5559730206871575e-05, "loss": 0.5102, "step": 35724 }, { "epoch": 2.16, "learning_rate": 3.5558221296757355e-05, "loss": 0.6056, "step": 35726 }, { "epoch": 2.16, "learning_rate": 3.5556712386643135e-05, "loss": 0.5806, "step": 35728 }, { "epoch": 2.16, "learning_rate": 3.55552034765289e-05, "loss": 0.6761, "step": 35730 }, { "epoch": 2.16, "learning_rate": 3.555369456641468e-05, "loss": 0.5772, "step": 35732 }, { "epoch": 2.16, "learning_rate": 3.555218565630046e-05, "loss": 0.6288, "step": 35734 }, { "epoch": 2.16, "learning_rate": 3.555067674618623e-05, "loss": 0.4506, "step": 35736 }, { "epoch": 2.16, "learning_rate": 3.554916783607201e-05, "loss": 0.5176, "step": 35738 }, { "epoch": 2.16, "learning_rate": 3.554765892595778e-05, "loss": 0.69, "step": 35740 }, { "epoch": 2.16, "learning_rate": 3.554615001584356e-05, "loss": 0.7603, "step": 35742 }, { "epoch": 2.16, "learning_rate": 3.554464110572933e-05, "loss": 0.3502, "step": 35744 }, { "epoch": 2.16, "learning_rate": 3.5543132195615106e-05, "loss": 0.6717, "step": 35746 }, { "epoch": 2.16, "learning_rate": 3.5541623285500886e-05, "loss": 0.5177, "step": 35748 }, { "epoch": 2.16, "learning_rate": 3.554011437538666e-05, "loss": 0.5896, "step": 35750 }, { "epoch": 2.16, "learning_rate": 3.553860546527243e-05, "loss": 0.4797, "step": 35752 }, { "epoch": 2.16, "learning_rate": 3.553709655515821e-05, "loss": 0.731, "step": 35754 }, { "epoch": 2.16, "learning_rate": 3.5535587645043985e-05, "loss": 0.6752, "step": 35756 }, { "epoch": 2.16, "learning_rate": 3.553407873492976e-05, "loss": 0.5457, "step": 35758 }, { "epoch": 2.16, "learning_rate": 3.553256982481554e-05, "loss": 0.6626, "step": 35760 }, { "epoch": 2.16, "learning_rate": 3.553106091470131e-05, "loss": 0.6632, "step": 35762 }, { "epoch": 2.16, "learning_rate": 3.552955200458709e-05, "loss": 0.8282, "step": 35764 }, { "epoch": 2.16, "learning_rate": 3.5528043094472865e-05, "loss": 0.8096, "step": 35766 }, { "epoch": 2.16, "learning_rate": 3.552653418435864e-05, "loss": 0.5676, "step": 35768 }, { "epoch": 2.16, "learning_rate": 3.552502527424442e-05, "loss": 0.4834, "step": 35770 }, { "epoch": 2.16, "learning_rate": 3.552351636413019e-05, "loss": 0.5059, "step": 35772 }, { "epoch": 2.16, "learning_rate": 3.5522007454015964e-05, "loss": 0.7038, "step": 35774 }, { "epoch": 2.16, "learning_rate": 3.5520498543901744e-05, "loss": 0.8572, "step": 35776 }, { "epoch": 2.16, "learning_rate": 3.551898963378752e-05, "loss": 0.4374, "step": 35778 }, { "epoch": 2.16, "learning_rate": 3.551748072367329e-05, "loss": 0.4452, "step": 35780 }, { "epoch": 2.16, "learning_rate": 3.551597181355907e-05, "loss": 0.6734, "step": 35782 }, { "epoch": 2.16, "learning_rate": 3.551446290344484e-05, "loss": 0.4687, "step": 35784 }, { "epoch": 2.16, "learning_rate": 3.5512953993330616e-05, "loss": 0.5055, "step": 35786 }, { "epoch": 2.16, "learning_rate": 3.5511445083216396e-05, "loss": 0.3379, "step": 35788 }, { "epoch": 2.16, "learning_rate": 3.550993617310217e-05, "loss": 0.5878, "step": 35790 }, { "epoch": 2.16, "learning_rate": 3.550842726298795e-05, "loss": 0.5424, "step": 35792 }, { "epoch": 2.16, "learning_rate": 3.550691835287372e-05, "loss": 0.5232, "step": 35794 }, { "epoch": 2.16, "learning_rate": 3.5505409442759495e-05, "loss": 0.6419, "step": 35796 }, { "epoch": 2.16, "learning_rate": 3.5503900532645275e-05, "loss": 0.7041, "step": 35798 }, { "epoch": 2.16, "learning_rate": 3.550239162253104e-05, "loss": 0.4808, "step": 35800 }, { "epoch": 2.16, "learning_rate": 3.550088271241682e-05, "loss": 0.9015, "step": 35802 }, { "epoch": 2.16, "learning_rate": 3.54993738023026e-05, "loss": 0.638, "step": 35804 }, { "epoch": 2.16, "learning_rate": 3.5497864892188375e-05, "loss": 0.6862, "step": 35806 }, { "epoch": 2.16, "learning_rate": 3.549635598207415e-05, "loss": 0.5363, "step": 35808 }, { "epoch": 2.16, "learning_rate": 3.549484707195993e-05, "loss": 0.5043, "step": 35810 }, { "epoch": 2.16, "learning_rate": 3.54933381618457e-05, "loss": 0.5436, "step": 35812 }, { "epoch": 2.16, "learning_rate": 3.549182925173148e-05, "loss": 0.7124, "step": 35814 }, { "epoch": 2.16, "learning_rate": 3.549032034161725e-05, "loss": 0.3605, "step": 35816 }, { "epoch": 2.16, "learning_rate": 3.548881143150303e-05, "loss": 0.5178, "step": 35818 }, { "epoch": 2.16, "learning_rate": 3.548730252138881e-05, "loss": 0.5866, "step": 35820 }, { "epoch": 2.16, "learning_rate": 3.548579361127457e-05, "loss": 0.7561, "step": 35822 }, { "epoch": 2.16, "learning_rate": 3.548428470116035e-05, "loss": 0.684, "step": 35824 }, { "epoch": 2.16, "learning_rate": 3.548277579104613e-05, "loss": 0.5412, "step": 35826 }, { "epoch": 2.16, "learning_rate": 3.5481266880931906e-05, "loss": 0.5026, "step": 35828 }, { "epoch": 2.16, "learning_rate": 3.547975797081768e-05, "loss": 0.7606, "step": 35830 }, { "epoch": 2.16, "learning_rate": 3.547824906070345e-05, "loss": 0.5134, "step": 35832 }, { "epoch": 2.16, "learning_rate": 3.547674015058923e-05, "loss": 0.5042, "step": 35834 }, { "epoch": 2.16, "learning_rate": 3.5475231240475005e-05, "loss": 0.5465, "step": 35836 }, { "epoch": 2.16, "learning_rate": 3.547372233036078e-05, "loss": 0.4247, "step": 35838 }, { "epoch": 2.16, "learning_rate": 3.547221342024656e-05, "loss": 0.6715, "step": 35840 }, { "epoch": 2.16, "learning_rate": 3.547070451013234e-05, "loss": 0.4645, "step": 35842 }, { "epoch": 2.16, "learning_rate": 3.5469195600018105e-05, "loss": 0.6582, "step": 35844 }, { "epoch": 2.16, "learning_rate": 3.5467686689903885e-05, "loss": 0.7001, "step": 35846 }, { "epoch": 2.16, "learning_rate": 3.5466177779789664e-05, "loss": 0.534, "step": 35848 }, { "epoch": 2.16, "learning_rate": 3.546466886967543e-05, "loss": 0.5648, "step": 35850 }, { "epoch": 2.16, "learning_rate": 3.546315995956121e-05, "loss": 0.5215, "step": 35852 }, { "epoch": 2.16, "learning_rate": 3.5461651049446984e-05, "loss": 0.59, "step": 35854 }, { "epoch": 2.16, "learning_rate": 3.5460142139332764e-05, "loss": 0.5525, "step": 35856 }, { "epoch": 2.16, "learning_rate": 3.545863322921854e-05, "loss": 0.5094, "step": 35858 }, { "epoch": 2.16, "learning_rate": 3.545712431910431e-05, "loss": 0.7589, "step": 35860 }, { "epoch": 2.16, "learning_rate": 3.545561540899009e-05, "loss": 0.7705, "step": 35862 }, { "epoch": 2.16, "learning_rate": 3.545410649887587e-05, "loss": 0.5871, "step": 35864 }, { "epoch": 2.16, "learning_rate": 3.5452597588761636e-05, "loss": 0.4091, "step": 35866 }, { "epoch": 2.16, "learning_rate": 3.5451088678647416e-05, "loss": 0.4653, "step": 35868 }, { "epoch": 2.17, "learning_rate": 3.544957976853319e-05, "loss": 0.4485, "step": 35870 }, { "epoch": 2.17, "learning_rate": 3.544807085841896e-05, "loss": 0.4865, "step": 35872 }, { "epoch": 2.17, "learning_rate": 3.544656194830474e-05, "loss": 0.6259, "step": 35874 }, { "epoch": 2.17, "learning_rate": 3.5445053038190515e-05, "loss": 0.4703, "step": 35876 }, { "epoch": 2.17, "learning_rate": 3.5443544128076295e-05, "loss": 0.463, "step": 35878 }, { "epoch": 2.17, "learning_rate": 3.544203521796207e-05, "loss": 0.6636, "step": 35880 }, { "epoch": 2.17, "learning_rate": 3.544052630784784e-05, "loss": 0.7253, "step": 35882 }, { "epoch": 2.17, "learning_rate": 3.543901739773362e-05, "loss": 0.5701, "step": 35884 }, { "epoch": 2.17, "learning_rate": 3.5437508487619394e-05, "loss": 0.6763, "step": 35886 }, { "epoch": 2.17, "learning_rate": 3.543599957750517e-05, "loss": 0.6978, "step": 35888 }, { "epoch": 2.17, "learning_rate": 3.543449066739095e-05, "loss": 0.5045, "step": 35890 }, { "epoch": 2.17, "learning_rate": 3.543298175727672e-05, "loss": 0.3354, "step": 35892 }, { "epoch": 2.17, "learning_rate": 3.5431472847162494e-05, "loss": 0.6105, "step": 35894 }, { "epoch": 2.17, "learning_rate": 3.5429963937048274e-05, "loss": 0.616, "step": 35896 }, { "epoch": 2.17, "learning_rate": 3.542845502693405e-05, "loss": 0.6944, "step": 35898 }, { "epoch": 2.17, "learning_rate": 3.542694611681982e-05, "loss": 0.5966, "step": 35900 }, { "epoch": 2.17, "learning_rate": 3.54254372067056e-05, "loss": 0.5728, "step": 35902 }, { "epoch": 2.17, "learning_rate": 3.542392829659137e-05, "loss": 0.4065, "step": 35904 }, { "epoch": 2.17, "learning_rate": 3.542241938647715e-05, "loss": 0.4983, "step": 35906 }, { "epoch": 2.17, "learning_rate": 3.5420910476362926e-05, "loss": 0.786, "step": 35908 }, { "epoch": 2.17, "learning_rate": 3.54194015662487e-05, "loss": 0.6417, "step": 35910 }, { "epoch": 2.17, "learning_rate": 3.541789265613448e-05, "loss": 0.4701, "step": 35912 }, { "epoch": 2.17, "learning_rate": 3.5416383746020245e-05, "loss": 0.7294, "step": 35914 }, { "epoch": 2.17, "learning_rate": 3.5414874835906025e-05, "loss": 0.4128, "step": 35916 }, { "epoch": 2.17, "learning_rate": 3.5413365925791805e-05, "loss": 0.8333, "step": 35918 }, { "epoch": 2.17, "learning_rate": 3.541185701567758e-05, "loss": 0.5611, "step": 35920 }, { "epoch": 2.17, "learning_rate": 3.541034810556335e-05, "loss": 0.4306, "step": 35922 }, { "epoch": 2.17, "learning_rate": 3.540883919544913e-05, "loss": 0.7353, "step": 35924 }, { "epoch": 2.17, "learning_rate": 3.5407330285334904e-05, "loss": 0.4532, "step": 35926 }, { "epoch": 2.17, "learning_rate": 3.5405821375220684e-05, "loss": 0.3313, "step": 35928 }, { "epoch": 2.17, "learning_rate": 3.540431246510645e-05, "loss": 0.5897, "step": 35930 }, { "epoch": 2.17, "learning_rate": 3.540280355499223e-05, "loss": 0.4333, "step": 35932 }, { "epoch": 2.17, "learning_rate": 3.540129464487801e-05, "loss": 0.7299, "step": 35934 }, { "epoch": 2.17, "learning_rate": 3.539978573476378e-05, "loss": 0.731, "step": 35936 }, { "epoch": 2.17, "learning_rate": 3.539827682464956e-05, "loss": 0.4457, "step": 35938 }, { "epoch": 2.17, "learning_rate": 3.5396767914535337e-05, "loss": 0.6522, "step": 35940 }, { "epoch": 2.17, "learning_rate": 3.539525900442111e-05, "loss": 0.5087, "step": 35942 }, { "epoch": 2.17, "learning_rate": 3.539375009430688e-05, "loss": 0.5208, "step": 35944 }, { "epoch": 2.17, "learning_rate": 3.5392241184192656e-05, "loss": 0.7146, "step": 35946 }, { "epoch": 2.17, "learning_rate": 3.5390732274078436e-05, "loss": 0.8815, "step": 35948 }, { "epoch": 2.17, "learning_rate": 3.538922336396421e-05, "loss": 0.7497, "step": 35950 }, { "epoch": 2.17, "learning_rate": 3.538771445384998e-05, "loss": 0.484, "step": 35952 }, { "epoch": 2.17, "learning_rate": 3.538620554373576e-05, "loss": 0.4734, "step": 35954 }, { "epoch": 2.17, "learning_rate": 3.538469663362154e-05, "loss": 0.4588, "step": 35956 }, { "epoch": 2.17, "learning_rate": 3.538318772350731e-05, "loss": 0.6326, "step": 35958 }, { "epoch": 2.17, "learning_rate": 3.538167881339309e-05, "loss": 0.9101, "step": 35960 }, { "epoch": 2.17, "learning_rate": 3.538016990327887e-05, "loss": 0.6334, "step": 35962 }, { "epoch": 2.17, "learning_rate": 3.5378660993164634e-05, "loss": 0.5191, "step": 35964 }, { "epoch": 2.17, "learning_rate": 3.5377152083050414e-05, "loss": 0.6865, "step": 35966 }, { "epoch": 2.17, "learning_rate": 3.537564317293619e-05, "loss": 0.5597, "step": 35968 }, { "epoch": 2.17, "learning_rate": 3.537413426282197e-05, "loss": 0.6338, "step": 35970 }, { "epoch": 2.17, "learning_rate": 3.537262535270774e-05, "loss": 0.6631, "step": 35972 }, { "epoch": 2.17, "learning_rate": 3.5371116442593514e-05, "loss": 0.6471, "step": 35974 }, { "epoch": 2.17, "learning_rate": 3.5369607532479294e-05, "loss": 0.6699, "step": 35976 }, { "epoch": 2.17, "learning_rate": 3.5368098622365073e-05, "loss": 0.5058, "step": 35978 }, { "epoch": 2.17, "learning_rate": 3.536658971225084e-05, "loss": 0.647, "step": 35980 }, { "epoch": 2.17, "learning_rate": 3.536508080213662e-05, "loss": 0.676, "step": 35982 }, { "epoch": 2.17, "learning_rate": 3.536357189202239e-05, "loss": 0.5434, "step": 35984 }, { "epoch": 2.17, "learning_rate": 3.5362062981908166e-05, "loss": 0.554, "step": 35986 }, { "epoch": 2.17, "learning_rate": 3.5360554071793946e-05, "loss": 0.6087, "step": 35988 }, { "epoch": 2.17, "learning_rate": 3.535904516167972e-05, "loss": 0.6098, "step": 35990 }, { "epoch": 2.17, "learning_rate": 3.53575362515655e-05, "loss": 0.6684, "step": 35992 }, { "epoch": 2.17, "learning_rate": 3.535602734145127e-05, "loss": 0.7841, "step": 35994 }, { "epoch": 2.17, "learning_rate": 3.5354518431337045e-05, "loss": 0.8028, "step": 35996 }, { "epoch": 2.17, "learning_rate": 3.5353009521222825e-05, "loss": 0.4998, "step": 35998 }, { "epoch": 2.17, "learning_rate": 3.53515006111086e-05, "loss": 0.7927, "step": 36000 }, { "epoch": 2.17, "learning_rate": 3.534999170099437e-05, "loss": 0.765, "step": 36002 }, { "epoch": 2.17, "learning_rate": 3.534848279088015e-05, "loss": 0.6663, "step": 36004 }, { "epoch": 2.17, "learning_rate": 3.5346973880765924e-05, "loss": 0.6633, "step": 36006 }, { "epoch": 2.17, "learning_rate": 3.53454649706517e-05, "loss": 0.9455, "step": 36008 }, { "epoch": 2.17, "learning_rate": 3.534395606053748e-05, "loss": 0.6695, "step": 36010 }, { "epoch": 2.17, "learning_rate": 3.534244715042325e-05, "loss": 0.6626, "step": 36012 }, { "epoch": 2.17, "learning_rate": 3.5340938240309024e-05, "loss": 0.8266, "step": 36014 }, { "epoch": 2.17, "learning_rate": 3.5339429330194803e-05, "loss": 0.499, "step": 36016 }, { "epoch": 2.17, "learning_rate": 3.5337920420080577e-05, "loss": 0.5821, "step": 36018 }, { "epoch": 2.17, "learning_rate": 3.5336411509966356e-05, "loss": 0.6184, "step": 36020 }, { "epoch": 2.17, "learning_rate": 3.533490259985213e-05, "loss": 0.5896, "step": 36022 }, { "epoch": 2.17, "learning_rate": 3.53333936897379e-05, "loss": 0.6227, "step": 36024 }, { "epoch": 2.17, "learning_rate": 3.533188477962368e-05, "loss": 1.0594, "step": 36026 }, { "epoch": 2.17, "learning_rate": 3.5330375869509456e-05, "loss": 0.628, "step": 36028 }, { "epoch": 2.17, "learning_rate": 3.532886695939523e-05, "loss": 0.5107, "step": 36030 }, { "epoch": 2.17, "learning_rate": 3.532735804928101e-05, "loss": 0.5627, "step": 36032 }, { "epoch": 2.17, "learning_rate": 3.532584913916678e-05, "loss": 0.608, "step": 36034 }, { "epoch": 2.18, "learning_rate": 3.5324340229052555e-05, "loss": 0.4142, "step": 36036 }, { "epoch": 2.18, "learning_rate": 3.5322831318938335e-05, "loss": 0.5954, "step": 36038 }, { "epoch": 2.18, "learning_rate": 3.532132240882411e-05, "loss": 0.5064, "step": 36040 }, { "epoch": 2.18, "learning_rate": 3.531981349870989e-05, "loss": 0.5243, "step": 36042 }, { "epoch": 2.18, "learning_rate": 3.5318304588595654e-05, "loss": 0.5317, "step": 36044 }, { "epoch": 2.18, "learning_rate": 3.5316795678481434e-05, "loss": 0.6772, "step": 36046 }, { "epoch": 2.18, "learning_rate": 3.5315286768367214e-05, "loss": 0.8794, "step": 36048 }, { "epoch": 2.18, "learning_rate": 3.531377785825298e-05, "loss": 0.4967, "step": 36050 }, { "epoch": 2.18, "learning_rate": 3.531226894813876e-05, "loss": 0.5287, "step": 36052 }, { "epoch": 2.18, "learning_rate": 3.531076003802454e-05, "loss": 0.6127, "step": 36054 }, { "epoch": 2.18, "learning_rate": 3.5309251127910313e-05, "loss": 0.5584, "step": 36056 }, { "epoch": 2.18, "learning_rate": 3.5307742217796087e-05, "loss": 0.7665, "step": 36058 }, { "epoch": 2.18, "learning_rate": 3.530623330768186e-05, "loss": 0.6041, "step": 36060 }, { "epoch": 2.18, "learning_rate": 3.530472439756764e-05, "loss": 0.4944, "step": 36062 }, { "epoch": 2.18, "learning_rate": 3.530321548745341e-05, "loss": 0.4893, "step": 36064 }, { "epoch": 2.18, "learning_rate": 3.5301706577339186e-05, "loss": 0.6407, "step": 36066 }, { "epoch": 2.18, "learning_rate": 3.5300197667224966e-05, "loss": 0.41, "step": 36068 }, { "epoch": 2.18, "learning_rate": 3.5298688757110746e-05, "loss": 0.5715, "step": 36070 }, { "epoch": 2.18, "learning_rate": 3.529717984699651e-05, "loss": 0.672, "step": 36072 }, { "epoch": 2.18, "learning_rate": 3.529567093688229e-05, "loss": 0.8058, "step": 36074 }, { "epoch": 2.18, "learning_rate": 3.5294162026768065e-05, "loss": 0.5203, "step": 36076 }, { "epoch": 2.18, "learning_rate": 3.529265311665384e-05, "loss": 0.6027, "step": 36078 }, { "epoch": 2.18, "learning_rate": 3.529114420653962e-05, "loss": 0.5111, "step": 36080 }, { "epoch": 2.18, "learning_rate": 3.528963529642539e-05, "loss": 0.5645, "step": 36082 }, { "epoch": 2.18, "learning_rate": 3.528812638631117e-05, "loss": 0.5491, "step": 36084 }, { "epoch": 2.18, "learning_rate": 3.5286617476196944e-05, "loss": 0.505, "step": 36086 }, { "epoch": 2.18, "learning_rate": 3.528510856608272e-05, "loss": 0.5373, "step": 36088 }, { "epoch": 2.18, "learning_rate": 3.52835996559685e-05, "loss": 0.6646, "step": 36090 }, { "epoch": 2.18, "learning_rate": 3.528209074585428e-05, "loss": 0.6663, "step": 36092 }, { "epoch": 2.18, "learning_rate": 3.5280581835740043e-05, "loss": 0.5499, "step": 36094 }, { "epoch": 2.18, "learning_rate": 3.527907292562582e-05, "loss": 0.6065, "step": 36096 }, { "epoch": 2.18, "learning_rate": 3.5277564015511596e-05, "loss": 0.5877, "step": 36098 }, { "epoch": 2.18, "learning_rate": 3.527605510539737e-05, "loss": 0.4567, "step": 36100 }, { "epoch": 2.18, "learning_rate": 3.527454619528315e-05, "loss": 0.5629, "step": 36102 }, { "epoch": 2.18, "learning_rate": 3.527303728516892e-05, "loss": 0.5124, "step": 36104 }, { "epoch": 2.18, "learning_rate": 3.52715283750547e-05, "loss": 0.754, "step": 36106 }, { "epoch": 2.18, "learning_rate": 3.5270019464940476e-05, "loss": 0.9093, "step": 36108 }, { "epoch": 2.18, "learning_rate": 3.526851055482625e-05, "loss": 0.5379, "step": 36110 }, { "epoch": 2.18, "learning_rate": 3.526700164471203e-05, "loss": 0.6929, "step": 36112 }, { "epoch": 2.18, "learning_rate": 3.52654927345978e-05, "loss": 0.4955, "step": 36114 }, { "epoch": 2.18, "learning_rate": 3.5263983824483575e-05, "loss": 0.8438, "step": 36116 }, { "epoch": 2.18, "learning_rate": 3.5262474914369355e-05, "loss": 0.543, "step": 36118 }, { "epoch": 2.18, "learning_rate": 3.526096600425513e-05, "loss": 0.7419, "step": 36120 }, { "epoch": 2.18, "learning_rate": 3.52594570941409e-05, "loss": 0.5874, "step": 36122 }, { "epoch": 2.18, "learning_rate": 3.525794818402668e-05, "loss": 0.6693, "step": 36124 }, { "epoch": 2.18, "learning_rate": 3.5256439273912454e-05, "loss": 0.612, "step": 36126 }, { "epoch": 2.18, "learning_rate": 3.525493036379823e-05, "loss": 0.6629, "step": 36128 }, { "epoch": 2.18, "learning_rate": 3.525342145368401e-05, "loss": 0.7053, "step": 36130 }, { "epoch": 2.18, "learning_rate": 3.525191254356978e-05, "loss": 0.6997, "step": 36132 }, { "epoch": 2.18, "learning_rate": 3.525040363345556e-05, "loss": 0.6973, "step": 36134 }, { "epoch": 2.18, "learning_rate": 3.524889472334133e-05, "loss": 0.6329, "step": 36136 }, { "epoch": 2.18, "learning_rate": 3.5247385813227106e-05, "loss": 0.6728, "step": 36138 }, { "epoch": 2.18, "learning_rate": 3.5245876903112886e-05, "loss": 0.4003, "step": 36140 }, { "epoch": 2.18, "learning_rate": 3.524436799299866e-05, "loss": 0.5188, "step": 36142 }, { "epoch": 2.18, "learning_rate": 3.524285908288443e-05, "loss": 0.7583, "step": 36144 }, { "epoch": 2.18, "learning_rate": 3.524135017277021e-05, "loss": 0.7594, "step": 36146 }, { "epoch": 2.18, "learning_rate": 3.5239841262655986e-05, "loss": 0.8252, "step": 36148 }, { "epoch": 2.18, "learning_rate": 3.523833235254176e-05, "loss": 0.7473, "step": 36150 }, { "epoch": 2.18, "learning_rate": 3.523682344242754e-05, "loss": 0.4745, "step": 36152 }, { "epoch": 2.18, "learning_rate": 3.523531453231331e-05, "loss": 0.5205, "step": 36154 }, { "epoch": 2.18, "learning_rate": 3.523380562219909e-05, "loss": 0.4269, "step": 36156 }, { "epoch": 2.18, "learning_rate": 3.523229671208486e-05, "loss": 0.6485, "step": 36158 }, { "epoch": 2.18, "learning_rate": 3.523078780197064e-05, "loss": 0.9619, "step": 36160 }, { "epoch": 2.18, "learning_rate": 3.522927889185642e-05, "loss": 0.507, "step": 36162 }, { "epoch": 2.18, "learning_rate": 3.5227769981742184e-05, "loss": 0.5281, "step": 36164 }, { "epoch": 2.18, "learning_rate": 3.5226261071627964e-05, "loss": 0.4016, "step": 36166 }, { "epoch": 2.18, "learning_rate": 3.5224752161513744e-05, "loss": 0.7483, "step": 36168 }, { "epoch": 2.18, "learning_rate": 3.522324325139952e-05, "loss": 0.8186, "step": 36170 }, { "epoch": 2.18, "learning_rate": 3.522173434128529e-05, "loss": 0.641, "step": 36172 }, { "epoch": 2.18, "learning_rate": 3.522022543117106e-05, "loss": 0.6572, "step": 36174 }, { "epoch": 2.18, "learning_rate": 3.521871652105684e-05, "loss": 0.8213, "step": 36176 }, { "epoch": 2.18, "learning_rate": 3.5217207610942616e-05, "loss": 0.5557, "step": 36178 }, { "epoch": 2.18, "learning_rate": 3.521569870082839e-05, "loss": 0.6733, "step": 36180 }, { "epoch": 2.18, "learning_rate": 3.521418979071417e-05, "loss": 0.5683, "step": 36182 }, { "epoch": 2.18, "learning_rate": 3.521268088059995e-05, "loss": 0.5791, "step": 36184 }, { "epoch": 2.18, "learning_rate": 3.5211171970485716e-05, "loss": 0.683, "step": 36186 }, { "epoch": 2.18, "learning_rate": 3.5209663060371496e-05, "loss": 0.4575, "step": 36188 }, { "epoch": 2.18, "learning_rate": 3.520815415025727e-05, "loss": 0.5728, "step": 36190 }, { "epoch": 2.18, "learning_rate": 3.520664524014305e-05, "loss": 0.8269, "step": 36192 }, { "epoch": 2.18, "learning_rate": 3.520513633002882e-05, "loss": 0.6386, "step": 36194 }, { "epoch": 2.18, "learning_rate": 3.5203627419914595e-05, "loss": 0.5012, "step": 36196 }, { "epoch": 2.18, "learning_rate": 3.5202118509800375e-05, "loss": 0.4669, "step": 36198 }, { "epoch": 2.18, "learning_rate": 3.520060959968615e-05, "loss": 0.6053, "step": 36200 }, { "epoch": 2.19, "learning_rate": 3.519910068957192e-05, "loss": 0.6139, "step": 36202 }, { "epoch": 2.19, "learning_rate": 3.51975917794577e-05, "loss": 0.5797, "step": 36204 }, { "epoch": 2.19, "learning_rate": 3.519608286934348e-05, "loss": 0.6493, "step": 36206 }, { "epoch": 2.19, "learning_rate": 3.519457395922925e-05, "loss": 0.775, "step": 36208 }, { "epoch": 2.19, "learning_rate": 3.519306504911503e-05, "loss": 0.5911, "step": 36210 }, { "epoch": 2.19, "learning_rate": 3.51915561390008e-05, "loss": 0.3782, "step": 36212 }, { "epoch": 2.19, "learning_rate": 3.519004722888657e-05, "loss": 0.4951, "step": 36214 }, { "epoch": 2.19, "learning_rate": 3.518853831877235e-05, "loss": 0.4642, "step": 36216 }, { "epoch": 2.19, "learning_rate": 3.5187029408658126e-05, "loss": 0.4638, "step": 36218 }, { "epoch": 2.19, "learning_rate": 3.5185520498543906e-05, "loss": 0.5774, "step": 36220 }, { "epoch": 2.19, "learning_rate": 3.518401158842968e-05, "loss": 0.7414, "step": 36222 }, { "epoch": 2.19, "learning_rate": 3.518250267831545e-05, "loss": 0.3946, "step": 36224 }, { "epoch": 2.19, "learning_rate": 3.518099376820123e-05, "loss": 0.5212, "step": 36226 }, { "epoch": 2.19, "learning_rate": 3.5179484858087005e-05, "loss": 0.5588, "step": 36228 }, { "epoch": 2.19, "learning_rate": 3.517797594797278e-05, "loss": 0.7215, "step": 36230 }, { "epoch": 2.19, "learning_rate": 3.517646703785856e-05, "loss": 0.7563, "step": 36232 }, { "epoch": 2.19, "learning_rate": 3.517495812774433e-05, "loss": 0.4492, "step": 36234 }, { "epoch": 2.19, "learning_rate": 3.5173449217630105e-05, "loss": 0.7488, "step": 36236 }, { "epoch": 2.19, "learning_rate": 3.5171940307515885e-05, "loss": 0.7569, "step": 36238 }, { "epoch": 2.19, "learning_rate": 3.517043139740166e-05, "loss": 0.6713, "step": 36240 }, { "epoch": 2.19, "learning_rate": 3.516892248728743e-05, "loss": 0.5292, "step": 36242 }, { "epoch": 2.19, "learning_rate": 3.516741357717321e-05, "loss": 0.6343, "step": 36244 }, { "epoch": 2.19, "learning_rate": 3.5165904667058984e-05, "loss": 0.6669, "step": 36246 }, { "epoch": 2.19, "learning_rate": 3.5164395756944764e-05, "loss": 0.623, "step": 36248 }, { "epoch": 2.19, "learning_rate": 3.516288684683054e-05, "loss": 0.7554, "step": 36250 }, { "epoch": 2.19, "learning_rate": 3.516137793671631e-05, "loss": 0.6856, "step": 36252 }, { "epoch": 2.19, "learning_rate": 3.515986902660209e-05, "loss": 0.4591, "step": 36254 }, { "epoch": 2.19, "learning_rate": 3.515836011648786e-05, "loss": 0.6573, "step": 36256 }, { "epoch": 2.19, "learning_rate": 3.5156851206373636e-05, "loss": 0.5281, "step": 36258 }, { "epoch": 2.19, "learning_rate": 3.5155342296259416e-05, "loss": 0.5536, "step": 36260 }, { "epoch": 2.19, "learning_rate": 3.515383338614519e-05, "loss": 0.3361, "step": 36262 }, { "epoch": 2.19, "learning_rate": 3.515232447603096e-05, "loss": 0.4891, "step": 36264 }, { "epoch": 2.19, "learning_rate": 3.515081556591674e-05, "loss": 0.5151, "step": 36266 }, { "epoch": 2.19, "learning_rate": 3.5149306655802515e-05, "loss": 0.6687, "step": 36268 }, { "epoch": 2.19, "learning_rate": 3.5147797745688295e-05, "loss": 0.5033, "step": 36270 }, { "epoch": 2.19, "learning_rate": 3.514628883557406e-05, "loss": 0.9466, "step": 36272 }, { "epoch": 2.19, "learning_rate": 3.514477992545984e-05, "loss": 0.454, "step": 36274 }, { "epoch": 2.19, "learning_rate": 3.514327101534562e-05, "loss": 0.416, "step": 36276 }, { "epoch": 2.19, "learning_rate": 3.514176210523139e-05, "loss": 0.6661, "step": 36278 }, { "epoch": 2.19, "learning_rate": 3.514025319511717e-05, "loss": 0.6721, "step": 36280 }, { "epoch": 2.19, "learning_rate": 3.513874428500295e-05, "loss": 0.6488, "step": 36282 }, { "epoch": 2.19, "learning_rate": 3.513723537488872e-05, "loss": 0.5416, "step": 36284 }, { "epoch": 2.19, "learning_rate": 3.5135726464774494e-05, "loss": 0.5594, "step": 36286 }, { "epoch": 2.19, "learning_rate": 3.513421755466027e-05, "loss": 0.6717, "step": 36288 }, { "epoch": 2.19, "learning_rate": 3.513270864454605e-05, "loss": 0.5417, "step": 36290 }, { "epoch": 2.19, "learning_rate": 3.513119973443182e-05, "loss": 0.6645, "step": 36292 }, { "epoch": 2.19, "learning_rate": 3.512969082431759e-05, "loss": 0.6377, "step": 36294 }, { "epoch": 2.19, "learning_rate": 3.512818191420337e-05, "loss": 0.6963, "step": 36296 }, { "epoch": 2.19, "learning_rate": 3.512667300408915e-05, "loss": 0.6786, "step": 36298 }, { "epoch": 2.19, "learning_rate": 3.512516409397492e-05, "loss": 0.5031, "step": 36300 }, { "epoch": 2.19, "learning_rate": 3.51236551838607e-05, "loss": 0.7356, "step": 36302 }, { "epoch": 2.19, "learning_rate": 3.512214627374647e-05, "loss": 0.5775, "step": 36304 }, { "epoch": 2.19, "learning_rate": 3.512063736363225e-05, "loss": 0.4748, "step": 36306 }, { "epoch": 2.19, "learning_rate": 3.5119128453518025e-05, "loss": 0.634, "step": 36308 }, { "epoch": 2.19, "learning_rate": 3.51176195434038e-05, "loss": 0.6509, "step": 36310 }, { "epoch": 2.19, "learning_rate": 3.511611063328958e-05, "loss": 0.5157, "step": 36312 }, { "epoch": 2.19, "learning_rate": 3.511460172317535e-05, "loss": 0.7886, "step": 36314 }, { "epoch": 2.19, "learning_rate": 3.5113092813061125e-05, "loss": 0.5228, "step": 36316 }, { "epoch": 2.19, "learning_rate": 3.5111583902946905e-05, "loss": 0.5232, "step": 36318 }, { "epoch": 2.19, "learning_rate": 3.5110074992832684e-05, "loss": 0.3935, "step": 36320 }, { "epoch": 2.19, "learning_rate": 3.510856608271845e-05, "loss": 0.731, "step": 36322 }, { "epoch": 2.19, "learning_rate": 3.510705717260423e-05, "loss": 0.6221, "step": 36324 }, { "epoch": 2.19, "learning_rate": 3.5105548262490004e-05, "loss": 0.5352, "step": 36326 }, { "epoch": 2.19, "learning_rate": 3.510403935237578e-05, "loss": 0.5936, "step": 36328 }, { "epoch": 2.19, "learning_rate": 3.510253044226156e-05, "loss": 0.5666, "step": 36330 }, { "epoch": 2.19, "learning_rate": 3.510102153214733e-05, "loss": 0.5327, "step": 36332 }, { "epoch": 2.19, "learning_rate": 3.509951262203311e-05, "loss": 0.6955, "step": 36334 }, { "epoch": 2.19, "learning_rate": 3.509800371191888e-05, "loss": 0.7098, "step": 36336 }, { "epoch": 2.19, "learning_rate": 3.5096494801804656e-05, "loss": 0.5609, "step": 36338 }, { "epoch": 2.19, "learning_rate": 3.5094985891690436e-05, "loss": 0.5997, "step": 36340 }, { "epoch": 2.19, "learning_rate": 3.509347698157621e-05, "loss": 0.3785, "step": 36342 }, { "epoch": 2.19, "learning_rate": 3.509196807146198e-05, "loss": 0.6162, "step": 36344 }, { "epoch": 2.19, "learning_rate": 3.509045916134776e-05, "loss": 0.6092, "step": 36346 }, { "epoch": 2.19, "learning_rate": 3.5088950251233535e-05, "loss": 0.5909, "step": 36348 }, { "epoch": 2.19, "learning_rate": 3.508744134111931e-05, "loss": 0.7296, "step": 36350 }, { "epoch": 2.19, "learning_rate": 3.508593243100509e-05, "loss": 0.8962, "step": 36352 }, { "epoch": 2.19, "learning_rate": 3.508442352089086e-05, "loss": 0.561, "step": 36354 }, { "epoch": 2.19, "learning_rate": 3.508291461077664e-05, "loss": 0.506, "step": 36356 }, { "epoch": 2.19, "learning_rate": 3.5081405700662415e-05, "loss": 0.6987, "step": 36358 }, { "epoch": 2.19, "learning_rate": 3.507989679054819e-05, "loss": 0.5382, "step": 36360 }, { "epoch": 2.19, "learning_rate": 3.507838788043397e-05, "loss": 0.5302, "step": 36362 }, { "epoch": 2.19, "learning_rate": 3.507687897031974e-05, "loss": 0.5704, "step": 36364 }, { "epoch": 2.19, "learning_rate": 3.5075370060205514e-05, "loss": 0.7739, "step": 36366 }, { "epoch": 2.2, "learning_rate": 3.5073861150091294e-05, "loss": 0.5405, "step": 36368 }, { "epoch": 2.2, "learning_rate": 3.507235223997707e-05, "loss": 0.6525, "step": 36370 }, { "epoch": 2.2, "learning_rate": 3.507084332986284e-05, "loss": 0.4849, "step": 36372 }, { "epoch": 2.2, "learning_rate": 3.506933441974862e-05, "loss": 0.6248, "step": 36374 }, { "epoch": 2.2, "learning_rate": 3.506782550963439e-05, "loss": 0.606, "step": 36376 }, { "epoch": 2.2, "learning_rate": 3.5066316599520166e-05, "loss": 0.6129, "step": 36378 }, { "epoch": 2.2, "learning_rate": 3.5064807689405946e-05, "loss": 0.6313, "step": 36380 }, { "epoch": 2.2, "learning_rate": 3.506329877929172e-05, "loss": 0.6374, "step": 36382 }, { "epoch": 2.2, "learning_rate": 3.50617898691775e-05, "loss": 0.5363, "step": 36384 }, { "epoch": 2.2, "learning_rate": 3.5060280959063265e-05, "loss": 0.5845, "step": 36386 }, { "epoch": 2.2, "learning_rate": 3.5058772048949045e-05, "loss": 0.6393, "step": 36388 }, { "epoch": 2.2, "learning_rate": 3.5057263138834825e-05, "loss": 0.6577, "step": 36390 }, { "epoch": 2.2, "learning_rate": 3.505575422872059e-05, "loss": 0.4164, "step": 36392 }, { "epoch": 2.2, "learning_rate": 3.505424531860637e-05, "loss": 0.4702, "step": 36394 }, { "epoch": 2.2, "learning_rate": 3.505273640849215e-05, "loss": 0.5974, "step": 36396 }, { "epoch": 2.2, "learning_rate": 3.5051227498377924e-05, "loss": 0.3808, "step": 36398 }, { "epoch": 2.2, "learning_rate": 3.50497185882637e-05, "loss": 0.6069, "step": 36400 }, { "epoch": 2.2, "learning_rate": 3.504820967814947e-05, "loss": 0.6019, "step": 36402 }, { "epoch": 2.2, "learning_rate": 3.504670076803525e-05, "loss": 0.5131, "step": 36404 }, { "epoch": 2.2, "learning_rate": 3.5045191857921024e-05, "loss": 0.9637, "step": 36406 }, { "epoch": 2.2, "learning_rate": 3.50436829478068e-05, "loss": 0.5983, "step": 36408 }, { "epoch": 2.2, "learning_rate": 3.504217403769258e-05, "loss": 0.5517, "step": 36410 }, { "epoch": 2.2, "learning_rate": 3.504066512757836e-05, "loss": 0.7588, "step": 36412 }, { "epoch": 2.2, "learning_rate": 3.503915621746412e-05, "loss": 0.6545, "step": 36414 }, { "epoch": 2.2, "learning_rate": 3.50376473073499e-05, "loss": 0.601, "step": 36416 }, { "epoch": 2.2, "learning_rate": 3.5036138397235676e-05, "loss": 0.4927, "step": 36418 }, { "epoch": 2.2, "learning_rate": 3.5034629487121456e-05, "loss": 0.5175, "step": 36420 }, { "epoch": 2.2, "learning_rate": 3.503312057700723e-05, "loss": 0.5299, "step": 36422 }, { "epoch": 2.2, "learning_rate": 3.5031611666893e-05, "loss": 0.7083, "step": 36424 }, { "epoch": 2.2, "learning_rate": 3.503010275677878e-05, "loss": 0.5735, "step": 36426 }, { "epoch": 2.2, "learning_rate": 3.5028593846664555e-05, "loss": 0.6698, "step": 36428 }, { "epoch": 2.2, "learning_rate": 3.502708493655033e-05, "loss": 0.5437, "step": 36430 }, { "epoch": 2.2, "learning_rate": 3.502557602643611e-05, "loss": 0.5833, "step": 36432 }, { "epoch": 2.2, "learning_rate": 3.502406711632189e-05, "loss": 0.431, "step": 36434 }, { "epoch": 2.2, "learning_rate": 3.5022558206207654e-05, "loss": 0.5457, "step": 36436 }, { "epoch": 2.2, "learning_rate": 3.5021049296093434e-05, "loss": 0.5477, "step": 36438 }, { "epoch": 2.2, "learning_rate": 3.501954038597921e-05, "loss": 0.495, "step": 36440 }, { "epoch": 2.2, "learning_rate": 3.501803147586498e-05, "loss": 0.6037, "step": 36442 }, { "epoch": 2.2, "learning_rate": 3.501652256575076e-05, "loss": 0.4646, "step": 36444 }, { "epoch": 2.2, "learning_rate": 3.5015013655636534e-05, "loss": 0.6874, "step": 36446 }, { "epoch": 2.2, "learning_rate": 3.5013504745522314e-05, "loss": 0.7481, "step": 36448 }, { "epoch": 2.2, "learning_rate": 3.501199583540809e-05, "loss": 0.5058, "step": 36450 }, { "epoch": 2.2, "learning_rate": 3.501048692529386e-05, "loss": 0.5887, "step": 36452 }, { "epoch": 2.2, "learning_rate": 3.500897801517964e-05, "loss": 0.801, "step": 36454 }, { "epoch": 2.2, "learning_rate": 3.500746910506541e-05, "loss": 0.4373, "step": 36456 }, { "epoch": 2.2, "learning_rate": 3.5005960194951186e-05, "loss": 0.5561, "step": 36458 }, { "epoch": 2.2, "learning_rate": 3.5004451284836966e-05, "loss": 0.4679, "step": 36460 }, { "epoch": 2.2, "learning_rate": 3.500294237472274e-05, "loss": 0.5179, "step": 36462 }, { "epoch": 2.2, "learning_rate": 3.500143346460851e-05, "loss": 0.6528, "step": 36464 }, { "epoch": 2.2, "learning_rate": 3.499992455449429e-05, "loss": 0.6445, "step": 36466 }, { "epoch": 2.2, "learning_rate": 3.4998415644380065e-05, "loss": 0.8434, "step": 36468 }, { "epoch": 2.2, "learning_rate": 3.4996906734265845e-05, "loss": 0.4336, "step": 36470 }, { "epoch": 2.2, "learning_rate": 3.499539782415162e-05, "loss": 0.5132, "step": 36472 }, { "epoch": 2.2, "learning_rate": 3.499388891403739e-05, "loss": 0.6056, "step": 36474 }, { "epoch": 2.2, "learning_rate": 3.499238000392317e-05, "loss": 0.7911, "step": 36476 }, { "epoch": 2.2, "learning_rate": 3.4990871093808944e-05, "loss": 0.6066, "step": 36478 }, { "epoch": 2.2, "learning_rate": 3.498936218369472e-05, "loss": 0.6504, "step": 36480 }, { "epoch": 2.2, "learning_rate": 3.49878532735805e-05, "loss": 0.8642, "step": 36482 }, { "epoch": 2.2, "learning_rate": 3.498634436346627e-05, "loss": 0.8052, "step": 36484 }, { "epoch": 2.2, "learning_rate": 3.4984835453352044e-05, "loss": 0.5595, "step": 36486 }, { "epoch": 2.2, "learning_rate": 3.4983326543237824e-05, "loss": 0.6697, "step": 36488 }, { "epoch": 2.2, "learning_rate": 3.49818176331236e-05, "loss": 0.6887, "step": 36490 }, { "epoch": 2.2, "learning_rate": 3.498030872300937e-05, "loss": 0.8648, "step": 36492 }, { "epoch": 2.2, "learning_rate": 3.497879981289515e-05, "loss": 0.562, "step": 36494 }, { "epoch": 2.2, "learning_rate": 3.497729090278092e-05, "loss": 0.546, "step": 36496 }, { "epoch": 2.2, "learning_rate": 3.49757819926667e-05, "loss": 0.5447, "step": 36498 }, { "epoch": 2.2, "learning_rate": 3.497427308255247e-05, "loss": 0.6251, "step": 36500 }, { "epoch": 2.2, "learning_rate": 3.497276417243825e-05, "loss": 0.4096, "step": 36502 }, { "epoch": 2.2, "learning_rate": 3.497125526232403e-05, "loss": 0.7016, "step": 36504 }, { "epoch": 2.2, "learning_rate": 3.4969746352209795e-05, "loss": 0.7864, "step": 36506 }, { "epoch": 2.2, "learning_rate": 3.4968237442095575e-05, "loss": 0.5361, "step": 36508 }, { "epoch": 2.2, "learning_rate": 3.4966728531981355e-05, "loss": 0.6489, "step": 36510 }, { "epoch": 2.2, "learning_rate": 3.496521962186713e-05, "loss": 0.5842, "step": 36512 }, { "epoch": 2.2, "learning_rate": 3.49637107117529e-05, "loss": 0.5281, "step": 36514 }, { "epoch": 2.2, "learning_rate": 3.4962201801638674e-05, "loss": 0.6546, "step": 36516 }, { "epoch": 2.2, "learning_rate": 3.4960692891524454e-05, "loss": 0.658, "step": 36518 }, { "epoch": 2.2, "learning_rate": 3.4959183981410234e-05, "loss": 0.703, "step": 36520 }, { "epoch": 2.2, "learning_rate": 3.4957675071296e-05, "loss": 0.7176, "step": 36522 }, { "epoch": 2.2, "learning_rate": 3.495616616118178e-05, "loss": 0.4009, "step": 36524 }, { "epoch": 2.2, "learning_rate": 3.495465725106756e-05, "loss": 0.5539, "step": 36526 }, { "epoch": 2.2, "learning_rate": 3.495314834095333e-05, "loss": 0.6444, "step": 36528 }, { "epoch": 2.2, "learning_rate": 3.4951639430839107e-05, "loss": 0.4968, "step": 36530 }, { "epoch": 2.2, "learning_rate": 3.495013052072488e-05, "loss": 0.6541, "step": 36532 }, { "epoch": 2.21, "learning_rate": 3.494862161061066e-05, "loss": 0.3739, "step": 36534 }, { "epoch": 2.21, "learning_rate": 3.494711270049643e-05, "loss": 0.5397, "step": 36536 }, { "epoch": 2.21, "learning_rate": 3.4945603790382206e-05, "loss": 0.4189, "step": 36538 }, { "epoch": 2.21, "learning_rate": 3.4944094880267986e-05, "loss": 0.4739, "step": 36540 }, { "epoch": 2.21, "learning_rate": 3.494258597015376e-05, "loss": 0.5559, "step": 36542 }, { "epoch": 2.21, "learning_rate": 3.494107706003953e-05, "loss": 0.5632, "step": 36544 }, { "epoch": 2.21, "learning_rate": 3.493956814992531e-05, "loss": 0.5393, "step": 36546 }, { "epoch": 2.21, "learning_rate": 3.4938059239811085e-05, "loss": 0.7959, "step": 36548 }, { "epoch": 2.21, "learning_rate": 3.493655032969686e-05, "loss": 0.5268, "step": 36550 }, { "epoch": 2.21, "learning_rate": 3.493504141958264e-05, "loss": 0.5428, "step": 36552 }, { "epoch": 2.21, "learning_rate": 3.493353250946841e-05, "loss": 0.5462, "step": 36554 }, { "epoch": 2.21, "learning_rate": 3.4932023599354184e-05, "loss": 0.4432, "step": 36556 }, { "epoch": 2.21, "learning_rate": 3.4930514689239964e-05, "loss": 0.8977, "step": 36558 }, { "epoch": 2.21, "learning_rate": 3.492900577912574e-05, "loss": 0.6185, "step": 36560 }, { "epoch": 2.21, "learning_rate": 3.492749686901152e-05, "loss": 0.5741, "step": 36562 }, { "epoch": 2.21, "learning_rate": 3.492598795889729e-05, "loss": 0.647, "step": 36564 }, { "epoch": 2.21, "learning_rate": 3.4924479048783063e-05, "loss": 0.4663, "step": 36566 }, { "epoch": 2.21, "learning_rate": 3.4922970138668843e-05, "loss": 0.6712, "step": 36568 }, { "epoch": 2.21, "learning_rate": 3.4921461228554617e-05, "loss": 0.6711, "step": 36570 }, { "epoch": 2.21, "learning_rate": 3.491995231844039e-05, "loss": 0.665, "step": 36572 }, { "epoch": 2.21, "learning_rate": 3.491844340832617e-05, "loss": 0.4556, "step": 36574 }, { "epoch": 2.21, "learning_rate": 3.491693449821194e-05, "loss": 0.5741, "step": 36576 }, { "epoch": 2.21, "learning_rate": 3.4915425588097716e-05, "loss": 0.6869, "step": 36578 }, { "epoch": 2.21, "learning_rate": 3.4913916677983496e-05, "loss": 0.5544, "step": 36580 }, { "epoch": 2.21, "learning_rate": 3.491240776786927e-05, "loss": 0.5156, "step": 36582 }, { "epoch": 2.21, "learning_rate": 3.491089885775505e-05, "loss": 0.5545, "step": 36584 }, { "epoch": 2.21, "learning_rate": 3.490938994764082e-05, "loss": 0.7145, "step": 36586 }, { "epoch": 2.21, "learning_rate": 3.4907881037526595e-05, "loss": 0.6123, "step": 36588 }, { "epoch": 2.21, "learning_rate": 3.4906372127412375e-05, "loss": 0.574, "step": 36590 }, { "epoch": 2.21, "learning_rate": 3.490486321729815e-05, "loss": 0.5206, "step": 36592 }, { "epoch": 2.21, "learning_rate": 3.490335430718392e-05, "loss": 0.5382, "step": 36594 }, { "epoch": 2.21, "learning_rate": 3.49018453970697e-05, "loss": 0.7667, "step": 36596 }, { "epoch": 2.21, "learning_rate": 3.4900336486955474e-05, "loss": 0.5553, "step": 36598 }, { "epoch": 2.21, "learning_rate": 3.489882757684125e-05, "loss": 0.5719, "step": 36600 }, { "epoch": 2.21, "learning_rate": 3.489731866672703e-05, "loss": 0.7181, "step": 36602 }, { "epoch": 2.21, "learning_rate": 3.48958097566128e-05, "loss": 0.6884, "step": 36604 }, { "epoch": 2.21, "learning_rate": 3.4894300846498573e-05, "loss": 0.5711, "step": 36606 }, { "epoch": 2.21, "learning_rate": 3.489279193638435e-05, "loss": 0.8429, "step": 36608 }, { "epoch": 2.21, "learning_rate": 3.4891283026270126e-05, "loss": 0.5102, "step": 36610 }, { "epoch": 2.21, "learning_rate": 3.4889774116155906e-05, "loss": 0.7696, "step": 36612 }, { "epoch": 2.21, "learning_rate": 3.488826520604167e-05, "loss": 1.0898, "step": 36614 }, { "epoch": 2.21, "learning_rate": 3.488675629592745e-05, "loss": 0.8059, "step": 36616 }, { "epoch": 2.21, "learning_rate": 3.488524738581323e-05, "loss": 0.7342, "step": 36618 }, { "epoch": 2.21, "learning_rate": 3.4883738475699e-05, "loss": 0.5477, "step": 36620 }, { "epoch": 2.21, "learning_rate": 3.488222956558478e-05, "loss": 0.6176, "step": 36622 }, { "epoch": 2.21, "learning_rate": 3.488072065547056e-05, "loss": 0.6354, "step": 36624 }, { "epoch": 2.21, "learning_rate": 3.487921174535633e-05, "loss": 0.683, "step": 36626 }, { "epoch": 2.21, "learning_rate": 3.4877702835242105e-05, "loss": 0.4786, "step": 36628 }, { "epoch": 2.21, "learning_rate": 3.487619392512788e-05, "loss": 0.5127, "step": 36630 }, { "epoch": 2.21, "learning_rate": 3.487468501501366e-05, "loss": 0.5765, "step": 36632 }, { "epoch": 2.21, "learning_rate": 3.487317610489944e-05, "loss": 0.5345, "step": 36634 }, { "epoch": 2.21, "learning_rate": 3.4871667194785204e-05, "loss": 0.7165, "step": 36636 }, { "epoch": 2.21, "learning_rate": 3.4870158284670984e-05, "loss": 0.7834, "step": 36638 }, { "epoch": 2.21, "learning_rate": 3.4868649374556764e-05, "loss": 0.4874, "step": 36640 }, { "epoch": 2.21, "learning_rate": 3.486714046444253e-05, "loss": 0.6547, "step": 36642 }, { "epoch": 2.21, "learning_rate": 3.486563155432831e-05, "loss": 0.7112, "step": 36644 }, { "epoch": 2.21, "learning_rate": 3.486412264421408e-05, "loss": 0.8659, "step": 36646 }, { "epoch": 2.21, "learning_rate": 3.486261373409986e-05, "loss": 0.612, "step": 36648 }, { "epoch": 2.21, "learning_rate": 3.4861104823985636e-05, "loss": 0.5446, "step": 36650 }, { "epoch": 2.21, "learning_rate": 3.485959591387141e-05, "loss": 0.7483, "step": 36652 }, { "epoch": 2.21, "learning_rate": 3.485808700375719e-05, "loss": 0.4995, "step": 36654 }, { "epoch": 2.21, "learning_rate": 3.485657809364296e-05, "loss": 0.7145, "step": 36656 }, { "epoch": 2.21, "learning_rate": 3.4855069183528736e-05, "loss": 0.703, "step": 36658 }, { "epoch": 2.21, "learning_rate": 3.4853560273414516e-05, "loss": 0.7869, "step": 36660 }, { "epoch": 2.21, "learning_rate": 3.485205136330029e-05, "loss": 0.614, "step": 36662 }, { "epoch": 2.21, "learning_rate": 3.485054245318606e-05, "loss": 0.8572, "step": 36664 }, { "epoch": 2.21, "learning_rate": 3.484903354307184e-05, "loss": 0.6222, "step": 36666 }, { "epoch": 2.21, "learning_rate": 3.4847524632957615e-05, "loss": 0.3978, "step": 36668 }, { "epoch": 2.21, "learning_rate": 3.484601572284339e-05, "loss": 0.9425, "step": 36670 }, { "epoch": 2.21, "learning_rate": 3.484450681272917e-05, "loss": 0.5901, "step": 36672 }, { "epoch": 2.21, "learning_rate": 3.484299790261494e-05, "loss": 0.8823, "step": 36674 }, { "epoch": 2.21, "learning_rate": 3.484148899250072e-05, "loss": 0.6873, "step": 36676 }, { "epoch": 2.21, "learning_rate": 3.4839980082386494e-05, "loss": 0.6834, "step": 36678 }, { "epoch": 2.21, "learning_rate": 3.483847117227227e-05, "loss": 0.5812, "step": 36680 }, { "epoch": 2.21, "learning_rate": 3.483696226215805e-05, "loss": 0.6137, "step": 36682 }, { "epoch": 2.21, "learning_rate": 3.483545335204382e-05, "loss": 0.5412, "step": 36684 }, { "epoch": 2.21, "learning_rate": 3.483394444192959e-05, "loss": 0.7336, "step": 36686 }, { "epoch": 2.21, "learning_rate": 3.483243553181537e-05, "loss": 0.4485, "step": 36688 }, { "epoch": 2.21, "learning_rate": 3.4830926621701146e-05, "loss": 0.6579, "step": 36690 }, { "epoch": 2.21, "learning_rate": 3.482941771158692e-05, "loss": 0.5082, "step": 36692 }, { "epoch": 2.21, "learning_rate": 3.48279088014727e-05, "loss": 0.6395, "step": 36694 }, { "epoch": 2.21, "learning_rate": 3.482639989135847e-05, "loss": 0.7761, "step": 36696 }, { "epoch": 2.21, "learning_rate": 3.482489098124425e-05, "loss": 0.4458, "step": 36698 }, { "epoch": 2.22, "learning_rate": 3.4823382071130026e-05, "loss": 0.5117, "step": 36700 }, { "epoch": 2.22, "learning_rate": 3.48218731610158e-05, "loss": 0.6047, "step": 36702 }, { "epoch": 2.22, "learning_rate": 3.482036425090158e-05, "loss": 0.5654, "step": 36704 }, { "epoch": 2.22, "learning_rate": 3.481885534078735e-05, "loss": 0.5087, "step": 36706 }, { "epoch": 2.22, "learning_rate": 3.4817346430673125e-05, "loss": 0.6924, "step": 36708 }, { "epoch": 2.22, "learning_rate": 3.4815837520558905e-05, "loss": 0.4922, "step": 36710 }, { "epoch": 2.22, "learning_rate": 3.481432861044468e-05, "loss": 0.703, "step": 36712 }, { "epoch": 2.22, "learning_rate": 3.481281970033045e-05, "loss": 0.3989, "step": 36714 }, { "epoch": 2.22, "learning_rate": 3.481131079021623e-05, "loss": 0.6139, "step": 36716 }, { "epoch": 2.22, "learning_rate": 3.4809801880102004e-05, "loss": 0.5866, "step": 36718 }, { "epoch": 2.22, "learning_rate": 3.480829296998778e-05, "loss": 0.6019, "step": 36720 }, { "epoch": 2.22, "learning_rate": 3.480678405987356e-05, "loss": 0.4915, "step": 36722 }, { "epoch": 2.22, "learning_rate": 3.480527514975933e-05, "loss": 0.6711, "step": 36724 }, { "epoch": 2.22, "learning_rate": 3.480376623964511e-05, "loss": 0.5445, "step": 36726 }, { "epoch": 2.22, "learning_rate": 3.4802257329530876e-05, "loss": 0.6223, "step": 36728 }, { "epoch": 2.22, "learning_rate": 3.4800748419416656e-05, "loss": 0.4499, "step": 36730 }, { "epoch": 2.22, "learning_rate": 3.4799239509302436e-05, "loss": 0.7561, "step": 36732 }, { "epoch": 2.22, "learning_rate": 3.479773059918821e-05, "loss": 0.5201, "step": 36734 }, { "epoch": 2.22, "learning_rate": 3.479622168907398e-05, "loss": 0.7564, "step": 36736 }, { "epoch": 2.22, "learning_rate": 3.479471277895976e-05, "loss": 0.7227, "step": 36738 }, { "epoch": 2.22, "learning_rate": 3.4793203868845535e-05, "loss": 0.5821, "step": 36740 }, { "epoch": 2.22, "learning_rate": 3.479169495873131e-05, "loss": 0.5712, "step": 36742 }, { "epoch": 2.22, "learning_rate": 3.479018604861708e-05, "loss": 0.5662, "step": 36744 }, { "epoch": 2.22, "learning_rate": 3.478867713850286e-05, "loss": 0.7068, "step": 36746 }, { "epoch": 2.22, "learning_rate": 3.478716822838864e-05, "loss": 0.5864, "step": 36748 }, { "epoch": 2.22, "learning_rate": 3.478565931827441e-05, "loss": 0.632, "step": 36750 }, { "epoch": 2.22, "learning_rate": 3.478415040816019e-05, "loss": 0.4842, "step": 36752 }, { "epoch": 2.22, "learning_rate": 3.478264149804597e-05, "loss": 0.56, "step": 36754 }, { "epoch": 2.22, "learning_rate": 3.4781132587931734e-05, "loss": 0.844, "step": 36756 }, { "epoch": 2.22, "learning_rate": 3.4779623677817514e-05, "loss": 0.652, "step": 36758 }, { "epoch": 2.22, "learning_rate": 3.477811476770329e-05, "loss": 0.4807, "step": 36760 }, { "epoch": 2.22, "learning_rate": 3.477660585758907e-05, "loss": 0.4441, "step": 36762 }, { "epoch": 2.22, "learning_rate": 3.477509694747484e-05, "loss": 0.489, "step": 36764 }, { "epoch": 2.22, "learning_rate": 3.477358803736061e-05, "loss": 0.6647, "step": 36766 }, { "epoch": 2.22, "learning_rate": 3.477207912724639e-05, "loss": 0.7124, "step": 36768 }, { "epoch": 2.22, "learning_rate": 3.4770570217132166e-05, "loss": 0.5295, "step": 36770 }, { "epoch": 2.22, "learning_rate": 3.476906130701794e-05, "loss": 0.6554, "step": 36772 }, { "epoch": 2.22, "learning_rate": 3.476755239690372e-05, "loss": 0.7326, "step": 36774 }, { "epoch": 2.22, "learning_rate": 3.476604348678949e-05, "loss": 0.544, "step": 36776 }, { "epoch": 2.22, "learning_rate": 3.4764534576675266e-05, "loss": 0.6342, "step": 36778 }, { "epoch": 2.22, "learning_rate": 3.4763025666561045e-05, "loss": 0.8598, "step": 36780 }, { "epoch": 2.22, "learning_rate": 3.476151675644682e-05, "loss": 0.429, "step": 36782 }, { "epoch": 2.22, "learning_rate": 3.476000784633259e-05, "loss": 0.6239, "step": 36784 }, { "epoch": 2.22, "learning_rate": 3.475849893621837e-05, "loss": 0.5905, "step": 36786 }, { "epoch": 2.22, "learning_rate": 3.4756990026104145e-05, "loss": 0.6431, "step": 36788 }, { "epoch": 2.22, "learning_rate": 3.4755481115989925e-05, "loss": 0.5259, "step": 36790 }, { "epoch": 2.22, "learning_rate": 3.47539722058757e-05, "loss": 0.54, "step": 36792 }, { "epoch": 2.22, "learning_rate": 3.475246329576147e-05, "loss": 0.3066, "step": 36794 }, { "epoch": 2.22, "learning_rate": 3.475095438564725e-05, "loss": 0.6485, "step": 36796 }, { "epoch": 2.22, "learning_rate": 3.4749445475533024e-05, "loss": 0.8203, "step": 36798 }, { "epoch": 2.22, "learning_rate": 3.47479365654188e-05, "loss": 0.4786, "step": 36800 }, { "epoch": 2.22, "learning_rate": 3.474642765530458e-05, "loss": 0.6393, "step": 36802 }, { "epoch": 2.22, "learning_rate": 3.474491874519035e-05, "loss": 0.6342, "step": 36804 }, { "epoch": 2.22, "learning_rate": 3.474340983507612e-05, "loss": 0.5227, "step": 36806 }, { "epoch": 2.22, "learning_rate": 3.47419009249619e-05, "loss": 0.8448, "step": 36808 }, { "epoch": 2.22, "learning_rate": 3.4740392014847676e-05, "loss": 1.017, "step": 36810 }, { "epoch": 2.22, "learning_rate": 3.4738883104733456e-05, "loss": 0.6497, "step": 36812 }, { "epoch": 2.22, "learning_rate": 3.473737419461923e-05, "loss": 0.5407, "step": 36814 }, { "epoch": 2.22, "learning_rate": 3.4735865284505e-05, "loss": 0.5642, "step": 36816 }, { "epoch": 2.22, "learning_rate": 3.473435637439078e-05, "loss": 0.5802, "step": 36818 }, { "epoch": 2.22, "learning_rate": 3.4732847464276555e-05, "loss": 0.5993, "step": 36820 }, { "epoch": 2.22, "learning_rate": 3.473133855416233e-05, "loss": 0.5089, "step": 36822 }, { "epoch": 2.22, "learning_rate": 3.472982964404811e-05, "loss": 0.676, "step": 36824 }, { "epoch": 2.22, "learning_rate": 3.472832073393388e-05, "loss": 0.618, "step": 36826 }, { "epoch": 2.22, "learning_rate": 3.4726811823819655e-05, "loss": 0.4949, "step": 36828 }, { "epoch": 2.22, "learning_rate": 3.4725302913705435e-05, "loss": 0.6963, "step": 36830 }, { "epoch": 2.22, "learning_rate": 3.472379400359121e-05, "loss": 0.3584, "step": 36832 }, { "epoch": 2.22, "learning_rate": 3.472228509347698e-05, "loss": 0.5766, "step": 36834 }, { "epoch": 2.22, "learning_rate": 3.472077618336276e-05, "loss": 0.6643, "step": 36836 }, { "epoch": 2.22, "learning_rate": 3.4719267273248534e-05, "loss": 0.5867, "step": 36838 }, { "epoch": 2.22, "learning_rate": 3.4717758363134314e-05, "loss": 0.6413, "step": 36840 }, { "epoch": 2.22, "learning_rate": 3.471624945302008e-05, "loss": 0.5685, "step": 36842 }, { "epoch": 2.22, "learning_rate": 3.471474054290586e-05, "loss": 0.586, "step": 36844 }, { "epoch": 2.22, "learning_rate": 3.471323163279164e-05, "loss": 0.6436, "step": 36846 }, { "epoch": 2.22, "learning_rate": 3.471172272267741e-05, "loss": 0.476, "step": 36848 }, { "epoch": 2.22, "learning_rate": 3.4710213812563186e-05, "loss": 0.6062, "step": 36850 }, { "epoch": 2.22, "learning_rate": 3.4708704902448966e-05, "loss": 0.4812, "step": 36852 }, { "epoch": 2.22, "learning_rate": 3.470719599233474e-05, "loss": 0.5484, "step": 36854 }, { "epoch": 2.22, "learning_rate": 3.470568708222051e-05, "loss": 0.5432, "step": 36856 }, { "epoch": 2.22, "learning_rate": 3.4704178172106285e-05, "loss": 0.5418, "step": 36858 }, { "epoch": 2.22, "learning_rate": 3.4702669261992065e-05, "loss": 0.7099, "step": 36860 }, { "epoch": 2.22, "learning_rate": 3.4701160351877845e-05, "loss": 0.4626, "step": 36862 }, { "epoch": 2.23, "learning_rate": 3.469965144176361e-05, "loss": 0.6282, "step": 36864 }, { "epoch": 2.23, "learning_rate": 3.469814253164939e-05, "loss": 0.653, "step": 36866 }, { "epoch": 2.23, "learning_rate": 3.469663362153517e-05, "loss": 0.5395, "step": 36868 }, { "epoch": 2.23, "learning_rate": 3.469512471142094e-05, "loss": 0.6203, "step": 36870 }, { "epoch": 2.23, "learning_rate": 3.469361580130672e-05, "loss": 0.4696, "step": 36872 }, { "epoch": 2.23, "learning_rate": 3.469210689119249e-05, "loss": 0.4557, "step": 36874 }, { "epoch": 2.23, "learning_rate": 3.469059798107827e-05, "loss": 0.3658, "step": 36876 }, { "epoch": 2.23, "learning_rate": 3.4689089070964044e-05, "loss": 0.4925, "step": 36878 }, { "epoch": 2.23, "learning_rate": 3.468758016084982e-05, "loss": 0.6115, "step": 36880 }, { "epoch": 2.23, "learning_rate": 3.46860712507356e-05, "loss": 0.4772, "step": 36882 }, { "epoch": 2.23, "learning_rate": 3.468456234062137e-05, "loss": 0.768, "step": 36884 }, { "epoch": 2.23, "learning_rate": 3.468305343050714e-05, "loss": 0.606, "step": 36886 }, { "epoch": 2.23, "learning_rate": 3.468154452039292e-05, "loss": 0.4891, "step": 36888 }, { "epoch": 2.23, "learning_rate": 3.4680035610278696e-05, "loss": 0.5801, "step": 36890 }, { "epoch": 2.23, "learning_rate": 3.467852670016447e-05, "loss": 0.586, "step": 36892 }, { "epoch": 2.23, "learning_rate": 3.467701779005025e-05, "loss": 0.934, "step": 36894 }, { "epoch": 2.23, "learning_rate": 3.467550887993602e-05, "loss": 0.7558, "step": 36896 }, { "epoch": 2.23, "learning_rate": 3.46739999698218e-05, "loss": 0.4912, "step": 36898 }, { "epoch": 2.23, "learning_rate": 3.4672491059707575e-05, "loss": 0.8854, "step": 36900 }, { "epoch": 2.23, "learning_rate": 3.467098214959335e-05, "loss": 0.6015, "step": 36902 }, { "epoch": 2.23, "learning_rate": 3.466947323947913e-05, "loss": 0.535, "step": 36904 }, { "epoch": 2.23, "learning_rate": 3.46679643293649e-05, "loss": 0.5005, "step": 36906 }, { "epoch": 2.23, "learning_rate": 3.4666455419250675e-05, "loss": 0.5605, "step": 36908 }, { "epoch": 2.23, "learning_rate": 3.4664946509136454e-05, "loss": 0.6155, "step": 36910 }, { "epoch": 2.23, "learning_rate": 3.466343759902223e-05, "loss": 0.652, "step": 36912 }, { "epoch": 2.23, "learning_rate": 3.4661928688908e-05, "loss": 0.5473, "step": 36914 }, { "epoch": 2.23, "learning_rate": 3.466041977879378e-05, "loss": 0.3981, "step": 36916 }, { "epoch": 2.23, "learning_rate": 3.4658910868679554e-05, "loss": 0.5087, "step": 36918 }, { "epoch": 2.23, "learning_rate": 3.465740195856533e-05, "loss": 0.4577, "step": 36920 }, { "epoch": 2.23, "learning_rate": 3.465589304845111e-05, "loss": 0.6071, "step": 36922 }, { "epoch": 2.23, "learning_rate": 3.465438413833688e-05, "loss": 0.6951, "step": 36924 }, { "epoch": 2.23, "learning_rate": 3.465287522822266e-05, "loss": 0.6431, "step": 36926 }, { "epoch": 2.23, "learning_rate": 3.465136631810843e-05, "loss": 0.5497, "step": 36928 }, { "epoch": 2.23, "learning_rate": 3.4649857407994206e-05, "loss": 0.7815, "step": 36930 }, { "epoch": 2.23, "learning_rate": 3.4648348497879986e-05, "loss": 0.699, "step": 36932 }, { "epoch": 2.23, "learning_rate": 3.464683958776576e-05, "loss": 0.5366, "step": 36934 }, { "epoch": 2.23, "learning_rate": 3.464533067765153e-05, "loss": 0.6237, "step": 36936 }, { "epoch": 2.23, "learning_rate": 3.464382176753731e-05, "loss": 0.3307, "step": 36938 }, { "epoch": 2.23, "learning_rate": 3.4642312857423085e-05, "loss": 0.7908, "step": 36940 }, { "epoch": 2.23, "learning_rate": 3.464080394730886e-05, "loss": 0.4332, "step": 36942 }, { "epoch": 2.23, "learning_rate": 3.463929503719464e-05, "loss": 0.6352, "step": 36944 }, { "epoch": 2.23, "learning_rate": 3.463778612708041e-05, "loss": 0.7107, "step": 36946 }, { "epoch": 2.23, "learning_rate": 3.4636277216966184e-05, "loss": 0.4046, "step": 36948 }, { "epoch": 2.23, "learning_rate": 3.4634768306851964e-05, "loss": 0.6441, "step": 36950 }, { "epoch": 2.23, "learning_rate": 3.463325939673774e-05, "loss": 0.6726, "step": 36952 }, { "epoch": 2.23, "learning_rate": 3.463175048662352e-05, "loss": 0.8021, "step": 36954 }, { "epoch": 2.23, "learning_rate": 3.4630241576509284e-05, "loss": 0.4364, "step": 36956 }, { "epoch": 2.23, "learning_rate": 3.4628732666395064e-05, "loss": 0.7941, "step": 36958 }, { "epoch": 2.23, "learning_rate": 3.4627223756280844e-05, "loss": 0.5254, "step": 36960 }, { "epoch": 2.23, "learning_rate": 3.462571484616662e-05, "loss": 0.516, "step": 36962 }, { "epoch": 2.23, "learning_rate": 3.462420593605239e-05, "loss": 0.4012, "step": 36964 }, { "epoch": 2.23, "learning_rate": 3.462269702593817e-05, "loss": 0.4814, "step": 36966 }, { "epoch": 2.23, "learning_rate": 3.462118811582394e-05, "loss": 0.6457, "step": 36968 }, { "epoch": 2.23, "learning_rate": 3.4619679205709716e-05, "loss": 0.7646, "step": 36970 }, { "epoch": 2.23, "learning_rate": 3.461817029559549e-05, "loss": 0.5114, "step": 36972 }, { "epoch": 2.23, "learning_rate": 3.461666138548127e-05, "loss": 0.6001, "step": 36974 }, { "epoch": 2.23, "learning_rate": 3.461515247536705e-05, "loss": 0.5272, "step": 36976 }, { "epoch": 2.23, "learning_rate": 3.4613643565252815e-05, "loss": 0.4336, "step": 36978 }, { "epoch": 2.23, "learning_rate": 3.4612134655138595e-05, "loss": 0.522, "step": 36980 }, { "epoch": 2.23, "learning_rate": 3.4610625745024375e-05, "loss": 0.6603, "step": 36982 }, { "epoch": 2.23, "learning_rate": 3.460911683491014e-05, "loss": 0.582, "step": 36984 }, { "epoch": 2.23, "learning_rate": 3.460760792479592e-05, "loss": 0.607, "step": 36986 }, { "epoch": 2.23, "learning_rate": 3.4606099014681694e-05, "loss": 0.6822, "step": 36988 }, { "epoch": 2.23, "learning_rate": 3.4604590104567474e-05, "loss": 0.9151, "step": 36990 }, { "epoch": 2.23, "learning_rate": 3.460308119445325e-05, "loss": 0.6653, "step": 36992 }, { "epoch": 2.23, "learning_rate": 3.460157228433902e-05, "loss": 0.7165, "step": 36994 }, { "epoch": 2.23, "learning_rate": 3.46000633742248e-05, "loss": 0.538, "step": 36996 }, { "epoch": 2.23, "learning_rate": 3.4598554464110574e-05, "loss": 0.5114, "step": 36998 }, { "epoch": 2.23, "learning_rate": 3.459704555399635e-05, "loss": 0.5605, "step": 37000 }, { "epoch": 2.23, "learning_rate": 3.459553664388213e-05, "loss": 0.652, "step": 37002 }, { "epoch": 2.23, "learning_rate": 3.45940277337679e-05, "loss": 0.5145, "step": 37004 }, { "epoch": 2.23, "learning_rate": 3.459251882365367e-05, "loss": 0.59, "step": 37006 }, { "epoch": 2.23, "learning_rate": 3.459100991353945e-05, "loss": 0.6317, "step": 37008 }, { "epoch": 2.23, "learning_rate": 3.4589501003425226e-05, "loss": 0.6147, "step": 37010 }, { "epoch": 2.23, "learning_rate": 3.4587992093311006e-05, "loss": 0.5734, "step": 37012 }, { "epoch": 2.23, "learning_rate": 3.458648318319678e-05, "loss": 0.3829, "step": 37014 }, { "epoch": 2.23, "learning_rate": 3.458497427308255e-05, "loss": 0.5654, "step": 37016 }, { "epoch": 2.23, "learning_rate": 3.458346536296833e-05, "loss": 0.5685, "step": 37018 }, { "epoch": 2.23, "learning_rate": 3.4581956452854105e-05, "loss": 0.5839, "step": 37020 }, { "epoch": 2.23, "learning_rate": 3.458044754273988e-05, "loss": 0.5389, "step": 37022 }, { "epoch": 2.23, "learning_rate": 3.457893863262566e-05, "loss": 0.6904, "step": 37024 }, { "epoch": 2.23, "learning_rate": 3.457742972251143e-05, "loss": 0.6797, "step": 37026 }, { "epoch": 2.23, "learning_rate": 3.4575920812397204e-05, "loss": 0.5225, "step": 37028 }, { "epoch": 2.24, "learning_rate": 3.4574411902282984e-05, "loss": 0.5383, "step": 37030 }, { "epoch": 2.24, "learning_rate": 3.457290299216876e-05, "loss": 0.5632, "step": 37032 }, { "epoch": 2.24, "learning_rate": 3.457139408205453e-05, "loss": 0.6689, "step": 37034 }, { "epoch": 2.24, "learning_rate": 3.456988517194031e-05, "loss": 0.7251, "step": 37036 }, { "epoch": 2.24, "learning_rate": 3.4568376261826084e-05, "loss": 0.6563, "step": 37038 }, { "epoch": 2.24, "learning_rate": 3.4566867351711863e-05, "loss": 0.5634, "step": 37040 }, { "epoch": 2.24, "learning_rate": 3.4565358441597637e-05, "loss": 0.5021, "step": 37042 }, { "epoch": 2.24, "learning_rate": 3.456384953148341e-05, "loss": 0.5032, "step": 37044 }, { "epoch": 2.24, "learning_rate": 3.456234062136919e-05, "loss": 0.5177, "step": 37046 }, { "epoch": 2.24, "learning_rate": 3.456083171125496e-05, "loss": 0.7963, "step": 37048 }, { "epoch": 2.24, "learning_rate": 3.4559322801140736e-05, "loss": 0.7257, "step": 37050 }, { "epoch": 2.24, "learning_rate": 3.4557813891026516e-05, "loss": 0.6311, "step": 37052 }, { "epoch": 2.24, "learning_rate": 3.455630498091229e-05, "loss": 0.4979, "step": 37054 }, { "epoch": 2.24, "learning_rate": 3.455479607079806e-05, "loss": 0.5965, "step": 37056 }, { "epoch": 2.24, "learning_rate": 3.455328716068384e-05, "loss": 0.6493, "step": 37058 }, { "epoch": 2.24, "learning_rate": 3.4551778250569615e-05, "loss": 0.628, "step": 37060 }, { "epoch": 2.24, "learning_rate": 3.4550269340455395e-05, "loss": 0.4885, "step": 37062 }, { "epoch": 2.24, "learning_rate": 3.454876043034117e-05, "loss": 0.5897, "step": 37064 }, { "epoch": 2.24, "learning_rate": 3.454725152022694e-05, "loss": 0.7019, "step": 37066 }, { "epoch": 2.24, "learning_rate": 3.454574261011272e-05, "loss": 0.5829, "step": 37068 }, { "epoch": 2.24, "learning_rate": 3.454423369999849e-05, "loss": 0.8096, "step": 37070 }, { "epoch": 2.24, "learning_rate": 3.454272478988427e-05, "loss": 0.6029, "step": 37072 }, { "epoch": 2.24, "learning_rate": 3.454121587977005e-05, "loss": 0.6828, "step": 37074 }, { "epoch": 2.24, "learning_rate": 3.453970696965582e-05, "loss": 0.7138, "step": 37076 }, { "epoch": 2.24, "learning_rate": 3.4538198059541593e-05, "loss": 0.7066, "step": 37078 }, { "epoch": 2.24, "learning_rate": 3.4536689149427373e-05, "loss": 0.5276, "step": 37080 }, { "epoch": 2.24, "learning_rate": 3.4535180239313147e-05, "loss": 0.6112, "step": 37082 }, { "epoch": 2.24, "learning_rate": 3.453367132919892e-05, "loss": 0.7807, "step": 37084 }, { "epoch": 2.24, "learning_rate": 3.453216241908469e-05, "loss": 0.3788, "step": 37086 }, { "epoch": 2.24, "learning_rate": 3.453065350897047e-05, "loss": 0.4164, "step": 37088 }, { "epoch": 2.24, "learning_rate": 3.452914459885625e-05, "loss": 0.7997, "step": 37090 }, { "epoch": 2.24, "learning_rate": 3.452763568874202e-05, "loss": 0.5968, "step": 37092 }, { "epoch": 2.24, "learning_rate": 3.45261267786278e-05, "loss": 0.5618, "step": 37094 }, { "epoch": 2.24, "learning_rate": 3.452461786851358e-05, "loss": 0.5083, "step": 37096 }, { "epoch": 2.24, "learning_rate": 3.4523108958399345e-05, "loss": 0.981, "step": 37098 }, { "epoch": 2.24, "learning_rate": 3.4521600048285125e-05, "loss": 0.5146, "step": 37100 }, { "epoch": 2.24, "learning_rate": 3.45200911381709e-05, "loss": 0.4203, "step": 37102 }, { "epoch": 2.24, "learning_rate": 3.451858222805668e-05, "loss": 0.752, "step": 37104 }, { "epoch": 2.24, "learning_rate": 3.451707331794245e-05, "loss": 0.6574, "step": 37106 }, { "epoch": 2.24, "learning_rate": 3.4515564407828224e-05, "loss": 0.4929, "step": 37108 }, { "epoch": 2.24, "learning_rate": 3.4514055497714004e-05, "loss": 0.512, "step": 37110 }, { "epoch": 2.24, "learning_rate": 3.4512546587599784e-05, "loss": 0.5684, "step": 37112 }, { "epoch": 2.24, "learning_rate": 3.451103767748555e-05, "loss": 0.6037, "step": 37114 }, { "epoch": 2.24, "learning_rate": 3.450952876737133e-05, "loss": 0.4765, "step": 37116 }, { "epoch": 2.24, "learning_rate": 3.4508019857257103e-05, "loss": 0.8705, "step": 37118 }, { "epoch": 2.24, "learning_rate": 3.4506510947142877e-05, "loss": 0.7058, "step": 37120 }, { "epoch": 2.24, "learning_rate": 3.4505002037028656e-05, "loss": 0.6634, "step": 37122 }, { "epoch": 2.24, "learning_rate": 3.450349312691443e-05, "loss": 0.6807, "step": 37124 }, { "epoch": 2.24, "learning_rate": 3.450198421680021e-05, "loss": 0.9253, "step": 37126 }, { "epoch": 2.24, "learning_rate": 3.450047530668598e-05, "loss": 0.3912, "step": 37128 }, { "epoch": 2.24, "learning_rate": 3.4498966396571756e-05, "loss": 0.4906, "step": 37130 }, { "epoch": 2.24, "learning_rate": 3.4497457486457536e-05, "loss": 0.4754, "step": 37132 }, { "epoch": 2.24, "learning_rate": 3.449594857634331e-05, "loss": 0.7622, "step": 37134 }, { "epoch": 2.24, "learning_rate": 3.449443966622908e-05, "loss": 0.6212, "step": 37136 }, { "epoch": 2.24, "learning_rate": 3.449293075611486e-05, "loss": 0.3858, "step": 37138 }, { "epoch": 2.24, "learning_rate": 3.4491421846000635e-05, "loss": 0.5477, "step": 37140 }, { "epoch": 2.24, "learning_rate": 3.448991293588641e-05, "loss": 0.5442, "step": 37142 }, { "epoch": 2.24, "learning_rate": 3.448840402577219e-05, "loss": 0.6254, "step": 37144 }, { "epoch": 2.24, "learning_rate": 3.448689511565796e-05, "loss": 0.7544, "step": 37146 }, { "epoch": 2.24, "learning_rate": 3.4485386205543734e-05, "loss": 0.4609, "step": 37148 }, { "epoch": 2.24, "learning_rate": 3.4483877295429514e-05, "loss": 0.596, "step": 37150 }, { "epoch": 2.24, "learning_rate": 3.448236838531529e-05, "loss": 0.5456, "step": 37152 }, { "epoch": 2.24, "learning_rate": 3.448085947520107e-05, "loss": 0.6101, "step": 37154 }, { "epoch": 2.24, "learning_rate": 3.447935056508684e-05, "loss": 0.6207, "step": 37156 }, { "epoch": 2.24, "learning_rate": 3.447784165497261e-05, "loss": 0.5451, "step": 37158 }, { "epoch": 2.24, "learning_rate": 3.447633274485839e-05, "loss": 0.7641, "step": 37160 }, { "epoch": 2.24, "learning_rate": 3.4474823834744166e-05, "loss": 0.9401, "step": 37162 }, { "epoch": 2.24, "learning_rate": 3.447331492462994e-05, "loss": 0.5233, "step": 37164 }, { "epoch": 2.24, "learning_rate": 3.447180601451572e-05, "loss": 0.7826, "step": 37166 }, { "epoch": 2.24, "learning_rate": 3.447029710440149e-05, "loss": 0.6384, "step": 37168 }, { "epoch": 2.24, "learning_rate": 3.4468788194287266e-05, "loss": 0.5792, "step": 37170 }, { "epoch": 2.24, "learning_rate": 3.4467279284173046e-05, "loss": 0.5308, "step": 37172 }, { "epoch": 2.24, "learning_rate": 3.446577037405882e-05, "loss": 0.653, "step": 37174 }, { "epoch": 2.24, "learning_rate": 3.44642614639446e-05, "loss": 0.4422, "step": 37176 }, { "epoch": 2.24, "learning_rate": 3.446275255383037e-05, "loss": 0.5302, "step": 37178 }, { "epoch": 2.24, "learning_rate": 3.4461243643716145e-05, "loss": 0.7699, "step": 37180 }, { "epoch": 2.24, "learning_rate": 3.4459734733601925e-05, "loss": 0.4748, "step": 37182 }, { "epoch": 2.24, "learning_rate": 3.445822582348769e-05, "loss": 0.665, "step": 37184 }, { "epoch": 2.24, "learning_rate": 3.445671691337347e-05, "loss": 0.4738, "step": 37186 }, { "epoch": 2.24, "learning_rate": 3.445520800325925e-05, "loss": 0.4427, "step": 37188 }, { "epoch": 2.24, "learning_rate": 3.4453699093145024e-05, "loss": 0.7234, "step": 37190 }, { "epoch": 2.24, "learning_rate": 3.44521901830308e-05, "loss": 0.7171, "step": 37192 }, { "epoch": 2.24, "learning_rate": 3.445068127291658e-05, "loss": 0.6713, "step": 37194 }, { "epoch": 2.25, "learning_rate": 3.444917236280235e-05, "loss": 0.5444, "step": 37196 }, { "epoch": 2.25, "learning_rate": 3.444766345268812e-05, "loss": 0.6205, "step": 37198 }, { "epoch": 2.25, "learning_rate": 3.4446154542573896e-05, "loss": 0.5951, "step": 37200 }, { "epoch": 2.25, "learning_rate": 3.4444645632459676e-05, "loss": 0.4725, "step": 37202 }, { "epoch": 2.25, "learning_rate": 3.4443136722345456e-05, "loss": 0.4624, "step": 37204 }, { "epoch": 2.25, "learning_rate": 3.444162781223122e-05, "loss": 0.4676, "step": 37206 }, { "epoch": 2.25, "learning_rate": 3.4440118902117e-05, "loss": 0.9239, "step": 37208 }, { "epoch": 2.25, "learning_rate": 3.443860999200278e-05, "loss": 0.6982, "step": 37210 }, { "epoch": 2.25, "learning_rate": 3.443710108188855e-05, "loss": 0.5622, "step": 37212 }, { "epoch": 2.25, "learning_rate": 3.443559217177433e-05, "loss": 0.7246, "step": 37214 }, { "epoch": 2.25, "learning_rate": 3.44340832616601e-05, "loss": 0.5514, "step": 37216 }, { "epoch": 2.25, "learning_rate": 3.443257435154588e-05, "loss": 0.8932, "step": 37218 }, { "epoch": 2.25, "learning_rate": 3.4431065441431655e-05, "loss": 0.487, "step": 37220 }, { "epoch": 2.25, "learning_rate": 3.442955653131743e-05, "loss": 0.6239, "step": 37222 }, { "epoch": 2.25, "learning_rate": 3.442804762120321e-05, "loss": 0.67, "step": 37224 }, { "epoch": 2.25, "learning_rate": 3.442653871108899e-05, "loss": 0.6699, "step": 37226 }, { "epoch": 2.25, "learning_rate": 3.4425029800974754e-05, "loss": 0.5664, "step": 37228 }, { "epoch": 2.25, "learning_rate": 3.4423520890860534e-05, "loss": 0.4888, "step": 37230 }, { "epoch": 2.25, "learning_rate": 3.442201198074631e-05, "loss": 0.6768, "step": 37232 }, { "epoch": 2.25, "learning_rate": 3.442050307063208e-05, "loss": 0.6205, "step": 37234 }, { "epoch": 2.25, "learning_rate": 3.441899416051786e-05, "loss": 0.6258, "step": 37236 }, { "epoch": 2.25, "learning_rate": 3.441748525040363e-05, "loss": 0.6155, "step": 37238 }, { "epoch": 2.25, "learning_rate": 3.441597634028941e-05, "loss": 0.6336, "step": 37240 }, { "epoch": 2.25, "learning_rate": 3.4414467430175186e-05, "loss": 0.6245, "step": 37242 }, { "epoch": 2.25, "learning_rate": 3.441295852006096e-05, "loss": 0.4549, "step": 37244 }, { "epoch": 2.25, "learning_rate": 3.441144960994674e-05, "loss": 0.5642, "step": 37246 }, { "epoch": 2.25, "learning_rate": 3.440994069983251e-05, "loss": 0.457, "step": 37248 }, { "epoch": 2.25, "learning_rate": 3.4408431789718286e-05, "loss": 0.629, "step": 37250 }, { "epoch": 2.25, "learning_rate": 3.4406922879604065e-05, "loss": 0.3792, "step": 37252 }, { "epoch": 2.25, "learning_rate": 3.440541396948984e-05, "loss": 0.5646, "step": 37254 }, { "epoch": 2.25, "learning_rate": 3.440390505937561e-05, "loss": 0.5859, "step": 37256 }, { "epoch": 2.25, "learning_rate": 3.440239614926139e-05, "loss": 0.5395, "step": 37258 }, { "epoch": 2.25, "learning_rate": 3.4400887239147165e-05, "loss": 0.6246, "step": 37260 }, { "epoch": 2.25, "learning_rate": 3.439937832903294e-05, "loss": 0.4485, "step": 37262 }, { "epoch": 2.25, "learning_rate": 3.439786941891872e-05, "loss": 0.7574, "step": 37264 }, { "epoch": 2.25, "learning_rate": 3.439636050880449e-05, "loss": 0.4918, "step": 37266 }, { "epoch": 2.25, "learning_rate": 3.439485159869027e-05, "loss": 0.7803, "step": 37268 }, { "epoch": 2.25, "learning_rate": 3.4393342688576044e-05, "loss": 0.5019, "step": 37270 }, { "epoch": 2.25, "learning_rate": 3.439183377846182e-05, "loss": 0.5587, "step": 37272 }, { "epoch": 2.25, "learning_rate": 3.43903248683476e-05, "loss": 0.552, "step": 37274 }, { "epoch": 2.25, "learning_rate": 3.438881595823337e-05, "loss": 0.7106, "step": 37276 }, { "epoch": 2.25, "learning_rate": 3.438730704811914e-05, "loss": 0.685, "step": 37278 }, { "epoch": 2.25, "learning_rate": 3.438579813800492e-05, "loss": 0.6427, "step": 37280 }, { "epoch": 2.25, "learning_rate": 3.4384289227890696e-05, "loss": 0.6501, "step": 37282 }, { "epoch": 2.25, "learning_rate": 3.438278031777647e-05, "loss": 0.8167, "step": 37284 }, { "epoch": 2.25, "learning_rate": 3.438127140766225e-05, "loss": 0.6131, "step": 37286 }, { "epoch": 2.25, "learning_rate": 3.437976249754802e-05, "loss": 0.5451, "step": 37288 }, { "epoch": 2.25, "learning_rate": 3.43782535874338e-05, "loss": 0.6087, "step": 37290 }, { "epoch": 2.25, "learning_rate": 3.4376744677319575e-05, "loss": 0.5888, "step": 37292 }, { "epoch": 2.25, "learning_rate": 3.437523576720535e-05, "loss": 0.5759, "step": 37294 }, { "epoch": 2.25, "learning_rate": 3.437372685709113e-05, "loss": 0.764, "step": 37296 }, { "epoch": 2.25, "learning_rate": 3.4372217946976895e-05, "loss": 0.5708, "step": 37298 }, { "epoch": 2.25, "learning_rate": 3.4370709036862675e-05, "loss": 0.4151, "step": 37300 }, { "epoch": 2.25, "learning_rate": 3.4369200126748455e-05, "loss": 0.6832, "step": 37302 }, { "epoch": 2.25, "learning_rate": 3.436769121663423e-05, "loss": 0.7453, "step": 37304 }, { "epoch": 2.25, "learning_rate": 3.436618230652e-05, "loss": 0.7681, "step": 37306 }, { "epoch": 2.25, "learning_rate": 3.436467339640578e-05, "loss": 0.5325, "step": 37308 }, { "epoch": 2.25, "learning_rate": 3.4363164486291554e-05, "loss": 0.8368, "step": 37310 }, { "epoch": 2.25, "learning_rate": 3.436165557617733e-05, "loss": 0.5279, "step": 37312 }, { "epoch": 2.25, "learning_rate": 3.43601466660631e-05, "loss": 0.5876, "step": 37314 }, { "epoch": 2.25, "learning_rate": 3.435863775594888e-05, "loss": 0.5497, "step": 37316 }, { "epoch": 2.25, "learning_rate": 3.435712884583466e-05, "loss": 0.6481, "step": 37318 }, { "epoch": 2.25, "learning_rate": 3.4355619935720426e-05, "loss": 0.5425, "step": 37320 }, { "epoch": 2.25, "learning_rate": 3.4354111025606206e-05, "loss": 0.5092, "step": 37322 }, { "epoch": 2.25, "learning_rate": 3.4352602115491986e-05, "loss": 0.6847, "step": 37324 }, { "epoch": 2.25, "learning_rate": 3.435109320537775e-05, "loss": 0.7218, "step": 37326 }, { "epoch": 2.25, "learning_rate": 3.434958429526353e-05, "loss": 0.6999, "step": 37328 }, { "epoch": 2.25, "learning_rate": 3.4348075385149305e-05, "loss": 0.5252, "step": 37330 }, { "epoch": 2.25, "learning_rate": 3.4346566475035085e-05, "loss": 0.6345, "step": 37332 }, { "epoch": 2.25, "learning_rate": 3.434505756492086e-05, "loss": 0.6302, "step": 37334 }, { "epoch": 2.25, "learning_rate": 3.434354865480663e-05, "loss": 0.6129, "step": 37336 }, { "epoch": 2.25, "learning_rate": 3.434203974469241e-05, "loss": 0.7764, "step": 37338 }, { "epoch": 2.25, "learning_rate": 3.434053083457819e-05, "loss": 0.6847, "step": 37340 }, { "epoch": 2.25, "learning_rate": 3.433902192446396e-05, "loss": 0.6452, "step": 37342 }, { "epoch": 2.25, "learning_rate": 3.433751301434974e-05, "loss": 0.5381, "step": 37344 }, { "epoch": 2.25, "learning_rate": 3.433600410423551e-05, "loss": 0.6199, "step": 37346 }, { "epoch": 2.25, "learning_rate": 3.4334495194121284e-05, "loss": 0.5173, "step": 37348 }, { "epoch": 2.25, "learning_rate": 3.4332986284007064e-05, "loss": 0.6604, "step": 37350 }, { "epoch": 2.25, "learning_rate": 3.433147737389284e-05, "loss": 0.5364, "step": 37352 }, { "epoch": 2.25, "learning_rate": 3.432996846377862e-05, "loss": 0.5713, "step": 37354 }, { "epoch": 2.25, "learning_rate": 3.432845955366439e-05, "loss": 0.6518, "step": 37356 }, { "epoch": 2.25, "learning_rate": 3.432695064355016e-05, "loss": 0.8419, "step": 37358 }, { "epoch": 2.25, "learning_rate": 3.432544173343594e-05, "loss": 0.7481, "step": 37360 }, { "epoch": 2.26, "learning_rate": 3.4323932823321716e-05, "loss": 0.5911, "step": 37362 }, { "epoch": 2.26, "learning_rate": 3.432242391320749e-05, "loss": 0.7345, "step": 37364 }, { "epoch": 2.26, "learning_rate": 3.432091500309327e-05, "loss": 0.578, "step": 37366 }, { "epoch": 2.26, "learning_rate": 3.431940609297904e-05, "loss": 0.6775, "step": 37368 }, { "epoch": 2.26, "learning_rate": 3.4317897182864815e-05, "loss": 0.5718, "step": 37370 }, { "epoch": 2.26, "learning_rate": 3.4316388272750595e-05, "loss": 0.4174, "step": 37372 }, { "epoch": 2.26, "learning_rate": 3.431487936263637e-05, "loss": 0.5697, "step": 37374 }, { "epoch": 2.26, "learning_rate": 3.431337045252214e-05, "loss": 0.8512, "step": 37376 }, { "epoch": 2.26, "learning_rate": 3.431186154240792e-05, "loss": 0.4366, "step": 37378 }, { "epoch": 2.26, "learning_rate": 3.4310352632293695e-05, "loss": 0.4031, "step": 37380 }, { "epoch": 2.26, "learning_rate": 3.4308843722179474e-05, "loss": 0.4857, "step": 37382 }, { "epoch": 2.26, "learning_rate": 3.430733481206525e-05, "loss": 0.7133, "step": 37384 }, { "epoch": 2.26, "learning_rate": 3.430582590195102e-05, "loss": 0.6279, "step": 37386 }, { "epoch": 2.26, "learning_rate": 3.43043169918368e-05, "loss": 0.5467, "step": 37388 }, { "epoch": 2.26, "learning_rate": 3.4302808081722574e-05, "loss": 0.4657, "step": 37390 }, { "epoch": 2.26, "learning_rate": 3.430129917160835e-05, "loss": 0.628, "step": 37392 }, { "epoch": 2.26, "learning_rate": 3.429979026149413e-05, "loss": 0.581, "step": 37394 }, { "epoch": 2.26, "learning_rate": 3.42982813513799e-05, "loss": 0.732, "step": 37396 }, { "epoch": 2.26, "learning_rate": 3.429677244126567e-05, "loss": 0.3749, "step": 37398 }, { "epoch": 2.26, "learning_rate": 3.429526353115145e-05, "loss": 0.5788, "step": 37400 }, { "epoch": 2.26, "learning_rate": 3.4293754621037226e-05, "loss": 0.5016, "step": 37402 }, { "epoch": 2.26, "learning_rate": 3.4292245710923006e-05, "loss": 0.7189, "step": 37404 }, { "epoch": 2.26, "learning_rate": 3.429073680080878e-05, "loss": 0.574, "step": 37406 }, { "epoch": 2.26, "learning_rate": 3.428922789069455e-05, "loss": 0.4949, "step": 37408 }, { "epoch": 2.26, "learning_rate": 3.428771898058033e-05, "loss": 0.6974, "step": 37410 }, { "epoch": 2.26, "learning_rate": 3.42862100704661e-05, "loss": 0.3231, "step": 37412 }, { "epoch": 2.26, "learning_rate": 3.428470116035188e-05, "loss": 0.6399, "step": 37414 }, { "epoch": 2.26, "learning_rate": 3.428319225023766e-05, "loss": 0.4038, "step": 37416 }, { "epoch": 2.26, "learning_rate": 3.428168334012343e-05, "loss": 0.6782, "step": 37418 }, { "epoch": 2.26, "learning_rate": 3.4280174430009205e-05, "loss": 0.5477, "step": 37420 }, { "epoch": 2.26, "learning_rate": 3.4278665519894984e-05, "loss": 0.5431, "step": 37422 }, { "epoch": 2.26, "learning_rate": 3.427715660978076e-05, "loss": 0.5842, "step": 37424 }, { "epoch": 2.26, "learning_rate": 3.427564769966653e-05, "loss": 0.5124, "step": 37426 }, { "epoch": 2.26, "learning_rate": 3.4274138789552304e-05, "loss": 0.681, "step": 37428 }, { "epoch": 2.26, "learning_rate": 3.4272629879438084e-05, "loss": 0.6325, "step": 37430 }, { "epoch": 2.26, "learning_rate": 3.4271120969323864e-05, "loss": 0.5876, "step": 37432 }, { "epoch": 2.26, "learning_rate": 3.426961205920963e-05, "loss": 0.7858, "step": 37434 }, { "epoch": 2.26, "learning_rate": 3.426810314909541e-05, "loss": 0.6254, "step": 37436 }, { "epoch": 2.26, "learning_rate": 3.426659423898119e-05, "loss": 0.5484, "step": 37438 }, { "epoch": 2.26, "learning_rate": 3.426508532886696e-05, "loss": 0.4682, "step": 37440 }, { "epoch": 2.26, "learning_rate": 3.4263576418752736e-05, "loss": 0.6967, "step": 37442 }, { "epoch": 2.26, "learning_rate": 3.426206750863851e-05, "loss": 0.6117, "step": 37444 }, { "epoch": 2.26, "learning_rate": 3.426055859852429e-05, "loss": 0.7072, "step": 37446 }, { "epoch": 2.26, "learning_rate": 3.425904968841006e-05, "loss": 0.5445, "step": 37448 }, { "epoch": 2.26, "learning_rate": 3.4257540778295835e-05, "loss": 0.4897, "step": 37450 }, { "epoch": 2.26, "learning_rate": 3.4256031868181615e-05, "loss": 0.5733, "step": 37452 }, { "epoch": 2.26, "learning_rate": 3.4254522958067395e-05, "loss": 0.6099, "step": 37454 }, { "epoch": 2.26, "learning_rate": 3.425301404795316e-05, "loss": 0.6477, "step": 37456 }, { "epoch": 2.26, "learning_rate": 3.425150513783894e-05, "loss": 0.6947, "step": 37458 }, { "epoch": 2.26, "learning_rate": 3.4249996227724714e-05, "loss": 0.8382, "step": 37460 }, { "epoch": 2.26, "learning_rate": 3.424848731761049e-05, "loss": 0.738, "step": 37462 }, { "epoch": 2.26, "learning_rate": 3.424697840749627e-05, "loss": 0.4376, "step": 37464 }, { "epoch": 2.26, "learning_rate": 3.424546949738204e-05, "loss": 0.6481, "step": 37466 }, { "epoch": 2.26, "learning_rate": 3.424396058726782e-05, "loss": 0.4656, "step": 37468 }, { "epoch": 2.26, "learning_rate": 3.4242451677153594e-05, "loss": 0.4684, "step": 37470 }, { "epoch": 2.26, "learning_rate": 3.424094276703937e-05, "loss": 0.5386, "step": 37472 }, { "epoch": 2.26, "learning_rate": 3.423943385692515e-05, "loss": 0.3451, "step": 37474 }, { "epoch": 2.26, "learning_rate": 3.423792494681092e-05, "loss": 0.4402, "step": 37476 }, { "epoch": 2.26, "learning_rate": 3.423641603669669e-05, "loss": 0.756, "step": 37478 }, { "epoch": 2.26, "learning_rate": 3.423490712658247e-05, "loss": 0.6006, "step": 37480 }, { "epoch": 2.26, "learning_rate": 3.4233398216468246e-05, "loss": 0.5942, "step": 37482 }, { "epoch": 2.26, "learning_rate": 3.423188930635402e-05, "loss": 0.7312, "step": 37484 }, { "epoch": 2.26, "learning_rate": 3.42303803962398e-05, "loss": 0.6513, "step": 37486 }, { "epoch": 2.26, "learning_rate": 3.422887148612557e-05, "loss": 0.4871, "step": 37488 }, { "epoch": 2.26, "learning_rate": 3.4227362576011345e-05, "loss": 0.4847, "step": 37490 }, { "epoch": 2.26, "learning_rate": 3.4225853665897125e-05, "loss": 0.4567, "step": 37492 }, { "epoch": 2.26, "learning_rate": 3.42243447557829e-05, "loss": 0.6332, "step": 37494 }, { "epoch": 2.26, "learning_rate": 3.422283584566868e-05, "loss": 0.7305, "step": 37496 }, { "epoch": 2.26, "learning_rate": 3.422132693555445e-05, "loss": 0.646, "step": 37498 }, { "epoch": 2.26, "learning_rate": 3.4219818025440224e-05, "loss": 0.7029, "step": 37500 }, { "epoch": 2.26, "learning_rate": 3.4218309115326004e-05, "loss": 0.5649, "step": 37502 }, { "epoch": 2.26, "learning_rate": 3.421680020521178e-05, "loss": 0.5867, "step": 37504 }, { "epoch": 2.26, "learning_rate": 3.421529129509755e-05, "loss": 0.4029, "step": 37506 }, { "epoch": 2.26, "learning_rate": 3.421378238498333e-05, "loss": 0.5782, "step": 37508 }, { "epoch": 2.26, "learning_rate": 3.4212273474869104e-05, "loss": 0.6574, "step": 37510 }, { "epoch": 2.26, "learning_rate": 3.421076456475488e-05, "loss": 0.7973, "step": 37512 }, { "epoch": 2.26, "learning_rate": 3.420925565464066e-05, "loss": 0.5138, "step": 37514 }, { "epoch": 2.26, "learning_rate": 3.420774674452643e-05, "loss": 0.9615, "step": 37516 }, { "epoch": 2.26, "learning_rate": 3.420623783441221e-05, "loss": 0.5811, "step": 37518 }, { "epoch": 2.26, "learning_rate": 3.420472892429798e-05, "loss": 0.5697, "step": 37520 }, { "epoch": 2.26, "learning_rate": 3.4203220014183756e-05, "loss": 0.5489, "step": 37522 }, { "epoch": 2.26, "learning_rate": 3.4201711104069536e-05, "loss": 0.6164, "step": 37524 }, { "epoch": 2.26, "learning_rate": 3.42002021939553e-05, "loss": 0.6215, "step": 37526 }, { "epoch": 2.27, "learning_rate": 3.419869328384108e-05, "loss": 0.5497, "step": 37528 }, { "epoch": 2.27, "learning_rate": 3.419718437372686e-05, "loss": 0.5414, "step": 37530 }, { "epoch": 2.27, "learning_rate": 3.4195675463612635e-05, "loss": 0.4635, "step": 37532 }, { "epoch": 2.27, "learning_rate": 3.419416655349841e-05, "loss": 0.4676, "step": 37534 }, { "epoch": 2.27, "learning_rate": 3.419265764338419e-05, "loss": 0.6703, "step": 37536 }, { "epoch": 2.27, "learning_rate": 3.419114873326996e-05, "loss": 0.8047, "step": 37538 }, { "epoch": 2.27, "learning_rate": 3.4189639823155734e-05, "loss": 0.6958, "step": 37540 }, { "epoch": 2.27, "learning_rate": 3.418813091304151e-05, "loss": 0.6725, "step": 37542 }, { "epoch": 2.27, "learning_rate": 3.418662200292729e-05, "loss": 0.6635, "step": 37544 }, { "epoch": 2.27, "learning_rate": 3.418511309281307e-05, "loss": 0.7029, "step": 37546 }, { "epoch": 2.27, "learning_rate": 3.4183604182698834e-05, "loss": 0.633, "step": 37548 }, { "epoch": 2.27, "learning_rate": 3.4182095272584614e-05, "loss": 0.4287, "step": 37550 }, { "epoch": 2.27, "learning_rate": 3.4180586362470393e-05, "loss": 0.7154, "step": 37552 }, { "epoch": 2.27, "learning_rate": 3.4179077452356167e-05, "loss": 0.4267, "step": 37554 }, { "epoch": 2.27, "learning_rate": 3.417756854224194e-05, "loss": 0.829, "step": 37556 }, { "epoch": 2.27, "learning_rate": 3.417605963212771e-05, "loss": 0.6545, "step": 37558 }, { "epoch": 2.27, "learning_rate": 3.417455072201349e-05, "loss": 0.3611, "step": 37560 }, { "epoch": 2.27, "learning_rate": 3.4173041811899266e-05, "loss": 0.6337, "step": 37562 }, { "epoch": 2.27, "learning_rate": 3.417153290178504e-05, "loss": 0.5964, "step": 37564 }, { "epoch": 2.27, "learning_rate": 3.417002399167082e-05, "loss": 0.5984, "step": 37566 }, { "epoch": 2.27, "learning_rate": 3.41685150815566e-05, "loss": 0.4434, "step": 37568 }, { "epoch": 2.27, "learning_rate": 3.4167006171442365e-05, "loss": 0.589, "step": 37570 }, { "epoch": 2.27, "learning_rate": 3.4165497261328145e-05, "loss": 0.5558, "step": 37572 }, { "epoch": 2.27, "learning_rate": 3.416398835121392e-05, "loss": 0.6024, "step": 37574 }, { "epoch": 2.27, "learning_rate": 3.416247944109969e-05, "loss": 0.6604, "step": 37576 }, { "epoch": 2.27, "learning_rate": 3.416097053098547e-05, "loss": 0.3865, "step": 37578 }, { "epoch": 2.27, "learning_rate": 3.4159461620871244e-05, "loss": 0.6356, "step": 37580 }, { "epoch": 2.27, "learning_rate": 3.4157952710757024e-05, "loss": 0.6515, "step": 37582 }, { "epoch": 2.27, "learning_rate": 3.41564438006428e-05, "loss": 0.4881, "step": 37584 }, { "epoch": 2.27, "learning_rate": 3.415493489052857e-05, "loss": 0.5059, "step": 37586 }, { "epoch": 2.27, "learning_rate": 3.415342598041435e-05, "loss": 0.6816, "step": 37588 }, { "epoch": 2.27, "learning_rate": 3.4151917070300123e-05, "loss": 0.3115, "step": 37590 }, { "epoch": 2.27, "learning_rate": 3.4150408160185897e-05, "loss": 0.5084, "step": 37592 }, { "epoch": 2.27, "learning_rate": 3.4148899250071677e-05, "loss": 0.6572, "step": 37594 }, { "epoch": 2.27, "learning_rate": 3.414739033995745e-05, "loss": 0.6219, "step": 37596 }, { "epoch": 2.27, "learning_rate": 3.414588142984322e-05, "loss": 0.5299, "step": 37598 }, { "epoch": 2.27, "learning_rate": 3.4144372519729e-05, "loss": 0.5335, "step": 37600 }, { "epoch": 2.27, "learning_rate": 3.4142863609614776e-05, "loss": 0.6991, "step": 37602 }, { "epoch": 2.27, "learning_rate": 3.4141354699500556e-05, "loss": 0.6466, "step": 37604 }, { "epoch": 2.27, "learning_rate": 3.413984578938633e-05, "loss": 0.6385, "step": 37606 }, { "epoch": 2.27, "learning_rate": 3.41383368792721e-05, "loss": 0.5464, "step": 37608 }, { "epoch": 2.27, "learning_rate": 3.413682796915788e-05, "loss": 0.5726, "step": 37610 }, { "epoch": 2.27, "learning_rate": 3.4135319059043655e-05, "loss": 0.6914, "step": 37612 }, { "epoch": 2.27, "learning_rate": 3.413381014892943e-05, "loss": 0.7198, "step": 37614 }, { "epoch": 2.27, "learning_rate": 3.413230123881521e-05, "loss": 0.513, "step": 37616 }, { "epoch": 2.27, "learning_rate": 3.413079232870098e-05, "loss": 0.7151, "step": 37618 }, { "epoch": 2.27, "learning_rate": 3.4129283418586754e-05, "loss": 0.5549, "step": 37620 }, { "epoch": 2.27, "learning_rate": 3.4127774508472534e-05, "loss": 0.4418, "step": 37622 }, { "epoch": 2.27, "learning_rate": 3.412626559835831e-05, "loss": 0.514, "step": 37624 }, { "epoch": 2.27, "learning_rate": 3.412475668824408e-05, "loss": 0.5682, "step": 37626 }, { "epoch": 2.27, "learning_rate": 3.412324777812986e-05, "loss": 0.5644, "step": 37628 }, { "epoch": 2.27, "learning_rate": 3.4121738868015633e-05, "loss": 0.9241, "step": 37630 }, { "epoch": 2.27, "learning_rate": 3.412022995790141e-05, "loss": 0.7554, "step": 37632 }, { "epoch": 2.27, "learning_rate": 3.4118721047787186e-05, "loss": 0.4984, "step": 37634 }, { "epoch": 2.27, "learning_rate": 3.411721213767296e-05, "loss": 0.547, "step": 37636 }, { "epoch": 2.27, "learning_rate": 3.411570322755874e-05, "loss": 0.7475, "step": 37638 }, { "epoch": 2.27, "learning_rate": 3.4114194317444506e-05, "loss": 0.4997, "step": 37640 }, { "epoch": 2.27, "learning_rate": 3.4112685407330286e-05, "loss": 0.5971, "step": 37642 }, { "epoch": 2.27, "learning_rate": 3.4111176497216066e-05, "loss": 0.5874, "step": 37644 }, { "epoch": 2.27, "learning_rate": 3.410966758710184e-05, "loss": 0.395, "step": 37646 }, { "epoch": 2.27, "learning_rate": 3.410815867698761e-05, "loss": 0.7304, "step": 37648 }, { "epoch": 2.27, "learning_rate": 3.410664976687339e-05, "loss": 0.7113, "step": 37650 }, { "epoch": 2.27, "learning_rate": 3.4105140856759165e-05, "loss": 0.4676, "step": 37652 }, { "epoch": 2.27, "learning_rate": 3.4103631946644945e-05, "loss": 0.5897, "step": 37654 }, { "epoch": 2.27, "learning_rate": 3.410212303653071e-05, "loss": 0.4996, "step": 37656 }, { "epoch": 2.27, "learning_rate": 3.410061412641649e-05, "loss": 0.3466, "step": 37658 }, { "epoch": 2.27, "learning_rate": 3.409910521630227e-05, "loss": 0.6218, "step": 37660 }, { "epoch": 2.27, "learning_rate": 3.409759630618804e-05, "loss": 0.7431, "step": 37662 }, { "epoch": 2.27, "learning_rate": 3.409608739607382e-05, "loss": 0.6759, "step": 37664 }, { "epoch": 2.27, "learning_rate": 3.40945784859596e-05, "loss": 0.7905, "step": 37666 }, { "epoch": 2.27, "learning_rate": 3.409306957584537e-05, "loss": 0.6093, "step": 37668 }, { "epoch": 2.27, "learning_rate": 3.409156066573114e-05, "loss": 0.5711, "step": 37670 }, { "epoch": 2.27, "learning_rate": 3.4090051755616916e-05, "loss": 0.3962, "step": 37672 }, { "epoch": 2.27, "learning_rate": 3.4088542845502696e-05, "loss": 0.5149, "step": 37674 }, { "epoch": 2.27, "learning_rate": 3.408703393538847e-05, "loss": 0.4848, "step": 37676 }, { "epoch": 2.27, "learning_rate": 3.408552502527424e-05, "loss": 0.5273, "step": 37678 }, { "epoch": 2.27, "learning_rate": 3.408401611516002e-05, "loss": 0.5217, "step": 37680 }, { "epoch": 2.27, "learning_rate": 3.40825072050458e-05, "loss": 0.6044, "step": 37682 }, { "epoch": 2.27, "learning_rate": 3.408099829493157e-05, "loss": 0.6532, "step": 37684 }, { "epoch": 2.27, "learning_rate": 3.407948938481735e-05, "loss": 0.652, "step": 37686 }, { "epoch": 2.27, "learning_rate": 3.407798047470312e-05, "loss": 0.6138, "step": 37688 }, { "epoch": 2.27, "learning_rate": 3.4076471564588895e-05, "loss": 0.4892, "step": 37690 }, { "epoch": 2.27, "learning_rate": 3.4074962654474675e-05, "loss": 0.6835, "step": 37692 }, { "epoch": 2.28, "learning_rate": 3.407345374436045e-05, "loss": 0.7928, "step": 37694 }, { "epoch": 2.28, "learning_rate": 3.407194483424623e-05, "loss": 0.5625, "step": 37696 }, { "epoch": 2.28, "learning_rate": 3.4070435924132e-05, "loss": 0.6181, "step": 37698 }, { "epoch": 2.28, "learning_rate": 3.4068927014017774e-05, "loss": 0.541, "step": 37700 }, { "epoch": 2.28, "learning_rate": 3.4067418103903554e-05, "loss": 0.492, "step": 37702 }, { "epoch": 2.28, "learning_rate": 3.406590919378933e-05, "loss": 0.6923, "step": 37704 }, { "epoch": 2.28, "learning_rate": 3.40644002836751e-05, "loss": 0.712, "step": 37706 }, { "epoch": 2.28, "learning_rate": 3.406289137356088e-05, "loss": 0.7052, "step": 37708 }, { "epoch": 2.28, "learning_rate": 3.406138246344665e-05, "loss": 0.5687, "step": 37710 }, { "epoch": 2.28, "learning_rate": 3.4059873553332426e-05, "loss": 0.9327, "step": 37712 }, { "epoch": 2.28, "learning_rate": 3.4058364643218206e-05, "loss": 0.496, "step": 37714 }, { "epoch": 2.28, "learning_rate": 3.405685573310398e-05, "loss": 0.3702, "step": 37716 }, { "epoch": 2.28, "learning_rate": 3.405534682298976e-05, "loss": 0.6425, "step": 37718 }, { "epoch": 2.28, "learning_rate": 3.405383791287553e-05, "loss": 0.5956, "step": 37720 }, { "epoch": 2.28, "learning_rate": 3.4052329002761306e-05, "loss": 0.8944, "step": 37722 }, { "epoch": 2.28, "learning_rate": 3.4050820092647086e-05, "loss": 0.781, "step": 37724 }, { "epoch": 2.28, "learning_rate": 3.404931118253286e-05, "loss": 0.8032, "step": 37726 }, { "epoch": 2.28, "learning_rate": 3.404780227241863e-05, "loss": 0.6631, "step": 37728 }, { "epoch": 2.28, "learning_rate": 3.404629336230441e-05, "loss": 0.6739, "step": 37730 }, { "epoch": 2.28, "learning_rate": 3.4044784452190185e-05, "loss": 0.5601, "step": 37732 }, { "epoch": 2.28, "learning_rate": 3.404327554207596e-05, "loss": 0.4094, "step": 37734 }, { "epoch": 2.28, "learning_rate": 3.404176663196174e-05, "loss": 0.6966, "step": 37736 }, { "epoch": 2.28, "learning_rate": 3.404025772184751e-05, "loss": 0.6324, "step": 37738 }, { "epoch": 2.28, "learning_rate": 3.4038748811733284e-05, "loss": 0.5797, "step": 37740 }, { "epoch": 2.28, "learning_rate": 3.4037239901619064e-05, "loss": 0.481, "step": 37742 }, { "epoch": 2.28, "learning_rate": 3.403573099150484e-05, "loss": 0.6191, "step": 37744 }, { "epoch": 2.28, "learning_rate": 3.403422208139062e-05, "loss": 0.732, "step": 37746 }, { "epoch": 2.28, "learning_rate": 3.403271317127639e-05, "loss": 0.6209, "step": 37748 }, { "epoch": 2.28, "learning_rate": 3.403120426116216e-05, "loss": 0.8712, "step": 37750 }, { "epoch": 2.28, "learning_rate": 3.402969535104794e-05, "loss": 0.5872, "step": 37752 }, { "epoch": 2.28, "learning_rate": 3.402818644093371e-05, "loss": 0.6565, "step": 37754 }, { "epoch": 2.28, "learning_rate": 3.402667753081949e-05, "loss": 0.5216, "step": 37756 }, { "epoch": 2.28, "learning_rate": 3.402516862070527e-05, "loss": 0.552, "step": 37758 }, { "epoch": 2.28, "learning_rate": 3.402365971059104e-05, "loss": 0.4224, "step": 37760 }, { "epoch": 2.28, "learning_rate": 3.4022150800476816e-05, "loss": 0.5005, "step": 37762 }, { "epoch": 2.28, "learning_rate": 3.4020641890362595e-05, "loss": 0.6453, "step": 37764 }, { "epoch": 2.28, "learning_rate": 3.401913298024837e-05, "loss": 0.5794, "step": 37766 }, { "epoch": 2.28, "learning_rate": 3.401762407013415e-05, "loss": 0.4712, "step": 37768 }, { "epoch": 2.28, "learning_rate": 3.4016115160019915e-05, "loss": 0.5404, "step": 37770 }, { "epoch": 2.28, "learning_rate": 3.4014606249905695e-05, "loss": 0.5065, "step": 37772 }, { "epoch": 2.28, "learning_rate": 3.4013097339791475e-05, "loss": 0.6714, "step": 37774 }, { "epoch": 2.28, "learning_rate": 3.401158842967724e-05, "loss": 0.607, "step": 37776 }, { "epoch": 2.28, "learning_rate": 3.401007951956302e-05, "loss": 0.5835, "step": 37778 }, { "epoch": 2.28, "learning_rate": 3.40085706094488e-05, "loss": 0.4805, "step": 37780 }, { "epoch": 2.28, "learning_rate": 3.4007061699334574e-05, "loss": 0.6867, "step": 37782 }, { "epoch": 2.28, "learning_rate": 3.400555278922035e-05, "loss": 0.589, "step": 37784 }, { "epoch": 2.28, "learning_rate": 3.400404387910612e-05, "loss": 0.4143, "step": 37786 }, { "epoch": 2.28, "learning_rate": 3.40025349689919e-05, "loss": 0.6684, "step": 37788 }, { "epoch": 2.28, "learning_rate": 3.400102605887767e-05, "loss": 0.6589, "step": 37790 }, { "epoch": 2.28, "learning_rate": 3.3999517148763446e-05, "loss": 0.6538, "step": 37792 }, { "epoch": 2.28, "learning_rate": 3.3998008238649226e-05, "loss": 0.6765, "step": 37794 }, { "epoch": 2.28, "learning_rate": 3.3996499328535006e-05, "loss": 0.4686, "step": 37796 }, { "epoch": 2.28, "learning_rate": 3.399499041842077e-05, "loss": 0.6727, "step": 37798 }, { "epoch": 2.28, "learning_rate": 3.399348150830655e-05, "loss": 0.8749, "step": 37800 }, { "epoch": 2.28, "learning_rate": 3.3991972598192325e-05, "loss": 0.578, "step": 37802 }, { "epoch": 2.28, "learning_rate": 3.39904636880781e-05, "loss": 0.847, "step": 37804 }, { "epoch": 2.28, "learning_rate": 3.398895477796388e-05, "loss": 0.5918, "step": 37806 }, { "epoch": 2.28, "learning_rate": 3.398744586784965e-05, "loss": 0.6047, "step": 37808 }, { "epoch": 2.28, "learning_rate": 3.398593695773543e-05, "loss": 0.6363, "step": 37810 }, { "epoch": 2.28, "learning_rate": 3.3984428047621205e-05, "loss": 0.5775, "step": 37812 }, { "epoch": 2.28, "learning_rate": 3.398291913750698e-05, "loss": 0.6564, "step": 37814 }, { "epoch": 2.28, "learning_rate": 3.398141022739276e-05, "loss": 0.6103, "step": 37816 }, { "epoch": 2.28, "learning_rate": 3.397990131727853e-05, "loss": 0.4881, "step": 37818 }, { "epoch": 2.28, "learning_rate": 3.3978392407164304e-05, "loss": 0.6432, "step": 37820 }, { "epoch": 2.28, "learning_rate": 3.3976883497050084e-05, "loss": 0.3341, "step": 37822 }, { "epoch": 2.28, "learning_rate": 3.397537458693586e-05, "loss": 0.5451, "step": 37824 }, { "epoch": 2.28, "learning_rate": 3.397386567682163e-05, "loss": 0.5985, "step": 37826 }, { "epoch": 2.28, "learning_rate": 3.397235676670741e-05, "loss": 0.6763, "step": 37828 }, { "epoch": 2.28, "learning_rate": 3.397084785659318e-05, "loss": 0.946, "step": 37830 }, { "epoch": 2.28, "learning_rate": 3.396933894647896e-05, "loss": 0.8505, "step": 37832 }, { "epoch": 2.28, "learning_rate": 3.3967830036364736e-05, "loss": 0.6072, "step": 37834 }, { "epoch": 2.28, "learning_rate": 3.396632112625051e-05, "loss": 0.6227, "step": 37836 }, { "epoch": 2.28, "learning_rate": 3.396481221613629e-05, "loss": 0.4954, "step": 37838 }, { "epoch": 2.28, "learning_rate": 3.396330330602206e-05, "loss": 0.6744, "step": 37840 }, { "epoch": 2.28, "learning_rate": 3.3961794395907835e-05, "loss": 0.73, "step": 37842 }, { "epoch": 2.28, "learning_rate": 3.3960285485793615e-05, "loss": 0.7127, "step": 37844 }, { "epoch": 2.28, "learning_rate": 3.395877657567939e-05, "loss": 0.4614, "step": 37846 }, { "epoch": 2.28, "learning_rate": 3.395726766556516e-05, "loss": 0.5806, "step": 37848 }, { "epoch": 2.28, "learning_rate": 3.395575875545094e-05, "loss": 0.5528, "step": 37850 }, { "epoch": 2.28, "learning_rate": 3.3954249845336715e-05, "loss": 0.7236, "step": 37852 }, { "epoch": 2.28, "learning_rate": 3.395274093522249e-05, "loss": 0.6179, "step": 37854 }, { "epoch": 2.28, "learning_rate": 3.395123202510827e-05, "loss": 0.7573, "step": 37856 }, { "epoch": 2.29, "learning_rate": 3.394972311499404e-05, "loss": 0.6898, "step": 37858 }, { "epoch": 2.29, "learning_rate": 3.394821420487982e-05, "loss": 0.6939, "step": 37860 }, { "epoch": 2.29, "learning_rate": 3.3946705294765594e-05, "loss": 0.5625, "step": 37862 }, { "epoch": 2.29, "learning_rate": 3.394519638465137e-05, "loss": 0.663, "step": 37864 }, { "epoch": 2.29, "learning_rate": 3.394368747453715e-05, "loss": 0.801, "step": 37866 }, { "epoch": 2.29, "learning_rate": 3.394217856442291e-05, "loss": 0.5462, "step": 37868 }, { "epoch": 2.29, "learning_rate": 3.394066965430869e-05, "loss": 0.678, "step": 37870 }, { "epoch": 2.29, "learning_rate": 3.393916074419447e-05, "loss": 0.3899, "step": 37872 }, { "epoch": 2.29, "learning_rate": 3.3937651834080246e-05, "loss": 0.746, "step": 37874 }, { "epoch": 2.29, "learning_rate": 3.393614292396602e-05, "loss": 0.5908, "step": 37876 }, { "epoch": 2.29, "learning_rate": 3.39346340138518e-05, "loss": 0.6663, "step": 37878 }, { "epoch": 2.29, "learning_rate": 3.393312510373757e-05, "loss": 0.5956, "step": 37880 }, { "epoch": 2.29, "learning_rate": 3.393161619362335e-05, "loss": 0.542, "step": 37882 }, { "epoch": 2.29, "learning_rate": 3.393010728350912e-05, "loss": 0.5177, "step": 37884 }, { "epoch": 2.29, "learning_rate": 3.39285983733949e-05, "loss": 0.6772, "step": 37886 }, { "epoch": 2.29, "learning_rate": 3.392708946328068e-05, "loss": 0.5238, "step": 37888 }, { "epoch": 2.29, "learning_rate": 3.3925580553166445e-05, "loss": 0.8393, "step": 37890 }, { "epoch": 2.29, "learning_rate": 3.3924071643052225e-05, "loss": 0.4843, "step": 37892 }, { "epoch": 2.29, "learning_rate": 3.3922562732938004e-05, "loss": 0.6083, "step": 37894 }, { "epoch": 2.29, "learning_rate": 3.392105382282378e-05, "loss": 0.45, "step": 37896 }, { "epoch": 2.29, "learning_rate": 3.391954491270955e-05, "loss": 0.576, "step": 37898 }, { "epoch": 2.29, "learning_rate": 3.3918036002595324e-05, "loss": 0.5857, "step": 37900 }, { "epoch": 2.29, "learning_rate": 3.3916527092481104e-05, "loss": 0.6068, "step": 37902 }, { "epoch": 2.29, "learning_rate": 3.391501818236688e-05, "loss": 0.6902, "step": 37904 }, { "epoch": 2.29, "learning_rate": 3.391350927225265e-05, "loss": 0.6484, "step": 37906 }, { "epoch": 2.29, "learning_rate": 3.391200036213843e-05, "loss": 0.47, "step": 37908 }, { "epoch": 2.29, "learning_rate": 3.391049145202421e-05, "loss": 0.4859, "step": 37910 }, { "epoch": 2.29, "learning_rate": 3.3908982541909976e-05, "loss": 0.5226, "step": 37912 }, { "epoch": 2.29, "learning_rate": 3.3907473631795756e-05, "loss": 0.5324, "step": 37914 }, { "epoch": 2.29, "learning_rate": 3.390596472168153e-05, "loss": 0.3763, "step": 37916 }, { "epoch": 2.29, "learning_rate": 3.39044558115673e-05, "loss": 0.5772, "step": 37918 }, { "epoch": 2.29, "learning_rate": 3.390294690145308e-05, "loss": 0.6219, "step": 37920 }, { "epoch": 2.29, "learning_rate": 3.3901437991338855e-05, "loss": 0.6208, "step": 37922 }, { "epoch": 2.29, "learning_rate": 3.3899929081224635e-05, "loss": 0.6408, "step": 37924 }, { "epoch": 2.29, "learning_rate": 3.389842017111041e-05, "loss": 0.6377, "step": 37926 }, { "epoch": 2.29, "learning_rate": 3.389691126099618e-05, "loss": 0.491, "step": 37928 }, { "epoch": 2.29, "learning_rate": 3.389540235088196e-05, "loss": 0.7132, "step": 37930 }, { "epoch": 2.29, "learning_rate": 3.3893893440767735e-05, "loss": 0.6011, "step": 37932 }, { "epoch": 2.29, "learning_rate": 3.389238453065351e-05, "loss": 0.448, "step": 37934 }, { "epoch": 2.29, "learning_rate": 3.389087562053929e-05, "loss": 0.6568, "step": 37936 }, { "epoch": 2.29, "learning_rate": 3.388936671042506e-05, "loss": 0.6879, "step": 37938 }, { "epoch": 2.29, "learning_rate": 3.3887857800310834e-05, "loss": 0.6166, "step": 37940 }, { "epoch": 2.29, "learning_rate": 3.3886348890196614e-05, "loss": 0.5986, "step": 37942 }, { "epoch": 2.29, "learning_rate": 3.388483998008239e-05, "loss": 0.579, "step": 37944 }, { "epoch": 2.29, "learning_rate": 3.388333106996817e-05, "loss": 0.693, "step": 37946 }, { "epoch": 2.29, "learning_rate": 3.388182215985394e-05, "loss": 0.4896, "step": 37948 }, { "epoch": 2.29, "learning_rate": 3.388031324973971e-05, "loss": 0.5779, "step": 37950 }, { "epoch": 2.29, "learning_rate": 3.387880433962549e-05, "loss": 0.4701, "step": 37952 }, { "epoch": 2.29, "learning_rate": 3.3877295429511266e-05, "loss": 0.5649, "step": 37954 }, { "epoch": 2.29, "learning_rate": 3.387578651939704e-05, "loss": 0.6113, "step": 37956 }, { "epoch": 2.29, "learning_rate": 3.387427760928282e-05, "loss": 0.5276, "step": 37958 }, { "epoch": 2.29, "learning_rate": 3.387276869916859e-05, "loss": 0.6598, "step": 37960 }, { "epoch": 2.29, "learning_rate": 3.3871259789054365e-05, "loss": 0.4964, "step": 37962 }, { "epoch": 2.29, "learning_rate": 3.3869750878940145e-05, "loss": 0.6021, "step": 37964 }, { "epoch": 2.29, "learning_rate": 3.386824196882592e-05, "loss": 0.5251, "step": 37966 }, { "epoch": 2.29, "learning_rate": 3.386673305871169e-05, "loss": 0.5822, "step": 37968 }, { "epoch": 2.29, "learning_rate": 3.386522414859747e-05, "loss": 0.67, "step": 37970 }, { "epoch": 2.29, "learning_rate": 3.3863715238483244e-05, "loss": 0.7324, "step": 37972 }, { "epoch": 2.29, "learning_rate": 3.3862206328369024e-05, "loss": 0.5645, "step": 37974 }, { "epoch": 2.29, "learning_rate": 3.38606974182548e-05, "loss": 0.4616, "step": 37976 }, { "epoch": 2.29, "learning_rate": 3.385918850814057e-05, "loss": 0.5939, "step": 37978 }, { "epoch": 2.29, "learning_rate": 3.385767959802635e-05, "loss": 0.4838, "step": 37980 }, { "epoch": 2.29, "learning_rate": 3.3856170687912124e-05, "loss": 0.6653, "step": 37982 }, { "epoch": 2.29, "learning_rate": 3.38546617777979e-05, "loss": 0.4943, "step": 37984 }, { "epoch": 2.29, "learning_rate": 3.385315286768368e-05, "loss": 1.0268, "step": 37986 }, { "epoch": 2.29, "learning_rate": 3.385164395756945e-05, "loss": 0.4627, "step": 37988 }, { "epoch": 2.29, "learning_rate": 3.385013504745522e-05, "loss": 0.7393, "step": 37990 }, { "epoch": 2.29, "learning_rate": 3.3848626137341e-05, "loss": 0.7752, "step": 37992 }, { "epoch": 2.29, "learning_rate": 3.3847117227226776e-05, "loss": 0.598, "step": 37994 }, { "epoch": 2.29, "learning_rate": 3.3845608317112556e-05, "loss": 0.8139, "step": 37996 }, { "epoch": 2.29, "learning_rate": 3.384409940699832e-05, "loss": 0.6636, "step": 37998 }, { "epoch": 2.29, "learning_rate": 3.38425904968841e-05, "loss": 0.688, "step": 38000 }, { "epoch": 2.29, "learning_rate": 3.384108158676988e-05, "loss": 0.4084, "step": 38002 }, { "epoch": 2.29, "learning_rate": 3.383957267665565e-05, "loss": 0.631, "step": 38004 }, { "epoch": 2.29, "learning_rate": 3.383806376654143e-05, "loss": 0.6641, "step": 38006 }, { "epoch": 2.29, "learning_rate": 3.383655485642721e-05, "loss": 0.5817, "step": 38008 }, { "epoch": 2.29, "learning_rate": 3.383504594631298e-05, "loss": 0.608, "step": 38010 }, { "epoch": 2.29, "learning_rate": 3.3833537036198754e-05, "loss": 0.6867, "step": 38012 }, { "epoch": 2.29, "learning_rate": 3.383202812608453e-05, "loss": 0.6252, "step": 38014 }, { "epoch": 2.29, "learning_rate": 3.383051921597031e-05, "loss": 0.5259, "step": 38016 }, { "epoch": 2.29, "learning_rate": 3.382901030585608e-05, "loss": 0.7157, "step": 38018 }, { "epoch": 2.29, "learning_rate": 3.3827501395741854e-05, "loss": 0.9459, "step": 38020 }, { "epoch": 2.29, "learning_rate": 3.3825992485627634e-05, "loss": 0.5201, "step": 38022 }, { "epoch": 2.3, "learning_rate": 3.3824483575513413e-05, "loss": 0.3809, "step": 38024 }, { "epoch": 2.3, "learning_rate": 3.382297466539918e-05, "loss": 0.4822, "step": 38026 }, { "epoch": 2.3, "learning_rate": 3.382146575528496e-05, "loss": 0.5938, "step": 38028 }, { "epoch": 2.3, "learning_rate": 3.381995684517073e-05, "loss": 0.7382, "step": 38030 }, { "epoch": 2.3, "learning_rate": 3.3818447935056506e-05, "loss": 0.8541, "step": 38032 }, { "epoch": 2.3, "learning_rate": 3.3816939024942286e-05, "loss": 0.6607, "step": 38034 }, { "epoch": 2.3, "learning_rate": 3.381543011482806e-05, "loss": 0.5991, "step": 38036 }, { "epoch": 2.3, "learning_rate": 3.381392120471384e-05, "loss": 0.8512, "step": 38038 }, { "epoch": 2.3, "learning_rate": 3.381241229459961e-05, "loss": 0.5221, "step": 38040 }, { "epoch": 2.3, "learning_rate": 3.3810903384485385e-05, "loss": 0.9181, "step": 38042 }, { "epoch": 2.3, "learning_rate": 3.3809394474371165e-05, "loss": 0.6018, "step": 38044 }, { "epoch": 2.3, "learning_rate": 3.380788556425694e-05, "loss": 0.707, "step": 38046 }, { "epoch": 2.3, "learning_rate": 3.380637665414271e-05, "loss": 0.7135, "step": 38048 }, { "epoch": 2.3, "learning_rate": 3.380486774402849e-05, "loss": 0.5005, "step": 38050 }, { "epoch": 2.3, "learning_rate": 3.3803358833914264e-05, "loss": 0.4957, "step": 38052 }, { "epoch": 2.3, "learning_rate": 3.380184992380004e-05, "loss": 0.5641, "step": 38054 }, { "epoch": 2.3, "learning_rate": 3.380034101368582e-05, "loss": 0.5466, "step": 38056 }, { "epoch": 2.3, "learning_rate": 3.379883210357159e-05, "loss": 0.594, "step": 38058 }, { "epoch": 2.3, "learning_rate": 3.379732319345737e-05, "loss": 0.5876, "step": 38060 }, { "epoch": 2.3, "learning_rate": 3.3795814283343144e-05, "loss": 0.4949, "step": 38062 }, { "epoch": 2.3, "learning_rate": 3.379430537322892e-05, "loss": 0.6855, "step": 38064 }, { "epoch": 2.3, "learning_rate": 3.3792796463114697e-05, "loss": 0.5035, "step": 38066 }, { "epoch": 2.3, "learning_rate": 3.379128755300047e-05, "loss": 0.6275, "step": 38068 }, { "epoch": 2.3, "learning_rate": 3.378977864288624e-05, "loss": 0.3379, "step": 38070 }, { "epoch": 2.3, "learning_rate": 3.378826973277202e-05, "loss": 0.4721, "step": 38072 }, { "epoch": 2.3, "learning_rate": 3.3786760822657796e-05, "loss": 0.4531, "step": 38074 }, { "epoch": 2.3, "learning_rate": 3.378525191254357e-05, "loss": 0.5261, "step": 38076 }, { "epoch": 2.3, "learning_rate": 3.378374300242935e-05, "loss": 0.547, "step": 38078 }, { "epoch": 2.3, "learning_rate": 3.378223409231512e-05, "loss": 0.6993, "step": 38080 }, { "epoch": 2.3, "learning_rate": 3.3780725182200895e-05, "loss": 0.548, "step": 38082 }, { "epoch": 2.3, "learning_rate": 3.3779216272086675e-05, "loss": 0.6871, "step": 38084 }, { "epoch": 2.3, "learning_rate": 3.377770736197245e-05, "loss": 0.6315, "step": 38086 }, { "epoch": 2.3, "learning_rate": 3.377619845185823e-05, "loss": 0.4496, "step": 38088 }, { "epoch": 2.3, "learning_rate": 3.3774689541744e-05, "loss": 0.696, "step": 38090 }, { "epoch": 2.3, "learning_rate": 3.3773180631629774e-05, "loss": 0.7157, "step": 38092 }, { "epoch": 2.3, "learning_rate": 3.3771671721515554e-05, "loss": 0.3853, "step": 38094 }, { "epoch": 2.3, "learning_rate": 3.377016281140133e-05, "loss": 0.7126, "step": 38096 }, { "epoch": 2.3, "learning_rate": 3.37686539012871e-05, "loss": 0.7489, "step": 38098 }, { "epoch": 2.3, "learning_rate": 3.376714499117288e-05, "loss": 0.4319, "step": 38100 }, { "epoch": 2.3, "learning_rate": 3.3765636081058653e-05, "loss": 0.5222, "step": 38102 }, { "epoch": 2.3, "learning_rate": 3.3764127170944427e-05, "loss": 0.5814, "step": 38104 }, { "epoch": 2.3, "learning_rate": 3.3762618260830206e-05, "loss": 0.578, "step": 38106 }, { "epoch": 2.3, "learning_rate": 3.376110935071598e-05, "loss": 0.4944, "step": 38108 }, { "epoch": 2.3, "learning_rate": 3.375960044060176e-05, "loss": 0.5122, "step": 38110 }, { "epoch": 2.3, "learning_rate": 3.3758091530487526e-05, "loss": 0.7497, "step": 38112 }, { "epoch": 2.3, "learning_rate": 3.3756582620373306e-05, "loss": 0.6559, "step": 38114 }, { "epoch": 2.3, "learning_rate": 3.3755073710259086e-05, "loss": 0.8242, "step": 38116 }, { "epoch": 2.3, "learning_rate": 3.375356480014485e-05, "loss": 0.6673, "step": 38118 }, { "epoch": 2.3, "learning_rate": 3.375205589003063e-05, "loss": 0.6223, "step": 38120 }, { "epoch": 2.3, "learning_rate": 3.375054697991641e-05, "loss": 0.5435, "step": 38122 }, { "epoch": 2.3, "learning_rate": 3.3749038069802185e-05, "loss": 0.784, "step": 38124 }, { "epoch": 2.3, "learning_rate": 3.374752915968796e-05, "loss": 0.5169, "step": 38126 }, { "epoch": 2.3, "learning_rate": 3.374602024957373e-05, "loss": 0.6001, "step": 38128 }, { "epoch": 2.3, "learning_rate": 3.374451133945951e-05, "loss": 0.6063, "step": 38130 }, { "epoch": 2.3, "learning_rate": 3.3743002429345284e-05, "loss": 0.4996, "step": 38132 }, { "epoch": 2.3, "learning_rate": 3.374149351923106e-05, "loss": 0.5921, "step": 38134 }, { "epoch": 2.3, "learning_rate": 3.373998460911684e-05, "loss": 0.4394, "step": 38136 }, { "epoch": 2.3, "learning_rate": 3.373847569900262e-05, "loss": 0.4803, "step": 38138 }, { "epoch": 2.3, "learning_rate": 3.3736966788888384e-05, "loss": 0.6495, "step": 38140 }, { "epoch": 2.3, "learning_rate": 3.3735457878774163e-05, "loss": 0.5244, "step": 38142 }, { "epoch": 2.3, "learning_rate": 3.3733948968659937e-05, "loss": 0.8354, "step": 38144 }, { "epoch": 2.3, "learning_rate": 3.3732440058545716e-05, "loss": 0.4675, "step": 38146 }, { "epoch": 2.3, "learning_rate": 3.373093114843149e-05, "loss": 0.5996, "step": 38148 }, { "epoch": 2.3, "learning_rate": 3.372942223831726e-05, "loss": 0.5316, "step": 38150 }, { "epoch": 2.3, "learning_rate": 3.372791332820304e-05, "loss": 0.6174, "step": 38152 }, { "epoch": 2.3, "learning_rate": 3.3726404418088816e-05, "loss": 0.6746, "step": 38154 }, { "epoch": 2.3, "learning_rate": 3.372489550797459e-05, "loss": 0.6391, "step": 38156 }, { "epoch": 2.3, "learning_rate": 3.372338659786037e-05, "loss": 0.4746, "step": 38158 }, { "epoch": 2.3, "learning_rate": 3.372187768774614e-05, "loss": 0.4268, "step": 38160 }, { "epoch": 2.3, "learning_rate": 3.3720368777631915e-05, "loss": 0.528, "step": 38162 }, { "epoch": 2.3, "learning_rate": 3.3718859867517695e-05, "loss": 0.6127, "step": 38164 }, { "epoch": 2.3, "learning_rate": 3.371735095740347e-05, "loss": 0.7045, "step": 38166 }, { "epoch": 2.3, "learning_rate": 3.371584204728924e-05, "loss": 0.6741, "step": 38168 }, { "epoch": 2.3, "learning_rate": 3.371433313717502e-05, "loss": 0.6612, "step": 38170 }, { "epoch": 2.3, "learning_rate": 3.3712824227060794e-05, "loss": 0.8153, "step": 38172 }, { "epoch": 2.3, "learning_rate": 3.3711315316946574e-05, "loss": 0.4605, "step": 38174 }, { "epoch": 2.3, "learning_rate": 3.370980640683235e-05, "loss": 0.6001, "step": 38176 }, { "epoch": 2.3, "learning_rate": 3.370829749671812e-05, "loss": 0.4947, "step": 38178 }, { "epoch": 2.3, "learning_rate": 3.37067885866039e-05, "loss": 0.7094, "step": 38180 }, { "epoch": 2.3, "learning_rate": 3.370527967648967e-05, "loss": 0.5266, "step": 38182 }, { "epoch": 2.3, "learning_rate": 3.3703770766375446e-05, "loss": 0.8009, "step": 38184 }, { "epoch": 2.3, "learning_rate": 3.3702261856261226e-05, "loss": 0.6464, "step": 38186 }, { "epoch": 2.3, "learning_rate": 3.3700752946147e-05, "loss": 0.5856, "step": 38188 }, { "epoch": 2.31, "learning_rate": 3.369924403603277e-05, "loss": 0.5182, "step": 38190 }, { "epoch": 2.31, "learning_rate": 3.369773512591855e-05, "loss": 0.5644, "step": 38192 }, { "epoch": 2.31, "learning_rate": 3.3696226215804326e-05, "loss": 0.4719, "step": 38194 }, { "epoch": 2.31, "learning_rate": 3.3694717305690106e-05, "loss": 0.9141, "step": 38196 }, { "epoch": 2.31, "learning_rate": 3.369320839557588e-05, "loss": 0.4499, "step": 38198 }, { "epoch": 2.31, "learning_rate": 3.369169948546165e-05, "loss": 0.5733, "step": 38200 }, { "epoch": 2.31, "learning_rate": 3.369019057534743e-05, "loss": 0.2904, "step": 38202 }, { "epoch": 2.31, "learning_rate": 3.3688681665233205e-05, "loss": 0.5694, "step": 38204 }, { "epoch": 2.31, "learning_rate": 3.368717275511898e-05, "loss": 0.3928, "step": 38206 }, { "epoch": 2.31, "learning_rate": 3.368566384500476e-05, "loss": 0.478, "step": 38208 }, { "epoch": 2.31, "learning_rate": 3.368415493489053e-05, "loss": 0.5455, "step": 38210 }, { "epoch": 2.31, "learning_rate": 3.3682646024776304e-05, "loss": 0.6104, "step": 38212 }, { "epoch": 2.31, "learning_rate": 3.3681137114662084e-05, "loss": 0.6213, "step": 38214 }, { "epoch": 2.31, "learning_rate": 3.367962820454786e-05, "loss": 0.5392, "step": 38216 }, { "epoch": 2.31, "learning_rate": 3.367811929443363e-05, "loss": 0.5031, "step": 38218 }, { "epoch": 2.31, "learning_rate": 3.367661038431941e-05, "loss": 0.3901, "step": 38220 }, { "epoch": 2.31, "learning_rate": 3.367510147420518e-05, "loss": 0.4626, "step": 38222 }, { "epoch": 2.31, "learning_rate": 3.367359256409096e-05, "loss": 0.5394, "step": 38224 }, { "epoch": 2.31, "learning_rate": 3.367208365397673e-05, "loss": 0.6222, "step": 38226 }, { "epoch": 2.31, "learning_rate": 3.367057474386251e-05, "loss": 0.3939, "step": 38228 }, { "epoch": 2.31, "learning_rate": 3.366906583374829e-05, "loss": 0.6393, "step": 38230 }, { "epoch": 2.31, "learning_rate": 3.3667556923634056e-05, "loss": 0.4679, "step": 38232 }, { "epoch": 2.31, "learning_rate": 3.3666048013519836e-05, "loss": 0.5885, "step": 38234 }, { "epoch": 2.31, "learning_rate": 3.3664539103405616e-05, "loss": 0.4815, "step": 38236 }, { "epoch": 2.31, "learning_rate": 3.366303019329139e-05, "loss": 0.7015, "step": 38238 }, { "epoch": 2.31, "learning_rate": 3.366152128317716e-05, "loss": 0.4733, "step": 38240 }, { "epoch": 2.31, "learning_rate": 3.3660012373062935e-05, "loss": 0.415, "step": 38242 }, { "epoch": 2.31, "learning_rate": 3.3658503462948715e-05, "loss": 0.4077, "step": 38244 }, { "epoch": 2.31, "learning_rate": 3.365699455283449e-05, "loss": 0.6324, "step": 38246 }, { "epoch": 2.31, "learning_rate": 3.365548564272026e-05, "loss": 0.6829, "step": 38248 }, { "epoch": 2.31, "learning_rate": 3.365397673260604e-05, "loss": 0.714, "step": 38250 }, { "epoch": 2.31, "learning_rate": 3.365246782249182e-05, "loss": 0.5956, "step": 38252 }, { "epoch": 2.31, "learning_rate": 3.365095891237759e-05, "loss": 0.4964, "step": 38254 }, { "epoch": 2.31, "learning_rate": 3.364945000226337e-05, "loss": 0.7301, "step": 38256 }, { "epoch": 2.31, "learning_rate": 3.364794109214914e-05, "loss": 0.5386, "step": 38258 }, { "epoch": 2.31, "learning_rate": 3.364643218203492e-05, "loss": 0.5547, "step": 38260 }, { "epoch": 2.31, "learning_rate": 3.364492327192069e-05, "loss": 0.5551, "step": 38262 }, { "epoch": 2.31, "learning_rate": 3.3643414361806466e-05, "loss": 0.6874, "step": 38264 }, { "epoch": 2.31, "learning_rate": 3.3641905451692246e-05, "loss": 0.6687, "step": 38266 }, { "epoch": 2.31, "learning_rate": 3.364039654157802e-05, "loss": 0.4786, "step": 38268 }, { "epoch": 2.31, "learning_rate": 3.363888763146379e-05, "loss": 0.5198, "step": 38270 }, { "epoch": 2.31, "learning_rate": 3.363737872134957e-05, "loss": 0.5576, "step": 38272 }, { "epoch": 2.31, "learning_rate": 3.3635869811235346e-05, "loss": 0.6741, "step": 38274 }, { "epoch": 2.31, "learning_rate": 3.363436090112112e-05, "loss": 0.6153, "step": 38276 }, { "epoch": 2.31, "learning_rate": 3.36328519910069e-05, "loss": 0.5494, "step": 38278 }, { "epoch": 2.31, "learning_rate": 3.363134308089267e-05, "loss": 0.4936, "step": 38280 }, { "epoch": 2.31, "learning_rate": 3.3629834170778445e-05, "loss": 0.6271, "step": 38282 }, { "epoch": 2.31, "learning_rate": 3.3628325260664225e-05, "loss": 0.4808, "step": 38284 }, { "epoch": 2.31, "learning_rate": 3.362681635055e-05, "loss": 0.7428, "step": 38286 }, { "epoch": 2.31, "learning_rate": 3.362530744043578e-05, "loss": 0.6359, "step": 38288 }, { "epoch": 2.31, "learning_rate": 3.362379853032155e-05, "loss": 0.8715, "step": 38290 }, { "epoch": 2.31, "learning_rate": 3.3622289620207324e-05, "loss": 0.4424, "step": 38292 }, { "epoch": 2.31, "learning_rate": 3.3620780710093104e-05, "loss": 0.7091, "step": 38294 }, { "epoch": 2.31, "learning_rate": 3.361927179997888e-05, "loss": 0.6912, "step": 38296 }, { "epoch": 2.31, "learning_rate": 3.361776288986465e-05, "loss": 0.5558, "step": 38298 }, { "epoch": 2.31, "learning_rate": 3.361625397975043e-05, "loss": 0.4885, "step": 38300 }, { "epoch": 2.31, "learning_rate": 3.36147450696362e-05, "loss": 0.5545, "step": 38302 }, { "epoch": 2.31, "learning_rate": 3.3613236159521976e-05, "loss": 0.406, "step": 38304 }, { "epoch": 2.31, "learning_rate": 3.3611727249407756e-05, "loss": 0.6493, "step": 38306 }, { "epoch": 2.31, "learning_rate": 3.361021833929353e-05, "loss": 0.6107, "step": 38308 }, { "epoch": 2.31, "learning_rate": 3.360870942917931e-05, "loss": 0.7513, "step": 38310 }, { "epoch": 2.31, "learning_rate": 3.360720051906508e-05, "loss": 0.509, "step": 38312 }, { "epoch": 2.31, "learning_rate": 3.3605691608950855e-05, "loss": 0.7921, "step": 38314 }, { "epoch": 2.31, "learning_rate": 3.3604182698836635e-05, "loss": 0.5113, "step": 38316 }, { "epoch": 2.31, "learning_rate": 3.360267378872241e-05, "loss": 0.4323, "step": 38318 }, { "epoch": 2.31, "learning_rate": 3.360116487860818e-05, "loss": 0.4944, "step": 38320 }, { "epoch": 2.31, "learning_rate": 3.359965596849396e-05, "loss": 0.5922, "step": 38322 }, { "epoch": 2.31, "learning_rate": 3.3598147058379735e-05, "loss": 0.5092, "step": 38324 }, { "epoch": 2.31, "learning_rate": 3.359663814826551e-05, "loss": 0.5993, "step": 38326 }, { "epoch": 2.31, "learning_rate": 3.359512923815129e-05, "loss": 0.4349, "step": 38328 }, { "epoch": 2.31, "learning_rate": 3.359362032803706e-05, "loss": 0.3912, "step": 38330 }, { "epoch": 2.31, "learning_rate": 3.3592111417922834e-05, "loss": 0.5351, "step": 38332 }, { "epoch": 2.31, "learning_rate": 3.3590602507808614e-05, "loss": 0.4609, "step": 38334 }, { "epoch": 2.31, "learning_rate": 3.358909359769439e-05, "loss": 0.6764, "step": 38336 }, { "epoch": 2.31, "learning_rate": 3.358758468758017e-05, "loss": 0.7588, "step": 38338 }, { "epoch": 2.31, "learning_rate": 3.358607577746593e-05, "loss": 0.5205, "step": 38340 }, { "epoch": 2.31, "learning_rate": 3.358456686735171e-05, "loss": 0.7553, "step": 38342 }, { "epoch": 2.31, "learning_rate": 3.358305795723749e-05, "loss": 0.5234, "step": 38344 }, { "epoch": 2.31, "learning_rate": 3.358154904712326e-05, "loss": 0.6135, "step": 38346 }, { "epoch": 2.31, "learning_rate": 3.358004013700904e-05, "loss": 0.4078, "step": 38348 }, { "epoch": 2.31, "learning_rate": 3.357853122689482e-05, "loss": 0.6177, "step": 38350 }, { "epoch": 2.31, "learning_rate": 3.357702231678059e-05, "loss": 0.5135, "step": 38352 }, { "epoch": 2.31, "learning_rate": 3.3575513406666365e-05, "loss": 0.4665, "step": 38354 }, { "epoch": 2.32, "learning_rate": 3.357400449655214e-05, "loss": 0.3727, "step": 38356 }, { "epoch": 2.32, "learning_rate": 3.357249558643792e-05, "loss": 0.5713, "step": 38358 }, { "epoch": 2.32, "learning_rate": 3.35709866763237e-05, "loss": 0.5836, "step": 38360 }, { "epoch": 2.32, "learning_rate": 3.3569477766209465e-05, "loss": 0.4652, "step": 38362 }, { "epoch": 2.32, "learning_rate": 3.3567968856095245e-05, "loss": 0.6532, "step": 38364 }, { "epoch": 2.32, "learning_rate": 3.3566459945981025e-05, "loss": 0.712, "step": 38366 }, { "epoch": 2.32, "learning_rate": 3.356495103586679e-05, "loss": 0.5895, "step": 38368 }, { "epoch": 2.32, "learning_rate": 3.356344212575257e-05, "loss": 0.6962, "step": 38370 }, { "epoch": 2.32, "learning_rate": 3.3561933215638344e-05, "loss": 0.6162, "step": 38372 }, { "epoch": 2.32, "learning_rate": 3.3560424305524124e-05, "loss": 0.5366, "step": 38374 }, { "epoch": 2.32, "learning_rate": 3.35589153954099e-05, "loss": 0.4947, "step": 38376 }, { "epoch": 2.32, "learning_rate": 3.355740648529567e-05, "loss": 0.7901, "step": 38378 }, { "epoch": 2.32, "learning_rate": 3.355589757518145e-05, "loss": 0.5993, "step": 38380 }, { "epoch": 2.32, "learning_rate": 3.355438866506722e-05, "loss": 0.6596, "step": 38382 }, { "epoch": 2.32, "learning_rate": 3.3552879754952996e-05, "loss": 0.5686, "step": 38384 }, { "epoch": 2.32, "learning_rate": 3.3551370844838776e-05, "loss": 0.3615, "step": 38386 }, { "epoch": 2.32, "learning_rate": 3.354986193472455e-05, "loss": 0.5001, "step": 38388 }, { "epoch": 2.32, "learning_rate": 3.354835302461032e-05, "loss": 0.6019, "step": 38390 }, { "epoch": 2.32, "learning_rate": 3.35468441144961e-05, "loss": 0.669, "step": 38392 }, { "epoch": 2.32, "learning_rate": 3.3545335204381875e-05, "loss": 0.9359, "step": 38394 }, { "epoch": 2.32, "learning_rate": 3.354382629426765e-05, "loss": 0.6585, "step": 38396 }, { "epoch": 2.32, "learning_rate": 3.354231738415343e-05, "loss": 0.4339, "step": 38398 }, { "epoch": 2.32, "learning_rate": 3.35408084740392e-05, "loss": 0.5819, "step": 38400 }, { "epoch": 2.32, "learning_rate": 3.353929956392498e-05, "loss": 0.5144, "step": 38402 }, { "epoch": 2.32, "learning_rate": 3.3537790653810755e-05, "loss": 0.6735, "step": 38404 }, { "epoch": 2.32, "learning_rate": 3.353628174369653e-05, "loss": 0.4936, "step": 38406 }, { "epoch": 2.32, "learning_rate": 3.353477283358231e-05, "loss": 0.5279, "step": 38408 }, { "epoch": 2.32, "learning_rate": 3.353326392346808e-05, "loss": 0.52, "step": 38410 }, { "epoch": 2.32, "learning_rate": 3.3531755013353854e-05, "loss": 0.4362, "step": 38412 }, { "epoch": 2.32, "learning_rate": 3.3530246103239634e-05, "loss": 0.6415, "step": 38414 }, { "epoch": 2.32, "learning_rate": 3.352873719312541e-05, "loss": 0.7037, "step": 38416 }, { "epoch": 2.32, "learning_rate": 3.352722828301118e-05, "loss": 0.4654, "step": 38418 }, { "epoch": 2.32, "learning_rate": 3.352571937289696e-05, "loss": 0.4731, "step": 38420 }, { "epoch": 2.32, "learning_rate": 3.352421046278273e-05, "loss": 0.5066, "step": 38422 }, { "epoch": 2.32, "learning_rate": 3.352270155266851e-05, "loss": 0.616, "step": 38424 }, { "epoch": 2.32, "learning_rate": 3.3521192642554286e-05, "loss": 0.4675, "step": 38426 }, { "epoch": 2.32, "learning_rate": 3.351968373244006e-05, "loss": 0.4437, "step": 38428 }, { "epoch": 2.32, "learning_rate": 3.351817482232584e-05, "loss": 0.6127, "step": 38430 }, { "epoch": 2.32, "learning_rate": 3.351666591221161e-05, "loss": 0.6024, "step": 38432 }, { "epoch": 2.32, "learning_rate": 3.3515157002097385e-05, "loss": 0.5662, "step": 38434 }, { "epoch": 2.32, "learning_rate": 3.3513648091983165e-05, "loss": 0.8323, "step": 38436 }, { "epoch": 2.32, "learning_rate": 3.351213918186894e-05, "loss": 0.5126, "step": 38438 }, { "epoch": 2.32, "learning_rate": 3.351063027175471e-05, "loss": 0.4966, "step": 38440 }, { "epoch": 2.32, "learning_rate": 3.350912136164049e-05, "loss": 0.6211, "step": 38442 }, { "epoch": 2.32, "learning_rate": 3.3507612451526264e-05, "loss": 0.4057, "step": 38444 }, { "epoch": 2.32, "learning_rate": 3.350610354141204e-05, "loss": 0.5063, "step": 38446 }, { "epoch": 2.32, "learning_rate": 3.350459463129782e-05, "loss": 0.6004, "step": 38448 }, { "epoch": 2.32, "learning_rate": 3.350308572118359e-05, "loss": 0.6258, "step": 38450 }, { "epoch": 2.32, "learning_rate": 3.350157681106937e-05, "loss": 0.8407, "step": 38452 }, { "epoch": 2.32, "learning_rate": 3.350006790095514e-05, "loss": 0.71, "step": 38454 }, { "epoch": 2.32, "learning_rate": 3.349855899084092e-05, "loss": 0.5357, "step": 38456 }, { "epoch": 2.32, "learning_rate": 3.34970500807267e-05, "loss": 0.7075, "step": 38458 }, { "epoch": 2.32, "learning_rate": 3.349554117061246e-05, "loss": 0.5243, "step": 38460 }, { "epoch": 2.32, "learning_rate": 3.349403226049824e-05, "loss": 0.6884, "step": 38462 }, { "epoch": 2.32, "learning_rate": 3.349252335038402e-05, "loss": 0.9145, "step": 38464 }, { "epoch": 2.32, "learning_rate": 3.3491014440269796e-05, "loss": 0.5034, "step": 38466 }, { "epoch": 2.32, "learning_rate": 3.348950553015557e-05, "loss": 0.7438, "step": 38468 }, { "epoch": 2.32, "learning_rate": 3.348799662004134e-05, "loss": 0.4918, "step": 38470 }, { "epoch": 2.32, "learning_rate": 3.348648770992712e-05, "loss": 0.6893, "step": 38472 }, { "epoch": 2.32, "learning_rate": 3.34849787998129e-05, "loss": 0.5378, "step": 38474 }, { "epoch": 2.32, "learning_rate": 3.348346988969867e-05, "loss": 0.7623, "step": 38476 }, { "epoch": 2.32, "learning_rate": 3.348196097958445e-05, "loss": 0.5939, "step": 38478 }, { "epoch": 2.32, "learning_rate": 3.348045206947023e-05, "loss": 0.4737, "step": 38480 }, { "epoch": 2.32, "learning_rate": 3.3478943159355995e-05, "loss": 0.9657, "step": 38482 }, { "epoch": 2.32, "learning_rate": 3.3477434249241774e-05, "loss": 0.6059, "step": 38484 }, { "epoch": 2.32, "learning_rate": 3.347592533912755e-05, "loss": 0.5364, "step": 38486 }, { "epoch": 2.32, "learning_rate": 3.347441642901333e-05, "loss": 0.575, "step": 38488 }, { "epoch": 2.32, "learning_rate": 3.34729075188991e-05, "loss": 0.5054, "step": 38490 }, { "epoch": 2.32, "learning_rate": 3.3471398608784874e-05, "loss": 0.4897, "step": 38492 }, { "epoch": 2.32, "learning_rate": 3.3469889698670654e-05, "loss": 0.7321, "step": 38494 }, { "epoch": 2.32, "learning_rate": 3.346838078855643e-05, "loss": 0.6321, "step": 38496 }, { "epoch": 2.32, "learning_rate": 3.34668718784422e-05, "loss": 0.4355, "step": 38498 }, { "epoch": 2.32, "learning_rate": 3.346536296832798e-05, "loss": 0.658, "step": 38500 }, { "epoch": 2.32, "learning_rate": 3.346385405821375e-05, "loss": 0.6146, "step": 38502 }, { "epoch": 2.32, "learning_rate": 3.3462345148099526e-05, "loss": 0.5896, "step": 38504 }, { "epoch": 2.32, "learning_rate": 3.3460836237985306e-05, "loss": 0.4758, "step": 38506 }, { "epoch": 2.32, "learning_rate": 3.345932732787108e-05, "loss": 0.6632, "step": 38508 }, { "epoch": 2.32, "learning_rate": 3.345781841775685e-05, "loss": 0.6398, "step": 38510 }, { "epoch": 2.32, "learning_rate": 3.345630950764263e-05, "loss": 0.6593, "step": 38512 }, { "epoch": 2.32, "learning_rate": 3.3454800597528405e-05, "loss": 0.6294, "step": 38514 }, { "epoch": 2.32, "learning_rate": 3.3453291687414185e-05, "loss": 0.6058, "step": 38516 }, { "epoch": 2.32, "learning_rate": 3.345178277729996e-05, "loss": 0.638, "step": 38518 }, { "epoch": 2.32, "learning_rate": 3.345027386718573e-05, "loss": 0.6907, "step": 38520 }, { "epoch": 2.33, "learning_rate": 3.344876495707151e-05, "loss": 0.756, "step": 38522 }, { "epoch": 2.33, "learning_rate": 3.3447256046957284e-05, "loss": 0.7236, "step": 38524 }, { "epoch": 2.33, "learning_rate": 3.344574713684306e-05, "loss": 0.5708, "step": 38526 }, { "epoch": 2.33, "learning_rate": 3.344423822672884e-05, "loss": 0.4636, "step": 38528 }, { "epoch": 2.33, "learning_rate": 3.344272931661461e-05, "loss": 0.512, "step": 38530 }, { "epoch": 2.33, "learning_rate": 3.3441220406500384e-05, "loss": 0.5701, "step": 38532 }, { "epoch": 2.33, "learning_rate": 3.3439711496386164e-05, "loss": 0.5214, "step": 38534 }, { "epoch": 2.33, "learning_rate": 3.343820258627194e-05, "loss": 0.539, "step": 38536 }, { "epoch": 2.33, "learning_rate": 3.3436693676157717e-05, "loss": 0.5598, "step": 38538 }, { "epoch": 2.33, "learning_rate": 3.343518476604349e-05, "loss": 0.6749, "step": 38540 }, { "epoch": 2.33, "learning_rate": 3.343367585592926e-05, "loss": 0.8043, "step": 38542 }, { "epoch": 2.33, "learning_rate": 3.343216694581504e-05, "loss": 0.5749, "step": 38544 }, { "epoch": 2.33, "learning_rate": 3.3430658035700816e-05, "loss": 0.8206, "step": 38546 }, { "epoch": 2.33, "learning_rate": 3.342914912558659e-05, "loss": 0.5652, "step": 38548 }, { "epoch": 2.33, "learning_rate": 3.342764021547237e-05, "loss": 0.575, "step": 38550 }, { "epoch": 2.33, "learning_rate": 3.342613130535814e-05, "loss": 0.6174, "step": 38552 }, { "epoch": 2.33, "learning_rate": 3.3424622395243915e-05, "loss": 0.468, "step": 38554 }, { "epoch": 2.33, "learning_rate": 3.3423113485129695e-05, "loss": 0.5006, "step": 38556 }, { "epoch": 2.33, "learning_rate": 3.342160457501547e-05, "loss": 0.4831, "step": 38558 }, { "epoch": 2.33, "learning_rate": 3.342009566490124e-05, "loss": 0.5503, "step": 38560 }, { "epoch": 2.33, "learning_rate": 3.341858675478702e-05, "loss": 1.1339, "step": 38562 }, { "epoch": 2.33, "learning_rate": 3.3417077844672794e-05, "loss": 0.6749, "step": 38564 }, { "epoch": 2.33, "learning_rate": 3.3415568934558574e-05, "loss": 0.4342, "step": 38566 }, { "epoch": 2.33, "learning_rate": 3.341406002444434e-05, "loss": 0.382, "step": 38568 }, { "epoch": 2.33, "learning_rate": 3.341255111433012e-05, "loss": 0.3937, "step": 38570 }, { "epoch": 2.33, "learning_rate": 3.34110422042159e-05, "loss": 0.533, "step": 38572 }, { "epoch": 2.33, "learning_rate": 3.340953329410167e-05, "loss": 0.4655, "step": 38574 }, { "epoch": 2.33, "learning_rate": 3.340802438398745e-05, "loss": 0.5667, "step": 38576 }, { "epoch": 2.33, "learning_rate": 3.3406515473873227e-05, "loss": 0.5848, "step": 38578 }, { "epoch": 2.33, "learning_rate": 3.3405006563759e-05, "loss": 0.6453, "step": 38580 }, { "epoch": 2.33, "learning_rate": 3.340349765364477e-05, "loss": 0.923, "step": 38582 }, { "epoch": 2.33, "learning_rate": 3.3401988743530546e-05, "loss": 0.6241, "step": 38584 }, { "epoch": 2.33, "learning_rate": 3.3400479833416326e-05, "loss": 0.6094, "step": 38586 }, { "epoch": 2.33, "learning_rate": 3.3398970923302106e-05, "loss": 0.4652, "step": 38588 }, { "epoch": 2.33, "learning_rate": 3.339746201318787e-05, "loss": 0.7816, "step": 38590 }, { "epoch": 2.33, "learning_rate": 3.339595310307365e-05, "loss": 0.6953, "step": 38592 }, { "epoch": 2.33, "learning_rate": 3.339444419295943e-05, "loss": 0.5184, "step": 38594 }, { "epoch": 2.33, "learning_rate": 3.33929352828452e-05, "loss": 0.6466, "step": 38596 }, { "epoch": 2.33, "learning_rate": 3.339142637273098e-05, "loss": 0.6026, "step": 38598 }, { "epoch": 2.33, "learning_rate": 3.338991746261675e-05, "loss": 0.6812, "step": 38600 }, { "epoch": 2.33, "learning_rate": 3.338840855250253e-05, "loss": 0.6778, "step": 38602 }, { "epoch": 2.33, "learning_rate": 3.3386899642388304e-05, "loss": 0.54, "step": 38604 }, { "epoch": 2.33, "learning_rate": 3.338539073227408e-05, "loss": 0.8065, "step": 38606 }, { "epoch": 2.33, "learning_rate": 3.338388182215986e-05, "loss": 0.7348, "step": 38608 }, { "epoch": 2.33, "learning_rate": 3.338237291204563e-05, "loss": 0.6905, "step": 38610 }, { "epoch": 2.33, "learning_rate": 3.3380864001931404e-05, "loss": 0.5905, "step": 38612 }, { "epoch": 2.33, "learning_rate": 3.3379355091817183e-05, "loss": 0.4625, "step": 38614 }, { "epoch": 2.33, "learning_rate": 3.3377846181702957e-05, "loss": 0.6603, "step": 38616 }, { "epoch": 2.33, "learning_rate": 3.337633727158873e-05, "loss": 0.7458, "step": 38618 }, { "epoch": 2.33, "learning_rate": 3.337482836147451e-05, "loss": 0.7365, "step": 38620 }, { "epoch": 2.33, "learning_rate": 3.337331945136028e-05, "loss": 0.4205, "step": 38622 }, { "epoch": 2.33, "learning_rate": 3.3371810541246056e-05, "loss": 0.6527, "step": 38624 }, { "epoch": 2.33, "learning_rate": 3.3370301631131836e-05, "loss": 0.5256, "step": 38626 }, { "epoch": 2.33, "learning_rate": 3.336879272101761e-05, "loss": 0.7169, "step": 38628 }, { "epoch": 2.33, "learning_rate": 3.336728381090339e-05, "loss": 0.6123, "step": 38630 }, { "epoch": 2.33, "learning_rate": 3.336577490078916e-05, "loss": 0.4112, "step": 38632 }, { "epoch": 2.33, "learning_rate": 3.3364265990674935e-05, "loss": 0.6257, "step": 38634 }, { "epoch": 2.33, "learning_rate": 3.3362757080560715e-05, "loss": 0.6993, "step": 38636 }, { "epoch": 2.33, "learning_rate": 3.336124817044649e-05, "loss": 0.4492, "step": 38638 }, { "epoch": 2.33, "learning_rate": 3.335973926033226e-05, "loss": 0.5468, "step": 38640 }, { "epoch": 2.33, "learning_rate": 3.335823035021804e-05, "loss": 0.6607, "step": 38642 }, { "epoch": 2.33, "learning_rate": 3.3356721440103814e-05, "loss": 0.6049, "step": 38644 }, { "epoch": 2.33, "learning_rate": 3.335521252998959e-05, "loss": 0.4958, "step": 38646 }, { "epoch": 2.33, "learning_rate": 3.335370361987537e-05, "loss": 0.5635, "step": 38648 }, { "epoch": 2.33, "learning_rate": 3.335219470976114e-05, "loss": 0.5712, "step": 38650 }, { "epoch": 2.33, "learning_rate": 3.335068579964692e-05, "loss": 0.5765, "step": 38652 }, { "epoch": 2.33, "learning_rate": 3.3349176889532693e-05, "loss": 0.5211, "step": 38654 }, { "epoch": 2.33, "learning_rate": 3.3347667979418467e-05, "loss": 0.4919, "step": 38656 }, { "epoch": 2.33, "learning_rate": 3.3346159069304246e-05, "loss": 0.4553, "step": 38658 }, { "epoch": 2.33, "learning_rate": 3.334465015919002e-05, "loss": 0.5865, "step": 38660 }, { "epoch": 2.33, "learning_rate": 3.334314124907579e-05, "loss": 0.788, "step": 38662 }, { "epoch": 2.33, "learning_rate": 3.334163233896157e-05, "loss": 0.5036, "step": 38664 }, { "epoch": 2.33, "learning_rate": 3.3340123428847346e-05, "loss": 0.6874, "step": 38666 }, { "epoch": 2.33, "learning_rate": 3.333861451873312e-05, "loss": 0.5304, "step": 38668 }, { "epoch": 2.33, "learning_rate": 3.33371056086189e-05, "loss": 0.8837, "step": 38670 }, { "epoch": 2.33, "learning_rate": 3.333559669850467e-05, "loss": 0.8511, "step": 38672 }, { "epoch": 2.33, "learning_rate": 3.3334087788390445e-05, "loss": 0.6847, "step": 38674 }, { "epoch": 2.33, "learning_rate": 3.3332578878276225e-05, "loss": 0.7577, "step": 38676 }, { "epoch": 2.33, "learning_rate": 3.3331069968162e-05, "loss": 0.5033, "step": 38678 }, { "epoch": 2.33, "learning_rate": 3.332956105804778e-05, "loss": 0.4188, "step": 38680 }, { "epoch": 2.33, "learning_rate": 3.3328052147933544e-05, "loss": 0.7441, "step": 38682 }, { "epoch": 2.33, "learning_rate": 3.3326543237819324e-05, "loss": 0.5225, "step": 38684 }, { "epoch": 2.33, "learning_rate": 3.3325034327705104e-05, "loss": 0.6987, "step": 38686 }, { "epoch": 2.34, "learning_rate": 3.332352541759088e-05, "loss": 0.3191, "step": 38688 }, { "epoch": 2.34, "learning_rate": 3.332201650747665e-05, "loss": 0.415, "step": 38690 }, { "epoch": 2.34, "learning_rate": 3.332050759736243e-05, "loss": 0.4493, "step": 38692 }, { "epoch": 2.34, "learning_rate": 3.33189986872482e-05, "loss": 0.5516, "step": 38694 }, { "epoch": 2.34, "learning_rate": 3.3317489777133976e-05, "loss": 0.8064, "step": 38696 }, { "epoch": 2.34, "learning_rate": 3.331598086701975e-05, "loss": 0.5186, "step": 38698 }, { "epoch": 2.34, "learning_rate": 3.331447195690553e-05, "loss": 0.3528, "step": 38700 }, { "epoch": 2.34, "learning_rate": 3.331296304679131e-05, "loss": 0.4797, "step": 38702 }, { "epoch": 2.34, "learning_rate": 3.3311454136677076e-05, "loss": 0.4814, "step": 38704 }, { "epoch": 2.34, "learning_rate": 3.3309945226562856e-05, "loss": 0.6932, "step": 38706 }, { "epoch": 2.34, "learning_rate": 3.3308436316448636e-05, "loss": 0.5478, "step": 38708 }, { "epoch": 2.34, "learning_rate": 3.33069274063344e-05, "loss": 0.5699, "step": 38710 }, { "epoch": 2.34, "learning_rate": 3.330541849622018e-05, "loss": 0.583, "step": 38712 }, { "epoch": 2.34, "learning_rate": 3.3303909586105955e-05, "loss": 0.5634, "step": 38714 }, { "epoch": 2.34, "learning_rate": 3.3302400675991735e-05, "loss": 0.6617, "step": 38716 }, { "epoch": 2.34, "learning_rate": 3.330089176587751e-05, "loss": 0.5614, "step": 38718 }, { "epoch": 2.34, "learning_rate": 3.329938285576328e-05, "loss": 0.823, "step": 38720 }, { "epoch": 2.34, "learning_rate": 3.329787394564906e-05, "loss": 0.7294, "step": 38722 }, { "epoch": 2.34, "learning_rate": 3.3296365035534834e-05, "loss": 0.6787, "step": 38724 }, { "epoch": 2.34, "learning_rate": 3.329485612542061e-05, "loss": 0.5403, "step": 38726 }, { "epoch": 2.34, "learning_rate": 3.329334721530639e-05, "loss": 0.5324, "step": 38728 }, { "epoch": 2.34, "learning_rate": 3.329183830519216e-05, "loss": 0.9308, "step": 38730 }, { "epoch": 2.34, "learning_rate": 3.329032939507793e-05, "loss": 0.614, "step": 38732 }, { "epoch": 2.34, "learning_rate": 3.328957494002082e-05, "loss": 0.5279, "step": 38734 }, { "epoch": 2.34, "learning_rate": 3.32880660299066e-05, "loss": 0.6708, "step": 38736 }, { "epoch": 2.34, "learning_rate": 3.3286557119792376e-05, "loss": 0.5374, "step": 38738 }, { "epoch": 2.34, "learning_rate": 3.328504820967815e-05, "loss": 0.65, "step": 38740 }, { "epoch": 2.34, "learning_rate": 3.328353929956393e-05, "loss": 0.6068, "step": 38742 }, { "epoch": 2.34, "learning_rate": 3.32820303894497e-05, "loss": 0.5956, "step": 38744 }, { "epoch": 2.34, "learning_rate": 3.328052147933548e-05, "loss": 0.5574, "step": 38746 }, { "epoch": 2.34, "learning_rate": 3.327901256922125e-05, "loss": 0.8549, "step": 38748 }, { "epoch": 2.34, "learning_rate": 3.327750365910703e-05, "loss": 0.6306, "step": 38750 }, { "epoch": 2.34, "learning_rate": 3.327599474899281e-05, "loss": 0.657, "step": 38752 }, { "epoch": 2.34, "learning_rate": 3.3274485838878575e-05, "loss": 0.5179, "step": 38754 }, { "epoch": 2.34, "learning_rate": 3.3272976928764355e-05, "loss": 0.4855, "step": 38756 }, { "epoch": 2.34, "learning_rate": 3.3271468018650135e-05, "loss": 0.6891, "step": 38758 }, { "epoch": 2.34, "learning_rate": 3.326995910853591e-05, "loss": 0.5085, "step": 38760 }, { "epoch": 2.34, "learning_rate": 3.326845019842168e-05, "loss": 0.7217, "step": 38762 }, { "epoch": 2.34, "learning_rate": 3.3266941288307454e-05, "loss": 0.7802, "step": 38764 }, { "epoch": 2.34, "learning_rate": 3.3265432378193234e-05, "loss": 0.2611, "step": 38766 }, { "epoch": 2.34, "learning_rate": 3.326392346807901e-05, "loss": 0.3554, "step": 38768 }, { "epoch": 2.34, "learning_rate": 3.326241455796478e-05, "loss": 0.7427, "step": 38770 }, { "epoch": 2.34, "learning_rate": 3.326090564785056e-05, "loss": 0.4379, "step": 38772 }, { "epoch": 2.34, "learning_rate": 3.325939673773634e-05, "loss": 0.3856, "step": 38774 }, { "epoch": 2.34, "learning_rate": 3.3257887827622106e-05, "loss": 0.7747, "step": 38776 }, { "epoch": 2.34, "learning_rate": 3.3256378917507886e-05, "loss": 0.5931, "step": 38778 }, { "epoch": 2.34, "learning_rate": 3.325487000739366e-05, "loss": 0.4403, "step": 38780 }, { "epoch": 2.34, "learning_rate": 3.325336109727943e-05, "loss": 0.5883, "step": 38782 }, { "epoch": 2.34, "learning_rate": 3.325185218716521e-05, "loss": 0.3807, "step": 38784 }, { "epoch": 2.34, "learning_rate": 3.3250343277050986e-05, "loss": 0.6148, "step": 38786 }, { "epoch": 2.34, "learning_rate": 3.3248834366936765e-05, "loss": 0.7034, "step": 38788 }, { "epoch": 2.34, "learning_rate": 3.324732545682254e-05, "loss": 0.5727, "step": 38790 }, { "epoch": 2.34, "learning_rate": 3.324581654670831e-05, "loss": 0.6126, "step": 38792 }, { "epoch": 2.34, "learning_rate": 3.324430763659409e-05, "loss": 0.6112, "step": 38794 }, { "epoch": 2.34, "learning_rate": 3.3242798726479865e-05, "loss": 0.4012, "step": 38796 }, { "epoch": 2.34, "learning_rate": 3.324128981636564e-05, "loss": 0.6339, "step": 38798 }, { "epoch": 2.34, "learning_rate": 3.323978090625142e-05, "loss": 0.7731, "step": 38800 }, { "epoch": 2.34, "learning_rate": 3.323827199613719e-05, "loss": 0.8477, "step": 38802 }, { "epoch": 2.34, "learning_rate": 3.3236763086022964e-05, "loss": 0.7313, "step": 38804 }, { "epoch": 2.34, "learning_rate": 3.3235254175908744e-05, "loss": 0.5568, "step": 38806 }, { "epoch": 2.34, "learning_rate": 3.323374526579452e-05, "loss": 0.5038, "step": 38808 }, { "epoch": 2.34, "learning_rate": 3.32322363556803e-05, "loss": 0.6379, "step": 38810 }, { "epoch": 2.34, "learning_rate": 3.323072744556607e-05, "loss": 0.5137, "step": 38812 }, { "epoch": 2.34, "learning_rate": 3.322921853545184e-05, "loss": 0.5336, "step": 38814 }, { "epoch": 2.34, "learning_rate": 3.322770962533762e-05, "loss": 0.596, "step": 38816 }, { "epoch": 2.34, "learning_rate": 3.3226200715223396e-05, "loss": 0.4621, "step": 38818 }, { "epoch": 2.34, "learning_rate": 3.322469180510917e-05, "loss": 0.9835, "step": 38820 }, { "epoch": 2.34, "learning_rate": 3.322318289499495e-05, "loss": 0.6998, "step": 38822 }, { "epoch": 2.34, "learning_rate": 3.322167398488072e-05, "loss": 0.4741, "step": 38824 }, { "epoch": 2.34, "learning_rate": 3.3220165074766496e-05, "loss": 0.6781, "step": 38826 }, { "epoch": 2.34, "learning_rate": 3.3218656164652275e-05, "loss": 0.4192, "step": 38828 }, { "epoch": 2.34, "learning_rate": 3.321714725453805e-05, "loss": 0.6686, "step": 38830 }, { "epoch": 2.34, "learning_rate": 3.321563834442382e-05, "loss": 0.6229, "step": 38832 }, { "epoch": 2.34, "learning_rate": 3.32141294343096e-05, "loss": 0.6267, "step": 38834 }, { "epoch": 2.34, "learning_rate": 3.3212620524195375e-05, "loss": 0.6175, "step": 38836 }, { "epoch": 2.34, "learning_rate": 3.3211111614081155e-05, "loss": 0.743, "step": 38838 }, { "epoch": 2.34, "learning_rate": 3.320960270396693e-05, "loss": 0.6174, "step": 38840 }, { "epoch": 2.34, "learning_rate": 3.32080937938527e-05, "loss": 0.5545, "step": 38842 }, { "epoch": 2.34, "learning_rate": 3.320658488373848e-05, "loss": 0.7968, "step": 38844 }, { "epoch": 2.34, "learning_rate": 3.320507597362425e-05, "loss": 0.6819, "step": 38846 }, { "epoch": 2.34, "learning_rate": 3.320356706351003e-05, "loss": 0.5233, "step": 38848 }, { "epoch": 2.34, "learning_rate": 3.320205815339581e-05, "loss": 0.8057, "step": 38850 }, { "epoch": 2.35, "learning_rate": 3.320054924328158e-05, "loss": 0.6429, "step": 38852 }, { "epoch": 2.35, "learning_rate": 3.319904033316735e-05, "loss": 0.8395, "step": 38854 }, { "epoch": 2.35, "learning_rate": 3.319753142305313e-05, "loss": 0.5587, "step": 38856 }, { "epoch": 2.35, "learning_rate": 3.3196022512938906e-05, "loss": 0.4651, "step": 38858 }, { "epoch": 2.35, "learning_rate": 3.3194513602824686e-05, "loss": 0.4909, "step": 38860 }, { "epoch": 2.35, "learning_rate": 3.319300469271045e-05, "loss": 0.5953, "step": 38862 }, { "epoch": 2.35, "learning_rate": 3.319149578259623e-05, "loss": 0.5736, "step": 38864 }, { "epoch": 2.35, "learning_rate": 3.318998687248201e-05, "loss": 0.6594, "step": 38866 }, { "epoch": 2.35, "learning_rate": 3.318847796236778e-05, "loss": 0.6634, "step": 38868 }, { "epoch": 2.35, "learning_rate": 3.318696905225356e-05, "loss": 0.5281, "step": 38870 }, { "epoch": 2.35, "learning_rate": 3.318546014213934e-05, "loss": 0.8704, "step": 38872 }, { "epoch": 2.35, "learning_rate": 3.318395123202511e-05, "loss": 0.6912, "step": 38874 }, { "epoch": 2.35, "learning_rate": 3.3182442321910885e-05, "loss": 0.4094, "step": 38876 }, { "epoch": 2.35, "learning_rate": 3.318093341179666e-05, "loss": 0.6937, "step": 38878 }, { "epoch": 2.35, "learning_rate": 3.317942450168244e-05, "loss": 0.5981, "step": 38880 }, { "epoch": 2.35, "learning_rate": 3.317791559156821e-05, "loss": 0.7642, "step": 38882 }, { "epoch": 2.35, "learning_rate": 3.3176406681453984e-05, "loss": 0.6597, "step": 38884 }, { "epoch": 2.35, "learning_rate": 3.3174897771339764e-05, "loss": 0.5369, "step": 38886 }, { "epoch": 2.35, "learning_rate": 3.3173388861225544e-05, "loss": 0.4069, "step": 38888 }, { "epoch": 2.35, "learning_rate": 3.317187995111131e-05, "loss": 0.5448, "step": 38890 }, { "epoch": 2.35, "learning_rate": 3.317037104099709e-05, "loss": 0.7106, "step": 38892 }, { "epoch": 2.35, "learning_rate": 3.316886213088286e-05, "loss": 0.5681, "step": 38894 }, { "epoch": 2.35, "learning_rate": 3.3167353220768636e-05, "loss": 0.561, "step": 38896 }, { "epoch": 2.35, "learning_rate": 3.3165844310654416e-05, "loss": 0.6328, "step": 38898 }, { "epoch": 2.35, "learning_rate": 3.316433540054019e-05, "loss": 0.5408, "step": 38900 }, { "epoch": 2.35, "learning_rate": 3.316282649042597e-05, "loss": 0.5622, "step": 38902 }, { "epoch": 2.35, "learning_rate": 3.316131758031174e-05, "loss": 0.5829, "step": 38904 }, { "epoch": 2.35, "learning_rate": 3.3159808670197515e-05, "loss": 0.3962, "step": 38906 }, { "epoch": 2.35, "learning_rate": 3.3158299760083295e-05, "loss": 0.5811, "step": 38908 }, { "epoch": 2.35, "learning_rate": 3.315679084996907e-05, "loss": 0.5991, "step": 38910 }, { "epoch": 2.35, "learning_rate": 3.315528193985484e-05, "loss": 0.4912, "step": 38912 }, { "epoch": 2.35, "learning_rate": 3.315377302974062e-05, "loss": 0.3647, "step": 38914 }, { "epoch": 2.35, "learning_rate": 3.3152264119626395e-05, "loss": 0.6727, "step": 38916 }, { "epoch": 2.35, "learning_rate": 3.315075520951217e-05, "loss": 0.7379, "step": 38918 }, { "epoch": 2.35, "learning_rate": 3.314924629939795e-05, "loss": 0.6232, "step": 38920 }, { "epoch": 2.35, "learning_rate": 3.314773738928372e-05, "loss": 0.422, "step": 38922 }, { "epoch": 2.35, "learning_rate": 3.31462284791695e-05, "loss": 0.658, "step": 38924 }, { "epoch": 2.35, "learning_rate": 3.3144719569055274e-05, "loss": 0.5909, "step": 38926 }, { "epoch": 2.35, "learning_rate": 3.314321065894105e-05, "loss": 0.5015, "step": 38928 }, { "epoch": 2.35, "learning_rate": 3.314170174882683e-05, "loss": 0.6475, "step": 38930 }, { "epoch": 2.35, "learning_rate": 3.31401928387126e-05, "loss": 0.5764, "step": 38932 }, { "epoch": 2.35, "learning_rate": 3.313868392859837e-05, "loss": 0.5366, "step": 38934 }, { "epoch": 2.35, "learning_rate": 3.313717501848415e-05, "loss": 0.7308, "step": 38936 }, { "epoch": 2.35, "learning_rate": 3.3135666108369926e-05, "loss": 0.7026, "step": 38938 }, { "epoch": 2.35, "learning_rate": 3.31341571982557e-05, "loss": 0.5701, "step": 38940 }, { "epoch": 2.35, "learning_rate": 3.313264828814148e-05, "loss": 0.5334, "step": 38942 }, { "epoch": 2.35, "learning_rate": 3.313113937802725e-05, "loss": 0.598, "step": 38944 }, { "epoch": 2.35, "learning_rate": 3.3129630467913025e-05, "loss": 0.5242, "step": 38946 }, { "epoch": 2.35, "learning_rate": 3.3128121557798805e-05, "loss": 0.805, "step": 38948 }, { "epoch": 2.35, "learning_rate": 3.312661264768458e-05, "loss": 0.5289, "step": 38950 }, { "epoch": 2.35, "learning_rate": 3.312510373757036e-05, "loss": 0.6698, "step": 38952 }, { "epoch": 2.35, "learning_rate": 3.312359482745613e-05, "loss": 0.5541, "step": 38954 }, { "epoch": 2.35, "learning_rate": 3.3122085917341905e-05, "loss": 0.8302, "step": 38956 }, { "epoch": 2.35, "learning_rate": 3.3120577007227684e-05, "loss": 0.447, "step": 38958 }, { "epoch": 2.35, "learning_rate": 3.311906809711346e-05, "loss": 0.5298, "step": 38960 }, { "epoch": 2.35, "learning_rate": 3.311755918699923e-05, "loss": 0.5979, "step": 38962 }, { "epoch": 2.35, "learning_rate": 3.311605027688501e-05, "loss": 0.5556, "step": 38964 }, { "epoch": 2.35, "learning_rate": 3.3114541366770784e-05, "loss": 0.6557, "step": 38966 }, { "epoch": 2.35, "learning_rate": 3.311303245665656e-05, "loss": 0.6114, "step": 38968 }, { "epoch": 2.35, "learning_rate": 3.311152354654234e-05, "loss": 0.6498, "step": 38970 }, { "epoch": 2.35, "learning_rate": 3.311001463642811e-05, "loss": 0.6142, "step": 38972 }, { "epoch": 2.35, "learning_rate": 3.310850572631389e-05, "loss": 0.6523, "step": 38974 }, { "epoch": 2.35, "learning_rate": 3.3106996816199656e-05, "loss": 0.4539, "step": 38976 }, { "epoch": 2.35, "learning_rate": 3.3105487906085436e-05, "loss": 0.6361, "step": 38978 }, { "epoch": 2.35, "learning_rate": 3.3103978995971216e-05, "loss": 0.8687, "step": 38980 }, { "epoch": 2.35, "learning_rate": 3.310247008585698e-05, "loss": 0.6502, "step": 38982 }, { "epoch": 2.35, "learning_rate": 3.310096117574276e-05, "loss": 0.694, "step": 38984 }, { "epoch": 2.35, "learning_rate": 3.309945226562854e-05, "loss": 0.7138, "step": 38986 }, { "epoch": 2.35, "learning_rate": 3.3097943355514315e-05, "loss": 0.5727, "step": 38988 }, { "epoch": 2.35, "learning_rate": 3.309643444540009e-05, "loss": 0.5681, "step": 38990 }, { "epoch": 2.35, "learning_rate": 3.309492553528586e-05, "loss": 0.515, "step": 38992 }, { "epoch": 2.35, "learning_rate": 3.309341662517164e-05, "loss": 0.6963, "step": 38994 }, { "epoch": 2.35, "learning_rate": 3.3091907715057414e-05, "loss": 0.579, "step": 38996 }, { "epoch": 2.35, "learning_rate": 3.309039880494319e-05, "loss": 0.5835, "step": 38998 }, { "epoch": 2.35, "learning_rate": 3.308888989482897e-05, "loss": 0.5663, "step": 39000 }, { "epoch": 2.35, "learning_rate": 3.308738098471475e-05, "loss": 0.7554, "step": 39002 }, { "epoch": 2.35, "learning_rate": 3.3085872074600514e-05, "loss": 0.6635, "step": 39004 }, { "epoch": 2.35, "learning_rate": 3.3084363164486294e-05, "loss": 0.4083, "step": 39006 }, { "epoch": 2.35, "learning_rate": 3.308285425437207e-05, "loss": 0.5968, "step": 39008 }, { "epoch": 2.35, "learning_rate": 3.308134534425784e-05, "loss": 0.5912, "step": 39010 }, { "epoch": 2.35, "learning_rate": 3.307983643414362e-05, "loss": 0.5947, "step": 39012 }, { "epoch": 2.35, "learning_rate": 3.307832752402939e-05, "loss": 0.8414, "step": 39014 }, { "epoch": 2.35, "learning_rate": 3.307681861391517e-05, "loss": 0.5704, "step": 39016 }, { "epoch": 2.36, "learning_rate": 3.3075309703800946e-05, "loss": 0.7148, "step": 39018 }, { "epoch": 2.36, "learning_rate": 3.307380079368672e-05, "loss": 0.5336, "step": 39020 }, { "epoch": 2.36, "learning_rate": 3.30722918835725e-05, "loss": 0.5332, "step": 39022 }, { "epoch": 2.36, "learning_rate": 3.307078297345827e-05, "loss": 0.6237, "step": 39024 }, { "epoch": 2.36, "learning_rate": 3.3069274063344045e-05, "loss": 0.5482, "step": 39026 }, { "epoch": 2.36, "learning_rate": 3.3067765153229825e-05, "loss": 0.7424, "step": 39028 }, { "epoch": 2.36, "learning_rate": 3.30662562431156e-05, "loss": 0.6359, "step": 39030 }, { "epoch": 2.36, "learning_rate": 3.306474733300137e-05, "loss": 0.4185, "step": 39032 }, { "epoch": 2.36, "learning_rate": 3.306323842288715e-05, "loss": 0.5272, "step": 39034 }, { "epoch": 2.36, "learning_rate": 3.3061729512772924e-05, "loss": 0.506, "step": 39036 }, { "epoch": 2.36, "learning_rate": 3.3060220602658704e-05, "loss": 0.5915, "step": 39038 }, { "epoch": 2.36, "learning_rate": 3.305871169254448e-05, "loss": 0.5872, "step": 39040 }, { "epoch": 2.36, "learning_rate": 3.305720278243025e-05, "loss": 0.5761, "step": 39042 }, { "epoch": 2.36, "learning_rate": 3.305569387231603e-05, "loss": 0.7332, "step": 39044 }, { "epoch": 2.36, "learning_rate": 3.3054184962201804e-05, "loss": 0.668, "step": 39046 }, { "epoch": 2.36, "learning_rate": 3.305267605208758e-05, "loss": 0.5371, "step": 39048 }, { "epoch": 2.36, "learning_rate": 3.305116714197336e-05, "loss": 0.5742, "step": 39050 }, { "epoch": 2.36, "learning_rate": 3.304965823185913e-05, "loss": 0.7271, "step": 39052 }, { "epoch": 2.36, "learning_rate": 3.30481493217449e-05, "loss": 0.5674, "step": 39054 }, { "epoch": 2.36, "learning_rate": 3.304664041163068e-05, "loss": 0.5698, "step": 39056 }, { "epoch": 2.36, "learning_rate": 3.3045131501516456e-05, "loss": 0.537, "step": 39058 }, { "epoch": 2.36, "learning_rate": 3.304362259140223e-05, "loss": 0.4948, "step": 39060 }, { "epoch": 2.36, "learning_rate": 3.304211368128801e-05, "loss": 0.5064, "step": 39062 }, { "epoch": 2.36, "learning_rate": 3.304060477117378e-05, "loss": 0.7268, "step": 39064 }, { "epoch": 2.36, "learning_rate": 3.303909586105956e-05, "loss": 0.4585, "step": 39066 }, { "epoch": 2.36, "learning_rate": 3.3037586950945335e-05, "loss": 0.7621, "step": 39068 }, { "epoch": 2.36, "learning_rate": 3.303607804083111e-05, "loss": 0.5738, "step": 39070 }, { "epoch": 2.36, "learning_rate": 3.303456913071689e-05, "loss": 0.5749, "step": 39072 }, { "epoch": 2.36, "learning_rate": 3.303306022060266e-05, "loss": 0.6169, "step": 39074 }, { "epoch": 2.36, "learning_rate": 3.3031551310488434e-05, "loss": 0.7167, "step": 39076 }, { "epoch": 2.36, "learning_rate": 3.3030042400374214e-05, "loss": 0.5157, "step": 39078 }, { "epoch": 2.36, "learning_rate": 3.302853349025999e-05, "loss": 0.6398, "step": 39080 }, { "epoch": 2.36, "learning_rate": 3.302702458014576e-05, "loss": 0.7251, "step": 39082 }, { "epoch": 2.36, "learning_rate": 3.302551567003154e-05, "loss": 0.4767, "step": 39084 }, { "epoch": 2.36, "learning_rate": 3.3024006759917314e-05, "loss": 0.6157, "step": 39086 }, { "epoch": 2.36, "learning_rate": 3.3022497849803093e-05, "loss": 0.6038, "step": 39088 }, { "epoch": 2.36, "learning_rate": 3.302098893968886e-05, "loss": 0.5755, "step": 39090 }, { "epoch": 2.36, "learning_rate": 3.301948002957464e-05, "loss": 0.7658, "step": 39092 }, { "epoch": 2.36, "learning_rate": 3.301797111946042e-05, "loss": 0.5634, "step": 39094 }, { "epoch": 2.36, "learning_rate": 3.3016462209346186e-05, "loss": 0.4733, "step": 39096 }, { "epoch": 2.36, "learning_rate": 3.3014953299231966e-05, "loss": 0.5221, "step": 39098 }, { "epoch": 2.36, "learning_rate": 3.3013444389117746e-05, "loss": 0.6387, "step": 39100 }, { "epoch": 2.36, "learning_rate": 3.301193547900352e-05, "loss": 0.6821, "step": 39102 }, { "epoch": 2.36, "learning_rate": 3.301042656888929e-05, "loss": 0.6607, "step": 39104 }, { "epoch": 2.36, "learning_rate": 3.3008917658775065e-05, "loss": 0.6639, "step": 39106 }, { "epoch": 2.36, "learning_rate": 3.3007408748660845e-05, "loss": 0.4724, "step": 39108 }, { "epoch": 2.36, "learning_rate": 3.300589983854662e-05, "loss": 0.5858, "step": 39110 }, { "epoch": 2.36, "learning_rate": 3.300439092843239e-05, "loss": 0.5461, "step": 39112 }, { "epoch": 2.36, "learning_rate": 3.300288201831817e-05, "loss": 0.6012, "step": 39114 }, { "epoch": 2.36, "learning_rate": 3.300137310820395e-05, "loss": 0.752, "step": 39116 }, { "epoch": 2.36, "learning_rate": 3.299986419808972e-05, "loss": 0.527, "step": 39118 }, { "epoch": 2.36, "learning_rate": 3.29983552879755e-05, "loss": 0.8214, "step": 39120 }, { "epoch": 2.36, "learning_rate": 3.299684637786127e-05, "loss": 0.4065, "step": 39122 }, { "epoch": 2.36, "learning_rate": 3.299533746774705e-05, "loss": 0.7232, "step": 39124 }, { "epoch": 2.36, "learning_rate": 3.2993828557632823e-05, "loss": 0.5147, "step": 39126 }, { "epoch": 2.36, "learning_rate": 3.29923196475186e-05, "loss": 0.7516, "step": 39128 }, { "epoch": 2.36, "learning_rate": 3.2990810737404377e-05, "loss": 0.6938, "step": 39130 }, { "epoch": 2.36, "learning_rate": 3.298930182729015e-05, "loss": 0.6082, "step": 39132 }, { "epoch": 2.36, "learning_rate": 3.298779291717592e-05, "loss": 0.6193, "step": 39134 }, { "epoch": 2.36, "learning_rate": 3.29862840070617e-05, "loss": 0.4825, "step": 39136 }, { "epoch": 2.36, "learning_rate": 3.2984775096947476e-05, "loss": 0.5283, "step": 39138 }, { "epoch": 2.36, "learning_rate": 3.298326618683325e-05, "loss": 0.6401, "step": 39140 }, { "epoch": 2.36, "learning_rate": 3.298175727671903e-05, "loss": 0.6358, "step": 39142 }, { "epoch": 2.36, "learning_rate": 3.29802483666048e-05, "loss": 0.6343, "step": 39144 }, { "epoch": 2.36, "learning_rate": 3.2978739456490575e-05, "loss": 0.7108, "step": 39146 }, { "epoch": 2.36, "learning_rate": 3.2977230546376355e-05, "loss": 0.3617, "step": 39148 }, { "epoch": 2.36, "learning_rate": 3.297572163626213e-05, "loss": 0.51, "step": 39150 }, { "epoch": 2.36, "learning_rate": 3.297421272614791e-05, "loss": 0.5342, "step": 39152 }, { "epoch": 2.36, "learning_rate": 3.297270381603368e-05, "loss": 0.5948, "step": 39154 }, { "epoch": 2.36, "learning_rate": 3.2971194905919454e-05, "loss": 0.5617, "step": 39156 }, { "epoch": 2.36, "learning_rate": 3.2969685995805234e-05, "loss": 0.5825, "step": 39158 }, { "epoch": 2.36, "learning_rate": 3.296817708569101e-05, "loss": 0.4321, "step": 39160 }, { "epoch": 2.36, "learning_rate": 3.296666817557678e-05, "loss": 0.7936, "step": 39162 }, { "epoch": 2.36, "learning_rate": 3.296515926546256e-05, "loss": 0.7155, "step": 39164 }, { "epoch": 2.36, "learning_rate": 3.2963650355348333e-05, "loss": 0.4645, "step": 39166 }, { "epoch": 2.36, "learning_rate": 3.2962141445234107e-05, "loss": 0.412, "step": 39168 }, { "epoch": 2.36, "learning_rate": 3.2960632535119886e-05, "loss": 0.6313, "step": 39170 }, { "epoch": 2.36, "learning_rate": 3.295912362500566e-05, "loss": 0.5605, "step": 39172 }, { "epoch": 2.36, "learning_rate": 3.295761471489144e-05, "loss": 0.4949, "step": 39174 }, { "epoch": 2.36, "learning_rate": 3.295610580477721e-05, "loss": 0.6483, "step": 39176 }, { "epoch": 2.36, "learning_rate": 3.2954596894662986e-05, "loss": 0.6623, "step": 39178 }, { "epoch": 2.36, "learning_rate": 3.2953087984548766e-05, "loss": 0.7689, "step": 39180 }, { "epoch": 2.36, "learning_rate": 3.295157907443454e-05, "loss": 0.5788, "step": 39182 }, { "epoch": 2.37, "learning_rate": 3.295007016432031e-05, "loss": 0.4111, "step": 39184 }, { "epoch": 2.37, "learning_rate": 3.294856125420609e-05, "loss": 0.6077, "step": 39186 }, { "epoch": 2.37, "learning_rate": 3.2947052344091865e-05, "loss": 0.7822, "step": 39188 }, { "epoch": 2.37, "learning_rate": 3.294554343397764e-05, "loss": 0.3669, "step": 39190 }, { "epoch": 2.37, "learning_rate": 3.294403452386342e-05, "loss": 0.6244, "step": 39192 }, { "epoch": 2.37, "learning_rate": 3.294252561374919e-05, "loss": 0.7066, "step": 39194 }, { "epoch": 2.37, "learning_rate": 3.2941016703634964e-05, "loss": 0.7687, "step": 39196 }, { "epoch": 2.37, "learning_rate": 3.2939507793520744e-05, "loss": 0.7883, "step": 39198 }, { "epoch": 2.37, "learning_rate": 3.293799888340652e-05, "loss": 0.7849, "step": 39200 }, { "epoch": 2.37, "learning_rate": 3.29364899732923e-05, "loss": 0.7575, "step": 39202 }, { "epoch": 2.37, "learning_rate": 3.2934981063178063e-05, "loss": 0.6638, "step": 39204 }, { "epoch": 2.37, "learning_rate": 3.293347215306384e-05, "loss": 0.6052, "step": 39206 }, { "epoch": 2.37, "learning_rate": 3.293196324294962e-05, "loss": 0.5449, "step": 39208 }, { "epoch": 2.37, "learning_rate": 3.293045433283539e-05, "loss": 0.3783, "step": 39210 }, { "epoch": 2.37, "learning_rate": 3.292894542272117e-05, "loss": 0.6829, "step": 39212 }, { "epoch": 2.37, "learning_rate": 3.292743651260695e-05, "loss": 0.5909, "step": 39214 }, { "epoch": 2.37, "learning_rate": 3.292592760249272e-05, "loss": 0.5598, "step": 39216 }, { "epoch": 2.37, "learning_rate": 3.2924418692378496e-05, "loss": 0.6007, "step": 39218 }, { "epoch": 2.37, "learning_rate": 3.292290978226427e-05, "loss": 0.6066, "step": 39220 }, { "epoch": 2.37, "learning_rate": 3.292140087215005e-05, "loss": 0.4892, "step": 39222 }, { "epoch": 2.37, "learning_rate": 3.291989196203582e-05, "loss": 0.449, "step": 39224 }, { "epoch": 2.37, "learning_rate": 3.2918383051921595e-05, "loss": 0.5958, "step": 39226 }, { "epoch": 2.37, "learning_rate": 3.2916874141807375e-05, "loss": 0.5868, "step": 39228 }, { "epoch": 2.37, "learning_rate": 3.2915365231693155e-05, "loss": 0.6192, "step": 39230 }, { "epoch": 2.37, "learning_rate": 3.291385632157892e-05, "loss": 0.4225, "step": 39232 }, { "epoch": 2.37, "learning_rate": 3.29123474114647e-05, "loss": 0.5355, "step": 39234 }, { "epoch": 2.37, "learning_rate": 3.2910838501350474e-05, "loss": 0.6697, "step": 39236 }, { "epoch": 2.37, "learning_rate": 3.2909329591236254e-05, "loss": 0.7211, "step": 39238 }, { "epoch": 2.37, "learning_rate": 3.290782068112203e-05, "loss": 0.7099, "step": 39240 }, { "epoch": 2.37, "learning_rate": 3.29063117710078e-05, "loss": 0.4136, "step": 39242 }, { "epoch": 2.37, "learning_rate": 3.290480286089358e-05, "loss": 0.622, "step": 39244 }, { "epoch": 2.37, "learning_rate": 3.290329395077935e-05, "loss": 0.6028, "step": 39246 }, { "epoch": 2.37, "learning_rate": 3.2901785040665126e-05, "loss": 0.7432, "step": 39248 }, { "epoch": 2.37, "learning_rate": 3.2900276130550906e-05, "loss": 0.6101, "step": 39250 }, { "epoch": 2.37, "learning_rate": 3.289876722043668e-05, "loss": 0.4601, "step": 39252 }, { "epoch": 2.37, "learning_rate": 3.289725831032245e-05, "loss": 0.8774, "step": 39254 }, { "epoch": 2.37, "learning_rate": 3.289574940020823e-05, "loss": 0.7571, "step": 39256 }, { "epoch": 2.37, "learning_rate": 3.2894240490094006e-05, "loss": 0.585, "step": 39258 }, { "epoch": 2.37, "learning_rate": 3.289273157997978e-05, "loss": 0.4975, "step": 39260 }, { "epoch": 2.37, "learning_rate": 3.289122266986556e-05, "loss": 0.4309, "step": 39262 }, { "epoch": 2.37, "learning_rate": 3.288971375975133e-05, "loss": 0.4879, "step": 39264 }, { "epoch": 2.37, "learning_rate": 3.288820484963711e-05, "loss": 0.644, "step": 39266 }, { "epoch": 2.37, "learning_rate": 3.2886695939522885e-05, "loss": 0.52, "step": 39268 }, { "epoch": 2.37, "learning_rate": 3.288518702940866e-05, "loss": 0.6175, "step": 39270 }, { "epoch": 2.37, "learning_rate": 3.288367811929444e-05, "loss": 0.5829, "step": 39272 }, { "epoch": 2.37, "learning_rate": 3.288216920918021e-05, "loss": 0.5986, "step": 39274 }, { "epoch": 2.37, "learning_rate": 3.2880660299065984e-05, "loss": 0.4331, "step": 39276 }, { "epoch": 2.37, "learning_rate": 3.2879151388951764e-05, "loss": 0.4849, "step": 39278 }, { "epoch": 2.37, "learning_rate": 3.287764247883754e-05, "loss": 0.4043, "step": 39280 }, { "epoch": 2.37, "learning_rate": 3.287613356872331e-05, "loss": 0.5136, "step": 39282 }, { "epoch": 2.37, "learning_rate": 3.287462465860909e-05, "loss": 0.6107, "step": 39284 }, { "epoch": 2.37, "learning_rate": 3.287311574849486e-05, "loss": 0.4746, "step": 39286 }, { "epoch": 2.37, "learning_rate": 3.287160683838064e-05, "loss": 0.6032, "step": 39288 }, { "epoch": 2.37, "learning_rate": 3.2870097928266416e-05, "loss": 0.4222, "step": 39290 }, { "epoch": 2.37, "learning_rate": 3.286858901815219e-05, "loss": 0.5164, "step": 39292 }, { "epoch": 2.37, "learning_rate": 3.286708010803797e-05, "loss": 0.7212, "step": 39294 }, { "epoch": 2.37, "learning_rate": 3.286557119792374e-05, "loss": 0.8173, "step": 39296 }, { "epoch": 2.37, "learning_rate": 3.2864062287809516e-05, "loss": 0.4526, "step": 39298 }, { "epoch": 2.37, "learning_rate": 3.2862553377695295e-05, "loss": 0.5567, "step": 39300 }, { "epoch": 2.37, "learning_rate": 3.286104446758107e-05, "loss": 0.4908, "step": 39302 }, { "epoch": 2.37, "learning_rate": 3.285953555746684e-05, "loss": 0.6021, "step": 39304 }, { "epoch": 2.37, "learning_rate": 3.285802664735262e-05, "loss": 0.608, "step": 39306 }, { "epoch": 2.37, "learning_rate": 3.2856517737238395e-05, "loss": 0.641, "step": 39308 }, { "epoch": 2.37, "learning_rate": 3.285500882712417e-05, "loss": 0.5952, "step": 39310 }, { "epoch": 2.37, "learning_rate": 3.285349991700995e-05, "loss": 0.6042, "step": 39312 }, { "epoch": 2.37, "learning_rate": 3.285199100689572e-05, "loss": 0.5756, "step": 39314 }, { "epoch": 2.37, "learning_rate": 3.28504820967815e-05, "loss": 0.6123, "step": 39316 }, { "epoch": 2.37, "learning_rate": 3.284897318666727e-05, "loss": 0.563, "step": 39318 }, { "epoch": 2.37, "learning_rate": 3.284746427655305e-05, "loss": 0.5604, "step": 39320 }, { "epoch": 2.37, "learning_rate": 3.284595536643883e-05, "loss": 0.7674, "step": 39322 }, { "epoch": 2.37, "learning_rate": 3.284444645632459e-05, "loss": 0.4944, "step": 39324 }, { "epoch": 2.37, "learning_rate": 3.284293754621037e-05, "loss": 0.5676, "step": 39326 }, { "epoch": 2.37, "learning_rate": 3.284142863609615e-05, "loss": 0.5588, "step": 39328 }, { "epoch": 2.37, "learning_rate": 3.2839919725981926e-05, "loss": 0.5954, "step": 39330 }, { "epoch": 2.37, "learning_rate": 3.28384108158677e-05, "loss": 0.5582, "step": 39332 }, { "epoch": 2.37, "learning_rate": 3.283690190575347e-05, "loss": 0.5238, "step": 39334 }, { "epoch": 2.37, "learning_rate": 3.283539299563925e-05, "loss": 0.5099, "step": 39336 }, { "epoch": 2.37, "learning_rate": 3.283388408552503e-05, "loss": 0.446, "step": 39338 }, { "epoch": 2.37, "learning_rate": 3.28323751754108e-05, "loss": 0.5894, "step": 39340 }, { "epoch": 2.37, "learning_rate": 3.283086626529658e-05, "loss": 0.483, "step": 39342 }, { "epoch": 2.37, "learning_rate": 3.282935735518236e-05, "loss": 0.4649, "step": 39344 }, { "epoch": 2.37, "learning_rate": 3.2827848445068125e-05, "loss": 0.5289, "step": 39346 }, { "epoch": 2.37, "learning_rate": 3.2826339534953905e-05, "loss": 0.7422, "step": 39348 }, { "epoch": 2.38, "learning_rate": 3.282483062483968e-05, "loss": 0.5301, "step": 39350 }, { "epoch": 2.38, "learning_rate": 3.282332171472546e-05, "loss": 0.41, "step": 39352 }, { "epoch": 2.38, "learning_rate": 3.282181280461123e-05, "loss": 0.4675, "step": 39354 }, { "epoch": 2.38, "learning_rate": 3.2820303894497004e-05, "loss": 0.6273, "step": 39356 }, { "epoch": 2.38, "learning_rate": 3.2818794984382784e-05, "loss": 0.4589, "step": 39358 }, { "epoch": 2.38, "learning_rate": 3.281728607426856e-05, "loss": 0.3292, "step": 39360 }, { "epoch": 2.38, "learning_rate": 3.281577716415433e-05, "loss": 0.686, "step": 39362 }, { "epoch": 2.38, "learning_rate": 3.281426825404011e-05, "loss": 0.6636, "step": 39364 }, { "epoch": 2.38, "learning_rate": 3.281275934392588e-05, "loss": 0.4586, "step": 39366 }, { "epoch": 2.38, "learning_rate": 3.2811250433811656e-05, "loss": 0.5818, "step": 39368 }, { "epoch": 2.38, "learning_rate": 3.2809741523697436e-05, "loss": 0.3472, "step": 39370 }, { "epoch": 2.38, "learning_rate": 3.280823261358321e-05, "loss": 0.9023, "step": 39372 }, { "epoch": 2.38, "learning_rate": 3.280672370346898e-05, "loss": 0.6545, "step": 39374 }, { "epoch": 2.38, "learning_rate": 3.280521479335476e-05, "loss": 0.4256, "step": 39376 }, { "epoch": 2.38, "learning_rate": 3.2803705883240535e-05, "loss": 0.6163, "step": 39378 }, { "epoch": 2.38, "learning_rate": 3.2802196973126315e-05, "loss": 0.562, "step": 39380 }, { "epoch": 2.38, "learning_rate": 3.280068806301209e-05, "loss": 0.7094, "step": 39382 }, { "epoch": 2.38, "learning_rate": 3.279917915289786e-05, "loss": 0.5044, "step": 39384 }, { "epoch": 2.38, "learning_rate": 3.279767024278364e-05, "loss": 0.4684, "step": 39386 }, { "epoch": 2.38, "learning_rate": 3.2796161332669415e-05, "loss": 0.5791, "step": 39388 }, { "epoch": 2.38, "learning_rate": 3.279465242255519e-05, "loss": 0.836, "step": 39390 }, { "epoch": 2.38, "learning_rate": 3.279314351244097e-05, "loss": 0.3681, "step": 39392 }, { "epoch": 2.38, "learning_rate": 3.279163460232674e-05, "loss": 0.4976, "step": 39394 }, { "epoch": 2.38, "learning_rate": 3.2790125692212514e-05, "loss": 0.4072, "step": 39396 }, { "epoch": 2.38, "learning_rate": 3.2788616782098294e-05, "loss": 0.6693, "step": 39398 }, { "epoch": 2.38, "learning_rate": 3.278710787198407e-05, "loss": 0.5099, "step": 39400 }, { "epoch": 2.38, "learning_rate": 3.278559896186985e-05, "loss": 0.4607, "step": 39402 }, { "epoch": 2.38, "learning_rate": 3.278409005175562e-05, "loss": 0.8376, "step": 39404 }, { "epoch": 2.38, "learning_rate": 3.278258114164139e-05, "loss": 0.7255, "step": 39406 }, { "epoch": 2.38, "learning_rate": 3.278107223152717e-05, "loss": 0.5687, "step": 39408 }, { "epoch": 2.38, "learning_rate": 3.277956332141294e-05, "loss": 0.7877, "step": 39410 }, { "epoch": 2.38, "learning_rate": 3.277805441129872e-05, "loss": 0.7601, "step": 39412 }, { "epoch": 2.38, "learning_rate": 3.27765455011845e-05, "loss": 0.5874, "step": 39414 }, { "epoch": 2.38, "learning_rate": 3.277503659107027e-05, "loss": 0.599, "step": 39416 }, { "epoch": 2.38, "learning_rate": 3.2773527680956045e-05, "loss": 0.5275, "step": 39418 }, { "epoch": 2.38, "learning_rate": 3.2772018770841825e-05, "loss": 0.4946, "step": 39420 }, { "epoch": 2.38, "learning_rate": 3.27705098607276e-05, "loss": 0.667, "step": 39422 }, { "epoch": 2.38, "learning_rate": 3.276900095061337e-05, "loss": 0.3709, "step": 39424 }, { "epoch": 2.38, "learning_rate": 3.276749204049915e-05, "loss": 0.5964, "step": 39426 }, { "epoch": 2.38, "learning_rate": 3.2765983130384925e-05, "loss": 0.5518, "step": 39428 }, { "epoch": 2.38, "learning_rate": 3.2764474220270704e-05, "loss": 0.6529, "step": 39430 }, { "epoch": 2.38, "learning_rate": 3.276296531015647e-05, "loss": 0.4573, "step": 39432 }, { "epoch": 2.38, "learning_rate": 3.276145640004225e-05, "loss": 0.5904, "step": 39434 }, { "epoch": 2.38, "learning_rate": 3.275994748992803e-05, "loss": 0.4769, "step": 39436 }, { "epoch": 2.38, "learning_rate": 3.27584385798138e-05, "loss": 0.6983, "step": 39438 }, { "epoch": 2.38, "learning_rate": 3.275692966969958e-05, "loss": 0.5673, "step": 39440 }, { "epoch": 2.38, "learning_rate": 3.275542075958536e-05, "loss": 0.4107, "step": 39442 }, { "epoch": 2.38, "learning_rate": 3.275391184947113e-05, "loss": 0.4656, "step": 39444 }, { "epoch": 2.38, "learning_rate": 3.27524029393569e-05, "loss": 0.5734, "step": 39446 }, { "epoch": 2.38, "learning_rate": 3.2750894029242676e-05, "loss": 0.7851, "step": 39448 }, { "epoch": 2.38, "learning_rate": 3.2749385119128456e-05, "loss": 0.4313, "step": 39450 }, { "epoch": 2.38, "learning_rate": 3.2747876209014236e-05, "loss": 0.5517, "step": 39452 }, { "epoch": 2.38, "learning_rate": 3.27463672989e-05, "loss": 0.6329, "step": 39454 }, { "epoch": 2.38, "learning_rate": 3.274485838878578e-05, "loss": 0.6321, "step": 39456 }, { "epoch": 2.38, "learning_rate": 3.274334947867156e-05, "loss": 0.5616, "step": 39458 }, { "epoch": 2.38, "learning_rate": 3.274184056855733e-05, "loss": 0.4109, "step": 39460 }, { "epoch": 2.38, "learning_rate": 3.274033165844311e-05, "loss": 0.8019, "step": 39462 }, { "epoch": 2.38, "learning_rate": 3.273882274832888e-05, "loss": 0.603, "step": 39464 }, { "epoch": 2.38, "learning_rate": 3.273731383821466e-05, "loss": 0.8757, "step": 39466 }, { "epoch": 2.38, "learning_rate": 3.2735804928100435e-05, "loss": 0.5156, "step": 39468 }, { "epoch": 2.38, "learning_rate": 3.273429601798621e-05, "loss": 0.5025, "step": 39470 }, { "epoch": 2.38, "learning_rate": 3.273278710787199e-05, "loss": 0.4758, "step": 39472 }, { "epoch": 2.38, "learning_rate": 3.273127819775776e-05, "loss": 0.5887, "step": 39474 }, { "epoch": 2.38, "learning_rate": 3.2729769287643534e-05, "loss": 0.6858, "step": 39476 }, { "epoch": 2.38, "learning_rate": 3.2728260377529314e-05, "loss": 0.6187, "step": 39478 }, { "epoch": 2.38, "learning_rate": 3.272675146741509e-05, "loss": 0.5537, "step": 39480 }, { "epoch": 2.38, "learning_rate": 3.272524255730086e-05, "loss": 0.6346, "step": 39482 }, { "epoch": 2.38, "learning_rate": 3.272373364718664e-05, "loss": 0.4247, "step": 39484 }, { "epoch": 2.38, "learning_rate": 3.272222473707241e-05, "loss": 0.5348, "step": 39486 }, { "epoch": 2.38, "learning_rate": 3.2720715826958186e-05, "loss": 0.4583, "step": 39488 }, { "epoch": 2.38, "learning_rate": 3.2719206916843966e-05, "loss": 0.5669, "step": 39490 }, { "epoch": 2.38, "learning_rate": 3.271769800672974e-05, "loss": 0.4633, "step": 39492 }, { "epoch": 2.38, "learning_rate": 3.271618909661552e-05, "loss": 0.6935, "step": 39494 }, { "epoch": 2.38, "learning_rate": 3.271468018650129e-05, "loss": 0.5907, "step": 39496 }, { "epoch": 2.38, "learning_rate": 3.2713171276387065e-05, "loss": 0.4747, "step": 39498 }, { "epoch": 2.38, "learning_rate": 3.2711662366272845e-05, "loss": 0.589, "step": 39500 }, { "epoch": 2.38, "learning_rate": 3.271015345615862e-05, "loss": 1.0252, "step": 39502 }, { "epoch": 2.38, "learning_rate": 3.270864454604439e-05, "loss": 0.4717, "step": 39504 }, { "epoch": 2.38, "learning_rate": 3.270713563593017e-05, "loss": 0.5827, "step": 39506 }, { "epoch": 2.38, "learning_rate": 3.2705626725815944e-05, "loss": 0.6353, "step": 39508 }, { "epoch": 2.38, "learning_rate": 3.270411781570172e-05, "loss": 0.5708, "step": 39510 }, { "epoch": 2.38, "learning_rate": 3.27026089055875e-05, "loss": 0.6434, "step": 39512 }, { "epoch": 2.38, "learning_rate": 3.270109999547327e-05, "loss": 0.5624, "step": 39514 }, { "epoch": 2.39, "learning_rate": 3.269959108535905e-05, "loss": 0.8164, "step": 39516 }, { "epoch": 2.39, "learning_rate": 3.2698082175244824e-05, "loss": 0.5917, "step": 39518 }, { "epoch": 2.39, "learning_rate": 3.26965732651306e-05, "loss": 0.6266, "step": 39520 }, { "epoch": 2.39, "learning_rate": 3.269506435501638e-05, "loss": 0.6978, "step": 39522 }, { "epoch": 2.39, "learning_rate": 3.269355544490214e-05, "loss": 0.5814, "step": 39524 }, { "epoch": 2.39, "learning_rate": 3.269204653478792e-05, "loss": 0.4598, "step": 39526 }, { "epoch": 2.39, "learning_rate": 3.26905376246737e-05, "loss": 0.4707, "step": 39528 }, { "epoch": 2.39, "learning_rate": 3.2689028714559476e-05, "loss": 0.4917, "step": 39530 }, { "epoch": 2.39, "learning_rate": 3.268751980444525e-05, "loss": 0.6386, "step": 39532 }, { "epoch": 2.39, "learning_rate": 3.268601089433103e-05, "loss": 0.6363, "step": 39534 }, { "epoch": 2.39, "learning_rate": 3.26845019842168e-05, "loss": 0.5724, "step": 39536 }, { "epoch": 2.39, "learning_rate": 3.2682993074102575e-05, "loss": 0.4336, "step": 39538 }, { "epoch": 2.39, "learning_rate": 3.2681484163988355e-05, "loss": 0.6727, "step": 39540 }, { "epoch": 2.39, "learning_rate": 3.267997525387413e-05, "loss": 0.6919, "step": 39542 }, { "epoch": 2.39, "learning_rate": 3.267846634375991e-05, "loss": 0.6181, "step": 39544 }, { "epoch": 2.39, "learning_rate": 3.2676957433645675e-05, "loss": 0.614, "step": 39546 }, { "epoch": 2.39, "learning_rate": 3.2675448523531454e-05, "loss": 0.6753, "step": 39548 }, { "epoch": 2.39, "learning_rate": 3.2673939613417234e-05, "loss": 0.4695, "step": 39550 }, { "epoch": 2.39, "learning_rate": 3.2672430703303e-05, "loss": 0.851, "step": 39552 }, { "epoch": 2.39, "learning_rate": 3.267092179318878e-05, "loss": 0.4911, "step": 39554 }, { "epoch": 2.39, "learning_rate": 3.266941288307456e-05, "loss": 0.4138, "step": 39556 }, { "epoch": 2.39, "learning_rate": 3.2667903972960334e-05, "loss": 0.5685, "step": 39558 }, { "epoch": 2.39, "learning_rate": 3.266639506284611e-05, "loss": 0.4395, "step": 39560 }, { "epoch": 2.39, "learning_rate": 3.266488615273188e-05, "loss": 0.7365, "step": 39562 }, { "epoch": 2.39, "learning_rate": 3.266337724261766e-05, "loss": 0.4763, "step": 39564 }, { "epoch": 2.39, "learning_rate": 3.266186833250344e-05, "loss": 0.6293, "step": 39566 }, { "epoch": 2.39, "learning_rate": 3.2660359422389206e-05, "loss": 0.6325, "step": 39568 }, { "epoch": 2.39, "learning_rate": 3.2658850512274986e-05, "loss": 0.5451, "step": 39570 }, { "epoch": 2.39, "learning_rate": 3.2657341602160766e-05, "loss": 0.6295, "step": 39572 }, { "epoch": 2.39, "learning_rate": 3.265583269204653e-05, "loss": 0.445, "step": 39574 }, { "epoch": 2.39, "learning_rate": 3.265432378193231e-05, "loss": 0.7631, "step": 39576 }, { "epoch": 2.39, "learning_rate": 3.2652814871818085e-05, "loss": 0.4147, "step": 39578 }, { "epoch": 2.39, "learning_rate": 3.2651305961703865e-05, "loss": 0.6333, "step": 39580 }, { "epoch": 2.39, "learning_rate": 3.264979705158964e-05, "loss": 0.4471, "step": 39582 }, { "epoch": 2.39, "learning_rate": 3.264828814147541e-05, "loss": 0.7876, "step": 39584 }, { "epoch": 2.39, "learning_rate": 3.264677923136119e-05, "loss": 0.6404, "step": 39586 }, { "epoch": 2.39, "learning_rate": 3.2645270321246964e-05, "loss": 0.6069, "step": 39588 }, { "epoch": 2.39, "learning_rate": 3.264376141113274e-05, "loss": 0.4293, "step": 39590 }, { "epoch": 2.39, "learning_rate": 3.264225250101852e-05, "loss": 0.568, "step": 39592 }, { "epoch": 2.39, "learning_rate": 3.264074359090429e-05, "loss": 0.5536, "step": 39594 }, { "epoch": 2.39, "learning_rate": 3.2639234680790064e-05, "loss": 0.5835, "step": 39596 }, { "epoch": 2.39, "learning_rate": 3.2637725770675844e-05, "loss": 0.543, "step": 39598 }, { "epoch": 2.39, "learning_rate": 3.263621686056162e-05, "loss": 0.6281, "step": 39600 }, { "epoch": 2.39, "learning_rate": 3.263470795044739e-05, "loss": 0.7731, "step": 39602 }, { "epoch": 2.39, "learning_rate": 3.263319904033317e-05, "loss": 0.3672, "step": 39604 }, { "epoch": 2.39, "learning_rate": 3.263169013021894e-05, "loss": 0.6153, "step": 39606 }, { "epoch": 2.39, "learning_rate": 3.263018122010472e-05, "loss": 0.5393, "step": 39608 }, { "epoch": 2.39, "learning_rate": 3.2628672309990496e-05, "loss": 0.6218, "step": 39610 }, { "epoch": 2.39, "learning_rate": 3.262716339987627e-05, "loss": 0.4876, "step": 39612 }, { "epoch": 2.39, "learning_rate": 3.262565448976205e-05, "loss": 0.4493, "step": 39614 }, { "epoch": 2.39, "learning_rate": 3.262414557964782e-05, "loss": 0.4324, "step": 39616 }, { "epoch": 2.39, "learning_rate": 3.2622636669533595e-05, "loss": 0.8025, "step": 39618 }, { "epoch": 2.39, "learning_rate": 3.2621127759419375e-05, "loss": 0.6972, "step": 39620 }, { "epoch": 2.39, "learning_rate": 3.261961884930515e-05, "loss": 0.7002, "step": 39622 }, { "epoch": 2.39, "learning_rate": 3.261810993919092e-05, "loss": 0.5729, "step": 39624 }, { "epoch": 2.39, "learning_rate": 3.26166010290767e-05, "loss": 0.5495, "step": 39626 }, { "epoch": 2.39, "learning_rate": 3.2615092118962474e-05, "loss": 0.5307, "step": 39628 }, { "epoch": 2.39, "learning_rate": 3.2613583208848254e-05, "loss": 0.8242, "step": 39630 }, { "epoch": 2.39, "learning_rate": 3.261207429873403e-05, "loss": 0.4172, "step": 39632 }, { "epoch": 2.39, "learning_rate": 3.26105653886198e-05, "loss": 0.4466, "step": 39634 }, { "epoch": 2.39, "learning_rate": 3.260905647850558e-05, "loss": 0.3534, "step": 39636 }, { "epoch": 2.39, "learning_rate": 3.260754756839135e-05, "loss": 0.5655, "step": 39638 }, { "epoch": 2.39, "learning_rate": 3.2606038658277127e-05, "loss": 0.5692, "step": 39640 }, { "epoch": 2.39, "learning_rate": 3.2604529748162907e-05, "loss": 0.6973, "step": 39642 }, { "epoch": 2.39, "learning_rate": 3.260302083804868e-05, "loss": 0.5198, "step": 39644 }, { "epoch": 2.39, "learning_rate": 3.260151192793445e-05, "loss": 0.4318, "step": 39646 }, { "epoch": 2.39, "learning_rate": 3.260000301782023e-05, "loss": 0.7118, "step": 39648 }, { "epoch": 2.39, "learning_rate": 3.2598494107706006e-05, "loss": 0.5639, "step": 39650 }, { "epoch": 2.39, "learning_rate": 3.259698519759178e-05, "loss": 0.573, "step": 39652 }, { "epoch": 2.39, "learning_rate": 3.259547628747756e-05, "loss": 0.4298, "step": 39654 }, { "epoch": 2.39, "learning_rate": 3.259396737736333e-05, "loss": 0.5586, "step": 39656 }, { "epoch": 2.39, "learning_rate": 3.259245846724911e-05, "loss": 0.7877, "step": 39658 }, { "epoch": 2.39, "learning_rate": 3.259094955713488e-05, "loss": 0.3762, "step": 39660 }, { "epoch": 2.39, "learning_rate": 3.258944064702066e-05, "loss": 0.4739, "step": 39662 }, { "epoch": 2.39, "learning_rate": 3.258793173690644e-05, "loss": 0.5819, "step": 39664 }, { "epoch": 2.39, "learning_rate": 3.258642282679221e-05, "loss": 0.5105, "step": 39666 }, { "epoch": 2.39, "learning_rate": 3.2584913916677984e-05, "loss": 0.5685, "step": 39668 }, { "epoch": 2.39, "learning_rate": 3.2583405006563764e-05, "loss": 0.7416, "step": 39670 }, { "epoch": 2.39, "learning_rate": 3.258189609644954e-05, "loss": 0.5637, "step": 39672 }, { "epoch": 2.39, "learning_rate": 3.258038718633531e-05, "loss": 0.5562, "step": 39674 }, { "epoch": 2.39, "learning_rate": 3.2578878276221084e-05, "loss": 0.4586, "step": 39676 }, { "epoch": 2.39, "learning_rate": 3.2577369366106863e-05, "loss": 0.9963, "step": 39678 }, { "epoch": 2.39, "learning_rate": 3.257586045599264e-05, "loss": 0.5267, "step": 39680 }, { "epoch": 2.4, "learning_rate": 3.257435154587841e-05, "loss": 0.5015, "step": 39682 }, { "epoch": 2.4, "learning_rate": 3.257284263576419e-05, "loss": 0.5653, "step": 39684 }, { "epoch": 2.4, "learning_rate": 3.257133372564997e-05, "loss": 0.5745, "step": 39686 }, { "epoch": 2.4, "learning_rate": 3.2569824815535736e-05, "loss": 0.4099, "step": 39688 }, { "epoch": 2.4, "learning_rate": 3.2568315905421516e-05, "loss": 0.3638, "step": 39690 }, { "epoch": 2.4, "learning_rate": 3.256680699530729e-05, "loss": 0.55, "step": 39692 }, { "epoch": 2.4, "learning_rate": 3.256529808519307e-05, "loss": 0.7102, "step": 39694 }, { "epoch": 2.4, "learning_rate": 3.256378917507884e-05, "loss": 0.6693, "step": 39696 }, { "epoch": 2.4, "learning_rate": 3.2562280264964615e-05, "loss": 0.7442, "step": 39698 }, { "epoch": 2.4, "learning_rate": 3.2560771354850395e-05, "loss": 0.5384, "step": 39700 }, { "epoch": 2.4, "learning_rate": 3.255926244473617e-05, "loss": 0.5389, "step": 39702 }, { "epoch": 2.4, "learning_rate": 3.255775353462194e-05, "loss": 0.5757, "step": 39704 }, { "epoch": 2.4, "learning_rate": 3.255624462450772e-05, "loss": 0.701, "step": 39706 }, { "epoch": 2.4, "learning_rate": 3.2554735714393494e-05, "loss": 0.7161, "step": 39708 }, { "epoch": 2.4, "learning_rate": 3.255322680427927e-05, "loss": 0.7048, "step": 39710 }, { "epoch": 2.4, "learning_rate": 3.255171789416505e-05, "loss": 0.695, "step": 39712 }, { "epoch": 2.4, "learning_rate": 3.255020898405082e-05, "loss": 0.8264, "step": 39714 }, { "epoch": 2.4, "learning_rate": 3.25487000739366e-05, "loss": 0.6596, "step": 39716 }, { "epoch": 2.4, "learning_rate": 3.254719116382237e-05, "loss": 0.3128, "step": 39718 }, { "epoch": 2.4, "learning_rate": 3.2545682253708146e-05, "loss": 0.4569, "step": 39720 }, { "epoch": 2.4, "learning_rate": 3.2544173343593926e-05, "loss": 0.3495, "step": 39722 }, { "epoch": 2.4, "learning_rate": 3.25426644334797e-05, "loss": 0.7718, "step": 39724 }, { "epoch": 2.4, "learning_rate": 3.254115552336547e-05, "loss": 0.5255, "step": 39726 }, { "epoch": 2.4, "learning_rate": 3.253964661325125e-05, "loss": 0.8193, "step": 39728 }, { "epoch": 2.4, "learning_rate": 3.2538137703137026e-05, "loss": 0.6209, "step": 39730 }, { "epoch": 2.4, "learning_rate": 3.25366287930228e-05, "loss": 0.576, "step": 39732 }, { "epoch": 2.4, "learning_rate": 3.253511988290858e-05, "loss": 0.8121, "step": 39734 }, { "epoch": 2.4, "learning_rate": 3.253361097279435e-05, "loss": 0.6494, "step": 39736 }, { "epoch": 2.4, "learning_rate": 3.2532102062680125e-05, "loss": 0.6759, "step": 39738 }, { "epoch": 2.4, "learning_rate": 3.2530593152565905e-05, "loss": 0.5449, "step": 39740 }, { "epoch": 2.4, "learning_rate": 3.252908424245168e-05, "loss": 0.7316, "step": 39742 }, { "epoch": 2.4, "learning_rate": 3.252757533233746e-05, "loss": 0.5432, "step": 39744 }, { "epoch": 2.4, "learning_rate": 3.252606642222323e-05, "loss": 0.6562, "step": 39746 }, { "epoch": 2.4, "learning_rate": 3.2524557512109004e-05, "loss": 0.5351, "step": 39748 }, { "epoch": 2.4, "learning_rate": 3.2523048601994784e-05, "loss": 0.4503, "step": 39750 }, { "epoch": 2.4, "learning_rate": 3.252153969188055e-05, "loss": 0.7803, "step": 39752 }, { "epoch": 2.4, "learning_rate": 3.252003078176633e-05, "loss": 0.4378, "step": 39754 }, { "epoch": 2.4, "learning_rate": 3.251852187165211e-05, "loss": 1.0179, "step": 39756 }, { "epoch": 2.4, "learning_rate": 3.251701296153788e-05, "loss": 0.672, "step": 39758 }, { "epoch": 2.4, "learning_rate": 3.2515504051423656e-05, "loss": 0.4829, "step": 39760 }, { "epoch": 2.4, "learning_rate": 3.2513995141309436e-05, "loss": 0.729, "step": 39762 }, { "epoch": 2.4, "learning_rate": 3.251248623119521e-05, "loss": 0.5689, "step": 39764 }, { "epoch": 2.4, "learning_rate": 3.251097732108098e-05, "loss": 0.5561, "step": 39766 }, { "epoch": 2.4, "learning_rate": 3.250946841096676e-05, "loss": 0.5579, "step": 39768 }, { "epoch": 2.4, "learning_rate": 3.2507959500852536e-05, "loss": 0.5796, "step": 39770 }, { "epoch": 2.4, "learning_rate": 3.2506450590738316e-05, "loss": 0.679, "step": 39772 }, { "epoch": 2.4, "learning_rate": 3.250494168062408e-05, "loss": 0.7143, "step": 39774 }, { "epoch": 2.4, "learning_rate": 3.250343277050986e-05, "loss": 0.6867, "step": 39776 }, { "epoch": 2.4, "learning_rate": 3.250192386039564e-05, "loss": 0.6187, "step": 39778 }, { "epoch": 2.4, "learning_rate": 3.2500414950281415e-05, "loss": 0.3945, "step": 39780 }, { "epoch": 2.4, "learning_rate": 3.249890604016719e-05, "loss": 0.6154, "step": 39782 }, { "epoch": 2.4, "learning_rate": 3.249739713005297e-05, "loss": 0.6204, "step": 39784 }, { "epoch": 2.4, "learning_rate": 3.249588821993874e-05, "loss": 0.6641, "step": 39786 }, { "epoch": 2.4, "learning_rate": 3.2494379309824514e-05, "loss": 0.5259, "step": 39788 }, { "epoch": 2.4, "learning_rate": 3.249287039971029e-05, "loss": 0.4798, "step": 39790 }, { "epoch": 2.4, "learning_rate": 3.249136148959607e-05, "loss": 0.5354, "step": 39792 }, { "epoch": 2.4, "learning_rate": 3.248985257948185e-05, "loss": 0.3449, "step": 39794 }, { "epoch": 2.4, "learning_rate": 3.248834366936761e-05, "loss": 0.5925, "step": 39796 }, { "epoch": 2.4, "learning_rate": 3.248683475925339e-05, "loss": 0.618, "step": 39798 }, { "epoch": 2.4, "learning_rate": 3.248532584913917e-05, "loss": 0.4544, "step": 39800 }, { "epoch": 2.4, "learning_rate": 3.248381693902494e-05, "loss": 0.677, "step": 39802 }, { "epoch": 2.4, "learning_rate": 3.248230802891072e-05, "loss": 0.577, "step": 39804 }, { "epoch": 2.4, "learning_rate": 3.248079911879649e-05, "loss": 0.4773, "step": 39806 }, { "epoch": 2.4, "learning_rate": 3.247929020868227e-05, "loss": 0.6481, "step": 39808 }, { "epoch": 2.4, "learning_rate": 3.2477781298568046e-05, "loss": 0.5781, "step": 39810 }, { "epoch": 2.4, "learning_rate": 3.247627238845382e-05, "loss": 0.4002, "step": 39812 }, { "epoch": 2.4, "learning_rate": 3.24747634783396e-05, "loss": 0.637, "step": 39814 }, { "epoch": 2.4, "learning_rate": 3.247325456822537e-05, "loss": 0.5902, "step": 39816 }, { "epoch": 2.4, "learning_rate": 3.2471745658111145e-05, "loss": 0.5632, "step": 39818 }, { "epoch": 2.4, "learning_rate": 3.2470236747996925e-05, "loss": 0.5431, "step": 39820 }, { "epoch": 2.4, "learning_rate": 3.24687278378827e-05, "loss": 0.5019, "step": 39822 }, { "epoch": 2.4, "learning_rate": 3.246721892776847e-05, "loss": 0.7796, "step": 39824 }, { "epoch": 2.4, "learning_rate": 3.246571001765425e-05, "loss": 0.6181, "step": 39826 }, { "epoch": 2.4, "learning_rate": 3.2464201107540024e-05, "loss": 0.5788, "step": 39828 }, { "epoch": 2.4, "learning_rate": 3.2462692197425804e-05, "loss": 0.4766, "step": 39830 }, { "epoch": 2.4, "learning_rate": 3.246118328731158e-05, "loss": 0.5438, "step": 39832 }, { "epoch": 2.4, "learning_rate": 3.245967437719735e-05, "loss": 0.5384, "step": 39834 }, { "epoch": 2.4, "learning_rate": 3.245816546708313e-05, "loss": 0.572, "step": 39836 }, { "epoch": 2.4, "learning_rate": 3.24566565569689e-05, "loss": 0.6752, "step": 39838 }, { "epoch": 2.4, "learning_rate": 3.2455147646854676e-05, "loss": 0.6184, "step": 39840 }, { "epoch": 2.4, "learning_rate": 3.2453638736740456e-05, "loss": 0.5857, "step": 39842 }, { "epoch": 2.4, "learning_rate": 3.245212982662623e-05, "loss": 0.5589, "step": 39844 }, { "epoch": 2.4, "learning_rate": 3.2450620916512e-05, "loss": 0.417, "step": 39846 }, { "epoch": 2.41, "learning_rate": 3.244911200639778e-05, "loss": 0.4366, "step": 39848 }, { "epoch": 2.41, "learning_rate": 3.2447603096283556e-05, "loss": 0.8687, "step": 39850 }, { "epoch": 2.41, "learning_rate": 3.244609418616933e-05, "loss": 0.4281, "step": 39852 }, { "epoch": 2.41, "learning_rate": 3.244458527605511e-05, "loss": 0.6967, "step": 39854 }, { "epoch": 2.41, "learning_rate": 3.244307636594088e-05, "loss": 0.4404, "step": 39856 }, { "epoch": 2.41, "learning_rate": 3.244156745582666e-05, "loss": 0.5455, "step": 39858 }, { "epoch": 2.41, "learning_rate": 3.2440058545712435e-05, "loss": 0.5617, "step": 39860 }, { "epoch": 2.41, "learning_rate": 3.243854963559821e-05, "loss": 0.5738, "step": 39862 }, { "epoch": 2.41, "learning_rate": 3.243704072548399e-05, "loss": 0.3505, "step": 39864 }, { "epoch": 2.41, "learning_rate": 3.2435531815369754e-05, "loss": 0.5933, "step": 39866 }, { "epoch": 2.41, "learning_rate": 3.2434022905255534e-05, "loss": 0.4436, "step": 39868 }, { "epoch": 2.41, "learning_rate": 3.2432513995141314e-05, "loss": 0.6491, "step": 39870 }, { "epoch": 2.41, "learning_rate": 3.243100508502709e-05, "loss": 0.7331, "step": 39872 }, { "epoch": 2.41, "learning_rate": 3.242949617491286e-05, "loss": 0.5712, "step": 39874 }, { "epoch": 2.41, "learning_rate": 3.242798726479864e-05, "loss": 0.5077, "step": 39876 }, { "epoch": 2.41, "learning_rate": 3.242647835468441e-05, "loss": 0.4574, "step": 39878 }, { "epoch": 2.41, "learning_rate": 3.242496944457019e-05, "loss": 0.6446, "step": 39880 }, { "epoch": 2.41, "learning_rate": 3.242346053445596e-05, "loss": 0.4677, "step": 39882 }, { "epoch": 2.41, "learning_rate": 3.242195162434174e-05, "loss": 0.6161, "step": 39884 }, { "epoch": 2.41, "learning_rate": 3.242044271422752e-05, "loss": 0.6541, "step": 39886 }, { "epoch": 2.41, "learning_rate": 3.2418933804113286e-05, "loss": 0.6278, "step": 39888 }, { "epoch": 2.41, "learning_rate": 3.2417424893999065e-05, "loss": 0.4156, "step": 39890 }, { "epoch": 2.41, "learning_rate": 3.2415915983884845e-05, "loss": 0.7343, "step": 39892 }, { "epoch": 2.41, "learning_rate": 3.241440707377062e-05, "loss": 0.5151, "step": 39894 }, { "epoch": 2.41, "learning_rate": 3.241289816365639e-05, "loss": 0.516, "step": 39896 }, { "epoch": 2.41, "learning_rate": 3.241138925354217e-05, "loss": 0.8144, "step": 39898 }, { "epoch": 2.41, "learning_rate": 3.2409880343427945e-05, "loss": 0.5162, "step": 39900 }, { "epoch": 2.41, "learning_rate": 3.240837143331372e-05, "loss": 0.7194, "step": 39902 }, { "epoch": 2.41, "learning_rate": 3.240686252319949e-05, "loss": 0.41, "step": 39904 }, { "epoch": 2.41, "learning_rate": 3.240535361308527e-05, "loss": 0.5045, "step": 39906 }, { "epoch": 2.41, "learning_rate": 3.240384470297105e-05, "loss": 0.7563, "step": 39908 }, { "epoch": 2.41, "learning_rate": 3.240233579285682e-05, "loss": 0.6861, "step": 39910 }, { "epoch": 2.41, "learning_rate": 3.24008268827426e-05, "loss": 0.6581, "step": 39912 }, { "epoch": 2.41, "learning_rate": 3.239931797262838e-05, "loss": 0.5647, "step": 39914 }, { "epoch": 2.41, "learning_rate": 3.239780906251414e-05, "loss": 0.4798, "step": 39916 }, { "epoch": 2.41, "learning_rate": 3.239630015239992e-05, "loss": 0.5701, "step": 39918 }, { "epoch": 2.41, "learning_rate": 3.2394791242285696e-05, "loss": 0.4397, "step": 39920 }, { "epoch": 2.41, "learning_rate": 3.2393282332171476e-05, "loss": 0.4914, "step": 39922 }, { "epoch": 2.41, "learning_rate": 3.239177342205725e-05, "loss": 0.8389, "step": 39924 }, { "epoch": 2.41, "learning_rate": 3.239026451194302e-05, "loss": 0.4447, "step": 39926 }, { "epoch": 2.41, "learning_rate": 3.23887556018288e-05, "loss": 0.5879, "step": 39928 }, { "epoch": 2.41, "learning_rate": 3.2387246691714575e-05, "loss": 0.7412, "step": 39930 }, { "epoch": 2.41, "learning_rate": 3.238573778160035e-05, "loss": 0.5798, "step": 39932 }, { "epoch": 2.41, "learning_rate": 3.238422887148613e-05, "loss": 0.5384, "step": 39934 }, { "epoch": 2.41, "learning_rate": 3.23827199613719e-05, "loss": 0.4726, "step": 39936 }, { "epoch": 2.41, "learning_rate": 3.2381211051257675e-05, "loss": 0.4803, "step": 39938 }, { "epoch": 2.41, "learning_rate": 3.2379702141143455e-05, "loss": 0.5436, "step": 39940 }, { "epoch": 2.41, "learning_rate": 3.237819323102923e-05, "loss": 0.4079, "step": 39942 }, { "epoch": 2.41, "learning_rate": 3.237668432091501e-05, "loss": 0.5967, "step": 39944 }, { "epoch": 2.41, "learning_rate": 3.237517541080078e-05, "loss": 0.3582, "step": 39946 }, { "epoch": 2.41, "learning_rate": 3.2373666500686554e-05, "loss": 0.5414, "step": 39948 }, { "epoch": 2.41, "learning_rate": 3.2372157590572334e-05, "loss": 0.5753, "step": 39950 }, { "epoch": 2.41, "learning_rate": 3.237064868045811e-05, "loss": 0.5315, "step": 39952 }, { "epoch": 2.41, "learning_rate": 3.236913977034388e-05, "loss": 0.5231, "step": 39954 }, { "epoch": 2.41, "learning_rate": 3.236763086022966e-05, "loss": 0.5322, "step": 39956 }, { "epoch": 2.41, "learning_rate": 3.236612195011543e-05, "loss": 0.6059, "step": 39958 }, { "epoch": 2.41, "learning_rate": 3.2364613040001206e-05, "loss": 0.5765, "step": 39960 }, { "epoch": 2.41, "learning_rate": 3.2363104129886986e-05, "loss": 0.7606, "step": 39962 }, { "epoch": 2.41, "learning_rate": 3.236159521977276e-05, "loss": 0.7464, "step": 39964 }, { "epoch": 2.41, "learning_rate": 3.236008630965853e-05, "loss": 0.4627, "step": 39966 }, { "epoch": 2.41, "learning_rate": 3.235857739954431e-05, "loss": 0.4219, "step": 39968 }, { "epoch": 2.41, "learning_rate": 3.2357068489430085e-05, "loss": 0.5116, "step": 39970 }, { "epoch": 2.41, "learning_rate": 3.2355559579315865e-05, "loss": 0.6584, "step": 39972 }, { "epoch": 2.41, "learning_rate": 3.235405066920164e-05, "loss": 0.655, "step": 39974 }, { "epoch": 2.41, "learning_rate": 3.235254175908741e-05, "loss": 0.4528, "step": 39976 }, { "epoch": 2.41, "learning_rate": 3.235103284897319e-05, "loss": 0.6748, "step": 39978 }, { "epoch": 2.41, "learning_rate": 3.234952393885896e-05, "loss": 0.4986, "step": 39980 }, { "epoch": 2.41, "learning_rate": 3.234801502874474e-05, "loss": 0.6142, "step": 39982 }, { "epoch": 2.41, "learning_rate": 3.234650611863052e-05, "loss": 0.8125, "step": 39984 }, { "epoch": 2.41, "learning_rate": 3.234499720851629e-05, "loss": 0.4034, "step": 39986 }, { "epoch": 2.41, "learning_rate": 3.2343488298402064e-05, "loss": 0.5871, "step": 39988 }, { "epoch": 2.41, "learning_rate": 3.2341979388287844e-05, "loss": 0.5787, "step": 39990 }, { "epoch": 2.41, "learning_rate": 3.234047047817362e-05, "loss": 0.6619, "step": 39992 }, { "epoch": 2.41, "learning_rate": 3.23389615680594e-05, "loss": 0.5064, "step": 39994 }, { "epoch": 2.41, "learning_rate": 3.233745265794516e-05, "loss": 0.5205, "step": 39996 }, { "epoch": 2.41, "learning_rate": 3.233594374783094e-05, "loss": 0.4764, "step": 39998 }, { "epoch": 2.41, "learning_rate": 3.233443483771672e-05, "loss": 0.6932, "step": 40000 }, { "epoch": 2.41, "learning_rate": 3.233292592760249e-05, "loss": 0.6849, "step": 40002 }, { "epoch": 2.41, "learning_rate": 3.233141701748827e-05, "loss": 0.4328, "step": 40004 }, { "epoch": 2.41, "learning_rate": 3.232990810737405e-05, "loss": 0.8267, "step": 40006 }, { "epoch": 2.41, "learning_rate": 3.232839919725982e-05, "loss": 0.6003, "step": 40008 }, { "epoch": 2.41, "learning_rate": 3.2326890287145595e-05, "loss": 0.5206, "step": 40010 }, { "epoch": 2.42, "learning_rate": 3.2325381377031375e-05, "loss": 0.6263, "step": 40012 }, { "epoch": 2.42, "learning_rate": 3.232387246691715e-05, "loss": 0.6039, "step": 40014 }, { "epoch": 2.42, "learning_rate": 3.232236355680292e-05, "loss": 0.4718, "step": 40016 }, { "epoch": 2.42, "learning_rate": 3.2320854646688695e-05, "loss": 0.4076, "step": 40018 }, { "epoch": 2.42, "learning_rate": 3.2319345736574474e-05, "loss": 0.6123, "step": 40020 }, { "epoch": 2.42, "learning_rate": 3.2317836826460254e-05, "loss": 0.8306, "step": 40022 }, { "epoch": 2.42, "learning_rate": 3.231632791634602e-05, "loss": 0.4538, "step": 40024 }, { "epoch": 2.42, "learning_rate": 3.23148190062318e-05, "loss": 0.5349, "step": 40026 }, { "epoch": 2.42, "learning_rate": 3.231331009611758e-05, "loss": 0.5041, "step": 40028 }, { "epoch": 2.42, "learning_rate": 3.231180118600335e-05, "loss": 0.6019, "step": 40030 }, { "epoch": 2.42, "learning_rate": 3.231029227588913e-05, "loss": 0.374, "step": 40032 }, { "epoch": 2.42, "learning_rate": 3.23087833657749e-05, "loss": 0.6574, "step": 40034 }, { "epoch": 2.42, "learning_rate": 3.230727445566068e-05, "loss": 0.7875, "step": 40036 }, { "epoch": 2.42, "learning_rate": 3.230576554554645e-05, "loss": 0.5678, "step": 40038 }, { "epoch": 2.42, "learning_rate": 3.2304256635432226e-05, "loss": 0.6247, "step": 40040 }, { "epoch": 2.42, "learning_rate": 3.2302747725318006e-05, "loss": 0.529, "step": 40042 }, { "epoch": 2.42, "learning_rate": 3.2301238815203786e-05, "loss": 0.5545, "step": 40044 }, { "epoch": 2.42, "learning_rate": 3.229972990508955e-05, "loss": 0.4525, "step": 40046 }, { "epoch": 2.42, "learning_rate": 3.229822099497533e-05, "loss": 0.7098, "step": 40048 }, { "epoch": 2.42, "learning_rate": 3.2296712084861105e-05, "loss": 0.5744, "step": 40050 }, { "epoch": 2.42, "learning_rate": 3.229520317474688e-05, "loss": 0.6727, "step": 40052 }, { "epoch": 2.42, "learning_rate": 3.229369426463266e-05, "loss": 0.4222, "step": 40054 }, { "epoch": 2.42, "learning_rate": 3.229218535451843e-05, "loss": 0.6236, "step": 40056 }, { "epoch": 2.42, "learning_rate": 3.229067644440421e-05, "loss": 0.5815, "step": 40058 }, { "epoch": 2.42, "learning_rate": 3.2289167534289984e-05, "loss": 0.4578, "step": 40060 }, { "epoch": 2.42, "learning_rate": 3.228765862417576e-05, "loss": 0.5501, "step": 40062 }, { "epoch": 2.42, "learning_rate": 3.228614971406154e-05, "loss": 0.4884, "step": 40064 }, { "epoch": 2.42, "learning_rate": 3.228464080394731e-05, "loss": 0.8469, "step": 40066 }, { "epoch": 2.42, "learning_rate": 3.2283131893833084e-05, "loss": 0.3292, "step": 40068 }, { "epoch": 2.42, "learning_rate": 3.2281622983718864e-05, "loss": 0.5547, "step": 40070 }, { "epoch": 2.42, "learning_rate": 3.228011407360464e-05, "loss": 0.547, "step": 40072 }, { "epoch": 2.42, "learning_rate": 3.227860516349041e-05, "loss": 0.7322, "step": 40074 }, { "epoch": 2.42, "learning_rate": 3.227709625337619e-05, "loss": 0.6996, "step": 40076 }, { "epoch": 2.42, "learning_rate": 3.227558734326196e-05, "loss": 0.6053, "step": 40078 }, { "epoch": 2.42, "learning_rate": 3.2274078433147736e-05, "loss": 0.6017, "step": 40080 }, { "epoch": 2.42, "learning_rate": 3.2272569523033516e-05, "loss": 0.5814, "step": 40082 }, { "epoch": 2.42, "learning_rate": 3.227106061291929e-05, "loss": 0.5991, "step": 40084 }, { "epoch": 2.42, "learning_rate": 3.226955170280507e-05, "loss": 0.4412, "step": 40086 }, { "epoch": 2.42, "learning_rate": 3.226804279269084e-05, "loss": 0.5823, "step": 40088 }, { "epoch": 2.42, "learning_rate": 3.2266533882576615e-05, "loss": 0.5777, "step": 40090 }, { "epoch": 2.42, "learning_rate": 3.2265024972462395e-05, "loss": 0.8668, "step": 40092 }, { "epoch": 2.42, "learning_rate": 3.226351606234816e-05, "loss": 0.6893, "step": 40094 }, { "epoch": 2.42, "learning_rate": 3.226200715223394e-05, "loss": 0.7511, "step": 40096 }, { "epoch": 2.42, "learning_rate": 3.226049824211972e-05, "loss": 0.4658, "step": 40098 }, { "epoch": 2.42, "learning_rate": 3.2258989332005494e-05, "loss": 0.5328, "step": 40100 }, { "epoch": 2.42, "learning_rate": 3.225748042189127e-05, "loss": 0.5213, "step": 40102 }, { "epoch": 2.42, "learning_rate": 3.225597151177705e-05, "loss": 0.4698, "step": 40104 }, { "epoch": 2.42, "learning_rate": 3.225446260166282e-05, "loss": 0.4374, "step": 40106 }, { "epoch": 2.42, "learning_rate": 3.22529536915486e-05, "loss": 0.5769, "step": 40108 }, { "epoch": 2.42, "learning_rate": 3.225144478143437e-05, "loss": 0.6747, "step": 40110 }, { "epoch": 2.42, "learning_rate": 3.224993587132015e-05, "loss": 0.6063, "step": 40112 }, { "epoch": 2.42, "learning_rate": 3.2248426961205927e-05, "loss": 0.6105, "step": 40114 }, { "epoch": 2.42, "learning_rate": 3.224691805109169e-05, "loss": 0.5952, "step": 40116 }, { "epoch": 2.42, "learning_rate": 3.224540914097747e-05, "loss": 0.6035, "step": 40118 }, { "epoch": 2.42, "learning_rate": 3.224390023086325e-05, "loss": 0.708, "step": 40120 }, { "epoch": 2.42, "learning_rate": 3.2242391320749026e-05, "loss": 0.6108, "step": 40122 }, { "epoch": 2.42, "learning_rate": 3.22408824106348e-05, "loss": 0.8399, "step": 40124 }, { "epoch": 2.42, "learning_rate": 3.223937350052058e-05, "loss": 0.4656, "step": 40126 }, { "epoch": 2.42, "learning_rate": 3.223786459040635e-05, "loss": 0.5701, "step": 40128 }, { "epoch": 2.42, "learning_rate": 3.2236355680292125e-05, "loss": 0.4997, "step": 40130 }, { "epoch": 2.42, "learning_rate": 3.22348467701779e-05, "loss": 0.4549, "step": 40132 }, { "epoch": 2.42, "learning_rate": 3.223333786006368e-05, "loss": 0.4206, "step": 40134 }, { "epoch": 2.42, "learning_rate": 3.223182894994946e-05, "loss": 0.5228, "step": 40136 }, { "epoch": 2.42, "learning_rate": 3.2230320039835224e-05, "loss": 0.6417, "step": 40138 }, { "epoch": 2.42, "learning_rate": 3.2228811129721004e-05, "loss": 0.677, "step": 40140 }, { "epoch": 2.42, "learning_rate": 3.2227302219606784e-05, "loss": 0.403, "step": 40142 }, { "epoch": 2.42, "learning_rate": 3.222579330949255e-05, "loss": 0.7893, "step": 40144 }, { "epoch": 2.42, "learning_rate": 3.222428439937833e-05, "loss": 0.436, "step": 40146 }, { "epoch": 2.42, "learning_rate": 3.2222775489264104e-05, "loss": 0.4488, "step": 40148 }, { "epoch": 2.42, "learning_rate": 3.2221266579149883e-05, "loss": 0.5344, "step": 40150 }, { "epoch": 2.42, "learning_rate": 3.2219757669035657e-05, "loss": 0.6815, "step": 40152 }, { "epoch": 2.42, "learning_rate": 3.221824875892143e-05, "loss": 0.4744, "step": 40154 }, { "epoch": 2.42, "learning_rate": 3.221673984880721e-05, "loss": 0.4841, "step": 40156 }, { "epoch": 2.42, "learning_rate": 3.221523093869299e-05, "loss": 0.6505, "step": 40158 }, { "epoch": 2.42, "learning_rate": 3.2213722028578756e-05, "loss": 0.6487, "step": 40160 }, { "epoch": 2.42, "learning_rate": 3.2212213118464536e-05, "loss": 0.6452, "step": 40162 }, { "epoch": 2.42, "learning_rate": 3.221070420835031e-05, "loss": 0.6708, "step": 40164 }, { "epoch": 2.42, "learning_rate": 3.220919529823608e-05, "loss": 0.6172, "step": 40166 }, { "epoch": 2.42, "learning_rate": 3.220768638812186e-05, "loss": 0.6707, "step": 40168 }, { "epoch": 2.42, "learning_rate": 3.2206177478007635e-05, "loss": 0.5936, "step": 40170 }, { "epoch": 2.42, "learning_rate": 3.2204668567893415e-05, "loss": 0.5652, "step": 40172 }, { "epoch": 2.42, "learning_rate": 3.220315965777919e-05, "loss": 0.4942, "step": 40174 }, { "epoch": 2.42, "learning_rate": 3.220165074766496e-05, "loss": 0.3208, "step": 40176 }, { "epoch": 2.43, "learning_rate": 3.220014183755074e-05, "loss": 0.5141, "step": 40178 }, { "epoch": 2.43, "learning_rate": 3.2198632927436514e-05, "loss": 0.4601, "step": 40180 }, { "epoch": 2.43, "learning_rate": 3.219712401732229e-05, "loss": 0.4676, "step": 40182 }, { "epoch": 2.43, "learning_rate": 3.219561510720807e-05, "loss": 0.6062, "step": 40184 }, { "epoch": 2.43, "learning_rate": 3.219410619709384e-05, "loss": 0.5711, "step": 40186 }, { "epoch": 2.43, "learning_rate": 3.2192597286979614e-05, "loss": 0.4288, "step": 40188 }, { "epoch": 2.43, "learning_rate": 3.2191088376865393e-05, "loss": 0.3881, "step": 40190 }, { "epoch": 2.43, "learning_rate": 3.2189579466751167e-05, "loss": 0.6972, "step": 40192 }, { "epoch": 2.43, "learning_rate": 3.218807055663694e-05, "loss": 0.5039, "step": 40194 }, { "epoch": 2.43, "learning_rate": 3.218656164652272e-05, "loss": 0.6557, "step": 40196 }, { "epoch": 2.43, "learning_rate": 3.218505273640849e-05, "loss": 0.672, "step": 40198 }, { "epoch": 2.43, "learning_rate": 3.218354382629427e-05, "loss": 0.7522, "step": 40200 }, { "epoch": 2.43, "learning_rate": 3.2182034916180046e-05, "loss": 0.4961, "step": 40202 }, { "epoch": 2.43, "learning_rate": 3.218052600606582e-05, "loss": 0.4269, "step": 40204 }, { "epoch": 2.43, "learning_rate": 3.21790170959516e-05, "loss": 0.4702, "step": 40206 }, { "epoch": 2.43, "learning_rate": 3.217750818583737e-05, "loss": 0.458, "step": 40208 }, { "epoch": 2.43, "learning_rate": 3.2175999275723145e-05, "loss": 0.3863, "step": 40210 }, { "epoch": 2.43, "learning_rate": 3.2174490365608925e-05, "loss": 0.6611, "step": 40212 }, { "epoch": 2.43, "learning_rate": 3.21729814554947e-05, "loss": 0.6313, "step": 40214 }, { "epoch": 2.43, "learning_rate": 3.217147254538047e-05, "loss": 0.5112, "step": 40216 }, { "epoch": 2.43, "learning_rate": 3.216996363526625e-05, "loss": 0.5874, "step": 40218 }, { "epoch": 2.43, "learning_rate": 3.2168454725152024e-05, "loss": 0.646, "step": 40220 }, { "epoch": 2.43, "learning_rate": 3.2166945815037804e-05, "loss": 0.6468, "step": 40222 }, { "epoch": 2.43, "learning_rate": 3.216543690492357e-05, "loss": 0.6834, "step": 40224 }, { "epoch": 2.43, "learning_rate": 3.216392799480935e-05, "loss": 0.636, "step": 40226 }, { "epoch": 2.43, "learning_rate": 3.216241908469513e-05, "loss": 0.4995, "step": 40228 }, { "epoch": 2.43, "learning_rate": 3.2160910174580897e-05, "loss": 0.7848, "step": 40230 }, { "epoch": 2.43, "learning_rate": 3.2159401264466676e-05, "loss": 0.3635, "step": 40232 }, { "epoch": 2.43, "learning_rate": 3.2157892354352456e-05, "loss": 0.405, "step": 40234 }, { "epoch": 2.43, "learning_rate": 3.215638344423823e-05, "loss": 0.7481, "step": 40236 }, { "epoch": 2.43, "learning_rate": 3.2154874534124e-05, "loss": 0.6563, "step": 40238 }, { "epoch": 2.43, "learning_rate": 3.2153365624009776e-05, "loss": 0.4915, "step": 40240 }, { "epoch": 2.43, "learning_rate": 3.2151856713895556e-05, "loss": 0.6026, "step": 40242 }, { "epoch": 2.43, "learning_rate": 3.215034780378133e-05, "loss": 0.4826, "step": 40244 }, { "epoch": 2.43, "learning_rate": 3.21488388936671e-05, "loss": 0.5245, "step": 40246 }, { "epoch": 2.43, "learning_rate": 3.214732998355288e-05, "loss": 0.5529, "step": 40248 }, { "epoch": 2.43, "learning_rate": 3.214582107343866e-05, "loss": 0.497, "step": 40250 }, { "epoch": 2.43, "learning_rate": 3.214431216332443e-05, "loss": 0.6312, "step": 40252 }, { "epoch": 2.43, "learning_rate": 3.214280325321021e-05, "loss": 0.648, "step": 40254 }, { "epoch": 2.43, "learning_rate": 3.214129434309599e-05, "loss": 0.3883, "step": 40256 }, { "epoch": 2.43, "learning_rate": 3.213978543298176e-05, "loss": 0.5952, "step": 40258 }, { "epoch": 2.43, "learning_rate": 3.2138276522867534e-05, "loss": 0.4486, "step": 40260 }, { "epoch": 2.43, "learning_rate": 3.213676761275331e-05, "loss": 0.5335, "step": 40262 }, { "epoch": 2.43, "learning_rate": 3.213525870263909e-05, "loss": 0.6152, "step": 40264 }, { "epoch": 2.43, "learning_rate": 3.213374979252486e-05, "loss": 0.6254, "step": 40266 }, { "epoch": 2.43, "learning_rate": 3.2132240882410633e-05, "loss": 0.3716, "step": 40268 }, { "epoch": 2.43, "learning_rate": 3.213073197229641e-05, "loss": 0.5156, "step": 40270 }, { "epoch": 2.43, "learning_rate": 3.212922306218219e-05, "loss": 0.5659, "step": 40272 }, { "epoch": 2.43, "learning_rate": 3.212771415206796e-05, "loss": 0.4899, "step": 40274 }, { "epoch": 2.43, "learning_rate": 3.212620524195374e-05, "loss": 0.5322, "step": 40276 }, { "epoch": 2.43, "learning_rate": 3.212469633183951e-05, "loss": 0.5829, "step": 40278 }, { "epoch": 2.43, "learning_rate": 3.2123187421725286e-05, "loss": 0.5414, "step": 40280 }, { "epoch": 2.43, "learning_rate": 3.2121678511611066e-05, "loss": 0.521, "step": 40282 }, { "epoch": 2.43, "learning_rate": 3.212016960149684e-05, "loss": 0.6103, "step": 40284 }, { "epoch": 2.43, "learning_rate": 3.211866069138262e-05, "loss": 0.4436, "step": 40286 }, { "epoch": 2.43, "learning_rate": 3.211715178126839e-05, "loss": 0.539, "step": 40288 }, { "epoch": 2.43, "learning_rate": 3.2115642871154165e-05, "loss": 0.775, "step": 40290 }, { "epoch": 2.43, "learning_rate": 3.2114133961039945e-05, "loss": 0.5919, "step": 40292 }, { "epoch": 2.43, "learning_rate": 3.211262505092572e-05, "loss": 0.8231, "step": 40294 }, { "epoch": 2.43, "learning_rate": 3.211111614081149e-05, "loss": 0.6802, "step": 40296 }, { "epoch": 2.43, "learning_rate": 3.210960723069727e-05, "loss": 0.5267, "step": 40298 }, { "epoch": 2.43, "learning_rate": 3.2108098320583044e-05, "loss": 0.4347, "step": 40300 }, { "epoch": 2.43, "learning_rate": 3.210658941046882e-05, "loss": 0.5761, "step": 40302 }, { "epoch": 2.43, "learning_rate": 3.21050805003546e-05, "loss": 0.4994, "step": 40304 }, { "epoch": 2.43, "learning_rate": 3.210357159024037e-05, "loss": 0.4457, "step": 40306 }, { "epoch": 2.43, "learning_rate": 3.210206268012614e-05, "loss": 0.5461, "step": 40308 }, { "epoch": 2.43, "learning_rate": 3.210055377001192e-05, "loss": 0.5024, "step": 40310 }, { "epoch": 2.43, "learning_rate": 3.2099044859897696e-05, "loss": 0.6447, "step": 40312 }, { "epoch": 2.43, "learning_rate": 3.2097535949783476e-05, "loss": 0.5723, "step": 40314 }, { "epoch": 2.43, "learning_rate": 3.209602703966925e-05, "loss": 0.5556, "step": 40316 }, { "epoch": 2.43, "learning_rate": 3.209451812955502e-05, "loss": 0.6143, "step": 40318 }, { "epoch": 2.43, "learning_rate": 3.20930092194408e-05, "loss": 0.4068, "step": 40320 }, { "epoch": 2.43, "learning_rate": 3.2091500309326576e-05, "loss": 0.7012, "step": 40322 }, { "epoch": 2.43, "learning_rate": 3.208999139921235e-05, "loss": 0.5943, "step": 40324 }, { "epoch": 2.43, "learning_rate": 3.208848248909813e-05, "loss": 0.524, "step": 40326 }, { "epoch": 2.43, "learning_rate": 3.20869735789839e-05, "loss": 0.556, "step": 40328 }, { "epoch": 2.43, "learning_rate": 3.2085464668869675e-05, "loss": 0.4769, "step": 40330 }, { "epoch": 2.43, "learning_rate": 3.2083955758755455e-05, "loss": 0.4389, "step": 40332 }, { "epoch": 2.43, "learning_rate": 3.208244684864123e-05, "loss": 0.7326, "step": 40334 }, { "epoch": 2.43, "learning_rate": 3.208093793852701e-05, "loss": 0.8393, "step": 40336 }, { "epoch": 2.43, "learning_rate": 3.2079429028412774e-05, "loss": 0.5213, "step": 40338 }, { "epoch": 2.43, "learning_rate": 3.2077920118298554e-05, "loss": 0.6047, "step": 40340 }, { "epoch": 2.43, "learning_rate": 3.2076411208184334e-05, "loss": 0.6931, "step": 40342 }, { "epoch": 2.44, "learning_rate": 3.20749022980701e-05, "loss": 0.6579, "step": 40344 }, { "epoch": 2.44, "learning_rate": 3.207339338795588e-05, "loss": 0.6844, "step": 40346 }, { "epoch": 2.44, "learning_rate": 3.207188447784166e-05, "loss": 0.6007, "step": 40348 }, { "epoch": 2.44, "learning_rate": 3.207037556772743e-05, "loss": 0.4487, "step": 40350 }, { "epoch": 2.44, "learning_rate": 3.2068866657613206e-05, "loss": 0.6461, "step": 40352 }, { "epoch": 2.44, "learning_rate": 3.206735774749898e-05, "loss": 0.6604, "step": 40354 }, { "epoch": 2.44, "learning_rate": 3.206584883738476e-05, "loss": 0.7621, "step": 40356 }, { "epoch": 2.44, "learning_rate": 3.206433992727053e-05, "loss": 0.5872, "step": 40358 }, { "epoch": 2.44, "learning_rate": 3.2062831017156306e-05, "loss": 0.4172, "step": 40360 }, { "epoch": 2.44, "learning_rate": 3.2061322107042085e-05, "loss": 0.531, "step": 40362 }, { "epoch": 2.44, "learning_rate": 3.2059813196927865e-05, "loss": 0.5372, "step": 40364 }, { "epoch": 2.44, "learning_rate": 3.205830428681363e-05, "loss": 0.564, "step": 40366 }, { "epoch": 2.44, "learning_rate": 3.205679537669941e-05, "loss": 0.5166, "step": 40368 }, { "epoch": 2.44, "learning_rate": 3.205528646658519e-05, "loss": 0.4725, "step": 40370 }, { "epoch": 2.44, "learning_rate": 3.2053777556470965e-05, "loss": 0.4929, "step": 40372 }, { "epoch": 2.44, "learning_rate": 3.205226864635674e-05, "loss": 0.6516, "step": 40374 }, { "epoch": 2.44, "learning_rate": 3.205075973624251e-05, "loss": 0.4534, "step": 40376 }, { "epoch": 2.44, "learning_rate": 3.204925082612829e-05, "loss": 0.6206, "step": 40378 }, { "epoch": 2.44, "learning_rate": 3.2047741916014064e-05, "loss": 0.5675, "step": 40380 }, { "epoch": 2.44, "learning_rate": 3.204623300589984e-05, "loss": 0.563, "step": 40382 }, { "epoch": 2.44, "learning_rate": 3.204472409578562e-05, "loss": 0.5918, "step": 40384 }, { "epoch": 2.44, "learning_rate": 3.20432151856714e-05, "loss": 0.7315, "step": 40386 }, { "epoch": 2.44, "learning_rate": 3.204170627555716e-05, "loss": 0.8166, "step": 40388 }, { "epoch": 2.44, "learning_rate": 3.204019736544294e-05, "loss": 0.5834, "step": 40390 }, { "epoch": 2.44, "learning_rate": 3.2038688455328716e-05, "loss": 0.5141, "step": 40392 }, { "epoch": 2.44, "learning_rate": 3.203717954521449e-05, "loss": 0.3907, "step": 40394 }, { "epoch": 2.44, "learning_rate": 3.203567063510027e-05, "loss": 0.5042, "step": 40396 }, { "epoch": 2.44, "learning_rate": 3.203416172498604e-05, "loss": 0.7135, "step": 40398 }, { "epoch": 2.44, "learning_rate": 3.203265281487182e-05, "loss": 0.4807, "step": 40400 }, { "epoch": 2.44, "learning_rate": 3.2031143904757595e-05, "loss": 0.7254, "step": 40402 }, { "epoch": 2.44, "learning_rate": 3.202963499464337e-05, "loss": 0.5151, "step": 40404 }, { "epoch": 2.44, "learning_rate": 3.202812608452915e-05, "loss": 0.3784, "step": 40406 }, { "epoch": 2.44, "learning_rate": 3.202661717441492e-05, "loss": 0.5955, "step": 40408 }, { "epoch": 2.44, "learning_rate": 3.2025108264300695e-05, "loss": 0.6957, "step": 40410 }, { "epoch": 2.44, "learning_rate": 3.2023599354186475e-05, "loss": 0.6347, "step": 40412 }, { "epoch": 2.44, "learning_rate": 3.202209044407225e-05, "loss": 0.5402, "step": 40414 }, { "epoch": 2.44, "learning_rate": 3.202058153395802e-05, "loss": 0.6064, "step": 40416 }, { "epoch": 2.44, "learning_rate": 3.20190726238438e-05, "loss": 0.329, "step": 40418 }, { "epoch": 2.44, "learning_rate": 3.2017563713729574e-05, "loss": 0.6666, "step": 40420 }, { "epoch": 2.44, "learning_rate": 3.2016054803615354e-05, "loss": 0.5118, "step": 40422 }, { "epoch": 2.44, "learning_rate": 3.201454589350113e-05, "loss": 0.587, "step": 40424 }, { "epoch": 2.44, "learning_rate": 3.20130369833869e-05, "loss": 0.629, "step": 40426 }, { "epoch": 2.44, "learning_rate": 3.201152807327268e-05, "loss": 0.5926, "step": 40428 }, { "epoch": 2.44, "learning_rate": 3.201001916315845e-05, "loss": 0.5412, "step": 40430 }, { "epoch": 2.44, "learning_rate": 3.2008510253044226e-05, "loss": 0.528, "step": 40432 }, { "epoch": 2.44, "learning_rate": 3.2007001342930006e-05, "loss": 0.6725, "step": 40434 }, { "epoch": 2.44, "learning_rate": 3.200549243281578e-05, "loss": 0.4531, "step": 40436 }, { "epoch": 2.44, "learning_rate": 3.200398352270155e-05, "loss": 0.6177, "step": 40438 }, { "epoch": 2.44, "learning_rate": 3.200247461258733e-05, "loss": 0.5787, "step": 40440 }, { "epoch": 2.44, "learning_rate": 3.2000965702473105e-05, "loss": 0.6466, "step": 40442 }, { "epoch": 2.44, "learning_rate": 3.199945679235888e-05, "loss": 0.8458, "step": 40444 }, { "epoch": 2.44, "learning_rate": 3.199794788224466e-05, "loss": 0.6847, "step": 40446 }, { "epoch": 2.44, "learning_rate": 3.199643897213043e-05, "loss": 0.6704, "step": 40448 }, { "epoch": 2.44, "learning_rate": 3.199493006201621e-05, "loss": 0.5502, "step": 40450 }, { "epoch": 2.44, "learning_rate": 3.199342115190198e-05, "loss": 0.5967, "step": 40452 }, { "epoch": 2.44, "learning_rate": 3.199191224178776e-05, "loss": 0.6057, "step": 40454 }, { "epoch": 2.44, "learning_rate": 3.199040333167354e-05, "loss": 0.6268, "step": 40456 }, { "epoch": 2.44, "learning_rate": 3.1988894421559304e-05, "loss": 0.487, "step": 40458 }, { "epoch": 2.44, "learning_rate": 3.1987385511445084e-05, "loss": 0.5168, "step": 40460 }, { "epoch": 2.44, "learning_rate": 3.1985876601330864e-05, "loss": 0.4947, "step": 40462 }, { "epoch": 2.44, "learning_rate": 3.198436769121664e-05, "loss": 0.4954, "step": 40464 }, { "epoch": 2.44, "learning_rate": 3.198285878110241e-05, "loss": 0.5454, "step": 40466 }, { "epoch": 2.44, "learning_rate": 3.198134987098818e-05, "loss": 0.632, "step": 40468 }, { "epoch": 2.44, "learning_rate": 3.197984096087396e-05, "loss": 0.4674, "step": 40470 }, { "epoch": 2.44, "learning_rate": 3.1978332050759736e-05, "loss": 0.5554, "step": 40472 }, { "epoch": 2.44, "learning_rate": 3.197682314064551e-05, "loss": 0.5071, "step": 40474 }, { "epoch": 2.44, "learning_rate": 3.197531423053129e-05, "loss": 0.7419, "step": 40476 }, { "epoch": 2.44, "learning_rate": 3.197380532041707e-05, "loss": 0.7355, "step": 40478 }, { "epoch": 2.44, "learning_rate": 3.1972296410302835e-05, "loss": 0.6053, "step": 40480 }, { "epoch": 2.44, "learning_rate": 3.1970787500188615e-05, "loss": 0.4135, "step": 40482 }, { "epoch": 2.44, "learning_rate": 3.1969278590074395e-05, "loss": 0.662, "step": 40484 }, { "epoch": 2.44, "learning_rate": 3.196776967996017e-05, "loss": 0.5453, "step": 40486 }, { "epoch": 2.44, "learning_rate": 3.196626076984594e-05, "loss": 0.6732, "step": 40488 }, { "epoch": 2.44, "learning_rate": 3.1964751859731715e-05, "loss": 0.2842, "step": 40490 }, { "epoch": 2.44, "learning_rate": 3.1963242949617495e-05, "loss": 0.626, "step": 40492 }, { "epoch": 2.44, "learning_rate": 3.196173403950327e-05, "loss": 0.6271, "step": 40494 }, { "epoch": 2.44, "learning_rate": 3.196022512938904e-05, "loss": 0.5447, "step": 40496 }, { "epoch": 2.44, "learning_rate": 3.195871621927482e-05, "loss": 0.5012, "step": 40498 }, { "epoch": 2.44, "learning_rate": 3.19572073091606e-05, "loss": 0.4945, "step": 40500 }, { "epoch": 2.44, "learning_rate": 3.195569839904637e-05, "loss": 0.415, "step": 40502 }, { "epoch": 2.44, "learning_rate": 3.195418948893215e-05, "loss": 0.4377, "step": 40504 }, { "epoch": 2.44, "learning_rate": 3.195268057881792e-05, "loss": 0.659, "step": 40506 }, { "epoch": 2.44, "learning_rate": 3.195117166870369e-05, "loss": 0.5735, "step": 40508 }, { "epoch": 2.45, "learning_rate": 3.194966275858947e-05, "loss": 0.4518, "step": 40510 }, { "epoch": 2.45, "learning_rate": 3.1948153848475246e-05, "loss": 0.4563, "step": 40512 }, { "epoch": 2.45, "learning_rate": 3.1946644938361026e-05, "loss": 0.4301, "step": 40514 }, { "epoch": 2.45, "learning_rate": 3.19451360282468e-05, "loss": 0.6694, "step": 40516 }, { "epoch": 2.45, "learning_rate": 3.194362711813257e-05, "loss": 0.6077, "step": 40518 }, { "epoch": 2.45, "learning_rate": 3.194211820801835e-05, "loss": 0.5076, "step": 40520 }, { "epoch": 2.45, "learning_rate": 3.1940609297904125e-05, "loss": 0.4753, "step": 40522 }, { "epoch": 2.45, "learning_rate": 3.19391003877899e-05, "loss": 0.6682, "step": 40524 }, { "epoch": 2.45, "learning_rate": 3.193759147767568e-05, "loss": 0.6572, "step": 40526 }, { "epoch": 2.45, "learning_rate": 3.193608256756145e-05, "loss": 0.7957, "step": 40528 }, { "epoch": 2.45, "learning_rate": 3.1934573657447225e-05, "loss": 0.5176, "step": 40530 }, { "epoch": 2.45, "learning_rate": 3.1933064747333004e-05, "loss": 0.6371, "step": 40532 }, { "epoch": 2.45, "learning_rate": 3.193155583721878e-05, "loss": 0.5561, "step": 40534 }, { "epoch": 2.45, "learning_rate": 3.193004692710456e-05, "loss": 0.5966, "step": 40536 }, { "epoch": 2.45, "learning_rate": 3.192853801699033e-05, "loss": 0.4728, "step": 40538 }, { "epoch": 2.45, "learning_rate": 3.1927029106876104e-05, "loss": 0.4539, "step": 40540 }, { "epoch": 2.45, "learning_rate": 3.1925520196761884e-05, "loss": 0.5798, "step": 40542 }, { "epoch": 2.45, "learning_rate": 3.192401128664766e-05, "loss": 0.7869, "step": 40544 }, { "epoch": 2.45, "learning_rate": 3.192250237653343e-05, "loss": 0.5543, "step": 40546 }, { "epoch": 2.45, "learning_rate": 3.192099346641921e-05, "loss": 0.5802, "step": 40548 }, { "epoch": 2.45, "learning_rate": 3.191948455630498e-05, "loss": 0.6228, "step": 40550 }, { "epoch": 2.45, "learning_rate": 3.1917975646190756e-05, "loss": 0.5203, "step": 40552 }, { "epoch": 2.45, "learning_rate": 3.1916466736076536e-05, "loss": 0.5436, "step": 40554 }, { "epoch": 2.45, "learning_rate": 3.191495782596231e-05, "loss": 0.5611, "step": 40556 }, { "epoch": 2.45, "learning_rate": 3.191344891584808e-05, "loss": 0.7654, "step": 40558 }, { "epoch": 2.45, "learning_rate": 3.191194000573386e-05, "loss": 0.6495, "step": 40560 }, { "epoch": 2.45, "learning_rate": 3.1910431095619635e-05, "loss": 0.4935, "step": 40562 }, { "epoch": 2.45, "learning_rate": 3.1908922185505415e-05, "loss": 0.527, "step": 40564 }, { "epoch": 2.45, "learning_rate": 3.190741327539118e-05, "loss": 0.5836, "step": 40566 }, { "epoch": 2.45, "learning_rate": 3.190590436527696e-05, "loss": 0.574, "step": 40568 }, { "epoch": 2.45, "learning_rate": 3.190439545516274e-05, "loss": 0.5173, "step": 40570 }, { "epoch": 2.45, "learning_rate": 3.190288654504851e-05, "loss": 0.5065, "step": 40572 }, { "epoch": 2.45, "learning_rate": 3.190137763493429e-05, "loss": 0.657, "step": 40574 }, { "epoch": 2.45, "learning_rate": 3.189986872482007e-05, "loss": 0.7947, "step": 40576 }, { "epoch": 2.45, "learning_rate": 3.189835981470584e-05, "loss": 0.6779, "step": 40578 }, { "epoch": 2.45, "learning_rate": 3.1896850904591614e-05, "loss": 0.3989, "step": 40580 }, { "epoch": 2.45, "learning_rate": 3.189534199447739e-05, "loss": 0.6355, "step": 40582 }, { "epoch": 2.45, "learning_rate": 3.189383308436317e-05, "loss": 0.3964, "step": 40584 }, { "epoch": 2.45, "learning_rate": 3.1892324174248947e-05, "loss": 0.3458, "step": 40586 }, { "epoch": 2.45, "learning_rate": 3.189081526413471e-05, "loss": 0.4689, "step": 40588 }, { "epoch": 2.45, "learning_rate": 3.188930635402049e-05, "loss": 0.7039, "step": 40590 }, { "epoch": 2.45, "learning_rate": 3.188779744390627e-05, "loss": 0.6583, "step": 40592 }, { "epoch": 2.45, "learning_rate": 3.188628853379204e-05, "loss": 0.5151, "step": 40594 }, { "epoch": 2.45, "learning_rate": 3.188477962367782e-05, "loss": 0.6174, "step": 40596 }, { "epoch": 2.45, "learning_rate": 3.18832707135636e-05, "loss": 0.6207, "step": 40598 }, { "epoch": 2.45, "learning_rate": 3.188176180344937e-05, "loss": 0.5726, "step": 40600 }, { "epoch": 2.45, "learning_rate": 3.1880252893335145e-05, "loss": 0.8155, "step": 40602 }, { "epoch": 2.45, "learning_rate": 3.187874398322092e-05, "loss": 0.5915, "step": 40604 }, { "epoch": 2.45, "learning_rate": 3.18772350731067e-05, "loss": 0.5123, "step": 40606 }, { "epoch": 2.45, "learning_rate": 3.187572616299247e-05, "loss": 0.3564, "step": 40608 }, { "epoch": 2.45, "learning_rate": 3.1874217252878244e-05, "loss": 0.5265, "step": 40610 }, { "epoch": 2.45, "learning_rate": 3.1872708342764024e-05, "loss": 0.4765, "step": 40612 }, { "epoch": 2.45, "learning_rate": 3.1871199432649804e-05, "loss": 0.5782, "step": 40614 }, { "epoch": 2.45, "learning_rate": 3.186969052253557e-05, "loss": 0.5127, "step": 40616 }, { "epoch": 2.45, "learning_rate": 3.186818161242135e-05, "loss": 0.5165, "step": 40618 }, { "epoch": 2.45, "learning_rate": 3.1866672702307124e-05, "loss": 0.5645, "step": 40620 }, { "epoch": 2.45, "learning_rate": 3.18651637921929e-05, "loss": 0.6527, "step": 40622 }, { "epoch": 2.45, "learning_rate": 3.186365488207868e-05, "loss": 0.6927, "step": 40624 }, { "epoch": 2.45, "learning_rate": 3.186214597196445e-05, "loss": 0.6845, "step": 40626 }, { "epoch": 2.45, "learning_rate": 3.186063706185023e-05, "loss": 0.559, "step": 40628 }, { "epoch": 2.45, "learning_rate": 3.1859128151736e-05, "loss": 0.5491, "step": 40630 }, { "epoch": 2.45, "learning_rate": 3.1857619241621776e-05, "loss": 0.5659, "step": 40632 }, { "epoch": 2.45, "learning_rate": 3.1856110331507556e-05, "loss": 0.5211, "step": 40634 }, { "epoch": 2.45, "learning_rate": 3.185460142139333e-05, "loss": 0.3898, "step": 40636 }, { "epoch": 2.45, "learning_rate": 3.18530925112791e-05, "loss": 0.3448, "step": 40638 }, { "epoch": 2.45, "learning_rate": 3.185158360116488e-05, "loss": 0.561, "step": 40640 }, { "epoch": 2.45, "learning_rate": 3.1850074691050655e-05, "loss": 0.4833, "step": 40642 }, { "epoch": 2.45, "learning_rate": 3.184856578093643e-05, "loss": 0.674, "step": 40644 }, { "epoch": 2.45, "learning_rate": 3.184705687082221e-05, "loss": 0.458, "step": 40646 }, { "epoch": 2.45, "learning_rate": 3.184554796070798e-05, "loss": 0.7246, "step": 40648 }, { "epoch": 2.45, "learning_rate": 3.184403905059376e-05, "loss": 0.5056, "step": 40650 }, { "epoch": 2.45, "learning_rate": 3.1842530140479534e-05, "loss": 0.704, "step": 40652 }, { "epoch": 2.45, "learning_rate": 3.184102123036531e-05, "loss": 0.6457, "step": 40654 }, { "epoch": 2.45, "learning_rate": 3.183951232025109e-05, "loss": 0.6483, "step": 40656 }, { "epoch": 2.45, "learning_rate": 3.183800341013686e-05, "loss": 0.468, "step": 40658 }, { "epoch": 2.45, "learning_rate": 3.1836494500022634e-05, "loss": 0.5696, "step": 40660 }, { "epoch": 2.45, "learning_rate": 3.1834985589908413e-05, "loss": 0.6767, "step": 40662 }, { "epoch": 2.45, "learning_rate": 3.1833476679794187e-05, "loss": 0.5113, "step": 40664 }, { "epoch": 2.45, "learning_rate": 3.183196776967996e-05, "loss": 0.4617, "step": 40666 }, { "epoch": 2.45, "learning_rate": 3.183045885956574e-05, "loss": 0.4562, "step": 40668 }, { "epoch": 2.45, "learning_rate": 3.182894994945151e-05, "loss": 0.5173, "step": 40670 }, { "epoch": 2.45, "learning_rate": 3.1827441039337286e-05, "loss": 0.4476, "step": 40672 }, { "epoch": 2.45, "learning_rate": 3.1825932129223066e-05, "loss": 0.5697, "step": 40674 }, { "epoch": 2.46, "learning_rate": 3.182442321910884e-05, "loss": 0.5655, "step": 40676 }, { "epoch": 2.46, "learning_rate": 3.182291430899462e-05, "loss": 0.7077, "step": 40678 }, { "epoch": 2.46, "learning_rate": 3.1821405398880385e-05, "loss": 0.6439, "step": 40680 }, { "epoch": 2.46, "learning_rate": 3.1819896488766165e-05, "loss": 0.5259, "step": 40682 }, { "epoch": 2.46, "learning_rate": 3.1818387578651945e-05, "loss": 0.6728, "step": 40684 }, { "epoch": 2.46, "learning_rate": 3.181687866853771e-05, "loss": 0.6673, "step": 40686 }, { "epoch": 2.46, "learning_rate": 3.181536975842349e-05, "loss": 0.5297, "step": 40688 }, { "epoch": 2.46, "learning_rate": 3.181386084830927e-05, "loss": 0.5511, "step": 40690 }, { "epoch": 2.46, "learning_rate": 3.1812351938195044e-05, "loss": 0.5397, "step": 40692 }, { "epoch": 2.46, "learning_rate": 3.181084302808082e-05, "loss": 0.5556, "step": 40694 }, { "epoch": 2.46, "learning_rate": 3.180933411796659e-05, "loss": 0.496, "step": 40696 }, { "epoch": 2.46, "learning_rate": 3.180782520785237e-05, "loss": 0.4646, "step": 40698 }, { "epoch": 2.46, "learning_rate": 3.180631629773815e-05, "loss": 0.5533, "step": 40700 }, { "epoch": 2.46, "learning_rate": 3.180480738762392e-05, "loss": 0.5051, "step": 40702 }, { "epoch": 2.46, "learning_rate": 3.1803298477509697e-05, "loss": 0.646, "step": 40704 }, { "epoch": 2.46, "learning_rate": 3.1801789567395476e-05, "loss": 0.7114, "step": 40706 }, { "epoch": 2.46, "learning_rate": 3.180028065728124e-05, "loss": 0.5543, "step": 40708 }, { "epoch": 2.46, "learning_rate": 3.179877174716702e-05, "loss": 0.6171, "step": 40710 }, { "epoch": 2.46, "learning_rate": 3.1797262837052796e-05, "loss": 0.3695, "step": 40712 }, { "epoch": 2.46, "learning_rate": 3.1795753926938576e-05, "loss": 0.3986, "step": 40714 }, { "epoch": 2.46, "learning_rate": 3.179424501682435e-05, "loss": 0.733, "step": 40716 }, { "epoch": 2.46, "learning_rate": 3.179273610671012e-05, "loss": 0.6825, "step": 40718 }, { "epoch": 2.46, "learning_rate": 3.17912271965959e-05, "loss": 1.2279, "step": 40720 }, { "epoch": 2.46, "learning_rate": 3.1789718286481675e-05, "loss": 0.7742, "step": 40722 }, { "epoch": 2.46, "learning_rate": 3.178820937636745e-05, "loss": 0.4213, "step": 40724 }, { "epoch": 2.46, "learning_rate": 3.178670046625323e-05, "loss": 0.6037, "step": 40726 }, { "epoch": 2.46, "learning_rate": 3.178519155613901e-05, "loss": 0.5197, "step": 40728 }, { "epoch": 2.46, "learning_rate": 3.1783682646024774e-05, "loss": 0.5546, "step": 40730 }, { "epoch": 2.46, "learning_rate": 3.1782173735910554e-05, "loss": 0.5749, "step": 40732 }, { "epoch": 2.46, "learning_rate": 3.178066482579633e-05, "loss": 0.5418, "step": 40734 }, { "epoch": 2.46, "learning_rate": 3.17791559156821e-05, "loss": 0.5596, "step": 40736 }, { "epoch": 2.46, "learning_rate": 3.177764700556788e-05, "loss": 0.5641, "step": 40738 }, { "epoch": 2.46, "learning_rate": 3.1776138095453653e-05, "loss": 0.5078, "step": 40740 }, { "epoch": 2.46, "learning_rate": 3.177462918533943e-05, "loss": 0.3111, "step": 40742 }, { "epoch": 2.46, "learning_rate": 3.1773120275225206e-05, "loss": 0.564, "step": 40744 }, { "epoch": 2.46, "learning_rate": 3.177161136511098e-05, "loss": 0.8331, "step": 40746 }, { "epoch": 2.46, "learning_rate": 3.177010245499676e-05, "loss": 0.6909, "step": 40748 }, { "epoch": 2.46, "learning_rate": 3.176859354488253e-05, "loss": 0.7984, "step": 40750 }, { "epoch": 2.46, "learning_rate": 3.1767084634768306e-05, "loss": 0.6273, "step": 40752 }, { "epoch": 2.46, "learning_rate": 3.1765575724654086e-05, "loss": 0.5978, "step": 40754 }, { "epoch": 2.46, "learning_rate": 3.176406681453986e-05, "loss": 0.7668, "step": 40756 }, { "epoch": 2.46, "learning_rate": 3.176255790442563e-05, "loss": 0.7395, "step": 40758 }, { "epoch": 2.46, "learning_rate": 3.176104899431141e-05, "loss": 0.5782, "step": 40760 }, { "epoch": 2.46, "learning_rate": 3.1759540084197185e-05, "loss": 0.482, "step": 40762 }, { "epoch": 2.46, "learning_rate": 3.1758031174082965e-05, "loss": 0.5685, "step": 40764 }, { "epoch": 2.46, "learning_rate": 3.175652226396874e-05, "loss": 0.4722, "step": 40766 }, { "epoch": 2.46, "learning_rate": 3.175501335385451e-05, "loss": 0.6064, "step": 40768 }, { "epoch": 2.46, "learning_rate": 3.175350444374029e-05, "loss": 0.7068, "step": 40770 }, { "epoch": 2.46, "learning_rate": 3.1751995533626064e-05, "loss": 0.4947, "step": 40772 }, { "epoch": 2.46, "learning_rate": 3.175048662351184e-05, "loss": 0.5103, "step": 40774 }, { "epoch": 2.46, "learning_rate": 3.174897771339762e-05, "loss": 0.564, "step": 40776 }, { "epoch": 2.46, "learning_rate": 3.174746880328339e-05, "loss": 0.5613, "step": 40778 }, { "epoch": 2.46, "learning_rate": 3.174595989316916e-05, "loss": 0.507, "step": 40780 }, { "epoch": 2.46, "learning_rate": 3.174445098305494e-05, "loss": 0.418, "step": 40782 }, { "epoch": 2.46, "learning_rate": 3.1742942072940716e-05, "loss": 0.5967, "step": 40784 }, { "epoch": 2.46, "learning_rate": 3.174143316282649e-05, "loss": 0.5929, "step": 40786 }, { "epoch": 2.46, "learning_rate": 3.173992425271227e-05, "loss": 0.651, "step": 40788 }, { "epoch": 2.46, "learning_rate": 3.173841534259804e-05, "loss": 0.5363, "step": 40790 }, { "epoch": 2.46, "learning_rate": 3.173690643248382e-05, "loss": 0.5532, "step": 40792 }, { "epoch": 2.46, "learning_rate": 3.173539752236959e-05, "loss": 0.6872, "step": 40794 }, { "epoch": 2.46, "learning_rate": 3.173388861225537e-05, "loss": 0.4505, "step": 40796 }, { "epoch": 2.46, "learning_rate": 3.173237970214115e-05, "loss": 0.579, "step": 40798 }, { "epoch": 2.46, "learning_rate": 3.1730870792026915e-05, "loss": 0.5948, "step": 40800 }, { "epoch": 2.46, "learning_rate": 3.1729361881912695e-05, "loss": 0.5591, "step": 40802 }, { "epoch": 2.46, "learning_rate": 3.1727852971798475e-05, "loss": 0.5203, "step": 40804 }, { "epoch": 2.46, "learning_rate": 3.172634406168425e-05, "loss": 0.4532, "step": 40806 }, { "epoch": 2.46, "learning_rate": 3.172483515157002e-05, "loss": 0.7407, "step": 40808 }, { "epoch": 2.46, "learning_rate": 3.1723326241455794e-05, "loss": 0.5693, "step": 40810 }, { "epoch": 2.46, "learning_rate": 3.1721817331341574e-05, "loss": 0.6755, "step": 40812 }, { "epoch": 2.46, "learning_rate": 3.1720308421227354e-05, "loss": 0.6395, "step": 40814 }, { "epoch": 2.46, "learning_rate": 3.171879951111312e-05, "loss": 0.8748, "step": 40816 }, { "epoch": 2.46, "learning_rate": 3.17172906009989e-05, "loss": 0.7962, "step": 40818 }, { "epoch": 2.46, "learning_rate": 3.171578169088468e-05, "loss": 0.498, "step": 40820 }, { "epoch": 2.46, "learning_rate": 3.1714272780770446e-05, "loss": 0.6337, "step": 40822 }, { "epoch": 2.46, "learning_rate": 3.1712763870656226e-05, "loss": 0.5318, "step": 40824 }, { "epoch": 2.46, "learning_rate": 3.1711254960542e-05, "loss": 0.5582, "step": 40826 }, { "epoch": 2.46, "learning_rate": 3.170974605042778e-05, "loss": 0.5405, "step": 40828 }, { "epoch": 2.46, "learning_rate": 3.170823714031355e-05, "loss": 0.5802, "step": 40830 }, { "epoch": 2.46, "learning_rate": 3.1706728230199326e-05, "loss": 0.6102, "step": 40832 }, { "epoch": 2.46, "learning_rate": 3.1705219320085106e-05, "loss": 0.484, "step": 40834 }, { "epoch": 2.46, "learning_rate": 3.170371040997088e-05, "loss": 0.4982, "step": 40836 }, { "epoch": 2.46, "learning_rate": 3.170220149985665e-05, "loss": 0.5723, "step": 40838 }, { "epoch": 2.46, "learning_rate": 3.170069258974243e-05, "loss": 0.3504, "step": 40840 }, { "epoch": 2.47, "learning_rate": 3.169918367962821e-05, "loss": 0.7701, "step": 40842 }, { "epoch": 2.47, "learning_rate": 3.169767476951398e-05, "loss": 0.6165, "step": 40844 }, { "epoch": 2.47, "learning_rate": 3.169616585939976e-05, "loss": 0.6481, "step": 40846 }, { "epoch": 2.47, "learning_rate": 3.169465694928553e-05, "loss": 0.6619, "step": 40848 }, { "epoch": 2.47, "learning_rate": 3.1693148039171304e-05, "loss": 0.8082, "step": 40850 }, { "epoch": 2.47, "learning_rate": 3.1691639129057084e-05, "loss": 0.5446, "step": 40852 }, { "epoch": 2.47, "learning_rate": 3.169013021894286e-05, "loss": 0.5014, "step": 40854 }, { "epoch": 2.47, "learning_rate": 3.168862130882864e-05, "loss": 0.5698, "step": 40856 }, { "epoch": 2.47, "learning_rate": 3.168711239871441e-05, "loss": 0.4725, "step": 40858 }, { "epoch": 2.47, "learning_rate": 3.168560348860018e-05, "loss": 0.6531, "step": 40860 }, { "epoch": 2.47, "learning_rate": 3.168409457848596e-05, "loss": 0.4381, "step": 40862 }, { "epoch": 2.47, "learning_rate": 3.1682585668371736e-05, "loss": 0.4495, "step": 40864 }, { "epoch": 2.47, "learning_rate": 3.168107675825751e-05, "loss": 0.5635, "step": 40866 }, { "epoch": 2.47, "learning_rate": 3.167956784814329e-05, "loss": 0.4876, "step": 40868 }, { "epoch": 2.47, "learning_rate": 3.167805893802906e-05, "loss": 0.5611, "step": 40870 }, { "epoch": 2.47, "learning_rate": 3.1676550027914836e-05, "loss": 0.5472, "step": 40872 }, { "epoch": 2.47, "learning_rate": 3.1675041117800615e-05, "loss": 0.6084, "step": 40874 }, { "epoch": 2.47, "learning_rate": 3.167353220768639e-05, "loss": 0.6196, "step": 40876 }, { "epoch": 2.47, "learning_rate": 3.167202329757217e-05, "loss": 0.4662, "step": 40878 }, { "epoch": 2.47, "learning_rate": 3.167051438745794e-05, "loss": 0.6907, "step": 40880 }, { "epoch": 2.47, "learning_rate": 3.1669005477343715e-05, "loss": 0.579, "step": 40882 }, { "epoch": 2.47, "learning_rate": 3.1667496567229495e-05, "loss": 0.7484, "step": 40884 }, { "epoch": 2.47, "learning_rate": 3.166598765711527e-05, "loss": 0.6265, "step": 40886 }, { "epoch": 2.47, "learning_rate": 3.166447874700104e-05, "loss": 0.6601, "step": 40888 }, { "epoch": 2.47, "learning_rate": 3.166296983688682e-05, "loss": 0.6522, "step": 40890 }, { "epoch": 2.47, "learning_rate": 3.1661460926772594e-05, "loss": 0.6463, "step": 40892 }, { "epoch": 2.47, "learning_rate": 3.165995201665837e-05, "loss": 0.6343, "step": 40894 }, { "epoch": 2.47, "learning_rate": 3.165844310654415e-05, "loss": 0.4756, "step": 40896 }, { "epoch": 2.47, "learning_rate": 3.165693419642992e-05, "loss": 0.7944, "step": 40898 }, { "epoch": 2.47, "learning_rate": 3.165542528631569e-05, "loss": 0.5833, "step": 40900 }, { "epoch": 2.47, "learning_rate": 3.165391637620147e-05, "loss": 0.6081, "step": 40902 }, { "epoch": 2.47, "learning_rate": 3.1652407466087246e-05, "loss": 0.5587, "step": 40904 }, { "epoch": 2.47, "learning_rate": 3.1650898555973026e-05, "loss": 0.5481, "step": 40906 }, { "epoch": 2.47, "learning_rate": 3.164938964585879e-05, "loss": 0.6693, "step": 40908 }, { "epoch": 2.47, "learning_rate": 3.164788073574457e-05, "loss": 0.6002, "step": 40910 }, { "epoch": 2.47, "learning_rate": 3.164637182563035e-05, "loss": 0.5495, "step": 40912 }, { "epoch": 2.47, "learning_rate": 3.1644862915516125e-05, "loss": 0.5781, "step": 40914 }, { "epoch": 2.47, "learning_rate": 3.16433540054019e-05, "loss": 0.4606, "step": 40916 }, { "epoch": 2.47, "learning_rate": 3.164184509528768e-05, "loss": 0.4442, "step": 40918 }, { "epoch": 2.47, "learning_rate": 3.164033618517345e-05, "loss": 0.5617, "step": 40920 }, { "epoch": 2.47, "learning_rate": 3.1638827275059225e-05, "loss": 0.5247, "step": 40922 }, { "epoch": 2.47, "learning_rate": 3.1637318364945e-05, "loss": 0.6547, "step": 40924 }, { "epoch": 2.47, "learning_rate": 3.163580945483078e-05, "loss": 0.4929, "step": 40926 }, { "epoch": 2.47, "learning_rate": 3.163430054471656e-05, "loss": 0.4998, "step": 40928 }, { "epoch": 2.47, "learning_rate": 3.1632791634602324e-05, "loss": 0.6957, "step": 40930 }, { "epoch": 2.47, "learning_rate": 3.1631282724488104e-05, "loss": 0.5009, "step": 40932 }, { "epoch": 2.47, "learning_rate": 3.1629773814373884e-05, "loss": 0.6464, "step": 40934 }, { "epoch": 2.47, "learning_rate": 3.162826490425965e-05, "loss": 0.4858, "step": 40936 }, { "epoch": 2.47, "learning_rate": 3.162675599414543e-05, "loss": 0.659, "step": 40938 }, { "epoch": 2.47, "learning_rate": 3.16252470840312e-05, "loss": 0.5961, "step": 40940 }, { "epoch": 2.47, "learning_rate": 3.162373817391698e-05, "loss": 0.6501, "step": 40942 }, { "epoch": 2.47, "learning_rate": 3.1622229263802756e-05, "loss": 0.5341, "step": 40944 }, { "epoch": 2.47, "learning_rate": 3.162072035368853e-05, "loss": 0.6394, "step": 40946 }, { "epoch": 2.47, "learning_rate": 3.161921144357431e-05, "loss": 0.5431, "step": 40948 }, { "epoch": 2.47, "learning_rate": 3.161770253346008e-05, "loss": 0.4505, "step": 40950 }, { "epoch": 2.47, "learning_rate": 3.1616193623345855e-05, "loss": 0.548, "step": 40952 }, { "epoch": 2.47, "learning_rate": 3.1614684713231635e-05, "loss": 0.5718, "step": 40954 }, { "epoch": 2.47, "learning_rate": 3.1613175803117415e-05, "loss": 0.4873, "step": 40956 }, { "epoch": 2.47, "learning_rate": 3.161166689300318e-05, "loss": 0.8069, "step": 40958 }, { "epoch": 2.47, "learning_rate": 3.161015798288896e-05, "loss": 0.5728, "step": 40960 }, { "epoch": 2.47, "learning_rate": 3.1608649072774735e-05, "loss": 0.6798, "step": 40962 }, { "epoch": 2.47, "learning_rate": 3.1607140162660515e-05, "loss": 0.7248, "step": 40964 }, { "epoch": 2.47, "learning_rate": 3.160563125254629e-05, "loss": 0.8793, "step": 40966 }, { "epoch": 2.47, "learning_rate": 3.160412234243206e-05, "loss": 0.6494, "step": 40968 }, { "epoch": 2.47, "learning_rate": 3.160261343231784e-05, "loss": 0.5507, "step": 40970 }, { "epoch": 2.47, "learning_rate": 3.1601104522203614e-05, "loss": 0.6174, "step": 40972 }, { "epoch": 2.47, "learning_rate": 3.159959561208939e-05, "loss": 0.6248, "step": 40974 }, { "epoch": 2.47, "learning_rate": 3.159808670197517e-05, "loss": 0.4071, "step": 40976 }, { "epoch": 2.47, "learning_rate": 3.159657779186094e-05, "loss": 0.4223, "step": 40978 }, { "epoch": 2.47, "learning_rate": 3.159506888174671e-05, "loss": 0.5473, "step": 40980 }, { "epoch": 2.47, "learning_rate": 3.159355997163249e-05, "loss": 0.4981, "step": 40982 }, { "epoch": 2.47, "learning_rate": 3.1592051061518266e-05, "loss": 0.6835, "step": 40984 }, { "epoch": 2.47, "learning_rate": 3.159054215140404e-05, "loss": 0.6253, "step": 40986 }, { "epoch": 2.47, "learning_rate": 3.158903324128982e-05, "loss": 0.7291, "step": 40988 }, { "epoch": 2.47, "learning_rate": 3.158752433117559e-05, "loss": 0.5778, "step": 40990 }, { "epoch": 2.47, "learning_rate": 3.158601542106137e-05, "loss": 0.6146, "step": 40992 }, { "epoch": 2.47, "learning_rate": 3.1584506510947145e-05, "loss": 0.6073, "step": 40994 }, { "epoch": 2.47, "learning_rate": 3.158299760083292e-05, "loss": 0.705, "step": 40996 }, { "epoch": 2.47, "learning_rate": 3.15814886907187e-05, "loss": 0.3862, "step": 40998 }, { "epoch": 2.47, "learning_rate": 3.157997978060447e-05, "loss": 0.4237, "step": 41000 }, { "epoch": 2.47, "learning_rate": 3.1578470870490245e-05, "loss": 0.4036, "step": 41002 }, { "epoch": 2.47, "learning_rate": 3.1576961960376024e-05, "loss": 0.3582, "step": 41004 }, { "epoch": 2.48, "learning_rate": 3.15754530502618e-05, "loss": 0.669, "step": 41006 }, { "epoch": 2.48, "learning_rate": 3.157394414014757e-05, "loss": 0.8694, "step": 41008 }, { "epoch": 2.48, "learning_rate": 3.157243523003335e-05, "loss": 0.5759, "step": 41010 }, { "epoch": 2.48, "learning_rate": 3.1570926319919124e-05, "loss": 0.7081, "step": 41012 }, { "epoch": 2.48, "learning_rate": 3.15694174098049e-05, "loss": 0.4437, "step": 41014 }, { "epoch": 2.48, "learning_rate": 3.156790849969068e-05, "loss": 0.513, "step": 41016 }, { "epoch": 2.48, "learning_rate": 3.156639958957645e-05, "loss": 0.7625, "step": 41018 }, { "epoch": 2.48, "learning_rate": 3.156489067946223e-05, "loss": 0.6988, "step": 41020 }, { "epoch": 2.48, "learning_rate": 3.1563381769347996e-05, "loss": 0.5953, "step": 41022 }, { "epoch": 2.48, "learning_rate": 3.1561872859233776e-05, "loss": 0.3156, "step": 41024 }, { "epoch": 2.48, "learning_rate": 3.1560363949119556e-05, "loss": 0.4382, "step": 41026 }, { "epoch": 2.48, "learning_rate": 3.155885503900533e-05, "loss": 0.5565, "step": 41028 }, { "epoch": 2.48, "learning_rate": 3.15573461288911e-05, "loss": 0.4832, "step": 41030 }, { "epoch": 2.48, "learning_rate": 3.155583721877688e-05, "loss": 0.3981, "step": 41032 }, { "epoch": 2.48, "learning_rate": 3.1554328308662655e-05, "loss": 0.4513, "step": 41034 }, { "epoch": 2.48, "learning_rate": 3.155281939854843e-05, "loss": 0.6322, "step": 41036 }, { "epoch": 2.48, "learning_rate": 3.15513104884342e-05, "loss": 0.4025, "step": 41038 }, { "epoch": 2.48, "learning_rate": 3.154980157831998e-05, "loss": 0.5467, "step": 41040 }, { "epoch": 2.48, "learning_rate": 3.154829266820576e-05, "loss": 0.4126, "step": 41042 }, { "epoch": 2.48, "learning_rate": 3.154678375809153e-05, "loss": 0.621, "step": 41044 }, { "epoch": 2.48, "learning_rate": 3.154527484797731e-05, "loss": 0.4963, "step": 41046 }, { "epoch": 2.48, "learning_rate": 3.154376593786309e-05, "loss": 0.447, "step": 41048 }, { "epoch": 2.48, "learning_rate": 3.1542257027748854e-05, "loss": 0.4628, "step": 41050 }, { "epoch": 2.48, "learning_rate": 3.1540748117634634e-05, "loss": 0.4463, "step": 41052 }, { "epoch": 2.48, "learning_rate": 3.153923920752041e-05, "loss": 0.637, "step": 41054 }, { "epoch": 2.48, "learning_rate": 3.153773029740619e-05, "loss": 0.5859, "step": 41056 }, { "epoch": 2.48, "learning_rate": 3.153622138729196e-05, "loss": 0.4561, "step": 41058 }, { "epoch": 2.48, "learning_rate": 3.153471247717773e-05, "loss": 0.6122, "step": 41060 }, { "epoch": 2.48, "learning_rate": 3.153320356706351e-05, "loss": 0.7326, "step": 41062 }, { "epoch": 2.48, "learning_rate": 3.1531694656949286e-05, "loss": 0.6659, "step": 41064 }, { "epoch": 2.48, "learning_rate": 3.153018574683506e-05, "loss": 0.5708, "step": 41066 }, { "epoch": 2.48, "learning_rate": 3.152867683672084e-05, "loss": 0.7995, "step": 41068 }, { "epoch": 2.48, "learning_rate": 3.152716792660661e-05, "loss": 0.4704, "step": 41070 }, { "epoch": 2.48, "learning_rate": 3.1525659016492385e-05, "loss": 0.6146, "step": 41072 }, { "epoch": 2.48, "learning_rate": 3.1524150106378165e-05, "loss": 0.5605, "step": 41074 }, { "epoch": 2.48, "learning_rate": 3.152264119626394e-05, "loss": 0.6417, "step": 41076 }, { "epoch": 2.48, "learning_rate": 3.152113228614972e-05, "loss": 0.7083, "step": 41078 }, { "epoch": 2.48, "learning_rate": 3.151962337603549e-05, "loss": 0.6749, "step": 41080 }, { "epoch": 2.48, "learning_rate": 3.1518114465921264e-05, "loss": 0.3752, "step": 41082 }, { "epoch": 2.48, "learning_rate": 3.1516605555807044e-05, "loss": 0.6451, "step": 41084 }, { "epoch": 2.48, "learning_rate": 3.151509664569282e-05, "loss": 0.5997, "step": 41086 }, { "epoch": 2.48, "learning_rate": 3.151358773557859e-05, "loss": 0.6806, "step": 41088 }, { "epoch": 2.48, "learning_rate": 3.151207882546437e-05, "loss": 0.5863, "step": 41090 }, { "epoch": 2.48, "learning_rate": 3.1510569915350144e-05, "loss": 0.4738, "step": 41092 }, { "epoch": 2.48, "learning_rate": 3.150906100523592e-05, "loss": 0.5697, "step": 41094 }, { "epoch": 2.48, "learning_rate": 3.15075520951217e-05, "loss": 0.8129, "step": 41096 }, { "epoch": 2.48, "learning_rate": 3.150604318500747e-05, "loss": 0.5556, "step": 41098 }, { "epoch": 2.48, "learning_rate": 3.150453427489324e-05, "loss": 0.5273, "step": 41100 }, { "epoch": 2.48, "learning_rate": 3.150302536477902e-05, "loss": 0.6157, "step": 41102 }, { "epoch": 2.48, "learning_rate": 3.1501516454664796e-05, "loss": 0.5585, "step": 41104 }, { "epoch": 2.48, "learning_rate": 3.1500007544550576e-05, "loss": 0.5482, "step": 41106 }, { "epoch": 2.48, "learning_rate": 3.149849863443635e-05, "loss": 0.5071, "step": 41108 }, { "epoch": 2.48, "learning_rate": 3.149698972432212e-05, "loss": 0.6716, "step": 41110 }, { "epoch": 2.48, "learning_rate": 3.14954808142079e-05, "loss": 0.5261, "step": 41112 }, { "epoch": 2.48, "learning_rate": 3.1493971904093675e-05, "loss": 0.649, "step": 41114 }, { "epoch": 2.48, "learning_rate": 3.149246299397945e-05, "loss": 0.5292, "step": 41116 }, { "epoch": 2.48, "learning_rate": 3.149095408386523e-05, "loss": 0.4979, "step": 41118 }, { "epoch": 2.48, "learning_rate": 3.1489445173751e-05, "loss": 0.4203, "step": 41120 }, { "epoch": 2.48, "learning_rate": 3.1487936263636774e-05, "loss": 0.7559, "step": 41122 }, { "epoch": 2.48, "learning_rate": 3.1486427353522554e-05, "loss": 0.5736, "step": 41124 }, { "epoch": 2.48, "learning_rate": 3.148491844340833e-05, "loss": 0.5309, "step": 41126 }, { "epoch": 2.48, "learning_rate": 3.148340953329411e-05, "loss": 0.4724, "step": 41128 }, { "epoch": 2.48, "learning_rate": 3.148190062317988e-05, "loss": 0.589, "step": 41130 }, { "epoch": 2.48, "learning_rate": 3.1480391713065654e-05, "loss": 0.4791, "step": 41132 }, { "epoch": 2.48, "learning_rate": 3.1478882802951434e-05, "loss": 0.487, "step": 41134 }, { "epoch": 2.48, "learning_rate": 3.14773738928372e-05, "loss": 0.7462, "step": 41136 }, { "epoch": 2.48, "learning_rate": 3.147586498272298e-05, "loss": 0.6362, "step": 41138 }, { "epoch": 2.48, "learning_rate": 3.147435607260876e-05, "loss": 0.4754, "step": 41140 }, { "epoch": 2.48, "learning_rate": 3.147284716249453e-05, "loss": 0.6702, "step": 41142 }, { "epoch": 2.48, "learning_rate": 3.1471338252380306e-05, "loss": 0.5903, "step": 41144 }, { "epoch": 2.48, "learning_rate": 3.1469829342266086e-05, "loss": 0.6017, "step": 41146 }, { "epoch": 2.48, "learning_rate": 3.146832043215186e-05, "loss": 0.467, "step": 41148 }, { "epoch": 2.48, "learning_rate": 3.146681152203763e-05, "loss": 0.555, "step": 41150 }, { "epoch": 2.48, "learning_rate": 3.1465302611923405e-05, "loss": 0.879, "step": 41152 }, { "epoch": 2.48, "learning_rate": 3.1463793701809185e-05, "loss": 0.4502, "step": 41154 }, { "epoch": 2.48, "learning_rate": 3.1462284791694965e-05, "loss": 0.4334, "step": 41156 }, { "epoch": 2.48, "learning_rate": 3.146077588158073e-05, "loss": 0.5952, "step": 41158 }, { "epoch": 2.48, "learning_rate": 3.145926697146651e-05, "loss": 0.648, "step": 41160 }, { "epoch": 2.48, "learning_rate": 3.145775806135229e-05, "loss": 0.6574, "step": 41162 }, { "epoch": 2.48, "learning_rate": 3.145624915123806e-05, "loss": 0.833, "step": 41164 }, { "epoch": 2.48, "learning_rate": 3.145474024112384e-05, "loss": 0.5928, "step": 41166 }, { "epoch": 2.48, "learning_rate": 3.145323133100961e-05, "loss": 0.5679, "step": 41168 }, { "epoch": 2.48, "learning_rate": 3.145172242089539e-05, "loss": 0.4833, "step": 41170 }, { "epoch": 2.49, "learning_rate": 3.1450213510781164e-05, "loss": 0.359, "step": 41172 }, { "epoch": 2.49, "learning_rate": 3.144870460066694e-05, "loss": 0.4871, "step": 41174 }, { "epoch": 2.49, "learning_rate": 3.1447195690552717e-05, "loss": 0.5604, "step": 41176 }, { "epoch": 2.49, "learning_rate": 3.144568678043849e-05, "loss": 0.2845, "step": 41178 }, { "epoch": 2.49, "learning_rate": 3.144417787032426e-05, "loss": 0.6782, "step": 41180 }, { "epoch": 2.49, "learning_rate": 3.144266896021004e-05, "loss": 0.9345, "step": 41182 }, { "epoch": 2.49, "learning_rate": 3.1441160050095816e-05, "loss": 0.505, "step": 41184 }, { "epoch": 2.49, "learning_rate": 3.143965113998159e-05, "loss": 0.656, "step": 41186 }, { "epoch": 2.49, "learning_rate": 3.143814222986737e-05, "loss": 0.6962, "step": 41188 }, { "epoch": 2.49, "learning_rate": 3.143663331975314e-05, "loss": 0.6018, "step": 41190 }, { "epoch": 2.49, "learning_rate": 3.143512440963892e-05, "loss": 0.4347, "step": 41192 }, { "epoch": 2.49, "learning_rate": 3.1433615499524695e-05, "loss": 0.5527, "step": 41194 }, { "epoch": 2.49, "learning_rate": 3.143210658941047e-05, "loss": 0.7647, "step": 41196 }, { "epoch": 2.49, "learning_rate": 3.143059767929625e-05, "loss": 0.4047, "step": 41198 }, { "epoch": 2.49, "learning_rate": 3.142908876918202e-05, "loss": 0.5925, "step": 41200 }, { "epoch": 2.49, "learning_rate": 3.1427579859067794e-05, "loss": 0.644, "step": 41202 }, { "epoch": 2.49, "learning_rate": 3.1426070948953574e-05, "loss": 0.5098, "step": 41204 }, { "epoch": 2.49, "learning_rate": 3.142456203883935e-05, "loss": 0.6384, "step": 41206 }, { "epoch": 2.49, "learning_rate": 3.142305312872512e-05, "loss": 0.5946, "step": 41208 }, { "epoch": 2.49, "learning_rate": 3.14215442186109e-05, "loss": 0.5944, "step": 41210 }, { "epoch": 2.49, "learning_rate": 3.1420035308496673e-05, "loss": 0.5748, "step": 41212 }, { "epoch": 2.49, "learning_rate": 3.1418526398382447e-05, "loss": 0.7711, "step": 41214 }, { "epoch": 2.49, "learning_rate": 3.1417017488268227e-05, "loss": 0.5675, "step": 41216 }, { "epoch": 2.49, "learning_rate": 3.1415508578154e-05, "loss": 0.5149, "step": 41218 }, { "epoch": 2.49, "learning_rate": 3.141399966803978e-05, "loss": 0.6605, "step": 41220 }, { "epoch": 2.49, "learning_rate": 3.141249075792555e-05, "loss": 0.4096, "step": 41222 }, { "epoch": 2.49, "learning_rate": 3.1410981847811326e-05, "loss": 0.5063, "step": 41224 }, { "epoch": 2.49, "learning_rate": 3.1409472937697106e-05, "loss": 0.6223, "step": 41226 }, { "epoch": 2.49, "learning_rate": 3.140796402758288e-05, "loss": 0.7587, "step": 41228 }, { "epoch": 2.49, "learning_rate": 3.140645511746865e-05, "loss": 0.4135, "step": 41230 }, { "epoch": 2.49, "learning_rate": 3.140494620735443e-05, "loss": 0.5631, "step": 41232 }, { "epoch": 2.49, "learning_rate": 3.1403437297240205e-05, "loss": 0.6502, "step": 41234 }, { "epoch": 2.49, "learning_rate": 3.140192838712598e-05, "loss": 0.5411, "step": 41236 }, { "epoch": 2.49, "learning_rate": 3.140041947701176e-05, "loss": 0.7066, "step": 41238 }, { "epoch": 2.49, "learning_rate": 3.139891056689753e-05, "loss": 0.4814, "step": 41240 }, { "epoch": 2.49, "learning_rate": 3.139740165678331e-05, "loss": 0.5359, "step": 41242 }, { "epoch": 2.49, "learning_rate": 3.1395892746669084e-05, "loss": 0.8018, "step": 41244 }, { "epoch": 2.49, "learning_rate": 3.139438383655486e-05, "loss": 0.4354, "step": 41246 }, { "epoch": 2.49, "learning_rate": 3.139287492644064e-05, "loss": 0.5787, "step": 41248 }, { "epoch": 2.49, "learning_rate": 3.1391366016326404e-05, "loss": 0.5602, "step": 41250 }, { "epoch": 2.49, "learning_rate": 3.1389857106212183e-05, "loss": 0.426, "step": 41252 }, { "epoch": 2.49, "learning_rate": 3.138834819609796e-05, "loss": 0.4, "step": 41254 }, { "epoch": 2.49, "learning_rate": 3.1386839285983736e-05, "loss": 0.7627, "step": 41256 }, { "epoch": 2.49, "learning_rate": 3.138533037586951e-05, "loss": 0.6009, "step": 41258 }, { "epoch": 2.49, "learning_rate": 3.138382146575529e-05, "loss": 0.7539, "step": 41260 }, { "epoch": 2.49, "learning_rate": 3.138231255564106e-05, "loss": 0.6617, "step": 41262 }, { "epoch": 2.49, "learning_rate": 3.1380803645526836e-05, "loss": 0.7841, "step": 41264 }, { "epoch": 2.49, "learning_rate": 3.137929473541261e-05, "loss": 0.3726, "step": 41266 }, { "epoch": 2.49, "learning_rate": 3.137778582529839e-05, "loss": 0.6018, "step": 41268 }, { "epoch": 2.49, "learning_rate": 3.137627691518417e-05, "loss": 0.6424, "step": 41270 }, { "epoch": 2.49, "learning_rate": 3.1374768005069935e-05, "loss": 0.6488, "step": 41272 }, { "epoch": 2.49, "learning_rate": 3.1373259094955715e-05, "loss": 0.5579, "step": 41274 }, { "epoch": 2.49, "learning_rate": 3.1371750184841495e-05, "loss": 0.6127, "step": 41276 }, { "epoch": 2.49, "learning_rate": 3.137024127472726e-05, "loss": 0.5575, "step": 41278 }, { "epoch": 2.49, "learning_rate": 3.136873236461304e-05, "loss": 0.5983, "step": 41280 }, { "epoch": 2.49, "learning_rate": 3.1367223454498814e-05, "loss": 0.6937, "step": 41282 }, { "epoch": 2.49, "learning_rate": 3.1365714544384594e-05, "loss": 0.5661, "step": 41284 }, { "epoch": 2.49, "learning_rate": 3.136420563427037e-05, "loss": 0.4211, "step": 41286 }, { "epoch": 2.49, "learning_rate": 3.136269672415614e-05, "loss": 0.6878, "step": 41288 }, { "epoch": 2.49, "learning_rate": 3.136118781404192e-05, "loss": 0.5787, "step": 41290 }, { "epoch": 2.49, "learning_rate": 3.13596789039277e-05, "loss": 0.5304, "step": 41292 }, { "epoch": 2.49, "learning_rate": 3.1358169993813466e-05, "loss": 0.6127, "step": 41294 }, { "epoch": 2.49, "learning_rate": 3.1356661083699246e-05, "loss": 0.5688, "step": 41296 }, { "epoch": 2.49, "learning_rate": 3.135515217358502e-05, "loss": 0.6592, "step": 41298 }, { "epoch": 2.49, "learning_rate": 3.135364326347079e-05, "loss": 0.6008, "step": 41300 }, { "epoch": 2.49, "learning_rate": 3.135213435335657e-05, "loss": 0.5777, "step": 41302 }, { "epoch": 2.49, "learning_rate": 3.1350625443242346e-05, "loss": 0.6225, "step": 41304 }, { "epoch": 2.49, "learning_rate": 3.1349116533128126e-05, "loss": 0.5563, "step": 41306 }, { "epoch": 2.49, "learning_rate": 3.13476076230139e-05, "loss": 0.5278, "step": 41308 }, { "epoch": 2.49, "learning_rate": 3.134609871289967e-05, "loss": 0.7148, "step": 41310 }, { "epoch": 2.49, "learning_rate": 3.134458980278545e-05, "loss": 0.4652, "step": 41312 }, { "epoch": 2.49, "learning_rate": 3.1343080892671225e-05, "loss": 0.4845, "step": 41314 }, { "epoch": 2.49, "learning_rate": 3.1341571982557e-05, "loss": 0.6306, "step": 41316 }, { "epoch": 2.49, "learning_rate": 3.134006307244278e-05, "loss": 0.602, "step": 41318 }, { "epoch": 2.49, "learning_rate": 3.133855416232855e-05, "loss": 0.6256, "step": 41320 }, { "epoch": 2.49, "learning_rate": 3.1337045252214324e-05, "loss": 0.57, "step": 41322 }, { "epoch": 2.49, "learning_rate": 3.1335536342100104e-05, "loss": 0.571, "step": 41324 }, { "epoch": 2.49, "learning_rate": 3.133402743198588e-05, "loss": 0.5104, "step": 41326 }, { "epoch": 2.49, "learning_rate": 3.133251852187165e-05, "loss": 0.7176, "step": 41328 }, { "epoch": 2.49, "learning_rate": 3.133100961175743e-05, "loss": 0.7653, "step": 41330 }, { "epoch": 2.49, "learning_rate": 3.13295007016432e-05, "loss": 0.6416, "step": 41332 }, { "epoch": 2.49, "learning_rate": 3.132799179152898e-05, "loss": 0.9297, "step": 41334 }, { "epoch": 2.49, "learning_rate": 3.1326482881414756e-05, "loss": 0.4679, "step": 41336 }, { "epoch": 2.5, "learning_rate": 3.132497397130053e-05, "loss": 0.6149, "step": 41338 }, { "epoch": 2.5, "learning_rate": 3.132346506118631e-05, "loss": 0.6975, "step": 41340 }, { "epoch": 2.5, "learning_rate": 3.132195615107208e-05, "loss": 0.5483, "step": 41342 }, { "epoch": 2.5, "learning_rate": 3.1320447240957856e-05, "loss": 0.5973, "step": 41344 }, { "epoch": 2.5, "learning_rate": 3.1318938330843636e-05, "loss": 0.4969, "step": 41346 }, { "epoch": 2.5, "learning_rate": 3.131742942072941e-05, "loss": 0.4561, "step": 41348 }, { "epoch": 2.5, "learning_rate": 3.131592051061518e-05, "loss": 0.6363, "step": 41350 }, { "epoch": 2.5, "learning_rate": 3.131441160050096e-05, "loss": 0.5844, "step": 41352 }, { "epoch": 2.5, "learning_rate": 3.1312902690386735e-05, "loss": 0.5591, "step": 41354 }, { "epoch": 2.5, "learning_rate": 3.1311393780272515e-05, "loss": 0.4128, "step": 41356 }, { "epoch": 2.5, "learning_rate": 3.130988487015829e-05, "loss": 0.5181, "step": 41358 }, { "epoch": 2.5, "learning_rate": 3.130837596004406e-05, "loss": 0.5013, "step": 41360 }, { "epoch": 2.5, "learning_rate": 3.130686704992984e-05, "loss": 0.4585, "step": 41362 }, { "epoch": 2.5, "learning_rate": 3.130535813981561e-05, "loss": 0.5912, "step": 41364 }, { "epoch": 2.5, "learning_rate": 3.130384922970139e-05, "loss": 0.7494, "step": 41366 }, { "epoch": 2.5, "learning_rate": 3.130234031958717e-05, "loss": 0.5492, "step": 41368 }, { "epoch": 2.5, "learning_rate": 3.130083140947294e-05, "loss": 0.5892, "step": 41370 }, { "epoch": 2.5, "learning_rate": 3.129932249935871e-05, "loss": 0.7463, "step": 41372 }, { "epoch": 2.5, "learning_rate": 3.129781358924449e-05, "loss": 0.5199, "step": 41374 }, { "epoch": 2.5, "learning_rate": 3.1296304679130266e-05, "loss": 0.6283, "step": 41376 }, { "epoch": 2.5, "learning_rate": 3.129479576901604e-05, "loss": 0.6342, "step": 41378 }, { "epoch": 2.5, "learning_rate": 3.129328685890181e-05, "loss": 0.4806, "step": 41380 }, { "epoch": 2.5, "learning_rate": 3.129177794878759e-05, "loss": 0.5539, "step": 41382 }, { "epoch": 2.5, "learning_rate": 3.129026903867337e-05, "loss": 0.3032, "step": 41384 }, { "epoch": 2.5, "learning_rate": 3.128876012855914e-05, "loss": 0.586, "step": 41386 }, { "epoch": 2.5, "learning_rate": 3.128725121844492e-05, "loss": 0.6886, "step": 41388 }, { "epoch": 2.5, "learning_rate": 3.12857423083307e-05, "loss": 0.6884, "step": 41390 }, { "epoch": 2.5, "learning_rate": 3.1284233398216465e-05, "loss": 0.5227, "step": 41392 }, { "epoch": 2.5, "learning_rate": 3.1282724488102245e-05, "loss": 0.4964, "step": 41394 }, { "epoch": 2.5, "learning_rate": 3.128121557798802e-05, "loss": 0.3879, "step": 41396 }, { "epoch": 2.5, "learning_rate": 3.12797066678738e-05, "loss": 0.587, "step": 41398 }, { "epoch": 2.5, "learning_rate": 3.127819775775957e-05, "loss": 0.474, "step": 41400 }, { "epoch": 2.5, "learning_rate": 3.1276688847645344e-05, "loss": 0.4904, "step": 41402 }, { "epoch": 2.5, "learning_rate": 3.1275179937531124e-05, "loss": 0.6146, "step": 41404 }, { "epoch": 2.5, "learning_rate": 3.1273671027416904e-05, "loss": 0.4442, "step": 41406 }, { "epoch": 2.5, "learning_rate": 3.127216211730267e-05, "loss": 0.4965, "step": 41408 }, { "epoch": 2.5, "learning_rate": 3.127065320718845e-05, "loss": 0.6082, "step": 41410 }, { "epoch": 2.5, "learning_rate": 3.126914429707422e-05, "loss": 0.6269, "step": 41412 }, { "epoch": 2.5, "learning_rate": 3.1267635386959996e-05, "loss": 0.5764, "step": 41414 }, { "epoch": 2.5, "learning_rate": 3.1266126476845776e-05, "loss": 0.3951, "step": 41416 }, { "epoch": 2.5, "learning_rate": 3.126461756673155e-05, "loss": 0.6025, "step": 41418 }, { "epoch": 2.5, "learning_rate": 3.126310865661733e-05, "loss": 0.5865, "step": 41420 }, { "epoch": 2.5, "learning_rate": 3.12615997465031e-05, "loss": 0.6956, "step": 41422 }, { "epoch": 2.5, "learning_rate": 3.1260090836388876e-05, "loss": 0.4307, "step": 41424 }, { "epoch": 2.5, "learning_rate": 3.1258581926274655e-05, "loss": 0.5806, "step": 41426 }, { "epoch": 2.5, "learning_rate": 3.125707301616043e-05, "loss": 0.5835, "step": 41428 }, { "epoch": 2.5, "learning_rate": 3.12555641060462e-05, "loss": 0.6023, "step": 41430 }, { "epoch": 2.5, "learning_rate": 3.125405519593198e-05, "loss": 0.5148, "step": 41432 }, { "epoch": 2.5, "learning_rate": 3.1252546285817755e-05, "loss": 0.5548, "step": 41434 }, { "epoch": 2.5, "learning_rate": 3.125103737570353e-05, "loss": 0.4822, "step": 41436 }, { "epoch": 2.5, "learning_rate": 3.124952846558931e-05, "loss": 0.7072, "step": 41438 }, { "epoch": 2.5, "learning_rate": 3.124801955547508e-05, "loss": 0.4807, "step": 41440 }, { "epoch": 2.5, "learning_rate": 3.1246510645360854e-05, "loss": 0.4937, "step": 41442 }, { "epoch": 2.5, "learning_rate": 3.1245001735246634e-05, "loss": 0.6113, "step": 41444 }, { "epoch": 2.5, "learning_rate": 3.124349282513241e-05, "loss": 0.4528, "step": 41446 }, { "epoch": 2.5, "learning_rate": 3.124198391501819e-05, "loss": 0.6329, "step": 41448 }, { "epoch": 2.5, "learning_rate": 3.124047500490396e-05, "loss": 0.5253, "step": 41450 }, { "epoch": 2.5, "learning_rate": 3.123896609478973e-05, "loss": 0.4827, "step": 41452 }, { "epoch": 2.5, "learning_rate": 3.123745718467551e-05, "loss": 0.5067, "step": 41454 }, { "epoch": 2.5, "learning_rate": 3.1235948274561286e-05, "loss": 0.5851, "step": 41456 }, { "epoch": 2.5, "learning_rate": 3.123443936444706e-05, "loss": 0.6034, "step": 41458 }, { "epoch": 2.5, "learning_rate": 3.123293045433284e-05, "loss": 0.4906, "step": 41460 }, { "epoch": 2.5, "learning_rate": 3.123142154421861e-05, "loss": 0.663, "step": 41462 }, { "epoch": 2.5, "learning_rate": 3.1229912634104385e-05, "loss": 0.6107, "step": 41464 }, { "epoch": 2.5, "learning_rate": 3.1228403723990165e-05, "loss": 0.5564, "step": 41466 }, { "epoch": 2.5, "learning_rate": 3.122689481387594e-05, "loss": 0.5176, "step": 41468 }, { "epoch": 2.5, "learning_rate": 3.122538590376172e-05, "loss": 0.6028, "step": 41470 }, { "epoch": 2.5, "learning_rate": 3.122387699364749e-05, "loss": 0.5172, "step": 41472 }, { "epoch": 2.5, "learning_rate": 3.1222368083533265e-05, "loss": 0.3446, "step": 41474 }, { "epoch": 2.5, "learning_rate": 3.1220859173419045e-05, "loss": 0.5637, "step": 41476 }, { "epoch": 2.5, "learning_rate": 3.121935026330481e-05, "loss": 0.608, "step": 41478 }, { "epoch": 2.5, "learning_rate": 3.121784135319059e-05, "loss": 0.4087, "step": 41480 }, { "epoch": 2.5, "learning_rate": 3.121633244307637e-05, "loss": 0.8203, "step": 41482 }, { "epoch": 2.5, "learning_rate": 3.1214823532962144e-05, "loss": 0.5775, "step": 41484 }, { "epoch": 2.5, "learning_rate": 3.121331462284792e-05, "loss": 0.5276, "step": 41486 }, { "epoch": 2.5, "learning_rate": 3.12118057127337e-05, "loss": 0.6024, "step": 41488 }, { "epoch": 2.5, "learning_rate": 3.121029680261947e-05, "loss": 0.6727, "step": 41490 }, { "epoch": 2.5, "learning_rate": 3.120878789250524e-05, "loss": 0.6073, "step": 41492 }, { "epoch": 2.5, "learning_rate": 3.1207278982391016e-05, "loss": 0.6482, "step": 41494 }, { "epoch": 2.5, "learning_rate": 3.1205770072276796e-05, "loss": 0.7368, "step": 41496 }, { "epoch": 2.5, "learning_rate": 3.1204261162162576e-05, "loss": 0.6192, "step": 41498 }, { "epoch": 2.5, "learning_rate": 3.120275225204834e-05, "loss": 0.5742, "step": 41500 }, { "epoch": 2.5, "learning_rate": 3.120124334193412e-05, "loss": 0.4438, "step": 41502 }, { "epoch": 2.51, "learning_rate": 3.11997344318199e-05, "loss": 0.5687, "step": 41504 }, { "epoch": 2.51, "learning_rate": 3.1198225521705675e-05, "loss": 0.4793, "step": 41506 }, { "epoch": 2.51, "learning_rate": 3.119671661159145e-05, "loss": 0.7404, "step": 41508 }, { "epoch": 2.51, "learning_rate": 3.119520770147722e-05, "loss": 0.5968, "step": 41510 }, { "epoch": 2.51, "learning_rate": 3.1193698791363e-05, "loss": 0.4038, "step": 41512 }, { "epoch": 2.51, "learning_rate": 3.1192189881248775e-05, "loss": 0.5032, "step": 41514 }, { "epoch": 2.51, "learning_rate": 3.119068097113455e-05, "loss": 0.4611, "step": 41516 }, { "epoch": 2.51, "learning_rate": 3.118917206102033e-05, "loss": 0.4106, "step": 41518 }, { "epoch": 2.51, "learning_rate": 3.118766315090611e-05, "loss": 0.5969, "step": 41520 }, { "epoch": 2.51, "learning_rate": 3.1186154240791874e-05, "loss": 0.5417, "step": 41522 }, { "epoch": 2.51, "learning_rate": 3.1184645330677654e-05, "loss": 0.6542, "step": 41524 }, { "epoch": 2.51, "learning_rate": 3.118313642056343e-05, "loss": 0.4407, "step": 41526 }, { "epoch": 2.51, "learning_rate": 3.11816275104492e-05, "loss": 0.5108, "step": 41528 }, { "epoch": 2.51, "learning_rate": 3.118011860033498e-05, "loss": 0.5711, "step": 41530 }, { "epoch": 2.51, "learning_rate": 3.117860969022075e-05, "loss": 0.8607, "step": 41532 }, { "epoch": 2.51, "learning_rate": 3.117710078010653e-05, "loss": 0.4415, "step": 41534 }, { "epoch": 2.51, "learning_rate": 3.1175591869992306e-05, "loss": 0.6421, "step": 41536 }, { "epoch": 2.51, "learning_rate": 3.117408295987808e-05, "loss": 0.5335, "step": 41538 }, { "epoch": 2.51, "learning_rate": 3.117257404976386e-05, "loss": 0.5404, "step": 41540 }, { "epoch": 2.51, "learning_rate": 3.117106513964963e-05, "loss": 0.6226, "step": 41542 }, { "epoch": 2.51, "learning_rate": 3.1169556229535405e-05, "loss": 0.4618, "step": 41544 }, { "epoch": 2.51, "learning_rate": 3.1168047319421185e-05, "loss": 0.525, "step": 41546 }, { "epoch": 2.51, "learning_rate": 3.116653840930696e-05, "loss": 0.5646, "step": 41548 }, { "epoch": 2.51, "learning_rate": 3.116502949919273e-05, "loss": 0.8315, "step": 41550 }, { "epoch": 2.51, "learning_rate": 3.116352058907851e-05, "loss": 0.497, "step": 41552 }, { "epoch": 2.51, "learning_rate": 3.1162011678964285e-05, "loss": 0.4503, "step": 41554 }, { "epoch": 2.51, "learning_rate": 3.116050276885006e-05, "loss": 1.037, "step": 41556 }, { "epoch": 2.51, "learning_rate": 3.115899385873584e-05, "loss": 0.568, "step": 41558 }, { "epoch": 2.51, "learning_rate": 3.115748494862161e-05, "loss": 0.4481, "step": 41560 }, { "epoch": 2.51, "learning_rate": 3.115597603850739e-05, "loss": 0.6003, "step": 41562 }, { "epoch": 2.51, "learning_rate": 3.1154467128393164e-05, "loss": 0.7218, "step": 41564 }, { "epoch": 2.51, "learning_rate": 3.115295821827894e-05, "loss": 0.768, "step": 41566 }, { "epoch": 2.51, "learning_rate": 3.115144930816472e-05, "loss": 0.7249, "step": 41568 }, { "epoch": 2.51, "learning_rate": 3.114994039805049e-05, "loss": 0.5537, "step": 41570 }, { "epoch": 2.51, "learning_rate": 3.114843148793626e-05, "loss": 0.5084, "step": 41572 }, { "epoch": 2.51, "learning_rate": 3.114692257782204e-05, "loss": 0.7008, "step": 41574 }, { "epoch": 2.51, "learning_rate": 3.1145413667707816e-05, "loss": 0.7805, "step": 41576 }, { "epoch": 2.51, "learning_rate": 3.114390475759359e-05, "loss": 0.4922, "step": 41578 }, { "epoch": 2.51, "learning_rate": 3.114239584747937e-05, "loss": 0.5718, "step": 41580 }, { "epoch": 2.51, "learning_rate": 3.114088693736514e-05, "loss": 0.5747, "step": 41582 }, { "epoch": 2.51, "learning_rate": 3.113937802725092e-05, "loss": 0.5012, "step": 41584 }, { "epoch": 2.51, "learning_rate": 3.1137869117136695e-05, "loss": 0.4451, "step": 41586 }, { "epoch": 2.51, "learning_rate": 3.113636020702247e-05, "loss": 0.3791, "step": 41588 }, { "epoch": 2.51, "learning_rate": 3.113485129690825e-05, "loss": 0.3495, "step": 41590 }, { "epoch": 2.51, "learning_rate": 3.1133342386794015e-05, "loss": 0.4307, "step": 41592 }, { "epoch": 2.51, "learning_rate": 3.1131833476679794e-05, "loss": 0.5856, "step": 41594 }, { "epoch": 2.51, "learning_rate": 3.1130324566565574e-05, "loss": 0.3644, "step": 41596 }, { "epoch": 2.51, "learning_rate": 3.112881565645135e-05, "loss": 0.5776, "step": 41598 }, { "epoch": 2.51, "learning_rate": 3.112730674633712e-05, "loss": 0.7324, "step": 41600 }, { "epoch": 2.51, "learning_rate": 3.11257978362229e-05, "loss": 0.6584, "step": 41602 }, { "epoch": 2.51, "learning_rate": 3.1124288926108674e-05, "loss": 0.5782, "step": 41604 }, { "epoch": 2.51, "learning_rate": 3.112278001599445e-05, "loss": 0.405, "step": 41606 }, { "epoch": 2.51, "learning_rate": 3.112127110588022e-05, "loss": 0.6137, "step": 41608 }, { "epoch": 2.51, "learning_rate": 3.1119762195766e-05, "loss": 0.3944, "step": 41610 }, { "epoch": 2.51, "learning_rate": 3.111825328565178e-05, "loss": 0.8301, "step": 41612 }, { "epoch": 2.51, "learning_rate": 3.1116744375537546e-05, "loss": 0.6279, "step": 41614 }, { "epoch": 2.51, "learning_rate": 3.1115235465423326e-05, "loss": 0.5816, "step": 41616 }, { "epoch": 2.51, "learning_rate": 3.1113726555309106e-05, "loss": 0.4099, "step": 41618 }, { "epoch": 2.51, "learning_rate": 3.111221764519488e-05, "loss": 0.5235, "step": 41620 }, { "epoch": 2.51, "learning_rate": 3.111070873508065e-05, "loss": 0.4502, "step": 41622 }, { "epoch": 2.51, "learning_rate": 3.1109199824966425e-05, "loss": 0.5794, "step": 41624 }, { "epoch": 2.51, "learning_rate": 3.1107690914852205e-05, "loss": 0.4927, "step": 41626 }, { "epoch": 2.51, "learning_rate": 3.110618200473798e-05, "loss": 0.4821, "step": 41628 }, { "epoch": 2.51, "learning_rate": 3.110467309462375e-05, "loss": 0.5224, "step": 41630 }, { "epoch": 2.51, "learning_rate": 3.110316418450953e-05, "loss": 0.541, "step": 41632 }, { "epoch": 2.51, "learning_rate": 3.110165527439531e-05, "loss": 0.4463, "step": 41634 }, { "epoch": 2.51, "learning_rate": 3.110014636428108e-05, "loss": 0.6183, "step": 41636 }, { "epoch": 2.51, "learning_rate": 3.109863745416686e-05, "loss": 0.4265, "step": 41638 }, { "epoch": 2.51, "learning_rate": 3.109712854405263e-05, "loss": 0.7897, "step": 41640 }, { "epoch": 2.51, "learning_rate": 3.1095619633938404e-05, "loss": 0.589, "step": 41642 }, { "epoch": 2.51, "learning_rate": 3.1094110723824184e-05, "loss": 0.5406, "step": 41644 }, { "epoch": 2.51, "learning_rate": 3.109260181370996e-05, "loss": 0.6269, "step": 41646 }, { "epoch": 2.51, "learning_rate": 3.109109290359574e-05, "loss": 0.7872, "step": 41648 }, { "epoch": 2.51, "learning_rate": 3.108958399348151e-05, "loss": 0.4608, "step": 41650 }, { "epoch": 2.51, "learning_rate": 3.108807508336728e-05, "loss": 0.4981, "step": 41652 }, { "epoch": 2.51, "learning_rate": 3.108656617325306e-05, "loss": 0.5074, "step": 41654 }, { "epoch": 2.51, "learning_rate": 3.1085057263138836e-05, "loss": 0.6386, "step": 41656 }, { "epoch": 2.51, "learning_rate": 3.108354835302461e-05, "loss": 0.4631, "step": 41658 }, { "epoch": 2.51, "learning_rate": 3.108203944291039e-05, "loss": 0.6859, "step": 41660 }, { "epoch": 2.51, "learning_rate": 3.108053053279616e-05, "loss": 0.5018, "step": 41662 }, { "epoch": 2.51, "learning_rate": 3.1079021622681935e-05, "loss": 0.5166, "step": 41664 }, { "epoch": 2.51, "learning_rate": 3.1077512712567715e-05, "loss": 0.5438, "step": 41666 }, { "epoch": 2.51, "learning_rate": 3.107600380245349e-05, "loss": 0.5888, "step": 41668 }, { "epoch": 2.52, "learning_rate": 3.107449489233927e-05, "loss": 0.5841, "step": 41670 }, { "epoch": 2.52, "learning_rate": 3.107298598222504e-05, "loss": 0.4952, "step": 41672 }, { "epoch": 2.52, "learning_rate": 3.1071477072110814e-05, "loss": 0.5266, "step": 41674 }, { "epoch": 2.52, "learning_rate": 3.1069968161996594e-05, "loss": 0.5626, "step": 41676 }, { "epoch": 2.52, "learning_rate": 3.106845925188237e-05, "loss": 0.5167, "step": 41678 }, { "epoch": 2.52, "learning_rate": 3.106695034176814e-05, "loss": 0.5729, "step": 41680 }, { "epoch": 2.52, "learning_rate": 3.106544143165392e-05, "loss": 0.7029, "step": 41682 }, { "epoch": 2.52, "learning_rate": 3.1063932521539694e-05, "loss": 0.54, "step": 41684 }, { "epoch": 2.52, "learning_rate": 3.106242361142547e-05, "loss": 0.4863, "step": 41686 }, { "epoch": 2.52, "learning_rate": 3.1060914701311247e-05, "loss": 0.5598, "step": 41688 }, { "epoch": 2.52, "learning_rate": 3.105940579119702e-05, "loss": 0.915, "step": 41690 }, { "epoch": 2.52, "learning_rate": 3.105789688108279e-05, "loss": 0.6869, "step": 41692 }, { "epoch": 2.52, "learning_rate": 3.105638797096857e-05, "loss": 0.5455, "step": 41694 }, { "epoch": 2.52, "learning_rate": 3.1054879060854346e-05, "loss": 0.4432, "step": 41696 }, { "epoch": 2.52, "learning_rate": 3.1053370150740126e-05, "loss": 0.5798, "step": 41698 }, { "epoch": 2.52, "learning_rate": 3.10518612406259e-05, "loss": 0.5967, "step": 41700 }, { "epoch": 2.52, "learning_rate": 3.105035233051167e-05, "loss": 0.5645, "step": 41702 }, { "epoch": 2.52, "learning_rate": 3.104884342039745e-05, "loss": 0.5179, "step": 41704 }, { "epoch": 2.52, "learning_rate": 3.104733451028322e-05, "loss": 0.4918, "step": 41706 }, { "epoch": 2.52, "learning_rate": 3.1045825600169e-05, "loss": 0.7435, "step": 41708 }, { "epoch": 2.52, "learning_rate": 3.104431669005478e-05, "loss": 0.717, "step": 41710 }, { "epoch": 2.52, "learning_rate": 3.104280777994055e-05, "loss": 0.2909, "step": 41712 }, { "epoch": 2.52, "learning_rate": 3.1041298869826324e-05, "loss": 0.3956, "step": 41714 }, { "epoch": 2.52, "learning_rate": 3.1039789959712104e-05, "loss": 0.5246, "step": 41716 }, { "epoch": 2.52, "learning_rate": 3.103828104959788e-05, "loss": 0.472, "step": 41718 }, { "epoch": 2.52, "learning_rate": 3.103677213948365e-05, "loss": 0.322, "step": 41720 }, { "epoch": 2.52, "learning_rate": 3.1035263229369424e-05, "loss": 0.7476, "step": 41722 }, { "epoch": 2.52, "learning_rate": 3.1033754319255203e-05, "loss": 0.3328, "step": 41724 }, { "epoch": 2.52, "learning_rate": 3.103224540914098e-05, "loss": 0.4497, "step": 41726 }, { "epoch": 2.52, "learning_rate": 3.103073649902675e-05, "loss": 0.451, "step": 41728 }, { "epoch": 2.52, "learning_rate": 3.102922758891253e-05, "loss": 0.3902, "step": 41730 }, { "epoch": 2.52, "learning_rate": 3.102771867879831e-05, "loss": 0.4744, "step": 41732 }, { "epoch": 2.52, "learning_rate": 3.102620976868408e-05, "loss": 0.6901, "step": 41734 }, { "epoch": 2.52, "learning_rate": 3.1024700858569856e-05, "loss": 0.8602, "step": 41736 }, { "epoch": 2.52, "learning_rate": 3.102319194845563e-05, "loss": 0.6953, "step": 41738 }, { "epoch": 2.52, "learning_rate": 3.102168303834141e-05, "loss": 0.6271, "step": 41740 }, { "epoch": 2.52, "learning_rate": 3.102017412822718e-05, "loss": 0.4773, "step": 41742 }, { "epoch": 2.52, "learning_rate": 3.1018665218112955e-05, "loss": 0.5914, "step": 41744 }, { "epoch": 2.52, "learning_rate": 3.1017156307998735e-05, "loss": 0.8365, "step": 41746 }, { "epoch": 2.52, "learning_rate": 3.1015647397884515e-05, "loss": 0.4221, "step": 41748 }, { "epoch": 2.52, "learning_rate": 3.101413848777028e-05, "loss": 0.6785, "step": 41750 }, { "epoch": 2.52, "learning_rate": 3.101262957765606e-05, "loss": 0.609, "step": 41752 }, { "epoch": 2.52, "learning_rate": 3.1011120667541834e-05, "loss": 0.4774, "step": 41754 }, { "epoch": 2.52, "learning_rate": 3.100961175742761e-05, "loss": 0.4124, "step": 41756 }, { "epoch": 2.52, "learning_rate": 3.100810284731339e-05, "loss": 0.7063, "step": 41758 }, { "epoch": 2.52, "learning_rate": 3.100659393719916e-05, "loss": 0.5674, "step": 41760 }, { "epoch": 2.52, "learning_rate": 3.100508502708494e-05, "loss": 0.5511, "step": 41762 }, { "epoch": 2.52, "learning_rate": 3.1003576116970713e-05, "loss": 0.5148, "step": 41764 }, { "epoch": 2.52, "learning_rate": 3.1002067206856487e-05, "loss": 0.6425, "step": 41766 }, { "epoch": 2.52, "learning_rate": 3.1000558296742266e-05, "loss": 0.5248, "step": 41768 }, { "epoch": 2.52, "learning_rate": 3.099904938662804e-05, "loss": 0.6011, "step": 41770 }, { "epoch": 2.52, "learning_rate": 3.099754047651381e-05, "loss": 0.4809, "step": 41772 }, { "epoch": 2.52, "learning_rate": 3.099603156639959e-05, "loss": 0.6103, "step": 41774 }, { "epoch": 2.52, "learning_rate": 3.0994522656285366e-05, "loss": 0.4818, "step": 41776 }, { "epoch": 2.52, "learning_rate": 3.099301374617114e-05, "loss": 0.5546, "step": 41778 }, { "epoch": 2.52, "learning_rate": 3.099150483605692e-05, "loss": 0.4744, "step": 41780 }, { "epoch": 2.52, "learning_rate": 3.098999592594269e-05, "loss": 0.6075, "step": 41782 }, { "epoch": 2.52, "learning_rate": 3.098848701582847e-05, "loss": 0.6164, "step": 41784 }, { "epoch": 2.52, "learning_rate": 3.0986978105714245e-05, "loss": 0.6685, "step": 41786 }, { "epoch": 2.52, "learning_rate": 3.098546919560002e-05, "loss": 0.7315, "step": 41788 }, { "epoch": 2.52, "learning_rate": 3.09839602854858e-05, "loss": 0.5064, "step": 41790 }, { "epoch": 2.52, "learning_rate": 3.098245137537157e-05, "loss": 0.5415, "step": 41792 }, { "epoch": 2.52, "learning_rate": 3.0980942465257344e-05, "loss": 0.7116, "step": 41794 }, { "epoch": 2.52, "learning_rate": 3.0979433555143124e-05, "loss": 0.6768, "step": 41796 }, { "epoch": 2.52, "learning_rate": 3.09779246450289e-05, "loss": 0.473, "step": 41798 }, { "epoch": 2.52, "learning_rate": 3.097641573491467e-05, "loss": 0.413, "step": 41800 }, { "epoch": 2.52, "learning_rate": 3.097490682480045e-05, "loss": 0.5207, "step": 41802 }, { "epoch": 2.52, "learning_rate": 3.097339791468622e-05, "loss": 0.414, "step": 41804 }, { "epoch": 2.52, "learning_rate": 3.0971889004571996e-05, "loss": 0.5129, "step": 41806 }, { "epoch": 2.52, "learning_rate": 3.0970380094457776e-05, "loss": 0.5831, "step": 41808 }, { "epoch": 2.52, "learning_rate": 3.096887118434355e-05, "loss": 0.454, "step": 41810 }, { "epoch": 2.52, "learning_rate": 3.096736227422933e-05, "loss": 0.3919, "step": 41812 }, { "epoch": 2.52, "learning_rate": 3.09658533641151e-05, "loss": 0.5064, "step": 41814 }, { "epoch": 2.52, "learning_rate": 3.0964344454000876e-05, "loss": 0.675, "step": 41816 }, { "epoch": 2.52, "learning_rate": 3.0962835543886656e-05, "loss": 0.4333, "step": 41818 }, { "epoch": 2.52, "learning_rate": 3.096132663377242e-05, "loss": 0.4273, "step": 41820 }, { "epoch": 2.52, "learning_rate": 3.09598177236582e-05, "loss": 0.6868, "step": 41822 }, { "epoch": 2.52, "learning_rate": 3.095830881354398e-05, "loss": 0.4925, "step": 41824 }, { "epoch": 2.52, "learning_rate": 3.0956799903429755e-05, "loss": 0.5782, "step": 41826 }, { "epoch": 2.52, "learning_rate": 3.095529099331553e-05, "loss": 0.4553, "step": 41828 }, { "epoch": 2.52, "learning_rate": 3.095378208320131e-05, "loss": 0.7048, "step": 41830 }, { "epoch": 2.52, "learning_rate": 3.095227317308708e-05, "loss": 0.5213, "step": 41832 }, { "epoch": 2.52, "learning_rate": 3.095076426297286e-05, "loss": 0.5354, "step": 41834 }, { "epoch": 2.53, "learning_rate": 3.094925535285863e-05, "loss": 0.569, "step": 41836 }, { "epoch": 2.53, "learning_rate": 3.094774644274441e-05, "loss": 0.3886, "step": 41838 }, { "epoch": 2.53, "learning_rate": 3.094623753263019e-05, "loss": 0.5534, "step": 41840 }, { "epoch": 2.53, "learning_rate": 3.0944728622515953e-05, "loss": 0.6395, "step": 41842 }, { "epoch": 2.53, "learning_rate": 3.094321971240173e-05, "loss": 0.6065, "step": 41844 }, { "epoch": 2.53, "learning_rate": 3.094171080228751e-05, "loss": 0.4017, "step": 41846 }, { "epoch": 2.53, "learning_rate": 3.0940201892173286e-05, "loss": 0.5479, "step": 41848 }, { "epoch": 2.53, "learning_rate": 3.093869298205906e-05, "loss": 0.5767, "step": 41850 }, { "epoch": 2.53, "learning_rate": 3.093718407194483e-05, "loss": 0.6458, "step": 41852 }, { "epoch": 2.53, "learning_rate": 3.093567516183061e-05, "loss": 0.4133, "step": 41854 }, { "epoch": 2.53, "learning_rate": 3.0934166251716386e-05, "loss": 0.4631, "step": 41856 }, { "epoch": 2.53, "learning_rate": 3.093265734160216e-05, "loss": 0.5041, "step": 41858 }, { "epoch": 2.53, "learning_rate": 3.093114843148794e-05, "loss": 0.711, "step": 41860 }, { "epoch": 2.53, "learning_rate": 3.092963952137372e-05, "loss": 0.4615, "step": 41862 }, { "epoch": 2.53, "learning_rate": 3.0928130611259485e-05, "loss": 0.4761, "step": 41864 }, { "epoch": 2.53, "learning_rate": 3.0926621701145265e-05, "loss": 0.5299, "step": 41866 }, { "epoch": 2.53, "learning_rate": 3.092511279103104e-05, "loss": 0.4318, "step": 41868 }, { "epoch": 2.53, "learning_rate": 3.092360388091681e-05, "loss": 0.5503, "step": 41870 }, { "epoch": 2.53, "learning_rate": 3.092209497080259e-05, "loss": 0.6622, "step": 41872 }, { "epoch": 2.53, "learning_rate": 3.0920586060688364e-05, "loss": 0.5273, "step": 41874 }, { "epoch": 2.53, "learning_rate": 3.0919077150574144e-05, "loss": 0.5375, "step": 41876 }, { "epoch": 2.53, "learning_rate": 3.091756824045992e-05, "loss": 0.7381, "step": 41878 }, { "epoch": 2.53, "learning_rate": 3.091605933034569e-05, "loss": 0.5646, "step": 41880 }, { "epoch": 2.53, "learning_rate": 3.091455042023147e-05, "loss": 0.5179, "step": 41882 }, { "epoch": 2.53, "learning_rate": 3.091304151011724e-05, "loss": 0.5512, "step": 41884 }, { "epoch": 2.53, "learning_rate": 3.0911532600003016e-05, "loss": 0.324, "step": 41886 }, { "epoch": 2.53, "learning_rate": 3.0910023689888796e-05, "loss": 0.5548, "step": 41888 }, { "epoch": 2.53, "learning_rate": 3.090851477977457e-05, "loss": 0.4867, "step": 41890 }, { "epoch": 2.53, "learning_rate": 3.090700586966034e-05, "loss": 0.4621, "step": 41892 }, { "epoch": 2.53, "learning_rate": 3.090549695954612e-05, "loss": 0.5511, "step": 41894 }, { "epoch": 2.53, "learning_rate": 3.0903988049431896e-05, "loss": 0.3744, "step": 41896 }, { "epoch": 2.53, "learning_rate": 3.0902479139317675e-05, "loss": 0.7364, "step": 41898 }, { "epoch": 2.53, "learning_rate": 3.090097022920345e-05, "loss": 0.7311, "step": 41900 }, { "epoch": 2.53, "learning_rate": 3.089946131908922e-05, "loss": 0.4077, "step": 41902 }, { "epoch": 2.53, "learning_rate": 3.0897952408975e-05, "loss": 0.5709, "step": 41904 }, { "epoch": 2.53, "learning_rate": 3.0896443498860775e-05, "loss": 0.544, "step": 41906 }, { "epoch": 2.53, "learning_rate": 3.089493458874655e-05, "loss": 0.4553, "step": 41908 }, { "epoch": 2.53, "learning_rate": 3.089342567863233e-05, "loss": 0.5526, "step": 41910 }, { "epoch": 2.53, "learning_rate": 3.08919167685181e-05, "loss": 0.596, "step": 41912 }, { "epoch": 2.53, "learning_rate": 3.0890407858403874e-05, "loss": 0.5937, "step": 41914 }, { "epoch": 2.53, "learning_rate": 3.0888898948289654e-05, "loss": 0.5863, "step": 41916 }, { "epoch": 2.53, "learning_rate": 3.088739003817543e-05, "loss": 0.455, "step": 41918 }, { "epoch": 2.53, "learning_rate": 3.08858811280612e-05, "loss": 0.588, "step": 41920 }, { "epoch": 2.53, "learning_rate": 3.088437221794698e-05, "loss": 0.7009, "step": 41922 }, { "epoch": 2.53, "learning_rate": 3.088286330783275e-05, "loss": 0.5311, "step": 41924 }, { "epoch": 2.53, "learning_rate": 3.088135439771853e-05, "loss": 0.5551, "step": 41926 }, { "epoch": 2.53, "learning_rate": 3.0879845487604306e-05, "loss": 0.717, "step": 41928 }, { "epoch": 2.53, "learning_rate": 3.087833657749008e-05, "loss": 0.5936, "step": 41930 }, { "epoch": 2.53, "learning_rate": 3.087682766737586e-05, "loss": 0.6546, "step": 41932 }, { "epoch": 2.53, "learning_rate": 3.0875318757261626e-05, "loss": 0.4329, "step": 41934 }, { "epoch": 2.53, "learning_rate": 3.0873809847147405e-05, "loss": 0.7041, "step": 41936 }, { "epoch": 2.53, "learning_rate": 3.0872300937033185e-05, "loss": 0.3161, "step": 41938 }, { "epoch": 2.53, "learning_rate": 3.087079202691896e-05, "loss": 0.6297, "step": 41940 }, { "epoch": 2.53, "learning_rate": 3.086928311680473e-05, "loss": 0.4866, "step": 41942 }, { "epoch": 2.53, "learning_rate": 3.086777420669051e-05, "loss": 0.4152, "step": 41944 }, { "epoch": 2.53, "learning_rate": 3.0866265296576285e-05, "loss": 0.6684, "step": 41946 }, { "epoch": 2.53, "learning_rate": 3.0864756386462065e-05, "loss": 0.5795, "step": 41948 }, { "epoch": 2.53, "learning_rate": 3.086324747634783e-05, "loss": 0.5886, "step": 41950 }, { "epoch": 2.53, "learning_rate": 3.086173856623361e-05, "loss": 0.4494, "step": 41952 }, { "epoch": 2.53, "learning_rate": 3.086022965611939e-05, "loss": 0.4508, "step": 41954 }, { "epoch": 2.53, "learning_rate": 3.085872074600516e-05, "loss": 0.5547, "step": 41956 }, { "epoch": 2.53, "learning_rate": 3.085721183589094e-05, "loss": 0.4516, "step": 41958 }, { "epoch": 2.53, "learning_rate": 3.085570292577672e-05, "loss": 0.6167, "step": 41960 }, { "epoch": 2.53, "learning_rate": 3.085419401566249e-05, "loss": 0.6877, "step": 41962 }, { "epoch": 2.53, "learning_rate": 3.085268510554826e-05, "loss": 0.5068, "step": 41964 }, { "epoch": 2.53, "learning_rate": 3.0851176195434036e-05, "loss": 0.573, "step": 41966 }, { "epoch": 2.53, "learning_rate": 3.0849667285319816e-05, "loss": 0.5007, "step": 41968 }, { "epoch": 2.53, "learning_rate": 3.084815837520559e-05, "loss": 0.6222, "step": 41970 }, { "epoch": 2.53, "learning_rate": 3.084664946509136e-05, "loss": 0.666, "step": 41972 }, { "epoch": 2.53, "learning_rate": 3.084514055497714e-05, "loss": 0.5864, "step": 41974 }, { "epoch": 2.53, "learning_rate": 3.084363164486292e-05, "loss": 0.6834, "step": 41976 }, { "epoch": 2.53, "learning_rate": 3.084212273474869e-05, "loss": 0.7271, "step": 41978 }, { "epoch": 2.53, "learning_rate": 3.084061382463447e-05, "loss": 0.3798, "step": 41980 }, { "epoch": 2.53, "learning_rate": 3.083910491452024e-05, "loss": 0.7576, "step": 41982 }, { "epoch": 2.53, "learning_rate": 3.0837596004406015e-05, "loss": 0.3881, "step": 41984 }, { "epoch": 2.53, "learning_rate": 3.0836087094291795e-05, "loss": 0.5715, "step": 41986 }, { "epoch": 2.53, "learning_rate": 3.083457818417757e-05, "loss": 0.6271, "step": 41988 }, { "epoch": 2.53, "learning_rate": 3.083306927406335e-05, "loss": 0.3461, "step": 41990 }, { "epoch": 2.53, "learning_rate": 3.083156036394912e-05, "loss": 0.5141, "step": 41992 }, { "epoch": 2.53, "learning_rate": 3.0830051453834894e-05, "loss": 0.4233, "step": 41994 }, { "epoch": 2.53, "learning_rate": 3.0828542543720674e-05, "loss": 0.4952, "step": 41996 }, { "epoch": 2.53, "learning_rate": 3.082703363360645e-05, "loss": 0.4623, "step": 41998 }, { "epoch": 2.54, "learning_rate": 3.082552472349222e-05, "loss": 0.6714, "step": 42000 }, { "epoch": 2.54, "learning_rate": 3.0824015813378e-05, "loss": 0.5707, "step": 42002 }, { "epoch": 2.54, "learning_rate": 3.082250690326377e-05, "loss": 0.6264, "step": 42004 }, { "epoch": 2.54, "learning_rate": 3.0820997993149546e-05, "loss": 0.8811, "step": 42006 }, { "epoch": 2.54, "learning_rate": 3.0819489083035326e-05, "loss": 0.6102, "step": 42008 }, { "epoch": 2.54, "learning_rate": 3.08179801729211e-05, "loss": 0.6327, "step": 42010 }, { "epoch": 2.54, "learning_rate": 3.081647126280688e-05, "loss": 0.512, "step": 42012 }, { "epoch": 2.54, "learning_rate": 3.081496235269265e-05, "loss": 0.4966, "step": 42014 }, { "epoch": 2.54, "learning_rate": 3.0813453442578425e-05, "loss": 0.5522, "step": 42016 }, { "epoch": 2.54, "learning_rate": 3.0811944532464205e-05, "loss": 0.4845, "step": 42018 }, { "epoch": 2.54, "learning_rate": 3.081043562234998e-05, "loss": 0.4797, "step": 42020 }, { "epoch": 2.54, "learning_rate": 3.080892671223575e-05, "loss": 0.505, "step": 42022 }, { "epoch": 2.54, "learning_rate": 3.080741780212153e-05, "loss": 0.5642, "step": 42024 }, { "epoch": 2.54, "learning_rate": 3.0805908892007305e-05, "loss": 0.6143, "step": 42026 }, { "epoch": 2.54, "learning_rate": 3.080439998189308e-05, "loss": 0.5717, "step": 42028 }, { "epoch": 2.54, "learning_rate": 3.080289107177886e-05, "loss": 0.5603, "step": 42030 }, { "epoch": 2.54, "learning_rate": 3.080138216166463e-05, "loss": 0.8072, "step": 42032 }, { "epoch": 2.54, "learning_rate": 3.0799873251550404e-05, "loss": 0.6067, "step": 42034 }, { "epoch": 2.54, "learning_rate": 3.0798364341436184e-05, "loss": 0.6367, "step": 42036 }, { "epoch": 2.54, "learning_rate": 3.079685543132196e-05, "loss": 0.4555, "step": 42038 }, { "epoch": 2.54, "learning_rate": 3.079534652120774e-05, "loss": 0.4939, "step": 42040 }, { "epoch": 2.54, "learning_rate": 3.079383761109351e-05, "loss": 0.6066, "step": 42042 }, { "epoch": 2.54, "learning_rate": 3.079232870097928e-05, "loss": 0.4646, "step": 42044 }, { "epoch": 2.54, "learning_rate": 3.079081979086506e-05, "loss": 0.534, "step": 42046 }, { "epoch": 2.54, "learning_rate": 3.0789310880750836e-05, "loss": 0.4251, "step": 42048 }, { "epoch": 2.54, "learning_rate": 3.078780197063661e-05, "loss": 0.2818, "step": 42050 }, { "epoch": 2.54, "learning_rate": 3.078629306052239e-05, "loss": 0.6162, "step": 42052 }, { "epoch": 2.54, "learning_rate": 3.078478415040816e-05, "loss": 0.4915, "step": 42054 }, { "epoch": 2.54, "learning_rate": 3.0783275240293935e-05, "loss": 0.4736, "step": 42056 }, { "epoch": 2.54, "learning_rate": 3.0781766330179715e-05, "loss": 0.4481, "step": 42058 }, { "epoch": 2.54, "learning_rate": 3.078025742006549e-05, "loss": 0.4994, "step": 42060 }, { "epoch": 2.54, "learning_rate": 3.077874850995127e-05, "loss": 0.516, "step": 42062 }, { "epoch": 2.54, "learning_rate": 3.0777239599837035e-05, "loss": 0.4786, "step": 42064 }, { "epoch": 2.54, "learning_rate": 3.0775730689722815e-05, "loss": 0.6676, "step": 42066 }, { "epoch": 2.54, "learning_rate": 3.0774221779608594e-05, "loss": 0.4044, "step": 42068 }, { "epoch": 2.54, "learning_rate": 3.077271286949436e-05, "loss": 0.5826, "step": 42070 }, { "epoch": 2.54, "learning_rate": 3.077120395938014e-05, "loss": 0.6414, "step": 42072 }, { "epoch": 2.54, "learning_rate": 3.076969504926592e-05, "loss": 0.4267, "step": 42074 }, { "epoch": 2.54, "learning_rate": 3.0768186139151694e-05, "loss": 0.5291, "step": 42076 }, { "epoch": 2.54, "learning_rate": 3.076667722903747e-05, "loss": 0.7834, "step": 42078 }, { "epoch": 2.54, "learning_rate": 3.076516831892324e-05, "loss": 0.5366, "step": 42080 }, { "epoch": 2.54, "learning_rate": 3.076365940880902e-05, "loss": 0.5462, "step": 42082 }, { "epoch": 2.54, "learning_rate": 3.076215049869479e-05, "loss": 0.5857, "step": 42084 }, { "epoch": 2.54, "learning_rate": 3.0760641588580566e-05, "loss": 0.697, "step": 42086 }, { "epoch": 2.54, "learning_rate": 3.0759132678466346e-05, "loss": 0.5241, "step": 42088 }, { "epoch": 2.54, "learning_rate": 3.0757623768352126e-05, "loss": 0.614, "step": 42090 }, { "epoch": 2.54, "learning_rate": 3.075611485823789e-05, "loss": 0.4526, "step": 42092 }, { "epoch": 2.54, "learning_rate": 3.075460594812367e-05, "loss": 0.6183, "step": 42094 }, { "epoch": 2.54, "learning_rate": 3.0753097038009445e-05, "loss": 0.619, "step": 42096 }, { "epoch": 2.54, "learning_rate": 3.075158812789522e-05, "loss": 0.6012, "step": 42098 }, { "epoch": 2.54, "learning_rate": 3.0750079217781e-05, "loss": 0.5371, "step": 42100 }, { "epoch": 2.54, "learning_rate": 3.074857030766677e-05, "loss": 0.8593, "step": 42102 }, { "epoch": 2.54, "learning_rate": 3.074706139755255e-05, "loss": 0.4666, "step": 42104 }, { "epoch": 2.54, "learning_rate": 3.0745552487438324e-05, "loss": 0.5113, "step": 42106 }, { "epoch": 2.54, "learning_rate": 3.07440435773241e-05, "loss": 0.6079, "step": 42108 }, { "epoch": 2.54, "learning_rate": 3.074253466720988e-05, "loss": 0.3627, "step": 42110 }, { "epoch": 2.54, "learning_rate": 3.074102575709565e-05, "loss": 0.3831, "step": 42112 }, { "epoch": 2.54, "learning_rate": 3.0739516846981424e-05, "loss": 0.5448, "step": 42114 }, { "epoch": 2.54, "learning_rate": 3.0738007936867204e-05, "loss": 0.5458, "step": 42116 }, { "epoch": 2.54, "learning_rate": 3.073649902675298e-05, "loss": 0.6831, "step": 42118 }, { "epoch": 2.54, "learning_rate": 3.073499011663875e-05, "loss": 0.6335, "step": 42120 }, { "epoch": 2.54, "learning_rate": 3.073348120652453e-05, "loss": 0.6855, "step": 42122 }, { "epoch": 2.54, "learning_rate": 3.07319722964103e-05, "loss": 0.6121, "step": 42124 }, { "epoch": 2.54, "learning_rate": 3.073046338629608e-05, "loss": 0.4082, "step": 42126 }, { "epoch": 2.54, "learning_rate": 3.0728954476181856e-05, "loss": 0.8293, "step": 42128 }, { "epoch": 2.54, "learning_rate": 3.072744556606763e-05, "loss": 0.5402, "step": 42130 }, { "epoch": 2.54, "learning_rate": 3.072593665595341e-05, "loss": 0.701, "step": 42132 }, { "epoch": 2.54, "learning_rate": 3.072442774583918e-05, "loss": 0.5125, "step": 42134 }, { "epoch": 2.54, "learning_rate": 3.0722918835724955e-05, "loss": 0.6408, "step": 42136 }, { "epoch": 2.54, "learning_rate": 3.0721409925610735e-05, "loss": 0.7167, "step": 42138 }, { "epoch": 2.54, "learning_rate": 3.071990101549651e-05, "loss": 0.6787, "step": 42140 }, { "epoch": 2.54, "learning_rate": 3.071839210538228e-05, "loss": 0.6099, "step": 42142 }, { "epoch": 2.54, "learning_rate": 3.071688319526806e-05, "loss": 0.5784, "step": 42144 }, { "epoch": 2.54, "learning_rate": 3.0715374285153834e-05, "loss": 0.5131, "step": 42146 }, { "epoch": 2.54, "learning_rate": 3.071386537503961e-05, "loss": 0.6645, "step": 42148 }, { "epoch": 2.54, "learning_rate": 3.071235646492539e-05, "loss": 0.5224, "step": 42150 }, { "epoch": 2.54, "learning_rate": 3.071084755481116e-05, "loss": 0.4556, "step": 42152 }, { "epoch": 2.54, "learning_rate": 3.070933864469694e-05, "loss": 0.4769, "step": 42154 }, { "epoch": 2.54, "learning_rate": 3.0707829734582714e-05, "loss": 0.4977, "step": 42156 }, { "epoch": 2.54, "learning_rate": 3.070632082446849e-05, "loss": 0.6385, "step": 42158 }, { "epoch": 2.54, "learning_rate": 3.0704811914354267e-05, "loss": 0.6011, "step": 42160 }, { "epoch": 2.54, "learning_rate": 3.070330300424004e-05, "loss": 0.6405, "step": 42162 }, { "epoch": 2.54, "learning_rate": 3.070179409412581e-05, "loss": 0.4414, "step": 42164 }, { "epoch": 2.55, "learning_rate": 3.070028518401159e-05, "loss": 0.5561, "step": 42166 }, { "epoch": 2.55, "learning_rate": 3.0698776273897366e-05, "loss": 0.4264, "step": 42168 }, { "epoch": 2.55, "learning_rate": 3.069726736378314e-05, "loss": 0.6444, "step": 42170 }, { "epoch": 2.55, "learning_rate": 3.069575845366892e-05, "loss": 0.7008, "step": 42172 }, { "epoch": 2.55, "learning_rate": 3.069424954355469e-05, "loss": 0.5223, "step": 42174 }, { "epoch": 2.55, "learning_rate": 3.069274063344047e-05, "loss": 0.6786, "step": 42176 }, { "epoch": 2.55, "learning_rate": 3.069123172332624e-05, "loss": 0.5012, "step": 42178 }, { "epoch": 2.55, "learning_rate": 3.068972281321202e-05, "loss": 0.5326, "step": 42180 }, { "epoch": 2.55, "learning_rate": 3.06882139030978e-05, "loss": 0.5577, "step": 42182 }, { "epoch": 2.55, "learning_rate": 3.0686704992983564e-05, "loss": 0.3828, "step": 42184 }, { "epoch": 2.55, "learning_rate": 3.0685196082869344e-05, "loss": 0.4846, "step": 42186 }, { "epoch": 2.55, "learning_rate": 3.0683687172755124e-05, "loss": 0.5912, "step": 42188 }, { "epoch": 2.55, "learning_rate": 3.06821782626409e-05, "loss": 0.4203, "step": 42190 }, { "epoch": 2.55, "learning_rate": 3.068066935252667e-05, "loss": 0.4512, "step": 42192 }, { "epoch": 2.55, "learning_rate": 3.0679160442412444e-05, "loss": 0.5049, "step": 42194 }, { "epoch": 2.55, "learning_rate": 3.0677651532298224e-05, "loss": 0.6985, "step": 42196 }, { "epoch": 2.55, "learning_rate": 3.0676142622184e-05, "loss": 0.5055, "step": 42198 }, { "epoch": 2.55, "learning_rate": 3.067463371206977e-05, "loss": 0.6614, "step": 42200 }, { "epoch": 2.55, "learning_rate": 3.067312480195555e-05, "loss": 0.5918, "step": 42202 }, { "epoch": 2.55, "learning_rate": 3.067161589184133e-05, "loss": 0.5581, "step": 42204 }, { "epoch": 2.55, "learning_rate": 3.0670106981727096e-05, "loss": 0.4994, "step": 42206 }, { "epoch": 2.55, "learning_rate": 3.0668598071612876e-05, "loss": 0.6206, "step": 42208 }, { "epoch": 2.55, "learning_rate": 3.066708916149865e-05, "loss": 0.5691, "step": 42210 }, { "epoch": 2.55, "learning_rate": 3.066558025138443e-05, "loss": 0.5284, "step": 42212 }, { "epoch": 2.55, "learning_rate": 3.06640713412702e-05, "loss": 0.6441, "step": 42214 }, { "epoch": 2.55, "learning_rate": 3.0662562431155975e-05, "loss": 0.4809, "step": 42216 }, { "epoch": 2.55, "learning_rate": 3.0661053521041755e-05, "loss": 0.4837, "step": 42218 }, { "epoch": 2.55, "learning_rate": 3.065954461092753e-05, "loss": 0.5139, "step": 42220 }, { "epoch": 2.55, "learning_rate": 3.06580357008133e-05, "loss": 0.5377, "step": 42222 }, { "epoch": 2.55, "learning_rate": 3.065652679069908e-05, "loss": 0.4528, "step": 42224 }, { "epoch": 2.55, "learning_rate": 3.0655017880584854e-05, "loss": 0.6524, "step": 42226 }, { "epoch": 2.55, "learning_rate": 3.065350897047063e-05, "loss": 0.5463, "step": 42228 }, { "epoch": 2.55, "learning_rate": 3.065200006035641e-05, "loss": 0.4267, "step": 42230 }, { "epoch": 2.55, "learning_rate": 3.065049115024218e-05, "loss": 0.684, "step": 42232 }, { "epoch": 2.55, "learning_rate": 3.0648982240127954e-05, "loss": 0.6297, "step": 42234 }, { "epoch": 2.55, "learning_rate": 3.0647473330013733e-05, "loss": 0.5951, "step": 42236 }, { "epoch": 2.55, "learning_rate": 3.0645964419899507e-05, "loss": 0.6481, "step": 42238 }, { "epoch": 2.55, "learning_rate": 3.0644455509785286e-05, "loss": 0.5349, "step": 42240 }, { "epoch": 2.55, "learning_rate": 3.064294659967106e-05, "loss": 0.4073, "step": 42242 }, { "epoch": 2.55, "learning_rate": 3.064143768955683e-05, "loss": 0.7264, "step": 42244 }, { "epoch": 2.55, "learning_rate": 3.063992877944261e-05, "loss": 0.5003, "step": 42246 }, { "epoch": 2.55, "learning_rate": 3.0638419869328386e-05, "loss": 0.429, "step": 42248 }, { "epoch": 2.55, "learning_rate": 3.063691095921416e-05, "loss": 0.5454, "step": 42250 }, { "epoch": 2.55, "learning_rate": 3.063540204909994e-05, "loss": 0.4831, "step": 42252 }, { "epoch": 2.55, "learning_rate": 3.063389313898571e-05, "loss": 0.4866, "step": 42254 }, { "epoch": 2.55, "learning_rate": 3.0632384228871485e-05, "loss": 0.486, "step": 42256 }, { "epoch": 2.55, "learning_rate": 3.0630875318757265e-05, "loss": 0.5316, "step": 42258 }, { "epoch": 2.55, "learning_rate": 3.062936640864304e-05, "loss": 0.5984, "step": 42260 }, { "epoch": 2.55, "learning_rate": 3.062785749852881e-05, "loss": 0.3833, "step": 42262 }, { "epoch": 2.55, "learning_rate": 3.062634858841459e-05, "loss": 0.82, "step": 42264 }, { "epoch": 2.55, "learning_rate": 3.0624839678300364e-05, "loss": 0.468, "step": 42266 }, { "epoch": 2.55, "learning_rate": 3.0623330768186144e-05, "loss": 0.6266, "step": 42268 }, { "epoch": 2.55, "learning_rate": 3.062182185807192e-05, "loss": 0.4829, "step": 42270 }, { "epoch": 2.55, "learning_rate": 3.062031294795769e-05, "loss": 0.4237, "step": 42272 }, { "epoch": 2.55, "learning_rate": 3.061880403784347e-05, "loss": 0.5128, "step": 42274 }, { "epoch": 2.55, "learning_rate": 3.0617295127729243e-05, "loss": 0.6405, "step": 42276 }, { "epoch": 2.55, "learning_rate": 3.0615786217615017e-05, "loss": 0.5439, "step": 42278 }, { "epoch": 2.55, "learning_rate": 3.0614277307500796e-05, "loss": 0.6786, "step": 42280 }, { "epoch": 2.55, "learning_rate": 3.061276839738657e-05, "loss": 0.4238, "step": 42282 }, { "epoch": 2.55, "learning_rate": 3.061125948727234e-05, "loss": 0.7097, "step": 42284 }, { "epoch": 2.55, "learning_rate": 3.060975057715812e-05, "loss": 0.9239, "step": 42286 }, { "epoch": 2.55, "learning_rate": 3.0608241667043896e-05, "loss": 0.6173, "step": 42288 }, { "epoch": 2.55, "learning_rate": 3.0606732756929676e-05, "loss": 0.508, "step": 42290 }, { "epoch": 2.55, "learning_rate": 3.060522384681544e-05, "loss": 0.4975, "step": 42292 }, { "epoch": 2.55, "learning_rate": 3.060371493670122e-05, "loss": 0.6453, "step": 42294 }, { "epoch": 2.55, "learning_rate": 3.0602206026587e-05, "loss": 0.646, "step": 42296 }, { "epoch": 2.55, "learning_rate": 3.060069711647277e-05, "loss": 0.6921, "step": 42298 }, { "epoch": 2.55, "learning_rate": 3.059918820635855e-05, "loss": 0.5419, "step": 42300 }, { "epoch": 2.55, "learning_rate": 3.059767929624433e-05, "loss": 0.6264, "step": 42302 }, { "epoch": 2.55, "learning_rate": 3.05961703861301e-05, "loss": 0.6988, "step": 42304 }, { "epoch": 2.55, "learning_rate": 3.0594661476015874e-05, "loss": 0.8147, "step": 42306 }, { "epoch": 2.55, "learning_rate": 3.059315256590165e-05, "loss": 0.774, "step": 42308 }, { "epoch": 2.55, "learning_rate": 3.059164365578743e-05, "loss": 0.673, "step": 42310 }, { "epoch": 2.55, "learning_rate": 3.05901347456732e-05, "loss": 0.4892, "step": 42312 }, { "epoch": 2.55, "learning_rate": 3.0588625835558973e-05, "loss": 0.6244, "step": 42314 }, { "epoch": 2.55, "learning_rate": 3.058711692544475e-05, "loss": 0.4772, "step": 42316 }, { "epoch": 2.55, "learning_rate": 3.058560801533053e-05, "loss": 0.6628, "step": 42318 }, { "epoch": 2.55, "learning_rate": 3.05840991052163e-05, "loss": 0.5548, "step": 42320 }, { "epoch": 2.55, "learning_rate": 3.058259019510208e-05, "loss": 0.3369, "step": 42322 }, { "epoch": 2.55, "learning_rate": 3.058108128498785e-05, "loss": 0.5409, "step": 42324 }, { "epoch": 2.55, "learning_rate": 3.057957237487363e-05, "loss": 0.4368, "step": 42326 }, { "epoch": 2.55, "learning_rate": 3.0578063464759406e-05, "loss": 0.3945, "step": 42328 }, { "epoch": 2.55, "learning_rate": 3.057655455464518e-05, "loss": 0.7689, "step": 42330 }, { "epoch": 2.56, "learning_rate": 3.057504564453096e-05, "loss": 0.7368, "step": 42332 }, { "epoch": 2.56, "learning_rate": 3.057353673441673e-05, "loss": 0.579, "step": 42334 }, { "epoch": 2.56, "learning_rate": 3.0572027824302505e-05, "loss": 0.5167, "step": 42336 }, { "epoch": 2.56, "learning_rate": 3.0570518914188285e-05, "loss": 0.4682, "step": 42338 }, { "epoch": 2.56, "learning_rate": 3.056901000407406e-05, "loss": 0.4905, "step": 42340 }, { "epoch": 2.56, "learning_rate": 3.056750109395983e-05, "loss": 0.4671, "step": 42342 }, { "epoch": 2.56, "learning_rate": 3.056599218384561e-05, "loss": 0.468, "step": 42344 }, { "epoch": 2.56, "learning_rate": 3.0564483273731384e-05, "loss": 0.4455, "step": 42346 }, { "epoch": 2.56, "learning_rate": 3.056297436361716e-05, "loss": 0.4757, "step": 42348 }, { "epoch": 2.56, "learning_rate": 3.056146545350294e-05, "loss": 0.4806, "step": 42350 }, { "epoch": 2.56, "learning_rate": 3.055995654338871e-05, "loss": 0.7754, "step": 42352 }, { "epoch": 2.56, "learning_rate": 3.055844763327449e-05, "loss": 0.5135, "step": 42354 }, { "epoch": 2.56, "learning_rate": 3.055693872316026e-05, "loss": 0.6055, "step": 42356 }, { "epoch": 2.56, "learning_rate": 3.0555429813046036e-05, "loss": 0.5513, "step": 42358 }, { "epoch": 2.56, "learning_rate": 3.0553920902931816e-05, "loss": 0.583, "step": 42360 }, { "epoch": 2.56, "learning_rate": 3.055241199281759e-05, "loss": 0.6683, "step": 42362 }, { "epoch": 2.56, "learning_rate": 3.055090308270336e-05, "loss": 0.6539, "step": 42364 }, { "epoch": 2.56, "learning_rate": 3.054939417258914e-05, "loss": 0.5577, "step": 42366 }, { "epoch": 2.56, "learning_rate": 3.0547885262474916e-05, "loss": 0.4669, "step": 42368 }, { "epoch": 2.56, "learning_rate": 3.054637635236069e-05, "loss": 0.3643, "step": 42370 }, { "epoch": 2.56, "learning_rate": 3.054486744224647e-05, "loss": 0.6815, "step": 42372 }, { "epoch": 2.56, "learning_rate": 3.054335853213224e-05, "loss": 0.5356, "step": 42374 }, { "epoch": 2.56, "learning_rate": 3.054184962201802e-05, "loss": 0.48, "step": 42376 }, { "epoch": 2.56, "learning_rate": 3.0540340711903795e-05, "loss": 0.6381, "step": 42378 }, { "epoch": 2.56, "learning_rate": 3.053883180178957e-05, "loss": 0.5954, "step": 42380 }, { "epoch": 2.56, "learning_rate": 3.053732289167535e-05, "loss": 0.5231, "step": 42382 }, { "epoch": 2.56, "learning_rate": 3.053581398156112e-05, "loss": 0.4053, "step": 42384 }, { "epoch": 2.56, "learning_rate": 3.0534305071446894e-05, "loss": 0.9233, "step": 42386 }, { "epoch": 2.56, "learning_rate": 3.0532796161332674e-05, "loss": 0.4862, "step": 42388 }, { "epoch": 2.56, "learning_rate": 3.053128725121845e-05, "loss": 0.6782, "step": 42390 }, { "epoch": 2.56, "learning_rate": 3.052977834110422e-05, "loss": 0.5623, "step": 42392 }, { "epoch": 2.56, "learning_rate": 3.052826943099e-05, "loss": 0.614, "step": 42394 }, { "epoch": 2.56, "learning_rate": 3.052676052087577e-05, "loss": 0.6405, "step": 42396 }, { "epoch": 2.56, "learning_rate": 3.0525251610761546e-05, "loss": 0.4931, "step": 42398 }, { "epoch": 2.56, "learning_rate": 3.0523742700647326e-05, "loss": 0.6047, "step": 42400 }, { "epoch": 2.56, "learning_rate": 3.05222337905331e-05, "loss": 0.6919, "step": 42402 }, { "epoch": 2.56, "learning_rate": 3.052072488041888e-05, "loss": 0.5845, "step": 42404 }, { "epoch": 2.56, "learning_rate": 3.0519215970304646e-05, "loss": 0.5145, "step": 42406 }, { "epoch": 2.56, "learning_rate": 3.0517707060190426e-05, "loss": 0.3743, "step": 42408 }, { "epoch": 2.56, "learning_rate": 3.0516198150076202e-05, "loss": 0.4881, "step": 42410 }, { "epoch": 2.56, "learning_rate": 3.0514689239961975e-05, "loss": 0.4932, "step": 42412 }, { "epoch": 2.56, "learning_rate": 3.0513180329847752e-05, "loss": 0.8697, "step": 42414 }, { "epoch": 2.56, "learning_rate": 3.0511671419733528e-05, "loss": 0.6895, "step": 42416 }, { "epoch": 2.56, "learning_rate": 3.0510162509619305e-05, "loss": 0.5817, "step": 42418 }, { "epoch": 2.56, "learning_rate": 3.0508653599505078e-05, "loss": 0.8748, "step": 42420 }, { "epoch": 2.56, "learning_rate": 3.0507144689390854e-05, "loss": 0.5265, "step": 42422 }, { "epoch": 2.56, "learning_rate": 3.050563577927663e-05, "loss": 0.4701, "step": 42424 }, { "epoch": 2.56, "learning_rate": 3.0504126869162404e-05, "loss": 0.6208, "step": 42426 }, { "epoch": 2.56, "learning_rate": 3.050261795904818e-05, "loss": 0.5233, "step": 42428 }, { "epoch": 2.56, "learning_rate": 3.0501109048933957e-05, "loss": 0.4976, "step": 42430 }, { "epoch": 2.56, "learning_rate": 3.0499600138819734e-05, "loss": 0.5885, "step": 42432 }, { "epoch": 2.56, "learning_rate": 3.0498091228705507e-05, "loss": 0.7441, "step": 42434 }, { "epoch": 2.56, "learning_rate": 3.0496582318591283e-05, "loss": 0.4025, "step": 42436 }, { "epoch": 2.56, "learning_rate": 3.049507340847706e-05, "loss": 0.4965, "step": 42438 }, { "epoch": 2.56, "learning_rate": 3.0493564498362836e-05, "loss": 0.4304, "step": 42440 }, { "epoch": 2.56, "learning_rate": 3.049205558824861e-05, "loss": 0.6858, "step": 42442 }, { "epoch": 2.56, "learning_rate": 3.0490546678134386e-05, "loss": 0.5369, "step": 42444 }, { "epoch": 2.56, "learning_rate": 3.0489037768020162e-05, "loss": 0.6106, "step": 42446 }, { "epoch": 2.56, "learning_rate": 3.0487528857905932e-05, "loss": 0.5868, "step": 42448 }, { "epoch": 2.56, "learning_rate": 3.0486019947791712e-05, "loss": 0.6378, "step": 42450 }, { "epoch": 2.56, "learning_rate": 3.048451103767749e-05, "loss": 0.6566, "step": 42452 }, { "epoch": 2.56, "learning_rate": 3.0483002127563265e-05, "loss": 0.7333, "step": 42454 }, { "epoch": 2.56, "learning_rate": 3.0481493217449035e-05, "loss": 0.5088, "step": 42456 }, { "epoch": 2.56, "learning_rate": 3.0479984307334815e-05, "loss": 0.7335, "step": 42458 }, { "epoch": 2.56, "learning_rate": 3.047847539722059e-05, "loss": 0.6317, "step": 42460 }, { "epoch": 2.56, "learning_rate": 3.047696648710636e-05, "loss": 0.6042, "step": 42462 }, { "epoch": 2.56, "learning_rate": 3.0475457576992137e-05, "loss": 0.5374, "step": 42464 }, { "epoch": 2.56, "learning_rate": 3.0473948666877917e-05, "loss": 0.6076, "step": 42466 }, { "epoch": 2.56, "learning_rate": 3.0472439756763694e-05, "loss": 0.5031, "step": 42468 }, { "epoch": 2.56, "learning_rate": 3.0470930846649464e-05, "loss": 0.6417, "step": 42470 }, { "epoch": 2.56, "learning_rate": 3.0469421936535243e-05, "loss": 0.3238, "step": 42472 }, { "epoch": 2.56, "learning_rate": 3.046791302642102e-05, "loss": 0.8116, "step": 42474 }, { "epoch": 2.56, "learning_rate": 3.046640411630679e-05, "loss": 0.6364, "step": 42476 }, { "epoch": 2.56, "learning_rate": 3.0464895206192566e-05, "loss": 0.5694, "step": 42478 }, { "epoch": 2.56, "learning_rate": 3.0463386296078346e-05, "loss": 0.7006, "step": 42480 }, { "epoch": 2.56, "learning_rate": 3.0461877385964123e-05, "loss": 0.4011, "step": 42482 }, { "epoch": 2.56, "learning_rate": 3.0460368475849892e-05, "loss": 0.5832, "step": 42484 }, { "epoch": 2.56, "learning_rate": 3.045885956573567e-05, "loss": 0.5546, "step": 42486 }, { "epoch": 2.56, "learning_rate": 3.045735065562145e-05, "loss": 0.5289, "step": 42488 }, { "epoch": 2.56, "learning_rate": 3.0455841745507225e-05, "loss": 0.5655, "step": 42490 }, { "epoch": 2.56, "learning_rate": 3.0454332835392995e-05, "loss": 0.4863, "step": 42492 }, { "epoch": 2.56, "learning_rate": 3.045282392527877e-05, "loss": 0.5343, "step": 42494 }, { "epoch": 2.56, "learning_rate": 3.045131501516455e-05, "loss": 0.442, "step": 42496 }, { "epoch": 2.57, "learning_rate": 3.044980610505032e-05, "loss": 0.5473, "step": 42498 }, { "epoch": 2.57, "learning_rate": 3.0448297194936098e-05, "loss": 0.5637, "step": 42500 }, { "epoch": 2.57, "learning_rate": 3.0446788284821874e-05, "loss": 0.6333, "step": 42502 }, { "epoch": 2.57, "learning_rate": 3.0445279374707654e-05, "loss": 0.4469, "step": 42504 }, { "epoch": 2.57, "learning_rate": 3.0443770464593424e-05, "loss": 0.6263, "step": 42506 }, { "epoch": 2.57, "learning_rate": 3.04422615544792e-05, "loss": 0.4592, "step": 42508 }, { "epoch": 2.57, "learning_rate": 3.0440752644364977e-05, "loss": 0.7288, "step": 42510 }, { "epoch": 2.57, "learning_rate": 3.043924373425075e-05, "loss": 0.5013, "step": 42512 }, { "epoch": 2.57, "learning_rate": 3.0437734824136527e-05, "loss": 0.5515, "step": 42514 }, { "epoch": 2.57, "learning_rate": 3.0436225914022303e-05, "loss": 0.3752, "step": 42516 }, { "epoch": 2.57, "learning_rate": 3.043471700390808e-05, "loss": 0.6256, "step": 42518 }, { "epoch": 2.57, "learning_rate": 3.0433208093793853e-05, "loss": 0.3409, "step": 42520 }, { "epoch": 2.57, "learning_rate": 3.043169918367963e-05, "loss": 0.6544, "step": 42522 }, { "epoch": 2.57, "learning_rate": 3.0430190273565406e-05, "loss": 0.7608, "step": 42524 }, { "epoch": 2.57, "learning_rate": 3.042868136345118e-05, "loss": 0.4956, "step": 42526 }, { "epoch": 2.57, "learning_rate": 3.0427172453336955e-05, "loss": 0.5659, "step": 42528 }, { "epoch": 2.57, "learning_rate": 3.0425663543222732e-05, "loss": 0.4705, "step": 42530 }, { "epoch": 2.57, "learning_rate": 3.042415463310851e-05, "loss": 0.5769, "step": 42532 }, { "epoch": 2.57, "learning_rate": 3.042264572299428e-05, "loss": 0.724, "step": 42534 }, { "epoch": 2.57, "learning_rate": 3.0421136812880058e-05, "loss": 0.5818, "step": 42536 }, { "epoch": 2.57, "learning_rate": 3.0419627902765835e-05, "loss": 0.6464, "step": 42538 }, { "epoch": 2.57, "learning_rate": 3.041811899265161e-05, "loss": 0.5823, "step": 42540 }, { "epoch": 2.57, "learning_rate": 3.0416610082537384e-05, "loss": 0.7, "step": 42542 }, { "epoch": 2.57, "learning_rate": 3.041510117242316e-05, "loss": 0.4088, "step": 42544 }, { "epoch": 2.57, "learning_rate": 3.0413592262308937e-05, "loss": 0.6235, "step": 42546 }, { "epoch": 2.57, "learning_rate": 3.041208335219471e-05, "loss": 0.529, "step": 42548 }, { "epoch": 2.57, "learning_rate": 3.0410574442080487e-05, "loss": 0.3889, "step": 42550 }, { "epoch": 2.57, "learning_rate": 3.0409065531966263e-05, "loss": 0.329, "step": 42552 }, { "epoch": 2.57, "learning_rate": 3.040755662185204e-05, "loss": 0.6153, "step": 42554 }, { "epoch": 2.57, "learning_rate": 3.0406047711737813e-05, "loss": 0.6866, "step": 42556 }, { "epoch": 2.57, "learning_rate": 3.040453880162359e-05, "loss": 0.5364, "step": 42558 }, { "epoch": 2.57, "learning_rate": 3.0403029891509366e-05, "loss": 0.5165, "step": 42560 }, { "epoch": 2.57, "learning_rate": 3.0401520981395136e-05, "loss": 0.4552, "step": 42562 }, { "epoch": 2.57, "learning_rate": 3.0400012071280916e-05, "loss": 0.7355, "step": 42564 }, { "epoch": 2.57, "learning_rate": 3.0398503161166692e-05, "loss": 0.7136, "step": 42566 }, { "epoch": 2.57, "learning_rate": 3.039699425105247e-05, "loss": 0.4927, "step": 42568 }, { "epoch": 2.57, "learning_rate": 3.039548534093824e-05, "loss": 0.5729, "step": 42570 }, { "epoch": 2.57, "learning_rate": 3.039397643082402e-05, "loss": 0.6299, "step": 42572 }, { "epoch": 2.57, "learning_rate": 3.0392467520709795e-05, "loss": 0.6089, "step": 42574 }, { "epoch": 2.57, "learning_rate": 3.0390958610595565e-05, "loss": 0.7405, "step": 42576 }, { "epoch": 2.57, "learning_rate": 3.038944970048134e-05, "loss": 0.5017, "step": 42578 }, { "epoch": 2.57, "learning_rate": 3.038794079036712e-05, "loss": 0.513, "step": 42580 }, { "epoch": 2.57, "learning_rate": 3.0386431880252898e-05, "loss": 0.7443, "step": 42582 }, { "epoch": 2.57, "learning_rate": 3.0384922970138667e-05, "loss": 0.4809, "step": 42584 }, { "epoch": 2.57, "learning_rate": 3.0383414060024444e-05, "loss": 0.4923, "step": 42586 }, { "epoch": 2.57, "learning_rate": 3.0381905149910224e-05, "loss": 0.5132, "step": 42588 }, { "epoch": 2.57, "learning_rate": 3.0380396239796e-05, "loss": 0.6106, "step": 42590 }, { "epoch": 2.57, "learning_rate": 3.037888732968177e-05, "loss": 0.5145, "step": 42592 }, { "epoch": 2.57, "learning_rate": 3.037737841956755e-05, "loss": 0.5223, "step": 42594 }, { "epoch": 2.57, "learning_rate": 3.0375869509453326e-05, "loss": 0.4507, "step": 42596 }, { "epoch": 2.57, "learning_rate": 3.0374360599339096e-05, "loss": 0.7174, "step": 42598 }, { "epoch": 2.57, "learning_rate": 3.0372851689224873e-05, "loss": 0.4435, "step": 42600 }, { "epoch": 2.57, "learning_rate": 3.0371342779110652e-05, "loss": 0.5926, "step": 42602 }, { "epoch": 2.57, "learning_rate": 3.036983386899643e-05, "loss": 0.538, "step": 42604 }, { "epoch": 2.57, "learning_rate": 3.03683249588822e-05, "loss": 0.5633, "step": 42606 }, { "epoch": 2.57, "learning_rate": 3.0366816048767975e-05, "loss": 0.7098, "step": 42608 }, { "epoch": 2.57, "learning_rate": 3.0365307138653755e-05, "loss": 0.4688, "step": 42610 }, { "epoch": 2.57, "learning_rate": 3.0363798228539525e-05, "loss": 0.6802, "step": 42612 }, { "epoch": 2.57, "learning_rate": 3.03622893184253e-05, "loss": 0.6971, "step": 42614 }, { "epoch": 2.57, "learning_rate": 3.0360780408311078e-05, "loss": 0.6444, "step": 42616 }, { "epoch": 2.57, "learning_rate": 3.0359271498196858e-05, "loss": 0.6049, "step": 42618 }, { "epoch": 2.57, "learning_rate": 3.0357762588082628e-05, "loss": 0.6277, "step": 42620 }, { "epoch": 2.57, "learning_rate": 3.0356253677968404e-05, "loss": 0.358, "step": 42622 }, { "epoch": 2.57, "learning_rate": 3.035474476785418e-05, "loss": 0.5384, "step": 42624 }, { "epoch": 2.57, "learning_rate": 3.0353235857739954e-05, "loss": 1.0, "step": 42626 }, { "epoch": 2.57, "learning_rate": 3.035172694762573e-05, "loss": 0.5059, "step": 42628 }, { "epoch": 2.57, "learning_rate": 3.0350218037511507e-05, "loss": 0.5695, "step": 42630 }, { "epoch": 2.57, "learning_rate": 3.0348709127397283e-05, "loss": 0.7021, "step": 42632 }, { "epoch": 2.57, "learning_rate": 3.0347200217283056e-05, "loss": 0.6342, "step": 42634 }, { "epoch": 2.57, "learning_rate": 3.0345691307168833e-05, "loss": 0.8035, "step": 42636 }, { "epoch": 2.57, "learning_rate": 3.034418239705461e-05, "loss": 0.5999, "step": 42638 }, { "epoch": 2.57, "learning_rate": 3.0342673486940383e-05, "loss": 0.4857, "step": 42640 }, { "epoch": 2.57, "learning_rate": 3.034116457682616e-05, "loss": 0.5986, "step": 42642 }, { "epoch": 2.57, "learning_rate": 3.0339655666711936e-05, "loss": 0.4805, "step": 42644 }, { "epoch": 2.57, "learning_rate": 3.0338146756597712e-05, "loss": 0.4925, "step": 42646 }, { "epoch": 2.57, "learning_rate": 3.0336637846483485e-05, "loss": 0.4835, "step": 42648 }, { "epoch": 2.57, "learning_rate": 3.0335128936369262e-05, "loss": 0.5368, "step": 42650 }, { "epoch": 2.57, "learning_rate": 3.0333620026255038e-05, "loss": 0.6319, "step": 42652 }, { "epoch": 2.57, "learning_rate": 3.0332111116140815e-05, "loss": 0.4949, "step": 42654 }, { "epoch": 2.57, "learning_rate": 3.0330602206026588e-05, "loss": 0.4904, "step": 42656 }, { "epoch": 2.57, "learning_rate": 3.0329093295912364e-05, "loss": 0.6075, "step": 42658 }, { "epoch": 2.57, "learning_rate": 3.032758438579814e-05, "loss": 0.7663, "step": 42660 }, { "epoch": 2.57, "learning_rate": 3.0326075475683914e-05, "loss": 0.6698, "step": 42662 }, { "epoch": 2.58, "learning_rate": 3.032456656556969e-05, "loss": 0.5242, "step": 42664 }, { "epoch": 2.58, "learning_rate": 3.0323057655455467e-05, "loss": 0.5029, "step": 42666 }, { "epoch": 2.58, "learning_rate": 3.0321548745341244e-05, "loss": 0.6482, "step": 42668 }, { "epoch": 2.58, "learning_rate": 3.0320039835227017e-05, "loss": 0.6966, "step": 42670 }, { "epoch": 2.58, "learning_rate": 3.0318530925112793e-05, "loss": 0.5236, "step": 42672 }, { "epoch": 2.58, "learning_rate": 3.031702201499857e-05, "loss": 0.7843, "step": 42674 }, { "epoch": 2.58, "learning_rate": 3.031551310488434e-05, "loss": 0.4636, "step": 42676 }, { "epoch": 2.58, "learning_rate": 3.031400419477012e-05, "loss": 0.3494, "step": 42678 }, { "epoch": 2.58, "learning_rate": 3.0312495284655896e-05, "loss": 0.5446, "step": 42680 }, { "epoch": 2.58, "learning_rate": 3.0310986374541672e-05, "loss": 0.4869, "step": 42682 }, { "epoch": 2.58, "learning_rate": 3.0309477464427442e-05, "loss": 0.595, "step": 42684 }, { "epoch": 2.58, "learning_rate": 3.0307968554313222e-05, "loss": 0.388, "step": 42686 }, { "epoch": 2.58, "learning_rate": 3.0306459644199e-05, "loss": 0.4733, "step": 42688 }, { "epoch": 2.58, "learning_rate": 3.0304950734084768e-05, "loss": 0.7009, "step": 42690 }, { "epoch": 2.58, "learning_rate": 3.0303441823970545e-05, "loss": 0.6019, "step": 42692 }, { "epoch": 2.58, "learning_rate": 3.0301932913856325e-05, "loss": 0.7333, "step": 42694 }, { "epoch": 2.58, "learning_rate": 3.03004240037421e-05, "loss": 0.6451, "step": 42696 }, { "epoch": 2.58, "learning_rate": 3.029891509362787e-05, "loss": 0.5988, "step": 42698 }, { "epoch": 2.58, "learning_rate": 3.0297406183513647e-05, "loss": 0.6228, "step": 42700 }, { "epoch": 2.58, "learning_rate": 3.0295897273399427e-05, "loss": 0.5756, "step": 42702 }, { "epoch": 2.58, "learning_rate": 3.0294388363285204e-05, "loss": 0.6189, "step": 42704 }, { "epoch": 2.58, "learning_rate": 3.0292879453170974e-05, "loss": 0.5543, "step": 42706 }, { "epoch": 2.58, "learning_rate": 3.0291370543056754e-05, "loss": 0.5766, "step": 42708 }, { "epoch": 2.58, "learning_rate": 3.028986163294253e-05, "loss": 0.5483, "step": 42710 }, { "epoch": 2.58, "learning_rate": 3.02883527228283e-05, "loss": 0.5794, "step": 42712 }, { "epoch": 2.58, "learning_rate": 3.0286843812714076e-05, "loss": 0.7252, "step": 42714 }, { "epoch": 2.58, "learning_rate": 3.0285334902599856e-05, "loss": 0.6992, "step": 42716 }, { "epoch": 2.58, "learning_rate": 3.0283825992485633e-05, "loss": 0.4996, "step": 42718 }, { "epoch": 2.58, "learning_rate": 3.0282317082371402e-05, "loss": 0.6163, "step": 42720 }, { "epoch": 2.58, "learning_rate": 3.028080817225718e-05, "loss": 0.7087, "step": 42722 }, { "epoch": 2.58, "learning_rate": 3.027929926214296e-05, "loss": 0.6859, "step": 42724 }, { "epoch": 2.58, "learning_rate": 3.027779035202873e-05, "loss": 0.5837, "step": 42726 }, { "epoch": 2.58, "learning_rate": 3.0276281441914505e-05, "loss": 0.4347, "step": 42728 }, { "epoch": 2.58, "learning_rate": 3.027477253180028e-05, "loss": 0.547, "step": 42730 }, { "epoch": 2.58, "learning_rate": 3.027326362168606e-05, "loss": 0.5369, "step": 42732 }, { "epoch": 2.58, "learning_rate": 3.027175471157183e-05, "loss": 0.5435, "step": 42734 }, { "epoch": 2.58, "learning_rate": 3.0270245801457608e-05, "loss": 0.5793, "step": 42736 }, { "epoch": 2.58, "learning_rate": 3.0268736891343384e-05, "loss": 0.5587, "step": 42738 }, { "epoch": 2.58, "learning_rate": 3.0267227981229157e-05, "loss": 0.46, "step": 42740 }, { "epoch": 2.58, "learning_rate": 3.0265719071114934e-05, "loss": 0.5293, "step": 42742 }, { "epoch": 2.58, "learning_rate": 3.026421016100071e-05, "loss": 0.6572, "step": 42744 }, { "epoch": 2.58, "learning_rate": 3.02634557059436e-05, "loss": 0.6467, "step": 42746 }, { "epoch": 2.58, "learning_rate": 3.026194679582937e-05, "loss": 0.5447, "step": 42748 }, { "epoch": 2.58, "learning_rate": 3.0260437885715147e-05, "loss": 0.6615, "step": 42750 }, { "epoch": 2.58, "learning_rate": 3.0258928975600927e-05, "loss": 0.5024, "step": 42752 }, { "epoch": 2.58, "learning_rate": 3.0257420065486703e-05, "loss": 0.5286, "step": 42754 }, { "epoch": 2.58, "learning_rate": 3.0255911155372473e-05, "loss": 0.4849, "step": 42756 }, { "epoch": 2.58, "learning_rate": 3.025440224525825e-05, "loss": 0.6464, "step": 42758 }, { "epoch": 2.58, "learning_rate": 3.025289333514403e-05, "loss": 0.4574, "step": 42760 }, { "epoch": 2.58, "learning_rate": 3.0251384425029806e-05, "loss": 0.5444, "step": 42762 }, { "epoch": 2.58, "learning_rate": 3.0249875514915575e-05, "loss": 0.5074, "step": 42764 }, { "epoch": 2.58, "learning_rate": 3.0248366604801352e-05, "loss": 0.5193, "step": 42766 }, { "epoch": 2.58, "learning_rate": 3.0246857694687132e-05, "loss": 0.4747, "step": 42768 }, { "epoch": 2.58, "learning_rate": 3.02453487845729e-05, "loss": 0.4444, "step": 42770 }, { "epoch": 2.58, "learning_rate": 3.0243839874458678e-05, "loss": 0.5427, "step": 42772 }, { "epoch": 2.58, "learning_rate": 3.0242330964344458e-05, "loss": 0.5628, "step": 42774 }, { "epoch": 2.58, "learning_rate": 3.0240822054230235e-05, "loss": 0.6248, "step": 42776 }, { "epoch": 2.58, "learning_rate": 3.0239313144116004e-05, "loss": 0.637, "step": 42778 }, { "epoch": 2.58, "learning_rate": 3.023780423400178e-05, "loss": 0.4594, "step": 42780 }, { "epoch": 2.58, "learning_rate": 3.023629532388756e-05, "loss": 0.5209, "step": 42782 }, { "epoch": 2.58, "learning_rate": 3.023478641377333e-05, "loss": 0.6043, "step": 42784 }, { "epoch": 2.58, "learning_rate": 3.0233277503659107e-05, "loss": 0.6696, "step": 42786 }, { "epoch": 2.58, "learning_rate": 3.0231768593544883e-05, "loss": 0.5736, "step": 42788 }, { "epoch": 2.58, "learning_rate": 3.0230259683430663e-05, "loss": 0.5481, "step": 42790 }, { "epoch": 2.58, "learning_rate": 3.0228750773316433e-05, "loss": 0.6163, "step": 42792 }, { "epoch": 2.58, "learning_rate": 3.022724186320221e-05, "loss": 0.5389, "step": 42794 }, { "epoch": 2.58, "learning_rate": 3.0225732953087986e-05, "loss": 0.6812, "step": 42796 }, { "epoch": 2.58, "learning_rate": 3.022422404297376e-05, "loss": 0.6168, "step": 42798 }, { "epoch": 2.58, "learning_rate": 3.0222715132859536e-05, "loss": 0.4977, "step": 42800 }, { "epoch": 2.58, "learning_rate": 3.0221206222745312e-05, "loss": 0.5204, "step": 42802 }, { "epoch": 2.58, "learning_rate": 3.021969731263109e-05, "loss": 0.4878, "step": 42804 }, { "epoch": 2.58, "learning_rate": 3.0218188402516862e-05, "loss": 0.387, "step": 42806 }, { "epoch": 2.58, "learning_rate": 3.021667949240264e-05, "loss": 0.6423, "step": 42808 }, { "epoch": 2.58, "learning_rate": 3.0215170582288415e-05, "loss": 0.4458, "step": 42810 }, { "epoch": 2.58, "learning_rate": 3.021366167217419e-05, "loss": 0.5082, "step": 42812 }, { "epoch": 2.58, "learning_rate": 3.0212152762059965e-05, "loss": 0.3758, "step": 42814 }, { "epoch": 2.58, "learning_rate": 3.021064385194574e-05, "loss": 0.7047, "step": 42816 }, { "epoch": 2.58, "learning_rate": 3.0209134941831518e-05, "loss": 0.5929, "step": 42818 }, { "epoch": 2.58, "learning_rate": 3.020762603171729e-05, "loss": 0.4942, "step": 42820 }, { "epoch": 2.58, "learning_rate": 3.0206117121603067e-05, "loss": 0.514, "step": 42822 }, { "epoch": 2.58, "learning_rate": 3.0204608211488844e-05, "loss": 0.5269, "step": 42824 }, { "epoch": 2.58, "learning_rate": 3.020309930137462e-05, "loss": 0.4665, "step": 42826 }, { "epoch": 2.58, "learning_rate": 3.0201590391260393e-05, "loss": 0.7583, "step": 42828 }, { "epoch": 2.59, "learning_rate": 3.020008148114617e-05, "loss": 0.3652, "step": 42830 }, { "epoch": 2.59, "learning_rate": 3.0198572571031946e-05, "loss": 0.3986, "step": 42832 }, { "epoch": 2.59, "learning_rate": 3.019706366091772e-05, "loss": 0.4832, "step": 42834 }, { "epoch": 2.59, "learning_rate": 3.0195554750803496e-05, "loss": 0.6828, "step": 42836 }, { "epoch": 2.59, "learning_rate": 3.0194045840689273e-05, "loss": 0.8099, "step": 42838 }, { "epoch": 2.59, "learning_rate": 3.019253693057505e-05, "loss": 0.4637, "step": 42840 }, { "epoch": 2.59, "learning_rate": 3.0191028020460822e-05, "loss": 0.7216, "step": 42842 }, { "epoch": 2.59, "learning_rate": 3.01895191103466e-05, "loss": 0.6551, "step": 42844 }, { "epoch": 2.59, "learning_rate": 3.0188010200232375e-05, "loss": 0.6543, "step": 42846 }, { "epoch": 2.59, "learning_rate": 3.0186501290118145e-05, "loss": 0.5293, "step": 42848 }, { "epoch": 2.59, "learning_rate": 3.0184992380003925e-05, "loss": 0.5971, "step": 42850 }, { "epoch": 2.59, "learning_rate": 3.01834834698897e-05, "loss": 0.8593, "step": 42852 }, { "epoch": 2.59, "learning_rate": 3.0181974559775478e-05, "loss": 0.5023, "step": 42854 }, { "epoch": 2.59, "learning_rate": 3.0180465649661248e-05, "loss": 0.419, "step": 42856 }, { "epoch": 2.59, "learning_rate": 3.0178956739547028e-05, "loss": 0.5829, "step": 42858 }, { "epoch": 2.59, "learning_rate": 3.0177447829432804e-05, "loss": 0.5217, "step": 42860 }, { "epoch": 2.59, "learning_rate": 3.017593891931858e-05, "loss": 0.6289, "step": 42862 }, { "epoch": 2.59, "learning_rate": 3.017443000920435e-05, "loss": 0.5565, "step": 42864 }, { "epoch": 2.59, "learning_rate": 3.017292109909013e-05, "loss": 0.7859, "step": 42866 }, { "epoch": 2.59, "learning_rate": 3.0171412188975907e-05, "loss": 0.453, "step": 42868 }, { "epoch": 2.59, "learning_rate": 3.0169903278861676e-05, "loss": 0.5202, "step": 42870 }, { "epoch": 2.59, "learning_rate": 3.0168394368747453e-05, "loss": 0.4859, "step": 42872 }, { "epoch": 2.59, "learning_rate": 3.0166885458633233e-05, "loss": 0.4317, "step": 42874 }, { "epoch": 2.59, "learning_rate": 3.016537654851901e-05, "loss": 0.536, "step": 42876 }, { "epoch": 2.59, "learning_rate": 3.016386763840478e-05, "loss": 0.6954, "step": 42878 }, { "epoch": 2.59, "learning_rate": 3.0162358728290556e-05, "loss": 0.5068, "step": 42880 }, { "epoch": 2.59, "learning_rate": 3.0160849818176336e-05, "loss": 0.581, "step": 42882 }, { "epoch": 2.59, "learning_rate": 3.0159340908062105e-05, "loss": 0.5071, "step": 42884 }, { "epoch": 2.59, "learning_rate": 3.0157831997947882e-05, "loss": 0.4931, "step": 42886 }, { "epoch": 2.59, "learning_rate": 3.015632308783366e-05, "loss": 0.5788, "step": 42888 }, { "epoch": 2.59, "learning_rate": 3.0154814177719438e-05, "loss": 0.6638, "step": 42890 }, { "epoch": 2.59, "learning_rate": 3.0153305267605208e-05, "loss": 0.6837, "step": 42892 }, { "epoch": 2.59, "learning_rate": 3.0151796357490984e-05, "loss": 0.5082, "step": 42894 }, { "epoch": 2.59, "learning_rate": 3.0150287447376764e-05, "loss": 0.4761, "step": 42896 }, { "epoch": 2.59, "learning_rate": 3.0148778537262534e-05, "loss": 0.6327, "step": 42898 }, { "epoch": 2.59, "learning_rate": 3.014726962714831e-05, "loss": 0.5478, "step": 42900 }, { "epoch": 2.59, "learning_rate": 3.0145760717034087e-05, "loss": 0.5549, "step": 42902 }, { "epoch": 2.59, "learning_rate": 3.0144251806919867e-05, "loss": 0.81, "step": 42904 }, { "epoch": 2.59, "learning_rate": 3.0142742896805637e-05, "loss": 0.5346, "step": 42906 }, { "epoch": 2.59, "learning_rate": 3.0141233986691413e-05, "loss": 0.5416, "step": 42908 }, { "epoch": 2.59, "learning_rate": 3.013972507657719e-05, "loss": 0.5752, "step": 42910 }, { "epoch": 2.59, "learning_rate": 3.0138216166462963e-05, "loss": 0.6156, "step": 42912 }, { "epoch": 2.59, "learning_rate": 3.013670725634874e-05, "loss": 0.5402, "step": 42914 }, { "epoch": 2.59, "learning_rate": 3.0135198346234516e-05, "loss": 0.4539, "step": 42916 }, { "epoch": 2.59, "learning_rate": 3.0133689436120292e-05, "loss": 0.5942, "step": 42918 }, { "epoch": 2.59, "learning_rate": 3.0132180526006066e-05, "loss": 0.4455, "step": 42920 }, { "epoch": 2.59, "learning_rate": 3.0130671615891842e-05, "loss": 0.6787, "step": 42922 }, { "epoch": 2.59, "learning_rate": 3.012916270577762e-05, "loss": 0.6779, "step": 42924 }, { "epoch": 2.59, "learning_rate": 3.0127653795663395e-05, "loss": 0.6323, "step": 42926 }, { "epoch": 2.59, "learning_rate": 3.0126144885549168e-05, "loss": 0.4592, "step": 42928 }, { "epoch": 2.59, "learning_rate": 3.0124635975434945e-05, "loss": 0.7905, "step": 42930 }, { "epoch": 2.59, "learning_rate": 3.012312706532072e-05, "loss": 0.414, "step": 42932 }, { "epoch": 2.59, "learning_rate": 3.0121618155206494e-05, "loss": 0.457, "step": 42934 }, { "epoch": 2.59, "learning_rate": 3.012010924509227e-05, "loss": 0.5232, "step": 42936 }, { "epoch": 2.59, "learning_rate": 3.0118600334978047e-05, "loss": 0.5429, "step": 42938 }, { "epoch": 2.59, "learning_rate": 3.0117091424863824e-05, "loss": 0.5818, "step": 42940 }, { "epoch": 2.59, "learning_rate": 3.0115582514749597e-05, "loss": 0.4782, "step": 42942 }, { "epoch": 2.59, "learning_rate": 3.0114073604635374e-05, "loss": 0.6853, "step": 42944 }, { "epoch": 2.59, "learning_rate": 3.011256469452115e-05, "loss": 0.755, "step": 42946 }, { "epoch": 2.59, "learning_rate": 3.0111055784406923e-05, "loss": 0.5054, "step": 42948 }, { "epoch": 2.59, "learning_rate": 3.01095468742927e-05, "loss": 0.7346, "step": 42950 }, { "epoch": 2.59, "learning_rate": 3.0108037964178476e-05, "loss": 0.5875, "step": 42952 }, { "epoch": 2.59, "learning_rate": 3.0106529054064253e-05, "loss": 0.6377, "step": 42954 }, { "epoch": 2.59, "learning_rate": 3.0105020143950026e-05, "loss": 0.4661, "step": 42956 }, { "epoch": 2.59, "learning_rate": 3.0103511233835802e-05, "loss": 0.5394, "step": 42958 }, { "epoch": 2.59, "learning_rate": 3.010200232372158e-05, "loss": 0.4876, "step": 42960 }, { "epoch": 2.59, "learning_rate": 3.010049341360735e-05, "loss": 0.7293, "step": 42962 }, { "epoch": 2.59, "learning_rate": 3.009898450349313e-05, "loss": 0.4962, "step": 42964 }, { "epoch": 2.59, "learning_rate": 3.0097475593378905e-05, "loss": 0.484, "step": 42966 }, { "epoch": 2.59, "learning_rate": 3.009596668326468e-05, "loss": 0.5113, "step": 42968 }, { "epoch": 2.59, "learning_rate": 3.009445777315045e-05, "loss": 0.5742, "step": 42970 }, { "epoch": 2.59, "learning_rate": 3.009294886303623e-05, "loss": 0.4221, "step": 42972 }, { "epoch": 2.59, "learning_rate": 3.0091439952922008e-05, "loss": 0.5336, "step": 42974 }, { "epoch": 2.59, "learning_rate": 3.0089931042807784e-05, "loss": 0.4959, "step": 42976 }, { "epoch": 2.59, "learning_rate": 3.0088422132693554e-05, "loss": 0.8002, "step": 42978 }, { "epoch": 2.59, "learning_rate": 3.0086913222579334e-05, "loss": 0.6656, "step": 42980 }, { "epoch": 2.59, "learning_rate": 3.008540431246511e-05, "loss": 0.8151, "step": 42982 }, { "epoch": 2.59, "learning_rate": 3.008389540235088e-05, "loss": 0.4046, "step": 42984 }, { "epoch": 2.59, "learning_rate": 3.0082386492236657e-05, "loss": 0.7279, "step": 42986 }, { "epoch": 2.59, "learning_rate": 3.0080877582122437e-05, "loss": 0.597, "step": 42988 }, { "epoch": 2.59, "learning_rate": 3.0079368672008213e-05, "loss": 0.5391, "step": 42990 }, { "epoch": 2.59, "learning_rate": 3.0077859761893983e-05, "loss": 0.4852, "step": 42992 }, { "epoch": 2.6, "learning_rate": 3.007635085177976e-05, "loss": 0.4989, "step": 42994 }, { "epoch": 2.6, "learning_rate": 3.007484194166554e-05, "loss": 0.332, "step": 42996 }, { "epoch": 2.6, "learning_rate": 3.007333303155131e-05, "loss": 0.6978, "step": 42998 }, { "epoch": 2.6, "learning_rate": 3.0071824121437085e-05, "loss": 0.3943, "step": 43000 }, { "epoch": 2.6, "learning_rate": 3.0070315211322862e-05, "loss": 0.748, "step": 43002 }, { "epoch": 2.6, "learning_rate": 3.0068806301208642e-05, "loss": 0.4799, "step": 43004 }, { "epoch": 2.6, "learning_rate": 3.006729739109441e-05, "loss": 0.5968, "step": 43006 }, { "epoch": 2.6, "learning_rate": 3.0065788480980188e-05, "loss": 0.4007, "step": 43008 }, { "epoch": 2.6, "learning_rate": 3.0064279570865968e-05, "loss": 0.6027, "step": 43010 }, { "epoch": 2.6, "learning_rate": 3.0062770660751738e-05, "loss": 0.2933, "step": 43012 }, { "epoch": 2.6, "learning_rate": 3.0061261750637514e-05, "loss": 0.5355, "step": 43014 }, { "epoch": 2.6, "learning_rate": 3.005975284052329e-05, "loss": 0.4489, "step": 43016 }, { "epoch": 2.6, "learning_rate": 3.005824393040907e-05, "loss": 0.3931, "step": 43018 }, { "epoch": 2.6, "learning_rate": 3.005673502029484e-05, "loss": 0.428, "step": 43020 }, { "epoch": 2.6, "learning_rate": 3.0055226110180617e-05, "loss": 0.6096, "step": 43022 }, { "epoch": 2.6, "learning_rate": 3.0053717200066393e-05, "loss": 0.3958, "step": 43024 }, { "epoch": 2.6, "learning_rate": 3.0052208289952173e-05, "loss": 0.6955, "step": 43026 }, { "epoch": 2.6, "learning_rate": 3.0050699379837943e-05, "loss": 0.4534, "step": 43028 }, { "epoch": 2.6, "learning_rate": 3.004919046972372e-05, "loss": 0.7281, "step": 43030 }, { "epoch": 2.6, "learning_rate": 3.0047681559609496e-05, "loss": 0.5492, "step": 43032 }, { "epoch": 2.6, "learning_rate": 3.004617264949527e-05, "loss": 0.2909, "step": 43034 }, { "epoch": 2.6, "learning_rate": 3.0044663739381046e-05, "loss": 0.5083, "step": 43036 }, { "epoch": 2.6, "learning_rate": 3.0043154829266822e-05, "loss": 0.4056, "step": 43038 }, { "epoch": 2.6, "learning_rate": 3.00416459191526e-05, "loss": 0.4712, "step": 43040 }, { "epoch": 2.6, "learning_rate": 3.0040137009038372e-05, "loss": 0.6818, "step": 43042 }, { "epoch": 2.6, "learning_rate": 3.003862809892415e-05, "loss": 0.7025, "step": 43044 }, { "epoch": 2.6, "learning_rate": 3.0037119188809925e-05, "loss": 0.5139, "step": 43046 }, { "epoch": 2.6, "learning_rate": 3.0035610278695698e-05, "loss": 0.4831, "step": 43048 }, { "epoch": 2.6, "learning_rate": 3.0034101368581475e-05, "loss": 0.5773, "step": 43050 }, { "epoch": 2.6, "learning_rate": 3.003259245846725e-05, "loss": 0.6607, "step": 43052 }, { "epoch": 2.6, "learning_rate": 3.0031083548353028e-05, "loss": 0.4406, "step": 43054 }, { "epoch": 2.6, "learning_rate": 3.00295746382388e-05, "loss": 0.6678, "step": 43056 }, { "epoch": 2.6, "learning_rate": 3.0028065728124577e-05, "loss": 0.9173, "step": 43058 }, { "epoch": 2.6, "learning_rate": 3.0026556818010354e-05, "loss": 0.4321, "step": 43060 }, { "epoch": 2.6, "learning_rate": 3.0025047907896127e-05, "loss": 0.7842, "step": 43062 }, { "epoch": 2.6, "learning_rate": 3.0023538997781903e-05, "loss": 0.5204, "step": 43064 }, { "epoch": 2.6, "learning_rate": 3.002203008766768e-05, "loss": 0.4259, "step": 43066 }, { "epoch": 2.6, "learning_rate": 3.0020521177553456e-05, "loss": 0.5517, "step": 43068 }, { "epoch": 2.6, "learning_rate": 3.001901226743923e-05, "loss": 0.5333, "step": 43070 }, { "epoch": 2.6, "learning_rate": 3.0017503357325006e-05, "loss": 0.628, "step": 43072 }, { "epoch": 2.6, "learning_rate": 3.0015994447210783e-05, "loss": 0.5024, "step": 43074 }, { "epoch": 2.6, "learning_rate": 3.0014485537096552e-05, "loss": 0.3793, "step": 43076 }, { "epoch": 2.6, "learning_rate": 3.0012976626982332e-05, "loss": 0.5201, "step": 43078 }, { "epoch": 2.6, "learning_rate": 3.001146771686811e-05, "loss": 0.4765, "step": 43080 }, { "epoch": 2.6, "learning_rate": 3.0009958806753885e-05, "loss": 0.4845, "step": 43082 }, { "epoch": 2.6, "learning_rate": 3.0008449896639655e-05, "loss": 0.904, "step": 43084 }, { "epoch": 2.6, "learning_rate": 3.0006940986525435e-05, "loss": 0.4399, "step": 43086 }, { "epoch": 2.6, "learning_rate": 3.000543207641121e-05, "loss": 0.6033, "step": 43088 }, { "epoch": 2.6, "learning_rate": 3.0003923166296988e-05, "loss": 0.5305, "step": 43090 }, { "epoch": 2.6, "learning_rate": 3.0002414256182758e-05, "loss": 0.4864, "step": 43092 }, { "epoch": 2.6, "learning_rate": 3.0000905346068538e-05, "loss": 0.5957, "step": 43094 }, { "epoch": 2.6, "learning_rate": 2.9999396435954314e-05, "loss": 0.547, "step": 43096 }, { "epoch": 2.6, "learning_rate": 2.9997887525840084e-05, "loss": 0.7212, "step": 43098 }, { "epoch": 2.6, "learning_rate": 2.999637861572586e-05, "loss": 0.5892, "step": 43100 }, { "epoch": 2.6, "learning_rate": 2.999486970561164e-05, "loss": 0.7234, "step": 43102 }, { "epoch": 2.6, "learning_rate": 2.9993360795497417e-05, "loss": 0.5391, "step": 43104 }, { "epoch": 2.6, "learning_rate": 2.9991851885383186e-05, "loss": 0.5512, "step": 43106 }, { "epoch": 2.6, "learning_rate": 2.9990342975268963e-05, "loss": 0.661, "step": 43108 }, { "epoch": 2.6, "learning_rate": 2.9988834065154743e-05, "loss": 0.6906, "step": 43110 }, { "epoch": 2.6, "learning_rate": 2.9987325155040513e-05, "loss": 0.5321, "step": 43112 }, { "epoch": 2.6, "learning_rate": 2.998581624492629e-05, "loss": 0.6061, "step": 43114 }, { "epoch": 2.6, "learning_rate": 2.9984307334812066e-05, "loss": 0.4908, "step": 43116 }, { "epoch": 2.6, "learning_rate": 2.9982798424697846e-05, "loss": 0.473, "step": 43118 }, { "epoch": 2.6, "learning_rate": 2.9981289514583615e-05, "loss": 0.5904, "step": 43120 }, { "epoch": 2.6, "learning_rate": 2.9979780604469392e-05, "loss": 0.3854, "step": 43122 }, { "epoch": 2.6, "learning_rate": 2.9978271694355172e-05, "loss": 0.4498, "step": 43124 }, { "epoch": 2.6, "learning_rate": 2.997676278424094e-05, "loss": 0.51, "step": 43126 }, { "epoch": 2.6, "learning_rate": 2.9975253874126718e-05, "loss": 0.6496, "step": 43128 }, { "epoch": 2.6, "learning_rate": 2.9973744964012494e-05, "loss": 0.6316, "step": 43130 }, { "epoch": 2.6, "learning_rate": 2.9972236053898274e-05, "loss": 0.6389, "step": 43132 }, { "epoch": 2.6, "learning_rate": 2.9970727143784044e-05, "loss": 0.6569, "step": 43134 }, { "epoch": 2.6, "learning_rate": 2.996921823366982e-05, "loss": 0.427, "step": 43136 }, { "epoch": 2.6, "learning_rate": 2.9967709323555597e-05, "loss": 0.6725, "step": 43138 }, { "epoch": 2.6, "learning_rate": 2.9966200413441377e-05, "loss": 0.4733, "step": 43140 }, { "epoch": 2.6, "learning_rate": 2.9964691503327147e-05, "loss": 0.3421, "step": 43142 }, { "epoch": 2.6, "learning_rate": 2.9963182593212923e-05, "loss": 0.4626, "step": 43144 }, { "epoch": 2.6, "learning_rate": 2.99616736830987e-05, "loss": 0.5386, "step": 43146 }, { "epoch": 2.6, "learning_rate": 2.9960164772984473e-05, "loss": 0.4274, "step": 43148 }, { "epoch": 2.6, "learning_rate": 2.995865586287025e-05, "loss": 0.6607, "step": 43150 }, { "epoch": 2.6, "learning_rate": 2.9957146952756026e-05, "loss": 0.8597, "step": 43152 }, { "epoch": 2.6, "learning_rate": 2.9955638042641802e-05, "loss": 0.4885, "step": 43154 }, { "epoch": 2.6, "learning_rate": 2.9954129132527576e-05, "loss": 0.6531, "step": 43156 }, { "epoch": 2.6, "learning_rate": 2.9952620222413352e-05, "loss": 0.3418, "step": 43158 }, { "epoch": 2.61, "learning_rate": 2.995111131229913e-05, "loss": 0.3137, "step": 43160 }, { "epoch": 2.61, "learning_rate": 2.9949602402184902e-05, "loss": 0.6145, "step": 43162 }, { "epoch": 2.61, "learning_rate": 2.9948093492070678e-05, "loss": 0.7794, "step": 43164 }, { "epoch": 2.61, "learning_rate": 2.9946584581956455e-05, "loss": 0.5191, "step": 43166 }, { "epoch": 2.61, "learning_rate": 2.994507567184223e-05, "loss": 0.6708, "step": 43168 }, { "epoch": 2.61, "learning_rate": 2.9943566761728004e-05, "loss": 0.5031, "step": 43170 }, { "epoch": 2.61, "learning_rate": 2.994205785161378e-05, "loss": 0.3295, "step": 43172 }, { "epoch": 2.61, "learning_rate": 2.9940548941499557e-05, "loss": 0.4356, "step": 43174 }, { "epoch": 2.61, "learning_rate": 2.993904003138533e-05, "loss": 0.5197, "step": 43176 }, { "epoch": 2.61, "learning_rate": 2.9937531121271107e-05, "loss": 0.5253, "step": 43178 }, { "epoch": 2.61, "learning_rate": 2.9936022211156884e-05, "loss": 0.4878, "step": 43180 }, { "epoch": 2.61, "learning_rate": 2.993451330104266e-05, "loss": 0.4919, "step": 43182 }, { "epoch": 2.61, "learning_rate": 2.9933004390928433e-05, "loss": 0.4197, "step": 43184 }, { "epoch": 2.61, "learning_rate": 2.993149548081421e-05, "loss": 0.5789, "step": 43186 }, { "epoch": 2.61, "learning_rate": 2.9929986570699986e-05, "loss": 0.3844, "step": 43188 }, { "epoch": 2.61, "learning_rate": 2.9928477660585763e-05, "loss": 0.4291, "step": 43190 }, { "epoch": 2.61, "learning_rate": 2.9926968750471536e-05, "loss": 0.6501, "step": 43192 }, { "epoch": 2.61, "learning_rate": 2.9925459840357312e-05, "loss": 0.6385, "step": 43194 }, { "epoch": 2.61, "learning_rate": 2.992395093024309e-05, "loss": 0.7443, "step": 43196 }, { "epoch": 2.61, "learning_rate": 2.992244202012886e-05, "loss": 0.4314, "step": 43198 }, { "epoch": 2.61, "learning_rate": 2.992093311001464e-05, "loss": 0.4974, "step": 43200 }, { "epoch": 2.61, "learning_rate": 2.9919424199900415e-05, "loss": 0.6366, "step": 43202 }, { "epoch": 2.61, "learning_rate": 2.991791528978619e-05, "loss": 0.5218, "step": 43204 }, { "epoch": 2.61, "learning_rate": 2.991640637967196e-05, "loss": 0.6298, "step": 43206 }, { "epoch": 2.61, "learning_rate": 2.991489746955774e-05, "loss": 0.3839, "step": 43208 }, { "epoch": 2.61, "learning_rate": 2.9913388559443518e-05, "loss": 0.4271, "step": 43210 }, { "epoch": 2.61, "learning_rate": 2.9911879649329287e-05, "loss": 0.8319, "step": 43212 }, { "epoch": 2.61, "learning_rate": 2.9910370739215064e-05, "loss": 0.7861, "step": 43214 }, { "epoch": 2.61, "learning_rate": 2.9908861829100844e-05, "loss": 0.4271, "step": 43216 }, { "epoch": 2.61, "learning_rate": 2.990735291898662e-05, "loss": 0.551, "step": 43218 }, { "epoch": 2.61, "learning_rate": 2.990584400887239e-05, "loss": 0.574, "step": 43220 }, { "epoch": 2.61, "learning_rate": 2.9904335098758167e-05, "loss": 0.6176, "step": 43222 }, { "epoch": 2.61, "learning_rate": 2.9902826188643947e-05, "loss": 0.4413, "step": 43224 }, { "epoch": 2.61, "learning_rate": 2.9901317278529716e-05, "loss": 0.6052, "step": 43226 }, { "epoch": 2.61, "learning_rate": 2.9899808368415493e-05, "loss": 0.5584, "step": 43228 }, { "epoch": 2.61, "learning_rate": 2.989829945830127e-05, "loss": 0.8158, "step": 43230 }, { "epoch": 2.61, "learning_rate": 2.989679054818705e-05, "loss": 0.5151, "step": 43232 }, { "epoch": 2.61, "learning_rate": 2.989528163807282e-05, "loss": 0.5066, "step": 43234 }, { "epoch": 2.61, "learning_rate": 2.9893772727958595e-05, "loss": 0.557, "step": 43236 }, { "epoch": 2.61, "learning_rate": 2.9892263817844372e-05, "loss": 0.5317, "step": 43238 }, { "epoch": 2.61, "learning_rate": 2.9890754907730145e-05, "loss": 0.5651, "step": 43240 }, { "epoch": 2.61, "learning_rate": 2.988924599761592e-05, "loss": 0.4397, "step": 43242 }, { "epoch": 2.61, "learning_rate": 2.9887737087501698e-05, "loss": 0.6516, "step": 43244 }, { "epoch": 2.61, "learning_rate": 2.9886228177387478e-05, "loss": 0.4019, "step": 43246 }, { "epoch": 2.61, "learning_rate": 2.9884719267273248e-05, "loss": 0.464, "step": 43248 }, { "epoch": 2.61, "learning_rate": 2.9883210357159024e-05, "loss": 0.6085, "step": 43250 }, { "epoch": 2.61, "learning_rate": 2.98817014470448e-05, "loss": 0.7578, "step": 43252 }, { "epoch": 2.61, "learning_rate": 2.988019253693058e-05, "loss": 0.5412, "step": 43254 }, { "epoch": 2.61, "learning_rate": 2.987868362681635e-05, "loss": 0.5498, "step": 43256 }, { "epoch": 2.61, "learning_rate": 2.9877174716702127e-05, "loss": 0.7651, "step": 43258 }, { "epoch": 2.61, "learning_rate": 2.9875665806587903e-05, "loss": 0.4001, "step": 43260 }, { "epoch": 2.61, "learning_rate": 2.9874156896473677e-05, "loss": 0.6301, "step": 43262 }, { "epoch": 2.61, "learning_rate": 2.9872647986359453e-05, "loss": 0.9192, "step": 43264 }, { "epoch": 2.61, "learning_rate": 2.987113907624523e-05, "loss": 0.6565, "step": 43266 }, { "epoch": 2.61, "learning_rate": 2.9869630166131006e-05, "loss": 0.4526, "step": 43268 }, { "epoch": 2.61, "learning_rate": 2.986812125601678e-05, "loss": 0.6068, "step": 43270 }, { "epoch": 2.61, "learning_rate": 2.9866612345902556e-05, "loss": 0.4408, "step": 43272 }, { "epoch": 2.61, "learning_rate": 2.9865103435788332e-05, "loss": 0.744, "step": 43274 }, { "epoch": 2.61, "learning_rate": 2.9863594525674105e-05, "loss": 0.4045, "step": 43276 }, { "epoch": 2.61, "learning_rate": 2.9862085615559882e-05, "loss": 0.6014, "step": 43278 }, { "epoch": 2.61, "learning_rate": 2.986057670544566e-05, "loss": 0.4631, "step": 43280 }, { "epoch": 2.61, "learning_rate": 2.9859067795331435e-05, "loss": 0.5808, "step": 43282 }, { "epoch": 2.61, "learning_rate": 2.9857558885217208e-05, "loss": 0.4117, "step": 43284 }, { "epoch": 2.61, "learning_rate": 2.9856049975102985e-05, "loss": 0.5127, "step": 43286 }, { "epoch": 2.61, "learning_rate": 2.985454106498876e-05, "loss": 0.4901, "step": 43288 }, { "epoch": 2.61, "learning_rate": 2.9853032154874534e-05, "loss": 0.5469, "step": 43290 }, { "epoch": 2.61, "learning_rate": 2.985152324476031e-05, "loss": 0.5091, "step": 43292 }, { "epoch": 2.61, "learning_rate": 2.9850014334646087e-05, "loss": 0.4246, "step": 43294 }, { "epoch": 2.61, "learning_rate": 2.9848505424531864e-05, "loss": 0.5794, "step": 43296 }, { "epoch": 2.61, "learning_rate": 2.9846996514417637e-05, "loss": 0.4503, "step": 43298 }, { "epoch": 2.61, "learning_rate": 2.9845487604303413e-05, "loss": 0.5579, "step": 43300 }, { "epoch": 2.61, "learning_rate": 2.984397869418919e-05, "loss": 0.7071, "step": 43302 }, { "epoch": 2.61, "learning_rate": 2.9842469784074966e-05, "loss": 0.5382, "step": 43304 }, { "epoch": 2.61, "learning_rate": 2.984096087396074e-05, "loss": 0.6352, "step": 43306 }, { "epoch": 2.61, "learning_rate": 2.9839451963846516e-05, "loss": 0.5024, "step": 43308 }, { "epoch": 2.61, "learning_rate": 2.9837943053732293e-05, "loss": 0.4686, "step": 43310 }, { "epoch": 2.61, "learning_rate": 2.9836434143618062e-05, "loss": 0.5831, "step": 43312 }, { "epoch": 2.61, "learning_rate": 2.9834925233503842e-05, "loss": 0.3469, "step": 43314 }, { "epoch": 2.61, "learning_rate": 2.983341632338962e-05, "loss": 0.3382, "step": 43316 }, { "epoch": 2.61, "learning_rate": 2.9831907413275395e-05, "loss": 0.5325, "step": 43318 }, { "epoch": 2.61, "learning_rate": 2.9830398503161165e-05, "loss": 0.4185, "step": 43320 }, { "epoch": 2.61, "learning_rate": 2.9828889593046945e-05, "loss": 0.3949, "step": 43322 }, { "epoch": 2.61, "learning_rate": 2.982738068293272e-05, "loss": 0.628, "step": 43324 }, { "epoch": 2.62, "learning_rate": 2.982587177281849e-05, "loss": 0.5089, "step": 43326 }, { "epoch": 2.62, "learning_rate": 2.9824362862704268e-05, "loss": 0.7185, "step": 43328 }, { "epoch": 2.62, "learning_rate": 2.9822853952590048e-05, "loss": 0.6348, "step": 43330 }, { "epoch": 2.62, "learning_rate": 2.9821345042475824e-05, "loss": 0.6193, "step": 43332 }, { "epoch": 2.62, "learning_rate": 2.9819836132361594e-05, "loss": 0.5228, "step": 43334 }, { "epoch": 2.62, "learning_rate": 2.981832722224737e-05, "loss": 0.4594, "step": 43336 }, { "epoch": 2.62, "learning_rate": 2.981681831213315e-05, "loss": 0.4962, "step": 43338 }, { "epoch": 2.62, "learning_rate": 2.981530940201892e-05, "loss": 0.5861, "step": 43340 }, { "epoch": 2.62, "learning_rate": 2.9813800491904697e-05, "loss": 0.628, "step": 43342 }, { "epoch": 2.62, "learning_rate": 2.9812291581790473e-05, "loss": 0.8571, "step": 43344 }, { "epoch": 2.62, "learning_rate": 2.9810782671676253e-05, "loss": 0.6001, "step": 43346 }, { "epoch": 2.62, "learning_rate": 2.9809273761562023e-05, "loss": 0.4984, "step": 43348 }, { "epoch": 2.62, "learning_rate": 2.98077648514478e-05, "loss": 0.5112, "step": 43350 }, { "epoch": 2.62, "learning_rate": 2.9806255941333576e-05, "loss": 0.4871, "step": 43352 }, { "epoch": 2.62, "learning_rate": 2.9804747031219356e-05, "loss": 0.6166, "step": 43354 }, { "epoch": 2.62, "learning_rate": 2.9803238121105125e-05, "loss": 0.5995, "step": 43356 }, { "epoch": 2.62, "learning_rate": 2.9801729210990902e-05, "loss": 0.6278, "step": 43358 }, { "epoch": 2.62, "learning_rate": 2.980022030087668e-05, "loss": 0.7661, "step": 43360 }, { "epoch": 2.62, "learning_rate": 2.979871139076245e-05, "loss": 0.458, "step": 43362 }, { "epoch": 2.62, "learning_rate": 2.9797202480648228e-05, "loss": 0.5734, "step": 43364 }, { "epoch": 2.62, "learning_rate": 2.9795693570534005e-05, "loss": 0.6468, "step": 43366 }, { "epoch": 2.62, "learning_rate": 2.9794184660419784e-05, "loss": 0.52, "step": 43368 }, { "epoch": 2.62, "learning_rate": 2.9792675750305554e-05, "loss": 0.7735, "step": 43370 }, { "epoch": 2.62, "learning_rate": 2.979116684019133e-05, "loss": 0.7053, "step": 43372 }, { "epoch": 2.62, "learning_rate": 2.9789657930077107e-05, "loss": 0.3187, "step": 43374 }, { "epoch": 2.62, "learning_rate": 2.978814901996288e-05, "loss": 0.6691, "step": 43376 }, { "epoch": 2.62, "learning_rate": 2.9786640109848657e-05, "loss": 0.5244, "step": 43378 }, { "epoch": 2.62, "learning_rate": 2.9785131199734433e-05, "loss": 0.5483, "step": 43380 }, { "epoch": 2.62, "learning_rate": 2.978362228962021e-05, "loss": 0.4278, "step": 43382 }, { "epoch": 2.62, "learning_rate": 2.9782113379505983e-05, "loss": 0.4656, "step": 43384 }, { "epoch": 2.62, "learning_rate": 2.978060446939176e-05, "loss": 0.7171, "step": 43386 }, { "epoch": 2.62, "learning_rate": 2.9779095559277536e-05, "loss": 0.3998, "step": 43388 }, { "epoch": 2.62, "learning_rate": 2.977758664916331e-05, "loss": 0.4367, "step": 43390 }, { "epoch": 2.62, "learning_rate": 2.9776077739049086e-05, "loss": 0.8599, "step": 43392 }, { "epoch": 2.62, "learning_rate": 2.9774568828934862e-05, "loss": 0.4321, "step": 43394 }, { "epoch": 2.62, "learning_rate": 2.977305991882064e-05, "loss": 0.596, "step": 43396 }, { "epoch": 2.62, "learning_rate": 2.9771551008706412e-05, "loss": 0.6155, "step": 43398 }, { "epoch": 2.62, "learning_rate": 2.9770042098592188e-05, "loss": 0.6032, "step": 43400 }, { "epoch": 2.62, "learning_rate": 2.9768533188477965e-05, "loss": 0.6944, "step": 43402 }, { "epoch": 2.62, "learning_rate": 2.9767024278363738e-05, "loss": 0.6711, "step": 43404 }, { "epoch": 2.62, "learning_rate": 2.9765515368249514e-05, "loss": 0.7479, "step": 43406 }, { "epoch": 2.62, "learning_rate": 2.976400645813529e-05, "loss": 0.5069, "step": 43408 }, { "epoch": 2.62, "learning_rate": 2.9762497548021067e-05, "loss": 0.7088, "step": 43410 }, { "epoch": 2.62, "learning_rate": 2.976098863790684e-05, "loss": 0.5117, "step": 43412 }, { "epoch": 2.62, "learning_rate": 2.9759479727792617e-05, "loss": 0.5397, "step": 43414 }, { "epoch": 2.62, "learning_rate": 2.9757970817678394e-05, "loss": 0.5776, "step": 43416 }, { "epoch": 2.62, "learning_rate": 2.975646190756417e-05, "loss": 0.4656, "step": 43418 }, { "epoch": 2.62, "learning_rate": 2.9754952997449943e-05, "loss": 0.4675, "step": 43420 }, { "epoch": 2.62, "learning_rate": 2.975344408733572e-05, "loss": 0.4275, "step": 43422 }, { "epoch": 2.62, "learning_rate": 2.9751935177221496e-05, "loss": 0.4757, "step": 43424 }, { "epoch": 2.62, "learning_rate": 2.9750426267107266e-05, "loss": 0.586, "step": 43426 }, { "epoch": 2.62, "learning_rate": 2.9748917356993046e-05, "loss": 0.7189, "step": 43428 }, { "epoch": 2.62, "learning_rate": 2.9747408446878822e-05, "loss": 0.4929, "step": 43430 }, { "epoch": 2.62, "learning_rate": 2.97458995367646e-05, "loss": 0.5837, "step": 43432 }, { "epoch": 2.62, "learning_rate": 2.974439062665037e-05, "loss": 0.4807, "step": 43434 }, { "epoch": 2.62, "learning_rate": 2.974288171653615e-05, "loss": 0.5501, "step": 43436 }, { "epoch": 2.62, "learning_rate": 2.9741372806421925e-05, "loss": 0.5869, "step": 43438 }, { "epoch": 2.62, "learning_rate": 2.9739863896307695e-05, "loss": 0.4173, "step": 43440 }, { "epoch": 2.62, "learning_rate": 2.973835498619347e-05, "loss": 0.6265, "step": 43442 }, { "epoch": 2.62, "learning_rate": 2.973684607607925e-05, "loss": 0.7091, "step": 43444 }, { "epoch": 2.62, "learning_rate": 2.9735337165965028e-05, "loss": 0.8131, "step": 43446 }, { "epoch": 2.62, "learning_rate": 2.9733828255850798e-05, "loss": 0.4071, "step": 43448 }, { "epoch": 2.62, "learning_rate": 2.9732319345736574e-05, "loss": 0.5184, "step": 43450 }, { "epoch": 2.62, "learning_rate": 2.9730810435622354e-05, "loss": 0.4936, "step": 43452 }, { "epoch": 2.62, "learning_rate": 2.9729301525508124e-05, "loss": 0.6458, "step": 43454 }, { "epoch": 2.62, "learning_rate": 2.97277926153939e-05, "loss": 0.6836, "step": 43456 }, { "epoch": 2.62, "learning_rate": 2.9726283705279677e-05, "loss": 0.4162, "step": 43458 }, { "epoch": 2.62, "learning_rate": 2.9724774795165457e-05, "loss": 0.4961, "step": 43460 }, { "epoch": 2.62, "learning_rate": 2.9723265885051226e-05, "loss": 0.5275, "step": 43462 }, { "epoch": 2.62, "learning_rate": 2.9721756974937003e-05, "loss": 0.4829, "step": 43464 }, { "epoch": 2.62, "learning_rate": 2.972024806482278e-05, "loss": 0.4996, "step": 43466 }, { "epoch": 2.62, "learning_rate": 2.971873915470856e-05, "loss": 0.519, "step": 43468 }, { "epoch": 2.62, "learning_rate": 2.971723024459433e-05, "loss": 0.4447, "step": 43470 }, { "epoch": 2.62, "learning_rate": 2.9715721334480106e-05, "loss": 0.5525, "step": 43472 }, { "epoch": 2.62, "learning_rate": 2.9714212424365882e-05, "loss": 0.4462, "step": 43474 }, { "epoch": 2.62, "learning_rate": 2.9712703514251655e-05, "loss": 0.4255, "step": 43476 }, { "epoch": 2.62, "learning_rate": 2.971119460413743e-05, "loss": 0.5171, "step": 43478 }, { "epoch": 2.62, "learning_rate": 2.9709685694023208e-05, "loss": 0.4653, "step": 43480 }, { "epoch": 2.62, "learning_rate": 2.9708176783908988e-05, "loss": 0.6513, "step": 43482 }, { "epoch": 2.62, "learning_rate": 2.9706667873794758e-05, "loss": 0.3644, "step": 43484 }, { "epoch": 2.62, "learning_rate": 2.9705158963680534e-05, "loss": 0.5838, "step": 43486 }, { "epoch": 2.62, "learning_rate": 2.970365005356631e-05, "loss": 0.6632, "step": 43488 }, { "epoch": 2.62, "learning_rate": 2.9702141143452084e-05, "loss": 0.5986, "step": 43490 }, { "epoch": 2.63, "learning_rate": 2.970063223333786e-05, "loss": 0.6078, "step": 43492 }, { "epoch": 2.63, "learning_rate": 2.9699123323223637e-05, "loss": 0.5091, "step": 43494 }, { "epoch": 2.63, "learning_rate": 2.9697614413109414e-05, "loss": 0.8256, "step": 43496 }, { "epoch": 2.63, "learning_rate": 2.9696105502995187e-05, "loss": 0.5829, "step": 43498 }, { "epoch": 2.63, "learning_rate": 2.9694596592880963e-05, "loss": 0.3529, "step": 43500 }, { "epoch": 2.63, "learning_rate": 2.969308768276674e-05, "loss": 0.6404, "step": 43502 }, { "epoch": 2.63, "learning_rate": 2.9691578772652513e-05, "loss": 0.4763, "step": 43504 }, { "epoch": 2.63, "learning_rate": 2.969006986253829e-05, "loss": 0.4949, "step": 43506 }, { "epoch": 2.63, "learning_rate": 2.9688560952424066e-05, "loss": 0.5444, "step": 43508 }, { "epoch": 2.63, "learning_rate": 2.9687052042309842e-05, "loss": 0.4783, "step": 43510 }, { "epoch": 2.63, "learning_rate": 2.9685543132195615e-05, "loss": 0.651, "step": 43512 }, { "epoch": 2.63, "learning_rate": 2.9684034222081392e-05, "loss": 0.6799, "step": 43514 }, { "epoch": 2.63, "learning_rate": 2.968252531196717e-05, "loss": 0.5181, "step": 43516 }, { "epoch": 2.63, "learning_rate": 2.9681016401852945e-05, "loss": 0.5122, "step": 43518 }, { "epoch": 2.63, "learning_rate": 2.9679507491738718e-05, "loss": 0.514, "step": 43520 }, { "epoch": 2.63, "learning_rate": 2.9677998581624495e-05, "loss": 0.6553, "step": 43522 }, { "epoch": 2.63, "learning_rate": 2.967648967151027e-05, "loss": 0.4626, "step": 43524 }, { "epoch": 2.63, "learning_rate": 2.9674980761396044e-05, "loss": 0.3214, "step": 43526 }, { "epoch": 2.63, "learning_rate": 2.967347185128182e-05, "loss": 0.5863, "step": 43528 }, { "epoch": 2.63, "learning_rate": 2.9671962941167597e-05, "loss": 0.4265, "step": 43530 }, { "epoch": 2.63, "learning_rate": 2.9670454031053374e-05, "loss": 0.5382, "step": 43532 }, { "epoch": 2.63, "learning_rate": 2.9668945120939147e-05, "loss": 0.4318, "step": 43534 }, { "epoch": 2.63, "learning_rate": 2.9667436210824923e-05, "loss": 0.4001, "step": 43536 }, { "epoch": 2.63, "learning_rate": 2.96659273007107e-05, "loss": 0.5652, "step": 43538 }, { "epoch": 2.63, "learning_rate": 2.966441839059647e-05, "loss": 0.4262, "step": 43540 }, { "epoch": 2.63, "learning_rate": 2.966290948048225e-05, "loss": 0.4212, "step": 43542 }, { "epoch": 2.63, "learning_rate": 2.9661400570368026e-05, "loss": 0.5948, "step": 43544 }, { "epoch": 2.63, "learning_rate": 2.9659891660253803e-05, "loss": 0.4165, "step": 43546 }, { "epoch": 2.63, "learning_rate": 2.9658382750139572e-05, "loss": 0.73, "step": 43548 }, { "epoch": 2.63, "learning_rate": 2.9656873840025352e-05, "loss": 0.684, "step": 43550 }, { "epoch": 2.63, "learning_rate": 2.965536492991113e-05, "loss": 0.5015, "step": 43552 }, { "epoch": 2.63, "learning_rate": 2.96538560197969e-05, "loss": 0.6916, "step": 43554 }, { "epoch": 2.63, "learning_rate": 2.9652347109682675e-05, "loss": 0.5782, "step": 43556 }, { "epoch": 2.63, "learning_rate": 2.9650838199568455e-05, "loss": 0.4455, "step": 43558 }, { "epoch": 2.63, "learning_rate": 2.964932928945423e-05, "loss": 0.6452, "step": 43560 }, { "epoch": 2.63, "learning_rate": 2.964782037934e-05, "loss": 0.49, "step": 43562 }, { "epoch": 2.63, "learning_rate": 2.9646311469225778e-05, "loss": 0.5598, "step": 43564 }, { "epoch": 2.63, "learning_rate": 2.9644802559111558e-05, "loss": 0.6249, "step": 43566 }, { "epoch": 2.63, "learning_rate": 2.9643293648997334e-05, "loss": 0.7143, "step": 43568 }, { "epoch": 2.63, "learning_rate": 2.9641784738883104e-05, "loss": 0.5264, "step": 43570 }, { "epoch": 2.63, "learning_rate": 2.964027582876888e-05, "loss": 0.6341, "step": 43572 }, { "epoch": 2.63, "learning_rate": 2.963876691865466e-05, "loss": 0.5617, "step": 43574 }, { "epoch": 2.63, "learning_rate": 2.963725800854043e-05, "loss": 0.4996, "step": 43576 }, { "epoch": 2.63, "learning_rate": 2.9635749098426207e-05, "loss": 0.5355, "step": 43578 }, { "epoch": 2.63, "learning_rate": 2.9634240188311983e-05, "loss": 0.5796, "step": 43580 }, { "epoch": 2.63, "learning_rate": 2.9632731278197763e-05, "loss": 0.6575, "step": 43582 }, { "epoch": 2.63, "learning_rate": 2.9631222368083533e-05, "loss": 0.43, "step": 43584 }, { "epoch": 2.63, "learning_rate": 2.962971345796931e-05, "loss": 0.4295, "step": 43586 }, { "epoch": 2.63, "learning_rate": 2.9628204547855086e-05, "loss": 0.7206, "step": 43588 }, { "epoch": 2.63, "learning_rate": 2.962669563774086e-05, "loss": 0.7229, "step": 43590 }, { "epoch": 2.63, "learning_rate": 2.9625186727626635e-05, "loss": 0.3287, "step": 43592 }, { "epoch": 2.63, "learning_rate": 2.9623677817512412e-05, "loss": 0.4775, "step": 43594 }, { "epoch": 2.63, "learning_rate": 2.962216890739819e-05, "loss": 0.4811, "step": 43596 }, { "epoch": 2.63, "learning_rate": 2.962065999728396e-05, "loss": 0.4513, "step": 43598 }, { "epoch": 2.63, "learning_rate": 2.9619151087169738e-05, "loss": 0.5979, "step": 43600 }, { "epoch": 2.63, "learning_rate": 2.9617642177055515e-05, "loss": 0.4875, "step": 43602 }, { "epoch": 2.63, "learning_rate": 2.9616133266941288e-05, "loss": 0.4627, "step": 43604 }, { "epoch": 2.63, "learning_rate": 2.9614624356827064e-05, "loss": 0.6181, "step": 43606 }, { "epoch": 2.63, "learning_rate": 2.961311544671284e-05, "loss": 0.7822, "step": 43608 }, { "epoch": 2.63, "learning_rate": 2.9611606536598617e-05, "loss": 0.432, "step": 43610 }, { "epoch": 2.63, "learning_rate": 2.961009762648439e-05, "loss": 0.61, "step": 43612 }, { "epoch": 2.63, "learning_rate": 2.9608588716370167e-05, "loss": 0.552, "step": 43614 }, { "epoch": 2.63, "learning_rate": 2.9607079806255943e-05, "loss": 0.5149, "step": 43616 }, { "epoch": 2.63, "learning_rate": 2.9605570896141716e-05, "loss": 0.423, "step": 43618 }, { "epoch": 2.63, "learning_rate": 2.9604061986027493e-05, "loss": 0.5406, "step": 43620 }, { "epoch": 2.63, "learning_rate": 2.960255307591327e-05, "loss": 0.4987, "step": 43622 }, { "epoch": 2.63, "learning_rate": 2.9601044165799046e-05, "loss": 0.5469, "step": 43624 }, { "epoch": 2.63, "learning_rate": 2.959953525568482e-05, "loss": 0.5069, "step": 43626 }, { "epoch": 2.63, "learning_rate": 2.9598026345570596e-05, "loss": 0.6011, "step": 43628 }, { "epoch": 2.63, "learning_rate": 2.9596517435456372e-05, "loss": 0.4954, "step": 43630 }, { "epoch": 2.63, "learning_rate": 2.959500852534215e-05, "loss": 0.5366, "step": 43632 }, { "epoch": 2.63, "learning_rate": 2.9593499615227922e-05, "loss": 0.6296, "step": 43634 }, { "epoch": 2.63, "learning_rate": 2.95919907051137e-05, "loss": 0.4686, "step": 43636 }, { "epoch": 2.63, "learning_rate": 2.9590481794999475e-05, "loss": 0.498, "step": 43638 }, { "epoch": 2.63, "learning_rate": 2.9588972884885248e-05, "loss": 0.5423, "step": 43640 }, { "epoch": 2.63, "learning_rate": 2.9587463974771024e-05, "loss": 0.9396, "step": 43642 }, { "epoch": 2.63, "learning_rate": 2.95859550646568e-05, "loss": 0.5002, "step": 43644 }, { "epoch": 2.63, "learning_rate": 2.9584446154542577e-05, "loss": 0.594, "step": 43646 }, { "epoch": 2.63, "learning_rate": 2.958293724442835e-05, "loss": 0.573, "step": 43648 }, { "epoch": 2.63, "learning_rate": 2.9581428334314127e-05, "loss": 0.4, "step": 43650 }, { "epoch": 2.63, "learning_rate": 2.9579919424199904e-05, "loss": 0.7541, "step": 43652 }, { "epoch": 2.63, "learning_rate": 2.9578410514085673e-05, "loss": 0.7337, "step": 43654 }, { "epoch": 2.63, "learning_rate": 2.9576901603971453e-05, "loss": 0.5673, "step": 43656 }, { "epoch": 2.64, "learning_rate": 2.957539269385723e-05, "loss": 0.5634, "step": 43658 }, { "epoch": 2.64, "learning_rate": 2.9573883783743006e-05, "loss": 0.4416, "step": 43660 }, { "epoch": 2.64, "learning_rate": 2.9572374873628776e-05, "loss": 0.5353, "step": 43662 }, { "epoch": 2.64, "learning_rate": 2.9570865963514556e-05, "loss": 0.5438, "step": 43664 }, { "epoch": 2.64, "learning_rate": 2.9569357053400332e-05, "loss": 0.4561, "step": 43666 }, { "epoch": 2.64, "learning_rate": 2.9567848143286102e-05, "loss": 0.4257, "step": 43668 }, { "epoch": 2.64, "learning_rate": 2.956633923317188e-05, "loss": 0.4512, "step": 43670 }, { "epoch": 2.64, "learning_rate": 2.956483032305766e-05, "loss": 0.6842, "step": 43672 }, { "epoch": 2.64, "learning_rate": 2.9563321412943435e-05, "loss": 0.5215, "step": 43674 }, { "epoch": 2.64, "learning_rate": 2.9561812502829205e-05, "loss": 0.5453, "step": 43676 }, { "epoch": 2.64, "learning_rate": 2.956030359271498e-05, "loss": 0.5928, "step": 43678 }, { "epoch": 2.64, "learning_rate": 2.955879468260076e-05, "loss": 0.6391, "step": 43680 }, { "epoch": 2.64, "learning_rate": 2.9557285772486538e-05, "loss": 0.4259, "step": 43682 }, { "epoch": 2.64, "learning_rate": 2.9555776862372308e-05, "loss": 0.5812, "step": 43684 }, { "epoch": 2.64, "learning_rate": 2.9554267952258084e-05, "loss": 0.6222, "step": 43686 }, { "epoch": 2.64, "learning_rate": 2.9552759042143864e-05, "loss": 0.6154, "step": 43688 }, { "epoch": 2.64, "learning_rate": 2.9551250132029634e-05, "loss": 0.4896, "step": 43690 }, { "epoch": 2.64, "learning_rate": 2.954974122191541e-05, "loss": 0.6646, "step": 43692 }, { "epoch": 2.64, "learning_rate": 2.9548232311801187e-05, "loss": 0.6771, "step": 43694 }, { "epoch": 2.64, "learning_rate": 2.9546723401686967e-05, "loss": 0.724, "step": 43696 }, { "epoch": 2.64, "learning_rate": 2.9545214491572736e-05, "loss": 0.4994, "step": 43698 }, { "epoch": 2.64, "learning_rate": 2.9543705581458513e-05, "loss": 0.2438, "step": 43700 }, { "epoch": 2.64, "learning_rate": 2.954219667134429e-05, "loss": 0.4212, "step": 43702 }, { "epoch": 2.64, "learning_rate": 2.9540687761230063e-05, "loss": 0.3262, "step": 43704 }, { "epoch": 2.64, "learning_rate": 2.953917885111584e-05, "loss": 0.5236, "step": 43706 }, { "epoch": 2.64, "learning_rate": 2.9537669941001616e-05, "loss": 0.5472, "step": 43708 }, { "epoch": 2.64, "learning_rate": 2.9536161030887392e-05, "loss": 0.5929, "step": 43710 }, { "epoch": 2.64, "learning_rate": 2.9534652120773165e-05, "loss": 0.7159, "step": 43712 }, { "epoch": 2.64, "learning_rate": 2.953314321065894e-05, "loss": 0.6772, "step": 43714 }, { "epoch": 2.64, "learning_rate": 2.9531634300544718e-05, "loss": 0.4171, "step": 43716 }, { "epoch": 2.64, "learning_rate": 2.953012539043049e-05, "loss": 0.856, "step": 43718 }, { "epoch": 2.64, "learning_rate": 2.9528616480316268e-05, "loss": 0.4777, "step": 43720 }, { "epoch": 2.64, "learning_rate": 2.9527107570202044e-05, "loss": 0.7147, "step": 43722 }, { "epoch": 2.64, "learning_rate": 2.952559866008782e-05, "loss": 0.5957, "step": 43724 }, { "epoch": 2.64, "learning_rate": 2.9524089749973594e-05, "loss": 0.5579, "step": 43726 }, { "epoch": 2.64, "learning_rate": 2.952258083985937e-05, "loss": 0.5691, "step": 43728 }, { "epoch": 2.64, "learning_rate": 2.9521071929745147e-05, "loss": 0.5718, "step": 43730 }, { "epoch": 2.64, "learning_rate": 2.9519563019630924e-05, "loss": 0.6435, "step": 43732 }, { "epoch": 2.64, "learning_rate": 2.9518054109516697e-05, "loss": 0.6583, "step": 43734 }, { "epoch": 2.64, "learning_rate": 2.9516545199402473e-05, "loss": 0.6951, "step": 43736 }, { "epoch": 2.64, "learning_rate": 2.951503628928825e-05, "loss": 0.4956, "step": 43738 }, { "epoch": 2.64, "learning_rate": 2.9513527379174023e-05, "loss": 0.6109, "step": 43740 }, { "epoch": 2.64, "learning_rate": 2.95120184690598e-05, "loss": 0.5588, "step": 43742 }, { "epoch": 2.64, "learning_rate": 2.9510509558945576e-05, "loss": 0.4913, "step": 43744 }, { "epoch": 2.64, "learning_rate": 2.9509000648831352e-05, "loss": 0.6233, "step": 43746 }, { "epoch": 2.64, "learning_rate": 2.9507491738717125e-05, "loss": 0.7236, "step": 43748 }, { "epoch": 2.64, "learning_rate": 2.9505982828602902e-05, "loss": 0.5918, "step": 43750 }, { "epoch": 2.64, "learning_rate": 2.950447391848868e-05, "loss": 0.7389, "step": 43752 }, { "epoch": 2.64, "learning_rate": 2.950296500837445e-05, "loss": 0.6404, "step": 43754 }, { "epoch": 2.64, "learning_rate": 2.9501456098260228e-05, "loss": 0.4344, "step": 43756 }, { "epoch": 2.64, "learning_rate": 2.9499947188146005e-05, "loss": 0.417, "step": 43758 }, { "epoch": 2.64, "learning_rate": 2.949843827803178e-05, "loss": 0.5468, "step": 43760 }, { "epoch": 2.64, "learning_rate": 2.9496929367917554e-05, "loss": 0.531, "step": 43762 }, { "epoch": 2.64, "learning_rate": 2.949542045780333e-05, "loss": 0.8128, "step": 43764 }, { "epoch": 2.64, "learning_rate": 2.9493911547689107e-05, "loss": 0.6838, "step": 43766 }, { "epoch": 2.64, "learning_rate": 2.9492402637574877e-05, "loss": 0.5537, "step": 43768 }, { "epoch": 2.64, "learning_rate": 2.9490893727460657e-05, "loss": 0.5589, "step": 43770 }, { "epoch": 2.64, "learning_rate": 2.9489384817346433e-05, "loss": 0.6548, "step": 43772 }, { "epoch": 2.64, "learning_rate": 2.948787590723221e-05, "loss": 0.6529, "step": 43774 }, { "epoch": 2.64, "learning_rate": 2.948636699711798e-05, "loss": 0.4645, "step": 43776 }, { "epoch": 2.64, "learning_rate": 2.948485808700376e-05, "loss": 0.5284, "step": 43778 }, { "epoch": 2.64, "learning_rate": 2.9483349176889536e-05, "loss": 0.4282, "step": 43780 }, { "epoch": 2.64, "learning_rate": 2.9481840266775306e-05, "loss": 0.5802, "step": 43782 }, { "epoch": 2.64, "learning_rate": 2.9480331356661082e-05, "loss": 0.5809, "step": 43784 }, { "epoch": 2.64, "learning_rate": 2.9478822446546862e-05, "loss": 0.4631, "step": 43786 }, { "epoch": 2.64, "learning_rate": 2.947731353643264e-05, "loss": 0.5994, "step": 43788 }, { "epoch": 2.64, "learning_rate": 2.947580462631841e-05, "loss": 0.5046, "step": 43790 }, { "epoch": 2.64, "learning_rate": 2.9474295716204185e-05, "loss": 0.5814, "step": 43792 }, { "epoch": 2.64, "learning_rate": 2.9472786806089965e-05, "loss": 0.5427, "step": 43794 }, { "epoch": 2.64, "learning_rate": 2.947127789597574e-05, "loss": 0.77, "step": 43796 }, { "epoch": 2.64, "learning_rate": 2.946976898586151e-05, "loss": 0.4728, "step": 43798 }, { "epoch": 2.64, "learning_rate": 2.9468260075747288e-05, "loss": 0.598, "step": 43800 }, { "epoch": 2.64, "learning_rate": 2.9466751165633068e-05, "loss": 0.4963, "step": 43802 }, { "epoch": 2.64, "learning_rate": 2.9465242255518837e-05, "loss": 0.7435, "step": 43804 }, { "epoch": 2.64, "learning_rate": 2.9463733345404614e-05, "loss": 0.5051, "step": 43806 }, { "epoch": 2.64, "learning_rate": 2.946222443529039e-05, "loss": 0.6804, "step": 43808 }, { "epoch": 2.64, "learning_rate": 2.946071552517617e-05, "loss": 0.4836, "step": 43810 }, { "epoch": 2.64, "learning_rate": 2.945920661506194e-05, "loss": 0.4779, "step": 43812 }, { "epoch": 2.64, "learning_rate": 2.9457697704947717e-05, "loss": 0.5882, "step": 43814 }, { "epoch": 2.64, "learning_rate": 2.9456188794833493e-05, "loss": 0.5435, "step": 43816 }, { "epoch": 2.64, "learning_rate": 2.9454679884719266e-05, "loss": 0.6391, "step": 43818 }, { "epoch": 2.64, "learning_rate": 2.9453170974605043e-05, "loss": 0.4615, "step": 43820 }, { "epoch": 2.64, "learning_rate": 2.945166206449082e-05, "loss": 0.6478, "step": 43822 }, { "epoch": 2.65, "learning_rate": 2.9450153154376596e-05, "loss": 0.4934, "step": 43824 }, { "epoch": 2.65, "learning_rate": 2.944864424426237e-05, "loss": 0.3479, "step": 43826 }, { "epoch": 2.65, "learning_rate": 2.9447135334148145e-05, "loss": 0.4355, "step": 43828 }, { "epoch": 2.65, "learning_rate": 2.9445626424033922e-05, "loss": 0.7267, "step": 43830 }, { "epoch": 2.65, "learning_rate": 2.9444117513919695e-05, "loss": 0.3706, "step": 43832 }, { "epoch": 2.65, "learning_rate": 2.944260860380547e-05, "loss": 0.4359, "step": 43834 }, { "epoch": 2.65, "learning_rate": 2.9441099693691248e-05, "loss": 0.5266, "step": 43836 }, { "epoch": 2.65, "learning_rate": 2.9439590783577025e-05, "loss": 0.5223, "step": 43838 }, { "epoch": 2.65, "learning_rate": 2.9438081873462798e-05, "loss": 0.5043, "step": 43840 }, { "epoch": 2.65, "learning_rate": 2.9436572963348574e-05, "loss": 0.4551, "step": 43842 }, { "epoch": 2.65, "learning_rate": 2.943506405323435e-05, "loss": 0.9059, "step": 43844 }, { "epoch": 2.65, "learning_rate": 2.9433555143120127e-05, "loss": 0.7416, "step": 43846 }, { "epoch": 2.65, "learning_rate": 2.94320462330059e-05, "loss": 0.5623, "step": 43848 }, { "epoch": 2.65, "learning_rate": 2.9430537322891677e-05, "loss": 0.5543, "step": 43850 }, { "epoch": 2.65, "learning_rate": 2.9429028412777453e-05, "loss": 0.7654, "step": 43852 }, { "epoch": 2.65, "learning_rate": 2.9427519502663226e-05, "loss": 0.6203, "step": 43854 }, { "epoch": 2.65, "learning_rate": 2.9426010592549003e-05, "loss": 0.5546, "step": 43856 }, { "epoch": 2.65, "learning_rate": 2.942450168243478e-05, "loss": 0.4765, "step": 43858 }, { "epoch": 2.65, "learning_rate": 2.9422992772320556e-05, "loss": 0.5145, "step": 43860 }, { "epoch": 2.65, "learning_rate": 2.942148386220633e-05, "loss": 0.6835, "step": 43862 }, { "epoch": 2.65, "learning_rate": 2.9419974952092106e-05, "loss": 0.6238, "step": 43864 }, { "epoch": 2.65, "learning_rate": 2.9418466041977882e-05, "loss": 0.5402, "step": 43866 }, { "epoch": 2.65, "learning_rate": 2.9416957131863655e-05, "loss": 0.7881, "step": 43868 }, { "epoch": 2.65, "learning_rate": 2.9415448221749432e-05, "loss": 0.4555, "step": 43870 }, { "epoch": 2.65, "learning_rate": 2.941393931163521e-05, "loss": 0.4393, "step": 43872 }, { "epoch": 2.65, "learning_rate": 2.9412430401520985e-05, "loss": 0.624, "step": 43874 }, { "epoch": 2.65, "learning_rate": 2.9410921491406758e-05, "loss": 0.6011, "step": 43876 }, { "epoch": 2.65, "learning_rate": 2.9409412581292534e-05, "loss": 0.6892, "step": 43878 }, { "epoch": 2.65, "learning_rate": 2.940790367117831e-05, "loss": 0.5451, "step": 43880 }, { "epoch": 2.65, "learning_rate": 2.940639476106408e-05, "loss": 0.668, "step": 43882 }, { "epoch": 2.65, "learning_rate": 2.940488585094986e-05, "loss": 0.7039, "step": 43884 }, { "epoch": 2.65, "learning_rate": 2.9403376940835637e-05, "loss": 0.6657, "step": 43886 }, { "epoch": 2.65, "learning_rate": 2.9401868030721414e-05, "loss": 0.584, "step": 43888 }, { "epoch": 2.65, "learning_rate": 2.9400359120607183e-05, "loss": 0.5277, "step": 43890 }, { "epoch": 2.65, "learning_rate": 2.9398850210492963e-05, "loss": 0.4698, "step": 43892 }, { "epoch": 2.65, "learning_rate": 2.939734130037874e-05, "loss": 0.5927, "step": 43894 }, { "epoch": 2.65, "learning_rate": 2.9395832390264516e-05, "loss": 0.5028, "step": 43896 }, { "epoch": 2.65, "learning_rate": 2.9394323480150286e-05, "loss": 0.5934, "step": 43898 }, { "epoch": 2.65, "learning_rate": 2.9392814570036066e-05, "loss": 0.4514, "step": 43900 }, { "epoch": 2.65, "learning_rate": 2.9391305659921842e-05, "loss": 0.4051, "step": 43902 }, { "epoch": 2.65, "learning_rate": 2.9389796749807612e-05, "loss": 0.4336, "step": 43904 }, { "epoch": 2.65, "learning_rate": 2.938828783969339e-05, "loss": 0.5488, "step": 43906 }, { "epoch": 2.65, "learning_rate": 2.938677892957917e-05, "loss": 0.4675, "step": 43908 }, { "epoch": 2.65, "learning_rate": 2.9385270019464945e-05, "loss": 0.4048, "step": 43910 }, { "epoch": 2.65, "learning_rate": 2.9383761109350715e-05, "loss": 0.4841, "step": 43912 }, { "epoch": 2.65, "learning_rate": 2.938225219923649e-05, "loss": 0.6302, "step": 43914 }, { "epoch": 2.65, "learning_rate": 2.938074328912227e-05, "loss": 0.426, "step": 43916 }, { "epoch": 2.65, "learning_rate": 2.937923437900804e-05, "loss": 0.6215, "step": 43918 }, { "epoch": 2.65, "learning_rate": 2.9377725468893818e-05, "loss": 0.6032, "step": 43920 }, { "epoch": 2.65, "learning_rate": 2.9376216558779594e-05, "loss": 0.7968, "step": 43922 }, { "epoch": 2.65, "learning_rate": 2.9374707648665374e-05, "loss": 0.5496, "step": 43924 }, { "epoch": 2.65, "learning_rate": 2.9373198738551144e-05, "loss": 0.628, "step": 43926 }, { "epoch": 2.65, "learning_rate": 2.937168982843692e-05, "loss": 0.6178, "step": 43928 }, { "epoch": 2.65, "learning_rate": 2.9370180918322697e-05, "loss": 0.5632, "step": 43930 }, { "epoch": 2.65, "learning_rate": 2.936867200820847e-05, "loss": 0.5746, "step": 43932 }, { "epoch": 2.65, "learning_rate": 2.9367163098094246e-05, "loss": 0.5658, "step": 43934 }, { "epoch": 2.65, "learning_rate": 2.9365654187980023e-05, "loss": 0.6063, "step": 43936 }, { "epoch": 2.65, "learning_rate": 2.93641452778658e-05, "loss": 0.56, "step": 43938 }, { "epoch": 2.65, "learning_rate": 2.9362636367751573e-05, "loss": 0.76, "step": 43940 }, { "epoch": 2.65, "learning_rate": 2.936112745763735e-05, "loss": 0.5421, "step": 43942 }, { "epoch": 2.65, "learning_rate": 2.9359618547523126e-05, "loss": 0.484, "step": 43944 }, { "epoch": 2.65, "learning_rate": 2.93581096374089e-05, "loss": 0.6049, "step": 43946 }, { "epoch": 2.65, "learning_rate": 2.9356600727294675e-05, "loss": 0.6618, "step": 43948 }, { "epoch": 2.65, "learning_rate": 2.9355091817180452e-05, "loss": 0.6324, "step": 43950 }, { "epoch": 2.65, "learning_rate": 2.9353582907066228e-05, "loss": 0.4295, "step": 43952 }, { "epoch": 2.65, "learning_rate": 2.9352073996952e-05, "loss": 0.6075, "step": 43954 }, { "epoch": 2.65, "learning_rate": 2.9350565086837778e-05, "loss": 0.5916, "step": 43956 }, { "epoch": 2.65, "learning_rate": 2.9349056176723554e-05, "loss": 0.4031, "step": 43958 }, { "epoch": 2.65, "learning_rate": 2.934754726660933e-05, "loss": 0.5074, "step": 43960 }, { "epoch": 2.65, "learning_rate": 2.9346038356495104e-05, "loss": 0.4823, "step": 43962 }, { "epoch": 2.65, "learning_rate": 2.934452944638088e-05, "loss": 0.665, "step": 43964 }, { "epoch": 2.65, "learning_rate": 2.9343020536266657e-05, "loss": 0.5005, "step": 43966 }, { "epoch": 2.65, "learning_rate": 2.934151162615243e-05, "loss": 0.6278, "step": 43968 }, { "epoch": 2.65, "learning_rate": 2.9340002716038207e-05, "loss": 0.4289, "step": 43970 }, { "epoch": 2.65, "learning_rate": 2.9338493805923983e-05, "loss": 0.8277, "step": 43972 }, { "epoch": 2.65, "learning_rate": 2.933698489580976e-05, "loss": 0.5929, "step": 43974 }, { "epoch": 2.65, "learning_rate": 2.9335475985695533e-05, "loss": 0.4309, "step": 43976 }, { "epoch": 2.65, "learning_rate": 2.933396707558131e-05, "loss": 0.4983, "step": 43978 }, { "epoch": 2.65, "learning_rate": 2.9332458165467086e-05, "loss": 0.4754, "step": 43980 }, { "epoch": 2.65, "learning_rate": 2.933094925535286e-05, "loss": 0.4485, "step": 43982 }, { "epoch": 2.65, "learning_rate": 2.9329440345238636e-05, "loss": 0.5391, "step": 43984 }, { "epoch": 2.65, "learning_rate": 2.9327931435124412e-05, "loss": 0.4812, "step": 43986 }, { "epoch": 2.65, "learning_rate": 2.932642252501019e-05, "loss": 0.6112, "step": 43988 }, { "epoch": 2.66, "learning_rate": 2.932491361489596e-05, "loss": 0.683, "step": 43990 }, { "epoch": 2.66, "learning_rate": 2.9323404704781738e-05, "loss": 0.7151, "step": 43992 }, { "epoch": 2.66, "learning_rate": 2.9321895794667515e-05, "loss": 0.5741, "step": 43994 }, { "epoch": 2.66, "learning_rate": 2.9320386884553284e-05, "loss": 0.5507, "step": 43996 }, { "epoch": 2.66, "learning_rate": 2.9318877974439064e-05, "loss": 0.624, "step": 43998 }, { "epoch": 2.66, "learning_rate": 2.931736906432484e-05, "loss": 0.4945, "step": 44000 }, { "epoch": 2.66, "learning_rate": 2.9315860154210617e-05, "loss": 0.6835, "step": 44002 }, { "epoch": 2.66, "learning_rate": 2.9314351244096387e-05, "loss": 0.6572, "step": 44004 }, { "epoch": 2.66, "learning_rate": 2.9312842333982167e-05, "loss": 0.7431, "step": 44006 }, { "epoch": 2.66, "learning_rate": 2.9311333423867944e-05, "loss": 0.5484, "step": 44008 }, { "epoch": 2.66, "learning_rate": 2.930982451375372e-05, "loss": 0.6133, "step": 44010 }, { "epoch": 2.66, "learning_rate": 2.930831560363949e-05, "loss": 0.4807, "step": 44012 }, { "epoch": 2.66, "learning_rate": 2.930680669352527e-05, "loss": 0.3517, "step": 44014 }, { "epoch": 2.66, "learning_rate": 2.9305297783411046e-05, "loss": 0.5374, "step": 44016 }, { "epoch": 2.66, "learning_rate": 2.9303788873296816e-05, "loss": 0.4969, "step": 44018 }, { "epoch": 2.66, "learning_rate": 2.9302279963182592e-05, "loss": 0.5179, "step": 44020 }, { "epoch": 2.66, "learning_rate": 2.9300771053068372e-05, "loss": 0.5423, "step": 44022 }, { "epoch": 2.66, "learning_rate": 2.929926214295415e-05, "loss": 0.4535, "step": 44024 }, { "epoch": 2.66, "learning_rate": 2.929775323283992e-05, "loss": 0.489, "step": 44026 }, { "epoch": 2.66, "learning_rate": 2.9296244322725695e-05, "loss": 0.6879, "step": 44028 }, { "epoch": 2.66, "learning_rate": 2.9294735412611475e-05, "loss": 0.4171, "step": 44030 }, { "epoch": 2.66, "learning_rate": 2.9293226502497245e-05, "loss": 0.3706, "step": 44032 }, { "epoch": 2.66, "learning_rate": 2.929171759238302e-05, "loss": 0.5512, "step": 44034 }, { "epoch": 2.66, "learning_rate": 2.9290208682268798e-05, "loss": 0.4324, "step": 44036 }, { "epoch": 2.66, "learning_rate": 2.9288699772154578e-05, "loss": 0.6438, "step": 44038 }, { "epoch": 2.66, "learning_rate": 2.9287190862040347e-05, "loss": 0.4394, "step": 44040 }, { "epoch": 2.66, "learning_rate": 2.9285681951926124e-05, "loss": 0.4754, "step": 44042 }, { "epoch": 2.66, "learning_rate": 2.92841730418119e-05, "loss": 0.5531, "step": 44044 }, { "epoch": 2.66, "learning_rate": 2.9282664131697674e-05, "loss": 0.6736, "step": 44046 }, { "epoch": 2.66, "learning_rate": 2.928115522158345e-05, "loss": 0.4085, "step": 44048 }, { "epoch": 2.66, "learning_rate": 2.9279646311469227e-05, "loss": 0.5537, "step": 44050 }, { "epoch": 2.66, "learning_rate": 2.9278137401355003e-05, "loss": 0.5201, "step": 44052 }, { "epoch": 2.66, "learning_rate": 2.9276628491240776e-05, "loss": 0.6478, "step": 44054 }, { "epoch": 2.66, "learning_rate": 2.9275119581126553e-05, "loss": 0.4298, "step": 44056 }, { "epoch": 2.66, "learning_rate": 2.927361067101233e-05, "loss": 0.5021, "step": 44058 }, { "epoch": 2.66, "learning_rate": 2.9272101760898106e-05, "loss": 0.4626, "step": 44060 }, { "epoch": 2.66, "learning_rate": 2.927059285078388e-05, "loss": 0.8062, "step": 44062 }, { "epoch": 2.66, "learning_rate": 2.9269083940669655e-05, "loss": 0.4991, "step": 44064 }, { "epoch": 2.66, "learning_rate": 2.9267575030555432e-05, "loss": 0.4466, "step": 44066 }, { "epoch": 2.66, "learning_rate": 2.9266066120441205e-05, "loss": 0.6651, "step": 44068 }, { "epoch": 2.66, "learning_rate": 2.926455721032698e-05, "loss": 0.4785, "step": 44070 }, { "epoch": 2.66, "learning_rate": 2.9263048300212758e-05, "loss": 0.5468, "step": 44072 }, { "epoch": 2.66, "learning_rate": 2.9261539390098535e-05, "loss": 0.4846, "step": 44074 }, { "epoch": 2.66, "learning_rate": 2.9260030479984308e-05, "loss": 0.7226, "step": 44076 }, { "epoch": 2.66, "learning_rate": 2.9258521569870084e-05, "loss": 0.6939, "step": 44078 }, { "epoch": 2.66, "learning_rate": 2.925701265975586e-05, "loss": 0.6931, "step": 44080 }, { "epoch": 2.66, "learning_rate": 2.9255503749641634e-05, "loss": 0.4667, "step": 44082 }, { "epoch": 2.66, "learning_rate": 2.925399483952741e-05, "loss": 0.6152, "step": 44084 }, { "epoch": 2.66, "learning_rate": 2.9252485929413187e-05, "loss": 0.4025, "step": 44086 }, { "epoch": 2.66, "learning_rate": 2.9250977019298963e-05, "loss": 0.5341, "step": 44088 }, { "epoch": 2.66, "learning_rate": 2.9249468109184737e-05, "loss": 0.5564, "step": 44090 }, { "epoch": 2.66, "learning_rate": 2.9247959199070513e-05, "loss": 0.5393, "step": 44092 }, { "epoch": 2.66, "learning_rate": 2.924645028895629e-05, "loss": 0.6449, "step": 44094 }, { "epoch": 2.66, "learning_rate": 2.9244941378842063e-05, "loss": 0.7058, "step": 44096 }, { "epoch": 2.66, "learning_rate": 2.924343246872784e-05, "loss": 0.4911, "step": 44098 }, { "epoch": 2.66, "learning_rate": 2.9241923558613616e-05, "loss": 0.6432, "step": 44100 }, { "epoch": 2.66, "learning_rate": 2.9240414648499392e-05, "loss": 0.4709, "step": 44102 }, { "epoch": 2.66, "learning_rate": 2.9238905738385165e-05, "loss": 0.3858, "step": 44104 }, { "epoch": 2.66, "learning_rate": 2.9237396828270942e-05, "loss": 0.5315, "step": 44106 }, { "epoch": 2.66, "learning_rate": 2.923588791815672e-05, "loss": 0.6291, "step": 44108 }, { "epoch": 2.66, "learning_rate": 2.9234379008042495e-05, "loss": 0.4846, "step": 44110 }, { "epoch": 2.66, "learning_rate": 2.9232870097928268e-05, "loss": 0.653, "step": 44112 }, { "epoch": 2.66, "learning_rate": 2.9231361187814045e-05, "loss": 0.5275, "step": 44114 }, { "epoch": 2.66, "learning_rate": 2.922985227769982e-05, "loss": 0.5739, "step": 44116 }, { "epoch": 2.66, "learning_rate": 2.922834336758559e-05, "loss": 0.5059, "step": 44118 }, { "epoch": 2.66, "learning_rate": 2.922683445747137e-05, "loss": 0.4112, "step": 44120 }, { "epoch": 2.66, "learning_rate": 2.9225325547357147e-05, "loss": 0.5465, "step": 44122 }, { "epoch": 2.66, "learning_rate": 2.9223816637242924e-05, "loss": 0.5196, "step": 44124 }, { "epoch": 2.66, "learning_rate": 2.9222307727128693e-05, "loss": 0.5413, "step": 44126 }, { "epoch": 2.66, "learning_rate": 2.9220798817014473e-05, "loss": 0.9349, "step": 44128 }, { "epoch": 2.66, "learning_rate": 2.921928990690025e-05, "loss": 0.5505, "step": 44130 }, { "epoch": 2.66, "learning_rate": 2.921778099678602e-05, "loss": 0.5429, "step": 44132 }, { "epoch": 2.66, "learning_rate": 2.9216272086671796e-05, "loss": 0.4238, "step": 44134 }, { "epoch": 2.66, "learning_rate": 2.9214763176557576e-05, "loss": 0.417, "step": 44136 }, { "epoch": 2.66, "learning_rate": 2.9213254266443353e-05, "loss": 0.488, "step": 44138 }, { "epoch": 2.66, "learning_rate": 2.9211745356329122e-05, "loss": 0.6349, "step": 44140 }, { "epoch": 2.66, "learning_rate": 2.92102364462149e-05, "loss": 0.5374, "step": 44142 }, { "epoch": 2.66, "learning_rate": 2.920872753610068e-05, "loss": 0.5165, "step": 44144 }, { "epoch": 2.66, "learning_rate": 2.920721862598645e-05, "loss": 0.6389, "step": 44146 }, { "epoch": 2.66, "learning_rate": 2.9205709715872225e-05, "loss": 0.4252, "step": 44148 }, { "epoch": 2.66, "learning_rate": 2.9204200805758e-05, "loss": 0.533, "step": 44150 }, { "epoch": 2.66, "learning_rate": 2.920269189564378e-05, "loss": 0.4415, "step": 44152 }, { "epoch": 2.67, "learning_rate": 2.920118298552955e-05, "loss": 0.6215, "step": 44154 }, { "epoch": 2.67, "learning_rate": 2.9199674075415328e-05, "loss": 0.5918, "step": 44156 }, { "epoch": 2.67, "learning_rate": 2.9198165165301104e-05, "loss": 0.4359, "step": 44158 }, { "epoch": 2.67, "learning_rate": 2.9196656255186877e-05, "loss": 0.4732, "step": 44160 }, { "epoch": 2.67, "learning_rate": 2.9195147345072654e-05, "loss": 0.4741, "step": 44162 }, { "epoch": 2.67, "learning_rate": 2.919363843495843e-05, "loss": 0.6176, "step": 44164 }, { "epoch": 2.67, "learning_rate": 2.9192129524844207e-05, "loss": 0.5864, "step": 44166 }, { "epoch": 2.67, "learning_rate": 2.919062061472998e-05, "loss": 0.4055, "step": 44168 }, { "epoch": 2.67, "learning_rate": 2.9189111704615756e-05, "loss": 0.5391, "step": 44170 }, { "epoch": 2.67, "learning_rate": 2.9187602794501533e-05, "loss": 0.6541, "step": 44172 }, { "epoch": 2.67, "learning_rate": 2.918609388438731e-05, "loss": 0.3941, "step": 44174 }, { "epoch": 2.67, "learning_rate": 2.9184584974273083e-05, "loss": 0.5673, "step": 44176 }, { "epoch": 2.67, "learning_rate": 2.918307606415886e-05, "loss": 0.5612, "step": 44178 }, { "epoch": 2.67, "learning_rate": 2.9181567154044636e-05, "loss": 0.5814, "step": 44180 }, { "epoch": 2.67, "learning_rate": 2.918005824393041e-05, "loss": 0.4409, "step": 44182 }, { "epoch": 2.67, "learning_rate": 2.9178549333816185e-05, "loss": 0.4616, "step": 44184 }, { "epoch": 2.67, "learning_rate": 2.9177040423701962e-05, "loss": 0.5597, "step": 44186 }, { "epoch": 2.67, "learning_rate": 2.9175531513587738e-05, "loss": 0.492, "step": 44188 }, { "epoch": 2.67, "learning_rate": 2.917402260347351e-05, "loss": 0.6598, "step": 44190 }, { "epoch": 2.67, "learning_rate": 2.9172513693359288e-05, "loss": 0.5535, "step": 44192 }, { "epoch": 2.67, "learning_rate": 2.9171004783245064e-05, "loss": 0.4928, "step": 44194 }, { "epoch": 2.67, "learning_rate": 2.9169495873130838e-05, "loss": 0.3532, "step": 44196 }, { "epoch": 2.67, "learning_rate": 2.9167986963016614e-05, "loss": 0.6341, "step": 44198 }, { "epoch": 2.67, "learning_rate": 2.916647805290239e-05, "loss": 0.4609, "step": 44200 }, { "epoch": 2.67, "learning_rate": 2.9164969142788167e-05, "loss": 0.5231, "step": 44202 }, { "epoch": 2.67, "learning_rate": 2.916346023267394e-05, "loss": 0.5779, "step": 44204 }, { "epoch": 2.67, "learning_rate": 2.9161951322559717e-05, "loss": 0.4747, "step": 44206 }, { "epoch": 2.67, "learning_rate": 2.9160442412445493e-05, "loss": 0.5378, "step": 44208 }, { "epoch": 2.67, "learning_rate": 2.9158933502331266e-05, "loss": 0.5665, "step": 44210 }, { "epoch": 2.67, "learning_rate": 2.9157424592217043e-05, "loss": 0.5283, "step": 44212 }, { "epoch": 2.67, "learning_rate": 2.915591568210282e-05, "loss": 0.7746, "step": 44214 }, { "epoch": 2.67, "learning_rate": 2.9154406771988596e-05, "loss": 0.4577, "step": 44216 }, { "epoch": 2.67, "learning_rate": 2.915289786187437e-05, "loss": 0.7502, "step": 44218 }, { "epoch": 2.67, "learning_rate": 2.9151388951760146e-05, "loss": 0.5892, "step": 44220 }, { "epoch": 2.67, "learning_rate": 2.9149880041645922e-05, "loss": 0.6484, "step": 44222 }, { "epoch": 2.67, "learning_rate": 2.91483711315317e-05, "loss": 0.7394, "step": 44224 }, { "epoch": 2.67, "learning_rate": 2.914686222141747e-05, "loss": 0.5617, "step": 44226 }, { "epoch": 2.67, "learning_rate": 2.9145353311303248e-05, "loss": 0.5553, "step": 44228 }, { "epoch": 2.67, "learning_rate": 2.9143844401189025e-05, "loss": 0.4983, "step": 44230 }, { "epoch": 2.67, "learning_rate": 2.9142335491074794e-05, "loss": 0.5107, "step": 44232 }, { "epoch": 2.67, "learning_rate": 2.9140826580960574e-05, "loss": 0.487, "step": 44234 }, { "epoch": 2.67, "learning_rate": 2.913931767084635e-05, "loss": 0.754, "step": 44236 }, { "epoch": 2.67, "learning_rate": 2.9137808760732127e-05, "loss": 0.4969, "step": 44238 }, { "epoch": 2.67, "learning_rate": 2.9136299850617897e-05, "loss": 0.4559, "step": 44240 }, { "epoch": 2.67, "learning_rate": 2.9134790940503677e-05, "loss": 0.5032, "step": 44242 }, { "epoch": 2.67, "learning_rate": 2.9133282030389454e-05, "loss": 0.5711, "step": 44244 }, { "epoch": 2.67, "learning_rate": 2.9131773120275223e-05, "loss": 0.5455, "step": 44246 }, { "epoch": 2.67, "learning_rate": 2.9130264210161e-05, "loss": 0.3392, "step": 44248 }, { "epoch": 2.67, "learning_rate": 2.912875530004678e-05, "loss": 0.4546, "step": 44250 }, { "epoch": 2.67, "learning_rate": 2.9127246389932556e-05, "loss": 0.4429, "step": 44252 }, { "epoch": 2.67, "learning_rate": 2.9125737479818326e-05, "loss": 0.4306, "step": 44254 }, { "epoch": 2.67, "learning_rate": 2.9124228569704102e-05, "loss": 0.6272, "step": 44256 }, { "epoch": 2.67, "learning_rate": 2.9122719659589882e-05, "loss": 0.6238, "step": 44258 }, { "epoch": 2.67, "learning_rate": 2.9121210749475652e-05, "loss": 0.4822, "step": 44260 }, { "epoch": 2.67, "learning_rate": 2.911970183936143e-05, "loss": 0.6073, "step": 44262 }, { "epoch": 2.67, "learning_rate": 2.9118192929247205e-05, "loss": 0.553, "step": 44264 }, { "epoch": 2.67, "learning_rate": 2.9116684019132985e-05, "loss": 0.4207, "step": 44266 }, { "epoch": 2.67, "learning_rate": 2.9115175109018755e-05, "loss": 0.5689, "step": 44268 }, { "epoch": 2.67, "learning_rate": 2.911366619890453e-05, "loss": 0.6032, "step": 44270 }, { "epoch": 2.67, "learning_rate": 2.9112157288790308e-05, "loss": 0.5161, "step": 44272 }, { "epoch": 2.67, "learning_rate": 2.9110648378676088e-05, "loss": 0.5105, "step": 44274 }, { "epoch": 2.67, "learning_rate": 2.9109139468561857e-05, "loss": 0.7154, "step": 44276 }, { "epoch": 2.67, "learning_rate": 2.9107630558447634e-05, "loss": 0.5932, "step": 44278 }, { "epoch": 2.67, "learning_rate": 2.910612164833341e-05, "loss": 0.5212, "step": 44280 }, { "epoch": 2.67, "learning_rate": 2.9104612738219184e-05, "loss": 0.4567, "step": 44282 }, { "epoch": 2.67, "learning_rate": 2.910310382810496e-05, "loss": 0.4177, "step": 44284 }, { "epoch": 2.67, "learning_rate": 2.9101594917990737e-05, "loss": 0.6087, "step": 44286 }, { "epoch": 2.67, "learning_rate": 2.9100086007876513e-05, "loss": 0.6583, "step": 44288 }, { "epoch": 2.67, "learning_rate": 2.9098577097762286e-05, "loss": 0.5803, "step": 44290 }, { "epoch": 2.67, "learning_rate": 2.9097068187648063e-05, "loss": 0.6403, "step": 44292 }, { "epoch": 2.67, "learning_rate": 2.909555927753384e-05, "loss": 0.4181, "step": 44294 }, { "epoch": 2.67, "learning_rate": 2.9094050367419612e-05, "loss": 0.6217, "step": 44296 }, { "epoch": 2.67, "learning_rate": 2.909254145730539e-05, "loss": 0.4778, "step": 44298 }, { "epoch": 2.67, "learning_rate": 2.9091032547191165e-05, "loss": 0.4765, "step": 44300 }, { "epoch": 2.67, "learning_rate": 2.9089523637076942e-05, "loss": 0.4535, "step": 44302 }, { "epoch": 2.67, "learning_rate": 2.9088014726962715e-05, "loss": 0.6332, "step": 44304 }, { "epoch": 2.67, "learning_rate": 2.908650581684849e-05, "loss": 0.4154, "step": 44306 }, { "epoch": 2.67, "learning_rate": 2.9084996906734268e-05, "loss": 0.5854, "step": 44308 }, { "epoch": 2.67, "learning_rate": 2.908348799662004e-05, "loss": 0.5239, "step": 44310 }, { "epoch": 2.67, "learning_rate": 2.9081979086505818e-05, "loss": 0.4656, "step": 44312 }, { "epoch": 2.67, "learning_rate": 2.9080470176391594e-05, "loss": 0.4555, "step": 44314 }, { "epoch": 2.67, "learning_rate": 2.907896126627737e-05, "loss": 0.9098, "step": 44316 }, { "epoch": 2.67, "learning_rate": 2.9077452356163144e-05, "loss": 0.5607, "step": 44318 }, { "epoch": 2.68, "learning_rate": 2.907594344604892e-05, "loss": 0.4906, "step": 44320 }, { "epoch": 2.68, "learning_rate": 2.9074434535934697e-05, "loss": 0.6322, "step": 44322 }, { "epoch": 2.68, "learning_rate": 2.907292562582047e-05, "loss": 0.6287, "step": 44324 }, { "epoch": 2.68, "learning_rate": 2.9071416715706247e-05, "loss": 0.4744, "step": 44326 }, { "epoch": 2.68, "learning_rate": 2.9069907805592023e-05, "loss": 0.3649, "step": 44328 }, { "epoch": 2.68, "learning_rate": 2.90683988954778e-05, "loss": 0.6448, "step": 44330 }, { "epoch": 2.68, "learning_rate": 2.9066889985363573e-05, "loss": 0.4172, "step": 44332 }, { "epoch": 2.68, "learning_rate": 2.906538107524935e-05, "loss": 0.5469, "step": 44334 }, { "epoch": 2.68, "learning_rate": 2.9063872165135126e-05, "loss": 0.4947, "step": 44336 }, { "epoch": 2.68, "learning_rate": 2.9062363255020902e-05, "loss": 0.724, "step": 44338 }, { "epoch": 2.68, "learning_rate": 2.9060854344906675e-05, "loss": 0.4967, "step": 44340 }, { "epoch": 2.68, "learning_rate": 2.9059345434792452e-05, "loss": 0.4767, "step": 44342 }, { "epoch": 2.68, "learning_rate": 2.905783652467823e-05, "loss": 0.491, "step": 44344 }, { "epoch": 2.68, "learning_rate": 2.9056327614563998e-05, "loss": 0.4511, "step": 44346 }, { "epoch": 2.68, "learning_rate": 2.9054818704449778e-05, "loss": 0.5363, "step": 44348 }, { "epoch": 2.68, "learning_rate": 2.9053309794335555e-05, "loss": 0.5701, "step": 44350 }, { "epoch": 2.68, "learning_rate": 2.905180088422133e-05, "loss": 0.4111, "step": 44352 }, { "epoch": 2.68, "learning_rate": 2.90502919741071e-05, "loss": 0.6581, "step": 44354 }, { "epoch": 2.68, "learning_rate": 2.904878306399288e-05, "loss": 0.7065, "step": 44356 }, { "epoch": 2.68, "learning_rate": 2.9047274153878657e-05, "loss": 0.4468, "step": 44358 }, { "epoch": 2.68, "learning_rate": 2.9045765243764427e-05, "loss": 0.7282, "step": 44360 }, { "epoch": 2.68, "learning_rate": 2.9044256333650203e-05, "loss": 0.6599, "step": 44362 }, { "epoch": 2.68, "learning_rate": 2.9042747423535983e-05, "loss": 0.339, "step": 44364 }, { "epoch": 2.68, "learning_rate": 2.904123851342176e-05, "loss": 0.6482, "step": 44366 }, { "epoch": 2.68, "learning_rate": 2.903972960330753e-05, "loss": 0.7276, "step": 44368 }, { "epoch": 2.68, "learning_rate": 2.9038220693193306e-05, "loss": 0.5125, "step": 44370 }, { "epoch": 2.68, "learning_rate": 2.9036711783079086e-05, "loss": 1.0418, "step": 44372 }, { "epoch": 2.68, "learning_rate": 2.9035202872964856e-05, "loss": 0.7883, "step": 44374 }, { "epoch": 2.68, "learning_rate": 2.9033693962850632e-05, "loss": 0.7546, "step": 44376 }, { "epoch": 2.68, "learning_rate": 2.903218505273641e-05, "loss": 0.6083, "step": 44378 }, { "epoch": 2.68, "learning_rate": 2.903067614262219e-05, "loss": 0.8076, "step": 44380 }, { "epoch": 2.68, "learning_rate": 2.902916723250796e-05, "loss": 0.4047, "step": 44382 }, { "epoch": 2.68, "learning_rate": 2.9027658322393735e-05, "loss": 0.5054, "step": 44384 }, { "epoch": 2.68, "learning_rate": 2.902614941227951e-05, "loss": 0.4771, "step": 44386 }, { "epoch": 2.68, "learning_rate": 2.902464050216529e-05, "loss": 0.4227, "step": 44388 }, { "epoch": 2.68, "learning_rate": 2.902313159205106e-05, "loss": 0.3938, "step": 44390 }, { "epoch": 2.68, "learning_rate": 2.9021622681936838e-05, "loss": 0.4585, "step": 44392 }, { "epoch": 2.68, "learning_rate": 2.9020113771822614e-05, "loss": 0.509, "step": 44394 }, { "epoch": 2.68, "learning_rate": 2.9018604861708387e-05, "loss": 0.5569, "step": 44396 }, { "epoch": 2.68, "learning_rate": 2.9017095951594164e-05, "loss": 0.4253, "step": 44398 }, { "epoch": 2.68, "learning_rate": 2.901558704147994e-05, "loss": 0.509, "step": 44400 }, { "epoch": 2.68, "learning_rate": 2.9014078131365717e-05, "loss": 0.4069, "step": 44402 }, { "epoch": 2.68, "learning_rate": 2.901256922125149e-05, "loss": 0.3371, "step": 44404 }, { "epoch": 2.68, "learning_rate": 2.9011060311137266e-05, "loss": 0.4541, "step": 44406 }, { "epoch": 2.68, "learning_rate": 2.9009551401023043e-05, "loss": 0.3677, "step": 44408 }, { "epoch": 2.68, "learning_rate": 2.9008042490908816e-05, "loss": 0.5634, "step": 44410 }, { "epoch": 2.68, "learning_rate": 2.9006533580794593e-05, "loss": 0.6239, "step": 44412 }, { "epoch": 2.68, "learning_rate": 2.900502467068037e-05, "loss": 0.3748, "step": 44414 }, { "epoch": 2.68, "learning_rate": 2.9003515760566146e-05, "loss": 0.5625, "step": 44416 }, { "epoch": 2.68, "learning_rate": 2.900200685045192e-05, "loss": 0.4747, "step": 44418 }, { "epoch": 2.68, "learning_rate": 2.9000497940337695e-05, "loss": 0.2676, "step": 44420 }, { "epoch": 2.68, "learning_rate": 2.8998989030223472e-05, "loss": 0.4863, "step": 44422 }, { "epoch": 2.68, "learning_rate": 2.8997480120109245e-05, "loss": 0.8238, "step": 44424 }, { "epoch": 2.68, "learning_rate": 2.899597120999502e-05, "loss": 0.3695, "step": 44426 }, { "epoch": 2.68, "learning_rate": 2.8994462299880798e-05, "loss": 0.521, "step": 44428 }, { "epoch": 2.68, "learning_rate": 2.8992953389766574e-05, "loss": 0.6453, "step": 44430 }, { "epoch": 2.68, "learning_rate": 2.8991444479652348e-05, "loss": 0.4418, "step": 44432 }, { "epoch": 2.68, "learning_rate": 2.8989935569538124e-05, "loss": 0.6399, "step": 44434 }, { "epoch": 2.68, "learning_rate": 2.89884266594239e-05, "loss": 0.4794, "step": 44436 }, { "epoch": 2.68, "learning_rate": 2.8986917749309677e-05, "loss": 0.4778, "step": 44438 }, { "epoch": 2.68, "learning_rate": 2.898540883919545e-05, "loss": 0.5996, "step": 44440 }, { "epoch": 2.68, "learning_rate": 2.8983899929081227e-05, "loss": 0.445, "step": 44442 }, { "epoch": 2.68, "learning_rate": 2.8982391018967003e-05, "loss": 0.7519, "step": 44444 }, { "epoch": 2.68, "learning_rate": 2.8980882108852776e-05, "loss": 0.5122, "step": 44446 }, { "epoch": 2.68, "learning_rate": 2.8979373198738553e-05, "loss": 0.6083, "step": 44448 }, { "epoch": 2.68, "learning_rate": 2.897786428862433e-05, "loss": 0.592, "step": 44450 }, { "epoch": 2.68, "learning_rate": 2.8976355378510106e-05, "loss": 0.5237, "step": 44452 }, { "epoch": 2.68, "learning_rate": 2.897484646839588e-05, "loss": 0.5562, "step": 44454 }, { "epoch": 2.68, "learning_rate": 2.8973337558281656e-05, "loss": 0.6753, "step": 44456 }, { "epoch": 2.68, "learning_rate": 2.8971828648167432e-05, "loss": 0.6503, "step": 44458 }, { "epoch": 2.68, "learning_rate": 2.8970319738053202e-05, "loss": 0.4078, "step": 44460 }, { "epoch": 2.68, "learning_rate": 2.8968810827938982e-05, "loss": 0.4954, "step": 44462 }, { "epoch": 2.68, "learning_rate": 2.8967301917824758e-05, "loss": 0.8508, "step": 44464 }, { "epoch": 2.68, "learning_rate": 2.8965793007710535e-05, "loss": 0.4595, "step": 44466 }, { "epoch": 2.68, "learning_rate": 2.8964284097596304e-05, "loss": 0.3868, "step": 44468 }, { "epoch": 2.68, "learning_rate": 2.8962775187482084e-05, "loss": 0.3281, "step": 44470 }, { "epoch": 2.68, "learning_rate": 2.896126627736786e-05, "loss": 0.4787, "step": 44472 }, { "epoch": 2.68, "learning_rate": 2.895975736725363e-05, "loss": 0.6475, "step": 44474 }, { "epoch": 2.68, "learning_rate": 2.8958248457139407e-05, "loss": 0.5848, "step": 44476 }, { "epoch": 2.68, "learning_rate": 2.8956739547025187e-05, "loss": 0.5295, "step": 44478 }, { "epoch": 2.68, "learning_rate": 2.8955230636910964e-05, "loss": 0.6375, "step": 44480 }, { "epoch": 2.68, "learning_rate": 2.8953721726796733e-05, "loss": 0.567, "step": 44482 }, { "epoch": 2.68, "learning_rate": 2.895221281668251e-05, "loss": 0.8591, "step": 44484 }, { "epoch": 2.69, "learning_rate": 2.895070390656829e-05, "loss": 0.4582, "step": 44486 }, { "epoch": 2.69, "learning_rate": 2.894919499645406e-05, "loss": 0.53, "step": 44488 }, { "epoch": 2.69, "learning_rate": 2.8947686086339836e-05, "loss": 0.4373, "step": 44490 }, { "epoch": 2.69, "learning_rate": 2.8946177176225612e-05, "loss": 0.4844, "step": 44492 }, { "epoch": 2.69, "learning_rate": 2.8944668266111392e-05, "loss": 0.4854, "step": 44494 }, { "epoch": 2.69, "learning_rate": 2.8943159355997162e-05, "loss": 0.4573, "step": 44496 }, { "epoch": 2.69, "learning_rate": 2.894165044588294e-05, "loss": 0.5251, "step": 44498 }, { "epoch": 2.69, "learning_rate": 2.8940141535768715e-05, "loss": 0.3975, "step": 44500 }, { "epoch": 2.69, "learning_rate": 2.8938632625654495e-05, "loss": 0.6081, "step": 44502 }, { "epoch": 2.69, "learning_rate": 2.8937123715540265e-05, "loss": 0.5042, "step": 44504 }, { "epoch": 2.69, "learning_rate": 2.893561480542604e-05, "loss": 0.5148, "step": 44506 }, { "epoch": 2.69, "learning_rate": 2.8934105895311818e-05, "loss": 0.4676, "step": 44508 }, { "epoch": 2.69, "learning_rate": 2.893259698519759e-05, "loss": 0.4876, "step": 44510 }, { "epoch": 2.69, "learning_rate": 2.8931088075083367e-05, "loss": 0.4572, "step": 44512 }, { "epoch": 2.69, "learning_rate": 2.8929579164969144e-05, "loss": 0.4326, "step": 44514 }, { "epoch": 2.69, "learning_rate": 2.892807025485492e-05, "loss": 0.5363, "step": 44516 }, { "epoch": 2.69, "learning_rate": 2.8926561344740694e-05, "loss": 0.5013, "step": 44518 }, { "epoch": 2.69, "learning_rate": 2.892505243462647e-05, "loss": 0.5683, "step": 44520 }, { "epoch": 2.69, "learning_rate": 2.8923543524512247e-05, "loss": 0.5454, "step": 44522 }, { "epoch": 2.69, "learning_rate": 2.892203461439802e-05, "loss": 0.6524, "step": 44524 }, { "epoch": 2.69, "learning_rate": 2.8920525704283796e-05, "loss": 0.4754, "step": 44526 }, { "epoch": 2.69, "learning_rate": 2.8919016794169573e-05, "loss": 0.5466, "step": 44528 }, { "epoch": 2.69, "learning_rate": 2.891750788405535e-05, "loss": 0.4781, "step": 44530 }, { "epoch": 2.69, "learning_rate": 2.8915998973941122e-05, "loss": 0.3411, "step": 44532 }, { "epoch": 2.69, "learning_rate": 2.89144900638269e-05, "loss": 0.4961, "step": 44534 }, { "epoch": 2.69, "learning_rate": 2.8912981153712675e-05, "loss": 0.4797, "step": 44536 }, { "epoch": 2.69, "learning_rate": 2.891147224359845e-05, "loss": 0.6025, "step": 44538 }, { "epoch": 2.69, "learning_rate": 2.8909963333484225e-05, "loss": 0.5553, "step": 44540 }, { "epoch": 2.69, "learning_rate": 2.890845442337e-05, "loss": 0.4561, "step": 44542 }, { "epoch": 2.69, "learning_rate": 2.8906945513255778e-05, "loss": 0.8171, "step": 44544 }, { "epoch": 2.69, "learning_rate": 2.890543660314155e-05, "loss": 0.5407, "step": 44546 }, { "epoch": 2.69, "learning_rate": 2.8903927693027328e-05, "loss": 0.3836, "step": 44548 }, { "epoch": 2.69, "learning_rate": 2.8902418782913104e-05, "loss": 0.4932, "step": 44550 }, { "epoch": 2.69, "learning_rate": 2.890090987279888e-05, "loss": 0.3731, "step": 44552 }, { "epoch": 2.69, "learning_rate": 2.8899400962684654e-05, "loss": 0.4612, "step": 44554 }, { "epoch": 2.69, "learning_rate": 2.889789205257043e-05, "loss": 0.465, "step": 44556 }, { "epoch": 2.69, "learning_rate": 2.8896383142456207e-05, "loss": 0.3569, "step": 44558 }, { "epoch": 2.69, "learning_rate": 2.889487423234198e-05, "loss": 0.6014, "step": 44560 }, { "epoch": 2.69, "learning_rate": 2.8893365322227757e-05, "loss": 0.6859, "step": 44562 }, { "epoch": 2.69, "learning_rate": 2.8891856412113533e-05, "loss": 0.6771, "step": 44564 }, { "epoch": 2.69, "learning_rate": 2.889034750199931e-05, "loss": 0.6076, "step": 44566 }, { "epoch": 2.69, "learning_rate": 2.8888838591885083e-05, "loss": 0.6725, "step": 44568 }, { "epoch": 2.69, "learning_rate": 2.888732968177086e-05, "loss": 0.3726, "step": 44570 }, { "epoch": 2.69, "learning_rate": 2.8885820771656636e-05, "loss": 0.4835, "step": 44572 }, { "epoch": 2.69, "learning_rate": 2.8884311861542405e-05, "loss": 0.4444, "step": 44574 }, { "epoch": 2.69, "learning_rate": 2.8882802951428185e-05, "loss": 0.4089, "step": 44576 }, { "epoch": 2.69, "learning_rate": 2.8881294041313962e-05, "loss": 0.4763, "step": 44578 }, { "epoch": 2.69, "learning_rate": 2.887978513119974e-05, "loss": 0.5003, "step": 44580 }, { "epoch": 2.69, "learning_rate": 2.8878276221085508e-05, "loss": 0.7141, "step": 44582 }, { "epoch": 2.69, "learning_rate": 2.8876767310971288e-05, "loss": 0.4446, "step": 44584 }, { "epoch": 2.69, "learning_rate": 2.8875258400857065e-05, "loss": 0.5387, "step": 44586 }, { "epoch": 2.69, "learning_rate": 2.8873749490742834e-05, "loss": 0.4877, "step": 44588 }, { "epoch": 2.69, "learning_rate": 2.887224058062861e-05, "loss": 0.4864, "step": 44590 }, { "epoch": 2.69, "learning_rate": 2.887073167051439e-05, "loss": 0.3801, "step": 44592 }, { "epoch": 2.69, "learning_rate": 2.8869222760400167e-05, "loss": 0.4824, "step": 44594 }, { "epoch": 2.69, "learning_rate": 2.8867713850285937e-05, "loss": 0.5269, "step": 44596 }, { "epoch": 2.69, "learning_rate": 2.8866204940171713e-05, "loss": 0.3406, "step": 44598 }, { "epoch": 2.69, "learning_rate": 2.8864696030057493e-05, "loss": 0.6071, "step": 44600 }, { "epoch": 2.69, "learning_rate": 2.886318711994327e-05, "loss": 0.5852, "step": 44602 }, { "epoch": 2.69, "learning_rate": 2.886167820982904e-05, "loss": 0.641, "step": 44604 }, { "epoch": 2.69, "learning_rate": 2.8860169299714816e-05, "loss": 0.4719, "step": 44606 }, { "epoch": 2.69, "learning_rate": 2.8858660389600596e-05, "loss": 0.72, "step": 44608 }, { "epoch": 2.69, "learning_rate": 2.8857151479486366e-05, "loss": 0.61, "step": 44610 }, { "epoch": 2.69, "learning_rate": 2.8855642569372142e-05, "loss": 0.573, "step": 44612 }, { "epoch": 2.69, "learning_rate": 2.885413365925792e-05, "loss": 0.5586, "step": 44614 }, { "epoch": 2.69, "learning_rate": 2.88526247491437e-05, "loss": 0.6179, "step": 44616 }, { "epoch": 2.69, "learning_rate": 2.885111583902947e-05, "loss": 0.7521, "step": 44618 }, { "epoch": 2.69, "learning_rate": 2.8849606928915245e-05, "loss": 0.4159, "step": 44620 }, { "epoch": 2.69, "learning_rate": 2.884809801880102e-05, "loss": 0.5257, "step": 44622 }, { "epoch": 2.69, "learning_rate": 2.8846589108686795e-05, "loss": 0.9571, "step": 44624 }, { "epoch": 2.69, "learning_rate": 2.884508019857257e-05, "loss": 0.4275, "step": 44626 }, { "epoch": 2.69, "learning_rate": 2.8843571288458348e-05, "loss": 0.8437, "step": 44628 }, { "epoch": 2.69, "learning_rate": 2.8842062378344124e-05, "loss": 0.4201, "step": 44630 }, { "epoch": 2.69, "learning_rate": 2.8840553468229897e-05, "loss": 0.5562, "step": 44632 }, { "epoch": 2.69, "learning_rate": 2.8839044558115674e-05, "loss": 0.7082, "step": 44634 }, { "epoch": 2.69, "learning_rate": 2.883753564800145e-05, "loss": 0.5646, "step": 44636 }, { "epoch": 2.69, "learning_rate": 2.8836026737887223e-05, "loss": 0.5594, "step": 44638 }, { "epoch": 2.69, "learning_rate": 2.8834517827773e-05, "loss": 0.5075, "step": 44640 }, { "epoch": 2.69, "learning_rate": 2.8833008917658776e-05, "loss": 0.4821, "step": 44642 }, { "epoch": 2.69, "learning_rate": 2.8831500007544553e-05, "loss": 0.5646, "step": 44644 }, { "epoch": 2.69, "learning_rate": 2.8829991097430326e-05, "loss": 0.3696, "step": 44646 }, { "epoch": 2.69, "learning_rate": 2.8828482187316103e-05, "loss": 0.4651, "step": 44648 }, { "epoch": 2.69, "learning_rate": 2.882697327720188e-05, "loss": 0.5381, "step": 44650 }, { "epoch": 2.7, "learning_rate": 2.8825464367087656e-05, "loss": 0.4575, "step": 44652 }, { "epoch": 2.7, "learning_rate": 2.882395545697343e-05, "loss": 0.4873, "step": 44654 }, { "epoch": 2.7, "learning_rate": 2.8822446546859205e-05, "loss": 0.4394, "step": 44656 }, { "epoch": 2.7, "learning_rate": 2.8820937636744982e-05, "loss": 0.7281, "step": 44658 }, { "epoch": 2.7, "learning_rate": 2.8819428726630755e-05, "loss": 0.333, "step": 44660 }, { "epoch": 2.7, "learning_rate": 2.881791981651653e-05, "loss": 0.6967, "step": 44662 }, { "epoch": 2.7, "learning_rate": 2.8816410906402308e-05, "loss": 0.6284, "step": 44664 }, { "epoch": 2.7, "learning_rate": 2.8814901996288084e-05, "loss": 0.4802, "step": 44666 }, { "epoch": 2.7, "learning_rate": 2.8813393086173858e-05, "loss": 0.3221, "step": 44668 }, { "epoch": 2.7, "learning_rate": 2.8811884176059634e-05, "loss": 0.4615, "step": 44670 }, { "epoch": 2.7, "learning_rate": 2.881037526594541e-05, "loss": 0.8215, "step": 44672 }, { "epoch": 2.7, "learning_rate": 2.8808866355831184e-05, "loss": 0.6385, "step": 44674 }, { "epoch": 2.7, "learning_rate": 2.880735744571696e-05, "loss": 0.5664, "step": 44676 }, { "epoch": 2.7, "learning_rate": 2.8805848535602737e-05, "loss": 0.6472, "step": 44678 }, { "epoch": 2.7, "learning_rate": 2.8804339625488513e-05, "loss": 0.4069, "step": 44680 }, { "epoch": 2.7, "learning_rate": 2.8802830715374286e-05, "loss": 0.508, "step": 44682 }, { "epoch": 2.7, "learning_rate": 2.8801321805260063e-05, "loss": 0.3939, "step": 44684 }, { "epoch": 2.7, "learning_rate": 2.879981289514584e-05, "loss": 0.7207, "step": 44686 }, { "epoch": 2.7, "learning_rate": 2.879830398503161e-05, "loss": 0.8058, "step": 44688 }, { "epoch": 2.7, "learning_rate": 2.879679507491739e-05, "loss": 0.5676, "step": 44690 }, { "epoch": 2.7, "learning_rate": 2.8795286164803166e-05, "loss": 0.5015, "step": 44692 }, { "epoch": 2.7, "learning_rate": 2.8793777254688942e-05, "loss": 0.5544, "step": 44694 }, { "epoch": 2.7, "learning_rate": 2.8792268344574712e-05, "loss": 0.6698, "step": 44696 }, { "epoch": 2.7, "learning_rate": 2.8790759434460492e-05, "loss": 0.6607, "step": 44698 }, { "epoch": 2.7, "learning_rate": 2.8789250524346268e-05, "loss": 0.6385, "step": 44700 }, { "epoch": 2.7, "learning_rate": 2.8787741614232038e-05, "loss": 0.5239, "step": 44702 }, { "epoch": 2.7, "learning_rate": 2.8786232704117814e-05, "loss": 0.4274, "step": 44704 }, { "epoch": 2.7, "learning_rate": 2.8784723794003594e-05, "loss": 0.7072, "step": 44706 }, { "epoch": 2.7, "learning_rate": 2.878321488388937e-05, "loss": 0.5628, "step": 44708 }, { "epoch": 2.7, "learning_rate": 2.878170597377514e-05, "loss": 0.8109, "step": 44710 }, { "epoch": 2.7, "learning_rate": 2.8780197063660917e-05, "loss": 0.5716, "step": 44712 }, { "epoch": 2.7, "learning_rate": 2.8778688153546697e-05, "loss": 0.4733, "step": 44714 }, { "epoch": 2.7, "learning_rate": 2.8777179243432474e-05, "loss": 0.7665, "step": 44716 }, { "epoch": 2.7, "learning_rate": 2.8775670333318243e-05, "loss": 0.4553, "step": 44718 }, { "epoch": 2.7, "learning_rate": 2.877416142320402e-05, "loss": 0.6462, "step": 44720 }, { "epoch": 2.7, "learning_rate": 2.87726525130898e-05, "loss": 0.6713, "step": 44722 }, { "epoch": 2.7, "learning_rate": 2.877114360297557e-05, "loss": 0.5013, "step": 44724 }, { "epoch": 2.7, "learning_rate": 2.8769634692861346e-05, "loss": 0.4992, "step": 44726 }, { "epoch": 2.7, "learning_rate": 2.8768125782747122e-05, "loss": 0.7478, "step": 44728 }, { "epoch": 2.7, "learning_rate": 2.8766616872632902e-05, "loss": 0.5204, "step": 44730 }, { "epoch": 2.7, "learning_rate": 2.8765107962518672e-05, "loss": 0.443, "step": 44732 }, { "epoch": 2.7, "learning_rate": 2.876359905240445e-05, "loss": 0.4962, "step": 44734 }, { "epoch": 2.7, "learning_rate": 2.8762090142290225e-05, "loss": 0.4395, "step": 44736 }, { "epoch": 2.7, "learning_rate": 2.8760581232175998e-05, "loss": 0.4343, "step": 44738 }, { "epoch": 2.7, "learning_rate": 2.8759072322061775e-05, "loss": 0.4696, "step": 44740 }, { "epoch": 2.7, "learning_rate": 2.875756341194755e-05, "loss": 0.4428, "step": 44742 }, { "epoch": 2.7, "learning_rate": 2.8756054501833328e-05, "loss": 0.5035, "step": 44744 }, { "epoch": 2.7, "learning_rate": 2.87545455917191e-05, "loss": 0.506, "step": 44746 }, { "epoch": 2.7, "learning_rate": 2.8753036681604877e-05, "loss": 0.4914, "step": 44748 }, { "epoch": 2.7, "learning_rate": 2.8751527771490654e-05, "loss": 0.6916, "step": 44750 }, { "epoch": 2.7, "learning_rate": 2.8750018861376427e-05, "loss": 0.5191, "step": 44752 }, { "epoch": 2.7, "learning_rate": 2.8748509951262204e-05, "loss": 0.59, "step": 44754 }, { "epoch": 2.7, "learning_rate": 2.874700104114798e-05, "loss": 0.3541, "step": 44756 }, { "epoch": 2.7, "learning_rate": 2.8745492131033757e-05, "loss": 0.5427, "step": 44758 }, { "epoch": 2.7, "learning_rate": 2.874398322091953e-05, "loss": 0.5807, "step": 44760 }, { "epoch": 2.7, "learning_rate": 2.8742474310805306e-05, "loss": 0.5238, "step": 44762 }, { "epoch": 2.7, "learning_rate": 2.8740965400691083e-05, "loss": 0.5985, "step": 44764 }, { "epoch": 2.7, "learning_rate": 2.873945649057686e-05, "loss": 0.4987, "step": 44766 }, { "epoch": 2.7, "learning_rate": 2.8737947580462632e-05, "loss": 0.5843, "step": 44768 }, { "epoch": 2.7, "learning_rate": 2.873643867034841e-05, "loss": 0.5043, "step": 44770 }, { "epoch": 2.7, "learning_rate": 2.8734929760234185e-05, "loss": 0.4874, "step": 44772 }, { "epoch": 2.7, "learning_rate": 2.873342085011996e-05, "loss": 0.5923, "step": 44774 }, { "epoch": 2.7, "learning_rate": 2.8731911940005735e-05, "loss": 0.6463, "step": 44776 }, { "epoch": 2.7, "learning_rate": 2.873040302989151e-05, "loss": 0.5917, "step": 44778 }, { "epoch": 2.7, "learning_rate": 2.8728894119777288e-05, "loss": 0.7607, "step": 44780 }, { "epoch": 2.7, "learning_rate": 2.872738520966306e-05, "loss": 0.4603, "step": 44782 }, { "epoch": 2.7, "learning_rate": 2.8725876299548838e-05, "loss": 0.4463, "step": 44784 }, { "epoch": 2.7, "learning_rate": 2.8724367389434614e-05, "loss": 0.454, "step": 44786 }, { "epoch": 2.7, "learning_rate": 2.8722858479320387e-05, "loss": 0.5874, "step": 44788 }, { "epoch": 2.7, "learning_rate": 2.8721349569206164e-05, "loss": 0.601, "step": 44790 }, { "epoch": 2.7, "learning_rate": 2.871984065909194e-05, "loss": 0.4099, "step": 44792 }, { "epoch": 2.7, "learning_rate": 2.8718331748977717e-05, "loss": 0.4453, "step": 44794 }, { "epoch": 2.7, "learning_rate": 2.871682283886349e-05, "loss": 0.5039, "step": 44796 }, { "epoch": 2.7, "learning_rate": 2.8715313928749267e-05, "loss": 0.5717, "step": 44798 }, { "epoch": 2.7, "learning_rate": 2.8713805018635043e-05, "loss": 0.5245, "step": 44800 }, { "epoch": 2.7, "learning_rate": 2.8712296108520813e-05, "loss": 0.4285, "step": 44802 }, { "epoch": 2.7, "learning_rate": 2.8710787198406593e-05, "loss": 0.7607, "step": 44804 }, { "epoch": 2.7, "learning_rate": 2.870927828829237e-05, "loss": 0.5427, "step": 44806 }, { "epoch": 2.7, "learning_rate": 2.8707769378178146e-05, "loss": 0.5161, "step": 44808 }, { "epoch": 2.7, "learning_rate": 2.8706260468063915e-05, "loss": 0.4208, "step": 44810 }, { "epoch": 2.7, "learning_rate": 2.8704751557949695e-05, "loss": 0.5692, "step": 44812 }, { "epoch": 2.7, "learning_rate": 2.8703242647835472e-05, "loss": 0.4062, "step": 44814 }, { "epoch": 2.7, "learning_rate": 2.870173373772125e-05, "loss": 0.4836, "step": 44816 }, { "epoch": 2.71, "learning_rate": 2.8700224827607018e-05, "loss": 0.6983, "step": 44818 }, { "epoch": 2.71, "learning_rate": 2.8698715917492798e-05, "loss": 0.4318, "step": 44820 }, { "epoch": 2.71, "learning_rate": 2.8697207007378575e-05, "loss": 0.4558, "step": 44822 }, { "epoch": 2.71, "learning_rate": 2.8695698097264344e-05, "loss": 0.7485, "step": 44824 }, { "epoch": 2.71, "learning_rate": 2.869418918715012e-05, "loss": 0.7426, "step": 44826 }, { "epoch": 2.71, "learning_rate": 2.86926802770359e-05, "loss": 0.6711, "step": 44828 }, { "epoch": 2.71, "learning_rate": 2.8691171366921677e-05, "loss": 0.6769, "step": 44830 }, { "epoch": 2.71, "learning_rate": 2.8689662456807447e-05, "loss": 0.5613, "step": 44832 }, { "epoch": 2.71, "learning_rate": 2.8688153546693223e-05, "loss": 0.5998, "step": 44834 }, { "epoch": 2.71, "learning_rate": 2.8686644636579003e-05, "loss": 0.5139, "step": 44836 }, { "epoch": 2.71, "learning_rate": 2.8685135726464773e-05, "loss": 0.4412, "step": 44838 }, { "epoch": 2.71, "learning_rate": 2.868362681635055e-05, "loss": 0.3159, "step": 44840 }, { "epoch": 2.71, "learning_rate": 2.8682117906236326e-05, "loss": 0.5178, "step": 44842 }, { "epoch": 2.71, "learning_rate": 2.8680608996122106e-05, "loss": 0.8642, "step": 44844 }, { "epoch": 2.71, "learning_rate": 2.8679100086007876e-05, "loss": 0.5643, "step": 44846 }, { "epoch": 2.71, "learning_rate": 2.8677591175893652e-05, "loss": 0.5562, "step": 44848 }, { "epoch": 2.71, "learning_rate": 2.867608226577943e-05, "loss": 0.3862, "step": 44850 }, { "epoch": 2.71, "learning_rate": 2.8674573355665202e-05, "loss": 0.4142, "step": 44852 }, { "epoch": 2.71, "learning_rate": 2.867306444555098e-05, "loss": 0.8024, "step": 44854 }, { "epoch": 2.71, "learning_rate": 2.8671555535436755e-05, "loss": 0.5375, "step": 44856 }, { "epoch": 2.71, "learning_rate": 2.867004662532253e-05, "loss": 0.4976, "step": 44858 }, { "epoch": 2.71, "learning_rate": 2.8668537715208305e-05, "loss": 0.3136, "step": 44860 }, { "epoch": 2.71, "learning_rate": 2.866702880509408e-05, "loss": 0.4998, "step": 44862 }, { "epoch": 2.71, "learning_rate": 2.8665519894979858e-05, "loss": 0.3165, "step": 44864 }, { "epoch": 2.71, "learning_rate": 2.866401098486563e-05, "loss": 0.5586, "step": 44866 }, { "epoch": 2.71, "learning_rate": 2.8662502074751407e-05, "loss": 0.4981, "step": 44868 }, { "epoch": 2.71, "learning_rate": 2.8660993164637184e-05, "loss": 0.4611, "step": 44870 }, { "epoch": 2.71, "learning_rate": 2.865948425452296e-05, "loss": 0.513, "step": 44872 }, { "epoch": 2.71, "learning_rate": 2.8657975344408733e-05, "loss": 0.5078, "step": 44874 }, { "epoch": 2.71, "learning_rate": 2.865646643429451e-05, "loss": 0.7313, "step": 44876 }, { "epoch": 2.71, "learning_rate": 2.8654957524180286e-05, "loss": 0.4573, "step": 44878 }, { "epoch": 2.71, "learning_rate": 2.8653448614066063e-05, "loss": 0.592, "step": 44880 }, { "epoch": 2.71, "learning_rate": 2.8651939703951836e-05, "loss": 0.6968, "step": 44882 }, { "epoch": 2.71, "learning_rate": 2.8650430793837613e-05, "loss": 0.4447, "step": 44884 }, { "epoch": 2.71, "learning_rate": 2.864892188372339e-05, "loss": 0.6257, "step": 44886 }, { "epoch": 2.71, "learning_rate": 2.8647412973609162e-05, "loss": 0.6435, "step": 44888 }, { "epoch": 2.71, "learning_rate": 2.864590406349494e-05, "loss": 0.7367, "step": 44890 }, { "epoch": 2.71, "learning_rate": 2.8644395153380715e-05, "loss": 0.5561, "step": 44892 }, { "epoch": 2.71, "learning_rate": 2.8642886243266492e-05, "loss": 0.4129, "step": 44894 }, { "epoch": 2.71, "learning_rate": 2.8641377333152265e-05, "loss": 0.5337, "step": 44896 }, { "epoch": 2.71, "learning_rate": 2.863986842303804e-05, "loss": 0.59, "step": 44898 }, { "epoch": 2.71, "learning_rate": 2.8638359512923818e-05, "loss": 0.5217, "step": 44900 }, { "epoch": 2.71, "learning_rate": 2.863685060280959e-05, "loss": 0.5744, "step": 44902 }, { "epoch": 2.71, "learning_rate": 2.8635341692695368e-05, "loss": 0.6396, "step": 44904 }, { "epoch": 2.71, "learning_rate": 2.8633832782581144e-05, "loss": 0.8259, "step": 44906 }, { "epoch": 2.71, "learning_rate": 2.863232387246692e-05, "loss": 0.5923, "step": 44908 }, { "epoch": 2.71, "learning_rate": 2.8630814962352694e-05, "loss": 0.2935, "step": 44910 }, { "epoch": 2.71, "learning_rate": 2.862930605223847e-05, "loss": 0.7898, "step": 44912 }, { "epoch": 2.71, "learning_rate": 2.8627797142124247e-05, "loss": 0.6763, "step": 44914 }, { "epoch": 2.71, "learning_rate": 2.8626288232010017e-05, "loss": 0.6085, "step": 44916 }, { "epoch": 2.71, "learning_rate": 2.8624779321895796e-05, "loss": 0.507, "step": 44918 }, { "epoch": 2.71, "learning_rate": 2.8623270411781573e-05, "loss": 0.5289, "step": 44920 }, { "epoch": 2.71, "learning_rate": 2.862176150166735e-05, "loss": 0.9425, "step": 44922 }, { "epoch": 2.71, "learning_rate": 2.862025259155312e-05, "loss": 0.627, "step": 44924 }, { "epoch": 2.71, "learning_rate": 2.86187436814389e-05, "loss": 0.5337, "step": 44926 }, { "epoch": 2.71, "learning_rate": 2.8617234771324676e-05, "loss": 0.4847, "step": 44928 }, { "epoch": 2.71, "learning_rate": 2.8615725861210452e-05, "loss": 0.6259, "step": 44930 }, { "epoch": 2.71, "learning_rate": 2.8614216951096222e-05, "loss": 0.4581, "step": 44932 }, { "epoch": 2.71, "learning_rate": 2.8612708040982002e-05, "loss": 0.742, "step": 44934 }, { "epoch": 2.71, "learning_rate": 2.8611199130867778e-05, "loss": 0.7311, "step": 44936 }, { "epoch": 2.71, "learning_rate": 2.8610444675810665e-05, "loss": 0.7455, "step": 44938 }, { "epoch": 2.71, "learning_rate": 2.8608935765696438e-05, "loss": 0.6708, "step": 44940 }, { "epoch": 2.71, "learning_rate": 2.8607426855582214e-05, "loss": 0.6487, "step": 44942 }, { "epoch": 2.71, "learning_rate": 2.860591794546799e-05, "loss": 0.5547, "step": 44944 }, { "epoch": 2.71, "learning_rate": 2.8604409035353764e-05, "loss": 0.508, "step": 44946 }, { "epoch": 2.71, "learning_rate": 2.860290012523954e-05, "loss": 0.4033, "step": 44948 }, { "epoch": 2.71, "learning_rate": 2.8601391215125317e-05, "loss": 0.5873, "step": 44950 }, { "epoch": 2.71, "learning_rate": 2.8599882305011094e-05, "loss": 0.6085, "step": 44952 }, { "epoch": 2.71, "learning_rate": 2.8598373394896867e-05, "loss": 0.6648, "step": 44954 }, { "epoch": 2.71, "learning_rate": 2.8596864484782643e-05, "loss": 0.4674, "step": 44956 }, { "epoch": 2.71, "learning_rate": 2.859535557466842e-05, "loss": 0.3219, "step": 44958 }, { "epoch": 2.71, "learning_rate": 2.8593846664554193e-05, "loss": 0.5921, "step": 44960 }, { "epoch": 2.71, "learning_rate": 2.859233775443997e-05, "loss": 0.4306, "step": 44962 }, { "epoch": 2.71, "learning_rate": 2.8590828844325746e-05, "loss": 0.2697, "step": 44964 }, { "epoch": 2.71, "learning_rate": 2.8589319934211522e-05, "loss": 0.5432, "step": 44966 }, { "epoch": 2.71, "learning_rate": 2.8587811024097296e-05, "loss": 0.5916, "step": 44968 }, { "epoch": 2.71, "learning_rate": 2.8586302113983072e-05, "loss": 0.4278, "step": 44970 }, { "epoch": 2.71, "learning_rate": 2.858479320386885e-05, "loss": 0.4699, "step": 44972 }, { "epoch": 2.71, "learning_rate": 2.858328429375462e-05, "loss": 0.4771, "step": 44974 }, { "epoch": 2.71, "learning_rate": 2.8581775383640398e-05, "loss": 0.6577, "step": 44976 }, { "epoch": 2.71, "learning_rate": 2.8580266473526175e-05, "loss": 0.4895, "step": 44978 }, { "epoch": 2.71, "learning_rate": 2.857875756341195e-05, "loss": 0.3859, "step": 44980 }, { "epoch": 2.71, "learning_rate": 2.857724865329772e-05, "loss": 0.5302, "step": 44982 }, { "epoch": 2.72, "learning_rate": 2.85757397431835e-05, "loss": 0.6941, "step": 44984 }, { "epoch": 2.72, "learning_rate": 2.8574230833069277e-05, "loss": 0.6987, "step": 44986 }, { "epoch": 2.72, "learning_rate": 2.8572721922955054e-05, "loss": 0.6466, "step": 44988 }, { "epoch": 2.72, "learning_rate": 2.8571213012840824e-05, "loss": 0.5514, "step": 44990 }, { "epoch": 2.72, "learning_rate": 2.8569704102726604e-05, "loss": 0.4224, "step": 44992 }, { "epoch": 2.72, "learning_rate": 2.856819519261238e-05, "loss": 0.5117, "step": 44994 }, { "epoch": 2.72, "learning_rate": 2.856668628249815e-05, "loss": 0.5642, "step": 44996 }, { "epoch": 2.72, "learning_rate": 2.8565177372383926e-05, "loss": 0.8122, "step": 44998 }, { "epoch": 2.72, "learning_rate": 2.8563668462269706e-05, "loss": 0.5276, "step": 45000 }, { "epoch": 2.72, "learning_rate": 2.8562159552155483e-05, "loss": 0.5028, "step": 45002 }, { "epoch": 2.72, "learning_rate": 2.8560650642041253e-05, "loss": 0.7975, "step": 45004 }, { "epoch": 2.72, "learning_rate": 2.855914173192703e-05, "loss": 0.4444, "step": 45006 }, { "epoch": 2.72, "learning_rate": 2.855763282181281e-05, "loss": 0.4197, "step": 45008 }, { "epoch": 2.72, "learning_rate": 2.855612391169858e-05, "loss": 0.5672, "step": 45010 }, { "epoch": 2.72, "learning_rate": 2.8554615001584355e-05, "loss": 0.6633, "step": 45012 }, { "epoch": 2.72, "learning_rate": 2.855310609147013e-05, "loss": 0.5059, "step": 45014 }, { "epoch": 2.72, "learning_rate": 2.855159718135591e-05, "loss": 0.6389, "step": 45016 }, { "epoch": 2.72, "learning_rate": 2.855008827124168e-05, "loss": 0.3745, "step": 45018 }, { "epoch": 2.72, "learning_rate": 2.8548579361127458e-05, "loss": 0.5685, "step": 45020 }, { "epoch": 2.72, "learning_rate": 2.8547070451013234e-05, "loss": 0.6954, "step": 45022 }, { "epoch": 2.72, "learning_rate": 2.8545561540899007e-05, "loss": 0.494, "step": 45024 }, { "epoch": 2.72, "learning_rate": 2.8544052630784784e-05, "loss": 0.4027, "step": 45026 }, { "epoch": 2.72, "learning_rate": 2.854254372067056e-05, "loss": 0.3923, "step": 45028 }, { "epoch": 2.72, "learning_rate": 2.8541034810556337e-05, "loss": 0.4983, "step": 45030 }, { "epoch": 2.72, "learning_rate": 2.853952590044211e-05, "loss": 0.6788, "step": 45032 }, { "epoch": 2.72, "learning_rate": 2.8538016990327887e-05, "loss": 0.5255, "step": 45034 }, { "epoch": 2.72, "learning_rate": 2.8536508080213663e-05, "loss": 0.5788, "step": 45036 }, { "epoch": 2.72, "learning_rate": 2.853499917009944e-05, "loss": 0.4756, "step": 45038 }, { "epoch": 2.72, "learning_rate": 2.8533490259985213e-05, "loss": 0.4172, "step": 45040 }, { "epoch": 2.72, "learning_rate": 2.853198134987099e-05, "loss": 0.4479, "step": 45042 }, { "epoch": 2.72, "learning_rate": 2.8530472439756766e-05, "loss": 0.3911, "step": 45044 }, { "epoch": 2.72, "learning_rate": 2.852896352964254e-05, "loss": 0.4979, "step": 45046 }, { "epoch": 2.72, "learning_rate": 2.8527454619528315e-05, "loss": 0.5684, "step": 45048 }, { "epoch": 2.72, "learning_rate": 2.8525945709414092e-05, "loss": 0.4028, "step": 45050 }, { "epoch": 2.72, "learning_rate": 2.852443679929987e-05, "loss": 0.6133, "step": 45052 }, { "epoch": 2.72, "learning_rate": 2.852292788918564e-05, "loss": 0.427, "step": 45054 }, { "epoch": 2.72, "learning_rate": 2.8521418979071418e-05, "loss": 0.8363, "step": 45056 }, { "epoch": 2.72, "learning_rate": 2.8519910068957195e-05, "loss": 0.4933, "step": 45058 }, { "epoch": 2.72, "learning_rate": 2.8518401158842968e-05, "loss": 0.6072, "step": 45060 }, { "epoch": 2.72, "learning_rate": 2.8516892248728744e-05, "loss": 0.595, "step": 45062 }, { "epoch": 2.72, "learning_rate": 2.851538333861452e-05, "loss": 0.6947, "step": 45064 }, { "epoch": 2.72, "learning_rate": 2.8513874428500297e-05, "loss": 0.5916, "step": 45066 }, { "epoch": 2.72, "learning_rate": 2.851236551838607e-05, "loss": 0.4709, "step": 45068 }, { "epoch": 2.72, "learning_rate": 2.8510856608271847e-05, "loss": 0.5022, "step": 45070 }, { "epoch": 2.72, "learning_rate": 2.8509347698157623e-05, "loss": 0.7084, "step": 45072 }, { "epoch": 2.72, "learning_rate": 2.8507838788043397e-05, "loss": 0.3926, "step": 45074 }, { "epoch": 2.72, "learning_rate": 2.8506329877929173e-05, "loss": 0.5461, "step": 45076 }, { "epoch": 2.72, "learning_rate": 2.850482096781495e-05, "loss": 0.5145, "step": 45078 }, { "epoch": 2.72, "learning_rate": 2.8503312057700726e-05, "loss": 0.5798, "step": 45080 }, { "epoch": 2.72, "learning_rate": 2.85018031475865e-05, "loss": 0.5595, "step": 45082 }, { "epoch": 2.72, "learning_rate": 2.8500294237472276e-05, "loss": 0.6049, "step": 45084 }, { "epoch": 2.72, "learning_rate": 2.8498785327358052e-05, "loss": 0.7285, "step": 45086 }, { "epoch": 2.72, "learning_rate": 2.849727641724383e-05, "loss": 0.471, "step": 45088 }, { "epoch": 2.72, "learning_rate": 2.8495767507129602e-05, "loss": 0.566, "step": 45090 }, { "epoch": 2.72, "learning_rate": 2.849425859701538e-05, "loss": 0.619, "step": 45092 }, { "epoch": 2.72, "learning_rate": 2.8492749686901155e-05, "loss": 0.6096, "step": 45094 }, { "epoch": 2.72, "learning_rate": 2.8491240776786925e-05, "loss": 0.4071, "step": 45096 }, { "epoch": 2.72, "learning_rate": 2.8489731866672705e-05, "loss": 0.6195, "step": 45098 }, { "epoch": 2.72, "learning_rate": 2.848822295655848e-05, "loss": 0.6277, "step": 45100 }, { "epoch": 2.72, "learning_rate": 2.8486714046444258e-05, "loss": 0.6806, "step": 45102 }, { "epoch": 2.72, "learning_rate": 2.8485205136330027e-05, "loss": 0.6196, "step": 45104 }, { "epoch": 2.72, "learning_rate": 2.8483696226215807e-05, "loss": 0.4677, "step": 45106 }, { "epoch": 2.72, "learning_rate": 2.8482187316101584e-05, "loss": 0.6219, "step": 45108 }, { "epoch": 2.72, "learning_rate": 2.8480678405987354e-05, "loss": 0.5178, "step": 45110 }, { "epoch": 2.72, "learning_rate": 2.847916949587313e-05, "loss": 0.5345, "step": 45112 }, { "epoch": 2.72, "learning_rate": 2.847766058575891e-05, "loss": 0.5283, "step": 45114 }, { "epoch": 2.72, "learning_rate": 2.8476151675644686e-05, "loss": 0.6575, "step": 45116 }, { "epoch": 2.72, "learning_rate": 2.8474642765530456e-05, "loss": 0.6584, "step": 45118 }, { "epoch": 2.72, "learning_rate": 2.8473133855416233e-05, "loss": 0.7959, "step": 45120 }, { "epoch": 2.72, "learning_rate": 2.8471624945302013e-05, "loss": 0.5204, "step": 45122 }, { "epoch": 2.72, "learning_rate": 2.8470116035187782e-05, "loss": 0.4696, "step": 45124 }, { "epoch": 2.72, "learning_rate": 2.846860712507356e-05, "loss": 0.5076, "step": 45126 }, { "epoch": 2.72, "learning_rate": 2.8467098214959335e-05, "loss": 0.5477, "step": 45128 }, { "epoch": 2.72, "learning_rate": 2.8465589304845115e-05, "loss": 0.576, "step": 45130 }, { "epoch": 2.72, "learning_rate": 2.8464080394730885e-05, "loss": 0.5358, "step": 45132 }, { "epoch": 2.72, "learning_rate": 2.846257148461666e-05, "loss": 0.6566, "step": 45134 }, { "epoch": 2.72, "learning_rate": 2.8461062574502438e-05, "loss": 0.5985, "step": 45136 }, { "epoch": 2.72, "learning_rate": 2.845955366438821e-05, "loss": 0.4022, "step": 45138 }, { "epoch": 2.72, "learning_rate": 2.8458044754273988e-05, "loss": 0.6728, "step": 45140 }, { "epoch": 2.72, "learning_rate": 2.8456535844159764e-05, "loss": 0.5886, "step": 45142 }, { "epoch": 2.72, "learning_rate": 2.845502693404554e-05, "loss": 0.3984, "step": 45144 }, { "epoch": 2.72, "learning_rate": 2.8453518023931314e-05, "loss": 0.5151, "step": 45146 }, { "epoch": 2.73, "learning_rate": 2.845200911381709e-05, "loss": 0.5769, "step": 45148 }, { "epoch": 2.73, "learning_rate": 2.8450500203702867e-05, "loss": 0.5033, "step": 45150 }, { "epoch": 2.73, "learning_rate": 2.8448991293588643e-05, "loss": 0.4967, "step": 45152 }, { "epoch": 2.73, "learning_rate": 2.8447482383474416e-05, "loss": 0.4515, "step": 45154 }, { "epoch": 2.73, "learning_rate": 2.8445973473360193e-05, "loss": 0.4441, "step": 45156 }, { "epoch": 2.73, "learning_rate": 2.844446456324597e-05, "loss": 0.5501, "step": 45158 }, { "epoch": 2.73, "learning_rate": 2.8442955653131743e-05, "loss": 0.7323, "step": 45160 }, { "epoch": 2.73, "learning_rate": 2.844144674301752e-05, "loss": 0.6118, "step": 45162 }, { "epoch": 2.73, "learning_rate": 2.8439937832903296e-05, "loss": 0.5306, "step": 45164 }, { "epoch": 2.73, "learning_rate": 2.8438428922789072e-05, "loss": 0.596, "step": 45166 }, { "epoch": 2.73, "learning_rate": 2.8436920012674845e-05, "loss": 0.5656, "step": 45168 }, { "epoch": 2.73, "learning_rate": 2.8435411102560622e-05, "loss": 0.5523, "step": 45170 }, { "epoch": 2.73, "learning_rate": 2.84339021924464e-05, "loss": 0.4483, "step": 45172 }, { "epoch": 2.73, "learning_rate": 2.843239328233217e-05, "loss": 0.5543, "step": 45174 }, { "epoch": 2.73, "learning_rate": 2.8430884372217948e-05, "loss": 0.563, "step": 45176 }, { "epoch": 2.73, "learning_rate": 2.8429375462103724e-05, "loss": 0.6002, "step": 45178 }, { "epoch": 2.73, "learning_rate": 2.84278665519895e-05, "loss": 0.6243, "step": 45180 }, { "epoch": 2.73, "learning_rate": 2.8426357641875274e-05, "loss": 0.5156, "step": 45182 }, { "epoch": 2.73, "learning_rate": 2.842484873176105e-05, "loss": 0.5021, "step": 45184 }, { "epoch": 2.73, "learning_rate": 2.8423339821646827e-05, "loss": 0.6379, "step": 45186 }, { "epoch": 2.73, "learning_rate": 2.84218309115326e-05, "loss": 0.6554, "step": 45188 }, { "epoch": 2.73, "learning_rate": 2.8420322001418377e-05, "loss": 0.4624, "step": 45190 }, { "epoch": 2.73, "learning_rate": 2.8418813091304153e-05, "loss": 0.8253, "step": 45192 }, { "epoch": 2.73, "learning_rate": 2.841730418118993e-05, "loss": 0.3885, "step": 45194 }, { "epoch": 2.73, "learning_rate": 2.8415795271075703e-05, "loss": 0.5031, "step": 45196 }, { "epoch": 2.73, "learning_rate": 2.841428636096148e-05, "loss": 0.4964, "step": 45198 }, { "epoch": 2.73, "learning_rate": 2.8412777450847256e-05, "loss": 0.3814, "step": 45200 }, { "epoch": 2.73, "learning_rate": 2.8411268540733032e-05, "loss": 0.5615, "step": 45202 }, { "epoch": 2.73, "learning_rate": 2.8409759630618806e-05, "loss": 0.5163, "step": 45204 }, { "epoch": 2.73, "learning_rate": 2.8408250720504582e-05, "loss": 0.3587, "step": 45206 }, { "epoch": 2.73, "learning_rate": 2.840674181039036e-05, "loss": 0.5715, "step": 45208 }, { "epoch": 2.73, "learning_rate": 2.840523290027613e-05, "loss": 0.4545, "step": 45210 }, { "epoch": 2.73, "learning_rate": 2.8403723990161908e-05, "loss": 0.6785, "step": 45212 }, { "epoch": 2.73, "learning_rate": 2.8402215080047685e-05, "loss": 0.4082, "step": 45214 }, { "epoch": 2.73, "learning_rate": 2.840070616993346e-05, "loss": 0.4119, "step": 45216 }, { "epoch": 2.73, "learning_rate": 2.839919725981923e-05, "loss": 0.5451, "step": 45218 }, { "epoch": 2.73, "learning_rate": 2.839768834970501e-05, "loss": 0.5101, "step": 45220 }, { "epoch": 2.73, "learning_rate": 2.8396179439590787e-05, "loss": 0.4897, "step": 45222 }, { "epoch": 2.73, "learning_rate": 2.8394670529476557e-05, "loss": 0.4304, "step": 45224 }, { "epoch": 2.73, "learning_rate": 2.8393161619362334e-05, "loss": 0.5154, "step": 45226 }, { "epoch": 2.73, "learning_rate": 2.8391652709248114e-05, "loss": 0.4946, "step": 45228 }, { "epoch": 2.73, "learning_rate": 2.839014379913389e-05, "loss": 0.7982, "step": 45230 }, { "epoch": 2.73, "learning_rate": 2.838863488901966e-05, "loss": 0.4093, "step": 45232 }, { "epoch": 2.73, "learning_rate": 2.8387125978905436e-05, "loss": 0.6923, "step": 45234 }, { "epoch": 2.73, "learning_rate": 2.8385617068791216e-05, "loss": 0.4191, "step": 45236 }, { "epoch": 2.73, "learning_rate": 2.8384108158676986e-05, "loss": 0.5432, "step": 45238 }, { "epoch": 2.73, "learning_rate": 2.8382599248562763e-05, "loss": 0.5254, "step": 45240 }, { "epoch": 2.73, "learning_rate": 2.838109033844854e-05, "loss": 0.6043, "step": 45242 }, { "epoch": 2.73, "learning_rate": 2.837958142833432e-05, "loss": 0.5666, "step": 45244 }, { "epoch": 2.73, "learning_rate": 2.837807251822009e-05, "loss": 0.4208, "step": 45246 }, { "epoch": 2.73, "learning_rate": 2.8376563608105865e-05, "loss": 0.555, "step": 45248 }, { "epoch": 2.73, "learning_rate": 2.8375054697991642e-05, "loss": 0.4511, "step": 45250 }, { "epoch": 2.73, "learning_rate": 2.837354578787742e-05, "loss": 0.5582, "step": 45252 }, { "epoch": 2.73, "learning_rate": 2.837203687776319e-05, "loss": 0.4548, "step": 45254 }, { "epoch": 2.73, "learning_rate": 2.8370527967648968e-05, "loss": 0.5307, "step": 45256 }, { "epoch": 2.73, "learning_rate": 2.8369019057534744e-05, "loss": 0.5125, "step": 45258 }, { "epoch": 2.73, "learning_rate": 2.8367510147420517e-05, "loss": 0.5134, "step": 45260 }, { "epoch": 2.73, "learning_rate": 2.8366001237306294e-05, "loss": 0.6339, "step": 45262 }, { "epoch": 2.73, "learning_rate": 2.836449232719207e-05, "loss": 0.4515, "step": 45264 }, { "epoch": 2.73, "learning_rate": 2.8362983417077847e-05, "loss": 0.5903, "step": 45266 }, { "epoch": 2.73, "learning_rate": 2.836147450696362e-05, "loss": 0.632, "step": 45268 }, { "epoch": 2.73, "learning_rate": 2.8359965596849397e-05, "loss": 0.532, "step": 45270 }, { "epoch": 2.73, "learning_rate": 2.8358456686735173e-05, "loss": 0.4846, "step": 45272 }, { "epoch": 2.73, "learning_rate": 2.8356947776620946e-05, "loss": 0.5077, "step": 45274 }, { "epoch": 2.73, "learning_rate": 2.8355438866506723e-05, "loss": 0.6511, "step": 45276 }, { "epoch": 2.73, "learning_rate": 2.83539299563925e-05, "loss": 0.6643, "step": 45278 }, { "epoch": 2.73, "learning_rate": 2.8352421046278276e-05, "loss": 0.5955, "step": 45280 }, { "epoch": 2.73, "learning_rate": 2.835091213616405e-05, "loss": 0.5242, "step": 45282 }, { "epoch": 2.73, "learning_rate": 2.8349403226049825e-05, "loss": 0.3225, "step": 45284 }, { "epoch": 2.73, "learning_rate": 2.8347894315935602e-05, "loss": 0.3811, "step": 45286 }, { "epoch": 2.73, "learning_rate": 2.8346385405821375e-05, "loss": 0.9225, "step": 45288 }, { "epoch": 2.73, "learning_rate": 2.834487649570715e-05, "loss": 0.4364, "step": 45290 }, { "epoch": 2.73, "learning_rate": 2.8343367585592928e-05, "loss": 0.4238, "step": 45292 }, { "epoch": 2.73, "learning_rate": 2.8341858675478705e-05, "loss": 0.5216, "step": 45294 }, { "epoch": 2.73, "learning_rate": 2.8340349765364478e-05, "loss": 0.4022, "step": 45296 }, { "epoch": 2.73, "learning_rate": 2.8338840855250254e-05, "loss": 0.5631, "step": 45298 }, { "epoch": 2.73, "learning_rate": 2.833733194513603e-05, "loss": 0.4357, "step": 45300 }, { "epoch": 2.73, "learning_rate": 2.8335823035021804e-05, "loss": 0.4886, "step": 45302 }, { "epoch": 2.73, "learning_rate": 2.833431412490758e-05, "loss": 0.6292, "step": 45304 }, { "epoch": 2.73, "learning_rate": 2.8332805214793357e-05, "loss": 0.599, "step": 45306 }, { "epoch": 2.73, "learning_rate": 2.8331296304679133e-05, "loss": 0.5868, "step": 45308 }, { "epoch": 2.73, "learning_rate": 2.8329787394564907e-05, "loss": 0.7004, "step": 45310 }, { "epoch": 2.73, "learning_rate": 2.8328278484450683e-05, "loss": 0.6019, "step": 45312 }, { "epoch": 2.74, "learning_rate": 2.832676957433646e-05, "loss": 0.4496, "step": 45314 }, { "epoch": 2.74, "learning_rate": 2.8325260664222236e-05, "loss": 0.5229, "step": 45316 }, { "epoch": 2.74, "learning_rate": 2.832375175410801e-05, "loss": 0.4827, "step": 45318 }, { "epoch": 2.74, "learning_rate": 2.8322242843993786e-05, "loss": 0.5079, "step": 45320 }, { "epoch": 2.74, "learning_rate": 2.8320733933879562e-05, "loss": 0.3967, "step": 45322 }, { "epoch": 2.74, "learning_rate": 2.8319225023765332e-05, "loss": 0.6275, "step": 45324 }, { "epoch": 2.74, "learning_rate": 2.8317716113651112e-05, "loss": 0.419, "step": 45326 }, { "epoch": 2.74, "learning_rate": 2.831620720353689e-05, "loss": 0.396, "step": 45328 }, { "epoch": 2.74, "learning_rate": 2.8314698293422665e-05, "loss": 0.5733, "step": 45330 }, { "epoch": 2.74, "learning_rate": 2.8313189383308435e-05, "loss": 0.3486, "step": 45332 }, { "epoch": 2.74, "learning_rate": 2.8311680473194215e-05, "loss": 0.4685, "step": 45334 }, { "epoch": 2.74, "learning_rate": 2.831017156307999e-05, "loss": 0.7103, "step": 45336 }, { "epoch": 2.74, "learning_rate": 2.830866265296576e-05, "loss": 0.5479, "step": 45338 }, { "epoch": 2.74, "learning_rate": 2.8307153742851537e-05, "loss": 0.483, "step": 45340 }, { "epoch": 2.74, "learning_rate": 2.8305644832737317e-05, "loss": 0.7096, "step": 45342 }, { "epoch": 2.74, "learning_rate": 2.8304135922623094e-05, "loss": 0.4762, "step": 45344 }, { "epoch": 2.74, "learning_rate": 2.8302627012508864e-05, "loss": 0.4025, "step": 45346 }, { "epoch": 2.74, "learning_rate": 2.830111810239464e-05, "loss": 0.6072, "step": 45348 }, { "epoch": 2.74, "learning_rate": 2.829960919228042e-05, "loss": 0.4283, "step": 45350 }, { "epoch": 2.74, "learning_rate": 2.829810028216619e-05, "loss": 0.3722, "step": 45352 }, { "epoch": 2.74, "learning_rate": 2.8296591372051966e-05, "loss": 0.4429, "step": 45354 }, { "epoch": 2.74, "learning_rate": 2.8295082461937743e-05, "loss": 0.5466, "step": 45356 }, { "epoch": 2.74, "learning_rate": 2.8293573551823523e-05, "loss": 0.3718, "step": 45358 }, { "epoch": 2.74, "learning_rate": 2.8292064641709292e-05, "loss": 0.6586, "step": 45360 }, { "epoch": 2.74, "learning_rate": 2.829055573159507e-05, "loss": 0.7292, "step": 45362 }, { "epoch": 2.74, "learning_rate": 2.8289046821480845e-05, "loss": 0.5689, "step": 45364 }, { "epoch": 2.74, "learning_rate": 2.8287537911366625e-05, "loss": 0.4335, "step": 45366 }, { "epoch": 2.74, "learning_rate": 2.8286029001252395e-05, "loss": 0.6673, "step": 45368 }, { "epoch": 2.74, "learning_rate": 2.828452009113817e-05, "loss": 0.4538, "step": 45370 }, { "epoch": 2.74, "learning_rate": 2.8283011181023948e-05, "loss": 0.3407, "step": 45372 }, { "epoch": 2.74, "learning_rate": 2.828150227090972e-05, "loss": 0.6523, "step": 45374 }, { "epoch": 2.74, "learning_rate": 2.8279993360795498e-05, "loss": 0.6053, "step": 45376 }, { "epoch": 2.74, "learning_rate": 2.8278484450681274e-05, "loss": 0.5464, "step": 45378 }, { "epoch": 2.74, "learning_rate": 2.827697554056705e-05, "loss": 0.3796, "step": 45380 }, { "epoch": 2.74, "learning_rate": 2.8275466630452824e-05, "loss": 0.5372, "step": 45382 }, { "epoch": 2.74, "learning_rate": 2.82739577203386e-05, "loss": 0.349, "step": 45384 }, { "epoch": 2.74, "learning_rate": 2.8272448810224377e-05, "loss": 0.3989, "step": 45386 }, { "epoch": 2.74, "learning_rate": 2.827093990011015e-05, "loss": 0.4302, "step": 45388 }, { "epoch": 2.74, "learning_rate": 2.8269430989995927e-05, "loss": 0.6019, "step": 45390 }, { "epoch": 2.74, "learning_rate": 2.8267922079881703e-05, "loss": 0.6714, "step": 45392 }, { "epoch": 2.74, "learning_rate": 2.826641316976748e-05, "loss": 0.5687, "step": 45394 }, { "epoch": 2.74, "learning_rate": 2.8264904259653253e-05, "loss": 0.5844, "step": 45396 }, { "epoch": 2.74, "learning_rate": 2.826339534953903e-05, "loss": 0.5368, "step": 45398 }, { "epoch": 2.74, "learning_rate": 2.8261886439424806e-05, "loss": 0.3825, "step": 45400 }, { "epoch": 2.74, "learning_rate": 2.826037752931058e-05, "loss": 0.4031, "step": 45402 }, { "epoch": 2.74, "learning_rate": 2.8258868619196355e-05, "loss": 0.4017, "step": 45404 }, { "epoch": 2.74, "learning_rate": 2.8257359709082132e-05, "loss": 0.5643, "step": 45406 }, { "epoch": 2.74, "learning_rate": 2.825585079896791e-05, "loss": 0.6842, "step": 45408 }, { "epoch": 2.74, "learning_rate": 2.825434188885368e-05, "loss": 0.4757, "step": 45410 }, { "epoch": 2.74, "learning_rate": 2.8252832978739458e-05, "loss": 0.3611, "step": 45412 }, { "epoch": 2.74, "learning_rate": 2.8251324068625235e-05, "loss": 0.5536, "step": 45414 }, { "epoch": 2.74, "learning_rate": 2.824981515851101e-05, "loss": 0.5975, "step": 45416 }, { "epoch": 2.74, "learning_rate": 2.8248306248396784e-05, "loss": 0.6036, "step": 45418 }, { "epoch": 2.74, "learning_rate": 2.824679733828256e-05, "loss": 0.4697, "step": 45420 }, { "epoch": 2.74, "learning_rate": 2.8245288428168337e-05, "loss": 0.6367, "step": 45422 }, { "epoch": 2.74, "learning_rate": 2.824377951805411e-05, "loss": 0.5778, "step": 45424 }, { "epoch": 2.74, "learning_rate": 2.8242270607939887e-05, "loss": 0.6269, "step": 45426 }, { "epoch": 2.74, "learning_rate": 2.8240761697825663e-05, "loss": 0.3781, "step": 45428 }, { "epoch": 2.74, "learning_rate": 2.823925278771144e-05, "loss": 0.3873, "step": 45430 }, { "epoch": 2.74, "learning_rate": 2.8237743877597213e-05, "loss": 0.5525, "step": 45432 }, { "epoch": 2.74, "learning_rate": 2.823623496748299e-05, "loss": 0.4635, "step": 45434 }, { "epoch": 2.74, "learning_rate": 2.8234726057368766e-05, "loss": 0.4005, "step": 45436 }, { "epoch": 2.74, "learning_rate": 2.8233217147254536e-05, "loss": 0.5688, "step": 45438 }, { "epoch": 2.74, "learning_rate": 2.8231708237140316e-05, "loss": 0.4854, "step": 45440 }, { "epoch": 2.74, "learning_rate": 2.8230199327026092e-05, "loss": 0.5755, "step": 45442 }, { "epoch": 2.74, "learning_rate": 2.822869041691187e-05, "loss": 0.4865, "step": 45444 }, { "epoch": 2.74, "learning_rate": 2.822718150679764e-05, "loss": 0.3327, "step": 45446 }, { "epoch": 2.74, "learning_rate": 2.8225672596683418e-05, "loss": 0.4524, "step": 45448 }, { "epoch": 2.74, "learning_rate": 2.8224163686569195e-05, "loss": 0.5349, "step": 45450 }, { "epoch": 2.74, "learning_rate": 2.8222654776454965e-05, "loss": 0.525, "step": 45452 }, { "epoch": 2.74, "learning_rate": 2.822114586634074e-05, "loss": 0.6892, "step": 45454 }, { "epoch": 2.74, "learning_rate": 2.821963695622652e-05, "loss": 0.8468, "step": 45456 }, { "epoch": 2.74, "learning_rate": 2.8218128046112297e-05, "loss": 0.4257, "step": 45458 }, { "epoch": 2.74, "learning_rate": 2.8216619135998067e-05, "loss": 0.5452, "step": 45460 }, { "epoch": 2.74, "learning_rate": 2.8215110225883844e-05, "loss": 0.4503, "step": 45462 }, { "epoch": 2.74, "learning_rate": 2.8213601315769624e-05, "loss": 0.7047, "step": 45464 }, { "epoch": 2.74, "learning_rate": 2.8212092405655393e-05, "loss": 0.5649, "step": 45466 }, { "epoch": 2.74, "learning_rate": 2.821058349554117e-05, "loss": 0.4941, "step": 45468 }, { "epoch": 2.74, "learning_rate": 2.8209074585426946e-05, "loss": 0.7087, "step": 45470 }, { "epoch": 2.74, "learning_rate": 2.8207565675312726e-05, "loss": 0.4201, "step": 45472 }, { "epoch": 2.74, "learning_rate": 2.8206056765198496e-05, "loss": 0.395, "step": 45474 }, { "epoch": 2.74, "learning_rate": 2.8204547855084273e-05, "loss": 0.5692, "step": 45476 }, { "epoch": 2.74, "learning_rate": 2.820303894497005e-05, "loss": 0.5137, "step": 45478 }, { "epoch": 2.75, "learning_rate": 2.820153003485583e-05, "loss": 0.562, "step": 45480 }, { "epoch": 2.75, "learning_rate": 2.82000211247416e-05, "loss": 0.6779, "step": 45482 }, { "epoch": 2.75, "learning_rate": 2.8198512214627375e-05, "loss": 0.8117, "step": 45484 }, { "epoch": 2.75, "learning_rate": 2.8197003304513152e-05, "loss": 0.597, "step": 45486 }, { "epoch": 2.75, "learning_rate": 2.8195494394398925e-05, "loss": 0.4771, "step": 45488 }, { "epoch": 2.75, "learning_rate": 2.81939854842847e-05, "loss": 0.6697, "step": 45490 }, { "epoch": 2.75, "learning_rate": 2.8192476574170478e-05, "loss": 0.5957, "step": 45492 }, { "epoch": 2.75, "learning_rate": 2.8190967664056254e-05, "loss": 0.5693, "step": 45494 }, { "epoch": 2.75, "learning_rate": 2.8189458753942028e-05, "loss": 0.4347, "step": 45496 }, { "epoch": 2.75, "learning_rate": 2.8187949843827804e-05, "loss": 0.5841, "step": 45498 }, { "epoch": 2.75, "learning_rate": 2.818644093371358e-05, "loss": 0.5863, "step": 45500 }, { "epoch": 2.75, "learning_rate": 2.8184932023599354e-05, "loss": 0.5771, "step": 45502 }, { "epoch": 2.75, "learning_rate": 2.818342311348513e-05, "loss": 0.5547, "step": 45504 }, { "epoch": 2.75, "learning_rate": 2.8181914203370907e-05, "loss": 0.4188, "step": 45506 }, { "epoch": 2.75, "learning_rate": 2.8180405293256683e-05, "loss": 0.3812, "step": 45508 }, { "epoch": 2.75, "learning_rate": 2.8178896383142456e-05, "loss": 0.6658, "step": 45510 }, { "epoch": 2.75, "learning_rate": 2.8177387473028233e-05, "loss": 0.7234, "step": 45512 }, { "epoch": 2.75, "learning_rate": 2.817587856291401e-05, "loss": 0.5283, "step": 45514 }, { "epoch": 2.75, "learning_rate": 2.8174369652799782e-05, "loss": 0.6708, "step": 45516 }, { "epoch": 2.75, "learning_rate": 2.817286074268556e-05, "loss": 0.4731, "step": 45518 }, { "epoch": 2.75, "learning_rate": 2.8171351832571336e-05, "loss": 0.4357, "step": 45520 }, { "epoch": 2.75, "learning_rate": 2.8169842922457112e-05, "loss": 0.4795, "step": 45522 }, { "epoch": 2.75, "learning_rate": 2.8168334012342885e-05, "loss": 0.5791, "step": 45524 }, { "epoch": 2.75, "learning_rate": 2.816682510222866e-05, "loss": 0.5428, "step": 45526 }, { "epoch": 2.75, "learning_rate": 2.8165316192114438e-05, "loss": 0.5018, "step": 45528 }, { "epoch": 2.75, "learning_rate": 2.8163807282000215e-05, "loss": 0.6307, "step": 45530 }, { "epoch": 2.75, "learning_rate": 2.8162298371885988e-05, "loss": 0.4818, "step": 45532 }, { "epoch": 2.75, "learning_rate": 2.8160789461771764e-05, "loss": 0.7055, "step": 45534 }, { "epoch": 2.75, "learning_rate": 2.815928055165754e-05, "loss": 0.4684, "step": 45536 }, { "epoch": 2.75, "learning_rate": 2.8157771641543314e-05, "loss": 0.7952, "step": 45538 }, { "epoch": 2.75, "learning_rate": 2.815626273142909e-05, "loss": 0.4352, "step": 45540 }, { "epoch": 2.75, "learning_rate": 2.8154753821314867e-05, "loss": 0.4098, "step": 45542 }, { "epoch": 2.75, "learning_rate": 2.8153244911200644e-05, "loss": 0.6422, "step": 45544 }, { "epoch": 2.75, "learning_rate": 2.8151736001086417e-05, "loss": 0.5795, "step": 45546 }, { "epoch": 2.75, "learning_rate": 2.8150227090972193e-05, "loss": 0.3502, "step": 45548 }, { "epoch": 2.75, "learning_rate": 2.814871818085797e-05, "loss": 0.5886, "step": 45550 }, { "epoch": 2.75, "learning_rate": 2.814720927074374e-05, "loss": 0.607, "step": 45552 }, { "epoch": 2.75, "learning_rate": 2.814570036062952e-05, "loss": 0.3588, "step": 45554 }, { "epoch": 2.75, "learning_rate": 2.8144191450515296e-05, "loss": 0.7465, "step": 45556 }, { "epoch": 2.75, "learning_rate": 2.8142682540401072e-05, "loss": 0.5676, "step": 45558 }, { "epoch": 2.75, "learning_rate": 2.8141173630286842e-05, "loss": 0.5458, "step": 45560 }, { "epoch": 2.75, "learning_rate": 2.8139664720172622e-05, "loss": 0.4679, "step": 45562 }, { "epoch": 2.75, "learning_rate": 2.81381558100584e-05, "loss": 0.5053, "step": 45564 }, { "epoch": 2.75, "learning_rate": 2.8136646899944168e-05, "loss": 0.4236, "step": 45566 }, { "epoch": 2.75, "learning_rate": 2.8135137989829945e-05, "loss": 0.3742, "step": 45568 }, { "epoch": 2.75, "learning_rate": 2.8133629079715725e-05, "loss": 0.4597, "step": 45570 }, { "epoch": 2.75, "learning_rate": 2.81321201696015e-05, "loss": 0.6477, "step": 45572 }, { "epoch": 2.75, "learning_rate": 2.813061125948727e-05, "loss": 0.6118, "step": 45574 }, { "epoch": 2.75, "learning_rate": 2.8129102349373047e-05, "loss": 0.5175, "step": 45576 }, { "epoch": 2.75, "learning_rate": 2.8127593439258827e-05, "loss": 0.8721, "step": 45578 }, { "epoch": 2.75, "learning_rate": 2.8126084529144604e-05, "loss": 0.6562, "step": 45580 }, { "epoch": 2.75, "learning_rate": 2.8124575619030374e-05, "loss": 0.5157, "step": 45582 }, { "epoch": 2.75, "learning_rate": 2.812306670891615e-05, "loss": 0.5859, "step": 45584 }, { "epoch": 2.75, "learning_rate": 2.812155779880193e-05, "loss": 0.6111, "step": 45586 }, { "epoch": 2.75, "learning_rate": 2.81200488886877e-05, "loss": 0.3007, "step": 45588 }, { "epoch": 2.75, "learning_rate": 2.8118539978573476e-05, "loss": 0.5978, "step": 45590 }, { "epoch": 2.75, "learning_rate": 2.8117031068459253e-05, "loss": 0.4681, "step": 45592 }, { "epoch": 2.75, "learning_rate": 2.8115522158345033e-05, "loss": 0.5793, "step": 45594 }, { "epoch": 2.75, "learning_rate": 2.8114013248230802e-05, "loss": 0.5669, "step": 45596 }, { "epoch": 2.75, "learning_rate": 2.811250433811658e-05, "loss": 0.4731, "step": 45598 }, { "epoch": 2.75, "learning_rate": 2.8110995428002355e-05, "loss": 0.6823, "step": 45600 }, { "epoch": 2.75, "learning_rate": 2.810948651788813e-05, "loss": 0.4956, "step": 45602 }, { "epoch": 2.75, "learning_rate": 2.8107977607773905e-05, "loss": 0.5485, "step": 45604 }, { "epoch": 2.75, "learning_rate": 2.810646869765968e-05, "loss": 0.6451, "step": 45606 }, { "epoch": 2.75, "learning_rate": 2.8104959787545458e-05, "loss": 0.681, "step": 45608 }, { "epoch": 2.75, "learning_rate": 2.810345087743123e-05, "loss": 0.5908, "step": 45610 }, { "epoch": 2.75, "learning_rate": 2.8101941967317008e-05, "loss": 0.6703, "step": 45612 }, { "epoch": 2.75, "learning_rate": 2.8100433057202784e-05, "loss": 0.425, "step": 45614 }, { "epoch": 2.75, "learning_rate": 2.8098924147088557e-05, "loss": 0.4502, "step": 45616 }, { "epoch": 2.75, "learning_rate": 2.8097415236974334e-05, "loss": 0.4443, "step": 45618 }, { "epoch": 2.75, "learning_rate": 2.809590632686011e-05, "loss": 0.4048, "step": 45620 }, { "epoch": 2.75, "learning_rate": 2.8094397416745887e-05, "loss": 0.6816, "step": 45622 }, { "epoch": 2.75, "learning_rate": 2.809288850663166e-05, "loss": 0.3988, "step": 45624 }, { "epoch": 2.75, "learning_rate": 2.8091379596517437e-05, "loss": 0.4733, "step": 45626 }, { "epoch": 2.75, "learning_rate": 2.8089870686403213e-05, "loss": 0.5067, "step": 45628 }, { "epoch": 2.75, "learning_rate": 2.808836177628899e-05, "loss": 0.6256, "step": 45630 }, { "epoch": 2.75, "learning_rate": 2.8086852866174763e-05, "loss": 0.4258, "step": 45632 }, { "epoch": 2.75, "learning_rate": 2.808534395606054e-05, "loss": 0.7765, "step": 45634 }, { "epoch": 2.75, "learning_rate": 2.8083835045946316e-05, "loss": 0.4605, "step": 45636 }, { "epoch": 2.75, "learning_rate": 2.808232613583209e-05, "loss": 0.5947, "step": 45638 }, { "epoch": 2.75, "learning_rate": 2.8080817225717865e-05, "loss": 0.6647, "step": 45640 }, { "epoch": 2.75, "learning_rate": 2.8079308315603642e-05, "loss": 0.6433, "step": 45642 }, { "epoch": 2.75, "learning_rate": 2.807779940548942e-05, "loss": 0.4708, "step": 45644 }, { "epoch": 2.76, "learning_rate": 2.807629049537519e-05, "loss": 0.7595, "step": 45646 }, { "epoch": 2.76, "learning_rate": 2.8074781585260968e-05, "loss": 0.9736, "step": 45648 }, { "epoch": 2.76, "learning_rate": 2.8073272675146745e-05, "loss": 0.5154, "step": 45650 }, { "epoch": 2.76, "learning_rate": 2.8071763765032518e-05, "loss": 0.6089, "step": 45652 }, { "epoch": 2.76, "learning_rate": 2.8070254854918294e-05, "loss": 0.3554, "step": 45654 }, { "epoch": 2.76, "learning_rate": 2.806874594480407e-05, "loss": 0.5694, "step": 45656 }, { "epoch": 2.76, "learning_rate": 2.8067237034689847e-05, "loss": 0.3893, "step": 45658 }, { "epoch": 2.76, "learning_rate": 2.806572812457562e-05, "loss": 0.3497, "step": 45660 }, { "epoch": 2.76, "learning_rate": 2.8064219214461397e-05, "loss": 0.6912, "step": 45662 }, { "epoch": 2.76, "learning_rate": 2.8062710304347173e-05, "loss": 0.4774, "step": 45664 }, { "epoch": 2.76, "learning_rate": 2.8061201394232943e-05, "loss": 0.3992, "step": 45666 }, { "epoch": 2.76, "learning_rate": 2.8059692484118723e-05, "loss": 0.4395, "step": 45668 }, { "epoch": 2.76, "learning_rate": 2.80581835740045e-05, "loss": 0.3475, "step": 45670 }, { "epoch": 2.76, "learning_rate": 2.8056674663890276e-05, "loss": 0.5096, "step": 45672 }, { "epoch": 2.76, "learning_rate": 2.8055165753776046e-05, "loss": 0.4959, "step": 45674 }, { "epoch": 2.76, "learning_rate": 2.8053656843661826e-05, "loss": 0.56, "step": 45676 }, { "epoch": 2.76, "learning_rate": 2.8052147933547602e-05, "loss": 0.5337, "step": 45678 }, { "epoch": 2.76, "learning_rate": 2.8050639023433372e-05, "loss": 0.5903, "step": 45680 }, { "epoch": 2.76, "learning_rate": 2.804913011331915e-05, "loss": 0.6889, "step": 45682 }, { "epoch": 2.76, "learning_rate": 2.804762120320493e-05, "loss": 0.6023, "step": 45684 }, { "epoch": 2.76, "learning_rate": 2.8046112293090705e-05, "loss": 0.5451, "step": 45686 }, { "epoch": 2.76, "learning_rate": 2.8044603382976475e-05, "loss": 0.4368, "step": 45688 }, { "epoch": 2.76, "learning_rate": 2.804309447286225e-05, "loss": 0.4763, "step": 45690 }, { "epoch": 2.76, "learning_rate": 2.804158556274803e-05, "loss": 0.8384, "step": 45692 }, { "epoch": 2.76, "learning_rate": 2.8040076652633807e-05, "loss": 0.4961, "step": 45694 }, { "epoch": 2.76, "learning_rate": 2.8038567742519577e-05, "loss": 0.6659, "step": 45696 }, { "epoch": 2.76, "learning_rate": 2.8037058832405354e-05, "loss": 0.4622, "step": 45698 }, { "epoch": 2.76, "learning_rate": 2.8035549922291134e-05, "loss": 0.6963, "step": 45700 }, { "epoch": 2.76, "learning_rate": 2.8034041012176903e-05, "loss": 0.7454, "step": 45702 }, { "epoch": 2.76, "learning_rate": 2.803253210206268e-05, "loss": 0.6578, "step": 45704 }, { "epoch": 2.76, "learning_rate": 2.8031023191948456e-05, "loss": 0.5195, "step": 45706 }, { "epoch": 2.76, "learning_rate": 2.8029514281834236e-05, "loss": 0.4857, "step": 45708 }, { "epoch": 2.76, "learning_rate": 2.8028005371720006e-05, "loss": 0.4557, "step": 45710 }, { "epoch": 2.76, "learning_rate": 2.8026496461605783e-05, "loss": 0.4087, "step": 45712 }, { "epoch": 2.76, "learning_rate": 2.802498755149156e-05, "loss": 0.4897, "step": 45714 }, { "epoch": 2.76, "learning_rate": 2.8023478641377332e-05, "loss": 0.4967, "step": 45716 }, { "epoch": 2.76, "learning_rate": 2.802196973126311e-05, "loss": 0.373, "step": 45718 }, { "epoch": 2.76, "learning_rate": 2.8020460821148885e-05, "loss": 0.4515, "step": 45720 }, { "epoch": 2.76, "learning_rate": 2.8018951911034662e-05, "loss": 0.6165, "step": 45722 }, { "epoch": 2.76, "learning_rate": 2.8017443000920435e-05, "loss": 0.5589, "step": 45724 }, { "epoch": 2.76, "learning_rate": 2.801593409080621e-05, "loss": 0.6009, "step": 45726 }, { "epoch": 2.76, "learning_rate": 2.8014425180691988e-05, "loss": 0.4311, "step": 45728 }, { "epoch": 2.76, "learning_rate": 2.801291627057776e-05, "loss": 0.5796, "step": 45730 }, { "epoch": 2.76, "learning_rate": 2.8011407360463538e-05, "loss": 0.6004, "step": 45732 }, { "epoch": 2.76, "learning_rate": 2.8009898450349314e-05, "loss": 0.5787, "step": 45734 }, { "epoch": 2.76, "learning_rate": 2.800838954023509e-05, "loss": 0.6821, "step": 45736 }, { "epoch": 2.76, "learning_rate": 2.8006880630120864e-05, "loss": 0.4955, "step": 45738 }, { "epoch": 2.76, "learning_rate": 2.800537172000664e-05, "loss": 0.4249, "step": 45740 }, { "epoch": 2.76, "learning_rate": 2.8003862809892417e-05, "loss": 0.3868, "step": 45742 }, { "epoch": 2.76, "learning_rate": 2.8002353899778193e-05, "loss": 0.4363, "step": 45744 }, { "epoch": 2.76, "learning_rate": 2.8000844989663966e-05, "loss": 0.6083, "step": 45746 }, { "epoch": 2.76, "learning_rate": 2.7999336079549743e-05, "loss": 0.4735, "step": 45748 }, { "epoch": 2.76, "learning_rate": 2.799782716943552e-05, "loss": 0.6427, "step": 45750 }, { "epoch": 2.76, "learning_rate": 2.7996318259321293e-05, "loss": 0.6723, "step": 45752 }, { "epoch": 2.76, "learning_rate": 2.799480934920707e-05, "loss": 0.501, "step": 45754 }, { "epoch": 2.76, "learning_rate": 2.7993300439092846e-05, "loss": 0.5365, "step": 45756 }, { "epoch": 2.76, "learning_rate": 2.7991791528978622e-05, "loss": 0.7888, "step": 45758 }, { "epoch": 2.76, "learning_rate": 2.7990282618864395e-05, "loss": 0.4973, "step": 45760 }, { "epoch": 2.76, "learning_rate": 2.798877370875017e-05, "loss": 0.4589, "step": 45762 }, { "epoch": 2.76, "learning_rate": 2.7987264798635948e-05, "loss": 0.4994, "step": 45764 }, { "epoch": 2.76, "learning_rate": 2.798575588852172e-05, "loss": 0.4845, "step": 45766 }, { "epoch": 2.76, "learning_rate": 2.7984246978407498e-05, "loss": 0.4869, "step": 45768 }, { "epoch": 2.76, "learning_rate": 2.7982738068293274e-05, "loss": 0.6214, "step": 45770 }, { "epoch": 2.76, "learning_rate": 2.798122915817905e-05, "loss": 0.6855, "step": 45772 }, { "epoch": 2.76, "learning_rate": 2.7979720248064824e-05, "loss": 0.6935, "step": 45774 }, { "epoch": 2.76, "learning_rate": 2.79782113379506e-05, "loss": 0.794, "step": 45776 }, { "epoch": 2.76, "learning_rate": 2.7976702427836377e-05, "loss": 0.5061, "step": 45778 }, { "epoch": 2.76, "learning_rate": 2.7975193517722147e-05, "loss": 0.6191, "step": 45780 }, { "epoch": 2.76, "learning_rate": 2.7973684607607927e-05, "loss": 0.471, "step": 45782 }, { "epoch": 2.76, "learning_rate": 2.7972175697493703e-05, "loss": 0.7918, "step": 45784 }, { "epoch": 2.76, "learning_rate": 2.797066678737948e-05, "loss": 0.5503, "step": 45786 }, { "epoch": 2.76, "learning_rate": 2.796915787726525e-05, "loss": 0.3853, "step": 45788 }, { "epoch": 2.76, "learning_rate": 2.796764896715103e-05, "loss": 0.7015, "step": 45790 }, { "epoch": 2.76, "learning_rate": 2.7966140057036806e-05, "loss": 0.5589, "step": 45792 }, { "epoch": 2.76, "learning_rate": 2.7964631146922582e-05, "loss": 0.3262, "step": 45794 }, { "epoch": 2.76, "learning_rate": 2.7963122236808352e-05, "loss": 0.5473, "step": 45796 }, { "epoch": 2.76, "learning_rate": 2.7961613326694132e-05, "loss": 0.5506, "step": 45798 }, { "epoch": 2.76, "learning_rate": 2.796010441657991e-05, "loss": 0.5009, "step": 45800 }, { "epoch": 2.76, "learning_rate": 2.7958595506465678e-05, "loss": 0.5954, "step": 45802 }, { "epoch": 2.76, "learning_rate": 2.7957086596351455e-05, "loss": 0.7892, "step": 45804 }, { "epoch": 2.76, "learning_rate": 2.7955577686237235e-05, "loss": 0.5417, "step": 45806 }, { "epoch": 2.76, "learning_rate": 2.795406877612301e-05, "loss": 0.5127, "step": 45808 }, { "epoch": 2.76, "learning_rate": 2.795255986600878e-05, "loss": 0.597, "step": 45810 }, { "epoch": 2.77, "learning_rate": 2.7951050955894557e-05, "loss": 0.5223, "step": 45812 }, { "epoch": 2.77, "learning_rate": 2.7949542045780337e-05, "loss": 0.6019, "step": 45814 }, { "epoch": 2.77, "learning_rate": 2.7948033135666107e-05, "loss": 0.4204, "step": 45816 }, { "epoch": 2.77, "learning_rate": 2.7946524225551884e-05, "loss": 0.7394, "step": 45818 }, { "epoch": 2.77, "learning_rate": 2.794501531543766e-05, "loss": 0.4753, "step": 45820 }, { "epoch": 2.77, "learning_rate": 2.794350640532344e-05, "loss": 0.4886, "step": 45822 }, { "epoch": 2.77, "learning_rate": 2.794199749520921e-05, "loss": 0.5128, "step": 45824 }, { "epoch": 2.77, "learning_rate": 2.7940488585094986e-05, "loss": 0.304, "step": 45826 }, { "epoch": 2.77, "learning_rate": 2.7938979674980763e-05, "loss": 0.4234, "step": 45828 }, { "epoch": 2.77, "learning_rate": 2.7937470764866536e-05, "loss": 0.6507, "step": 45830 }, { "epoch": 2.77, "learning_rate": 2.7935961854752312e-05, "loss": 0.589, "step": 45832 }, { "epoch": 2.77, "learning_rate": 2.793445294463809e-05, "loss": 0.5883, "step": 45834 }, { "epoch": 2.77, "learning_rate": 2.7932944034523865e-05, "loss": 0.4418, "step": 45836 }, { "epoch": 2.77, "learning_rate": 2.793143512440964e-05, "loss": 0.4227, "step": 45838 }, { "epoch": 2.77, "learning_rate": 2.7929926214295415e-05, "loss": 0.5446, "step": 45840 }, { "epoch": 2.77, "learning_rate": 2.792841730418119e-05, "loss": 0.389, "step": 45842 }, { "epoch": 2.77, "learning_rate": 2.7926908394066965e-05, "loss": 0.44, "step": 45844 }, { "epoch": 2.77, "learning_rate": 2.792539948395274e-05, "loss": 0.4322, "step": 45846 }, { "epoch": 2.77, "learning_rate": 2.7923890573838518e-05, "loss": 0.4744, "step": 45848 }, { "epoch": 2.77, "learning_rate": 2.7922381663724294e-05, "loss": 0.7476, "step": 45850 }, { "epoch": 2.77, "learning_rate": 2.7920872753610067e-05, "loss": 0.7134, "step": 45852 }, { "epoch": 2.77, "learning_rate": 2.7919363843495844e-05, "loss": 0.4044, "step": 45854 }, { "epoch": 2.77, "learning_rate": 2.791785493338162e-05, "loss": 0.673, "step": 45856 }, { "epoch": 2.77, "learning_rate": 2.7916346023267397e-05, "loss": 0.5052, "step": 45858 }, { "epoch": 2.77, "learning_rate": 2.791483711315317e-05, "loss": 0.6976, "step": 45860 }, { "epoch": 2.77, "learning_rate": 2.7913328203038947e-05, "loss": 0.4794, "step": 45862 }, { "epoch": 2.77, "learning_rate": 2.7911819292924723e-05, "loss": 0.4376, "step": 45864 }, { "epoch": 2.77, "learning_rate": 2.7910310382810496e-05, "loss": 0.3888, "step": 45866 }, { "epoch": 2.77, "learning_rate": 2.7908801472696273e-05, "loss": 0.5206, "step": 45868 }, { "epoch": 2.77, "learning_rate": 2.790729256258205e-05, "loss": 0.4774, "step": 45870 }, { "epoch": 2.77, "learning_rate": 2.7905783652467826e-05, "loss": 0.5382, "step": 45872 }, { "epoch": 2.77, "learning_rate": 2.79042747423536e-05, "loss": 0.4404, "step": 45874 }, { "epoch": 2.77, "learning_rate": 2.7902765832239375e-05, "loss": 0.3984, "step": 45876 }, { "epoch": 2.77, "learning_rate": 2.7901256922125152e-05, "loss": 0.6205, "step": 45878 }, { "epoch": 2.77, "learning_rate": 2.7899748012010925e-05, "loss": 0.4578, "step": 45880 }, { "epoch": 2.77, "learning_rate": 2.78982391018967e-05, "loss": 0.7651, "step": 45882 }, { "epoch": 2.77, "learning_rate": 2.7896730191782478e-05, "loss": 0.4456, "step": 45884 }, { "epoch": 2.77, "learning_rate": 2.7895221281668255e-05, "loss": 0.4537, "step": 45886 }, { "epoch": 2.77, "learning_rate": 2.7893712371554028e-05, "loss": 0.5825, "step": 45888 }, { "epoch": 2.77, "learning_rate": 2.7892203461439804e-05, "loss": 0.5729, "step": 45890 }, { "epoch": 2.77, "learning_rate": 2.789069455132558e-05, "loss": 0.502, "step": 45892 }, { "epoch": 2.77, "learning_rate": 2.788918564121135e-05, "loss": 0.6041, "step": 45894 }, { "epoch": 2.77, "learning_rate": 2.788767673109713e-05, "loss": 0.5705, "step": 45896 }, { "epoch": 2.77, "learning_rate": 2.7886167820982907e-05, "loss": 0.5234, "step": 45898 }, { "epoch": 2.77, "learning_rate": 2.7884658910868683e-05, "loss": 0.4439, "step": 45900 }, { "epoch": 2.77, "learning_rate": 2.7883150000754453e-05, "loss": 0.3923, "step": 45902 }, { "epoch": 2.77, "learning_rate": 2.7881641090640233e-05, "loss": 0.2888, "step": 45904 }, { "epoch": 2.77, "learning_rate": 2.788013218052601e-05, "loss": 0.5753, "step": 45906 }, { "epoch": 2.77, "learning_rate": 2.7878623270411786e-05, "loss": 0.5172, "step": 45908 }, { "epoch": 2.77, "learning_rate": 2.7877114360297556e-05, "loss": 0.4724, "step": 45910 }, { "epoch": 2.77, "learning_rate": 2.7875605450183336e-05, "loss": 0.4161, "step": 45912 }, { "epoch": 2.77, "learning_rate": 2.7874096540069112e-05, "loss": 0.6251, "step": 45914 }, { "epoch": 2.77, "learning_rate": 2.7872587629954882e-05, "loss": 0.5169, "step": 45916 }, { "epoch": 2.77, "learning_rate": 2.787107871984066e-05, "loss": 0.5596, "step": 45918 }, { "epoch": 2.77, "learning_rate": 2.786956980972644e-05, "loss": 0.6446, "step": 45920 }, { "epoch": 2.77, "learning_rate": 2.7868060899612215e-05, "loss": 0.5156, "step": 45922 }, { "epoch": 2.77, "learning_rate": 2.7866551989497985e-05, "loss": 0.3668, "step": 45924 }, { "epoch": 2.77, "learning_rate": 2.786504307938376e-05, "loss": 0.5113, "step": 45926 }, { "epoch": 2.77, "learning_rate": 2.786353416926954e-05, "loss": 0.6408, "step": 45928 }, { "epoch": 2.77, "learning_rate": 2.786202525915531e-05, "loss": 0.4747, "step": 45930 }, { "epoch": 2.77, "learning_rate": 2.7860516349041087e-05, "loss": 0.3097, "step": 45932 }, { "epoch": 2.77, "learning_rate": 2.7859007438926864e-05, "loss": 0.4392, "step": 45934 }, { "epoch": 2.77, "learning_rate": 2.7857498528812644e-05, "loss": 0.4187, "step": 45936 }, { "epoch": 2.77, "learning_rate": 2.7855989618698413e-05, "loss": 0.5208, "step": 45938 }, { "epoch": 2.77, "learning_rate": 2.785448070858419e-05, "loss": 0.5302, "step": 45940 }, { "epoch": 2.77, "learning_rate": 2.7852971798469966e-05, "loss": 0.5753, "step": 45942 }, { "epoch": 2.77, "learning_rate": 2.785146288835574e-05, "loss": 0.5617, "step": 45944 }, { "epoch": 2.77, "learning_rate": 2.7849953978241516e-05, "loss": 0.5271, "step": 45946 }, { "epoch": 2.77, "learning_rate": 2.7848445068127293e-05, "loss": 0.5252, "step": 45948 }, { "epoch": 2.77, "learning_rate": 2.784693615801307e-05, "loss": 0.3367, "step": 45950 }, { "epoch": 2.77, "learning_rate": 2.7845427247898842e-05, "loss": 0.6, "step": 45952 }, { "epoch": 2.77, "learning_rate": 2.784391833778462e-05, "loss": 0.4261, "step": 45954 }, { "epoch": 2.77, "learning_rate": 2.7842409427670395e-05, "loss": 0.5101, "step": 45956 }, { "epoch": 2.77, "learning_rate": 2.7840900517556172e-05, "loss": 0.7317, "step": 45958 }, { "epoch": 2.77, "learning_rate": 2.7839391607441945e-05, "loss": 0.5047, "step": 45960 }, { "epoch": 2.77, "learning_rate": 2.783788269732772e-05, "loss": 0.6257, "step": 45962 }, { "epoch": 2.77, "learning_rate": 2.7836373787213498e-05, "loss": 0.5092, "step": 45964 }, { "epoch": 2.77, "learning_rate": 2.783486487709927e-05, "loss": 0.537, "step": 45966 }, { "epoch": 2.77, "learning_rate": 2.7833355966985048e-05, "loss": 0.7748, "step": 45968 }, { "epoch": 2.77, "learning_rate": 2.7831847056870824e-05, "loss": 0.4422, "step": 45970 }, { "epoch": 2.77, "learning_rate": 2.78303381467566e-05, "loss": 0.3342, "step": 45972 }, { "epoch": 2.77, "learning_rate": 2.7828829236642374e-05, "loss": 0.5604, "step": 45974 }, { "epoch": 2.77, "learning_rate": 2.782732032652815e-05, "loss": 0.6263, "step": 45976 }, { "epoch": 2.78, "learning_rate": 2.7825811416413927e-05, "loss": 0.4298, "step": 45978 }, { "epoch": 2.78, "learning_rate": 2.78243025062997e-05, "loss": 0.4578, "step": 45980 }, { "epoch": 2.78, "learning_rate": 2.7822793596185476e-05, "loss": 0.6194, "step": 45982 }, { "epoch": 2.78, "learning_rate": 2.7821284686071253e-05, "loss": 0.4909, "step": 45984 }, { "epoch": 2.78, "learning_rate": 2.781977577595703e-05, "loss": 0.6379, "step": 45986 }, { "epoch": 2.78, "learning_rate": 2.7818266865842803e-05, "loss": 0.5127, "step": 45988 }, { "epoch": 2.78, "learning_rate": 2.781675795572858e-05, "loss": 0.7263, "step": 45990 }, { "epoch": 2.78, "learning_rate": 2.7815249045614356e-05, "loss": 0.5337, "step": 45992 }, { "epoch": 2.78, "learning_rate": 2.781374013550013e-05, "loss": 0.6178, "step": 45994 }, { "epoch": 2.78, "learning_rate": 2.7812231225385905e-05, "loss": 0.4216, "step": 45996 }, { "epoch": 2.78, "learning_rate": 2.7810722315271682e-05, "loss": 0.479, "step": 45998 }, { "epoch": 2.78, "learning_rate": 2.7809213405157458e-05, "loss": 0.4668, "step": 46000 }, { "epoch": 2.78, "learning_rate": 2.780770449504323e-05, "loss": 0.6636, "step": 46002 }, { "epoch": 2.78, "learning_rate": 2.7806195584929008e-05, "loss": 0.6121, "step": 46004 }, { "epoch": 2.78, "learning_rate": 2.7804686674814784e-05, "loss": 0.4393, "step": 46006 }, { "epoch": 2.78, "learning_rate": 2.7803177764700554e-05, "loss": 0.5649, "step": 46008 }, { "epoch": 2.78, "learning_rate": 2.7801668854586334e-05, "loss": 0.4271, "step": 46010 }, { "epoch": 2.78, "learning_rate": 2.780015994447211e-05, "loss": 0.683, "step": 46012 }, { "epoch": 2.78, "learning_rate": 2.7798651034357887e-05, "loss": 0.4776, "step": 46014 }, { "epoch": 2.78, "learning_rate": 2.7797142124243657e-05, "loss": 0.4552, "step": 46016 }, { "epoch": 2.78, "learning_rate": 2.7795633214129437e-05, "loss": 0.5239, "step": 46018 }, { "epoch": 2.78, "learning_rate": 2.7794124304015213e-05, "loss": 0.5846, "step": 46020 }, { "epoch": 2.78, "learning_rate": 2.779261539390099e-05, "loss": 0.5336, "step": 46022 }, { "epoch": 2.78, "learning_rate": 2.779110648378676e-05, "loss": 0.4349, "step": 46024 }, { "epoch": 2.78, "learning_rate": 2.778959757367254e-05, "loss": 0.4451, "step": 46026 }, { "epoch": 2.78, "learning_rate": 2.7788088663558316e-05, "loss": 0.5353, "step": 46028 }, { "epoch": 2.78, "learning_rate": 2.7786579753444086e-05, "loss": 0.6249, "step": 46030 }, { "epoch": 2.78, "learning_rate": 2.7785070843329862e-05, "loss": 0.6277, "step": 46032 }, { "epoch": 2.78, "learning_rate": 2.7783561933215642e-05, "loss": 0.5446, "step": 46034 }, { "epoch": 2.78, "learning_rate": 2.778205302310142e-05, "loss": 0.7307, "step": 46036 }, { "epoch": 2.78, "learning_rate": 2.7780544112987188e-05, "loss": 0.6405, "step": 46038 }, { "epoch": 2.78, "learning_rate": 2.7779035202872965e-05, "loss": 0.5783, "step": 46040 }, { "epoch": 2.78, "learning_rate": 2.7777526292758745e-05, "loss": 0.5432, "step": 46042 }, { "epoch": 2.78, "learning_rate": 2.7776017382644514e-05, "loss": 0.6289, "step": 46044 }, { "epoch": 2.78, "learning_rate": 2.777450847253029e-05, "loss": 0.5219, "step": 46046 }, { "epoch": 2.78, "learning_rate": 2.7772999562416067e-05, "loss": 0.5398, "step": 46048 }, { "epoch": 2.78, "learning_rate": 2.7771490652301847e-05, "loss": 0.4683, "step": 46050 }, { "epoch": 2.78, "learning_rate": 2.7769981742187617e-05, "loss": 0.4551, "step": 46052 }, { "epoch": 2.78, "learning_rate": 2.7768472832073394e-05, "loss": 0.569, "step": 46054 }, { "epoch": 2.78, "learning_rate": 2.776696392195917e-05, "loss": 0.4892, "step": 46056 }, { "epoch": 2.78, "learning_rate": 2.7765455011844943e-05, "loss": 0.4949, "step": 46058 }, { "epoch": 2.78, "learning_rate": 2.776394610173072e-05, "loss": 0.4673, "step": 46060 }, { "epoch": 2.78, "learning_rate": 2.7762437191616496e-05, "loss": 0.5881, "step": 46062 }, { "epoch": 2.78, "learning_rate": 2.7760928281502273e-05, "loss": 0.5277, "step": 46064 }, { "epoch": 2.78, "learning_rate": 2.7759419371388046e-05, "loss": 0.7445, "step": 46066 }, { "epoch": 2.78, "learning_rate": 2.7757910461273822e-05, "loss": 0.4783, "step": 46068 }, { "epoch": 2.78, "learning_rate": 2.77564015511596e-05, "loss": 0.5992, "step": 46070 }, { "epoch": 2.78, "learning_rate": 2.7754892641045375e-05, "loss": 0.4486, "step": 46072 }, { "epoch": 2.78, "learning_rate": 2.775338373093115e-05, "loss": 0.471, "step": 46074 }, { "epoch": 2.78, "learning_rate": 2.7751874820816925e-05, "loss": 0.5559, "step": 46076 }, { "epoch": 2.78, "learning_rate": 2.77503659107027e-05, "loss": 0.3589, "step": 46078 }, { "epoch": 2.78, "learning_rate": 2.7748857000588475e-05, "loss": 0.7159, "step": 46080 }, { "epoch": 2.78, "learning_rate": 2.774734809047425e-05, "loss": 0.5227, "step": 46082 }, { "epoch": 2.78, "learning_rate": 2.7745839180360028e-05, "loss": 0.507, "step": 46084 }, { "epoch": 2.78, "learning_rate": 2.7744330270245804e-05, "loss": 0.5454, "step": 46086 }, { "epoch": 2.78, "learning_rate": 2.7742821360131577e-05, "loss": 0.3169, "step": 46088 }, { "epoch": 2.78, "learning_rate": 2.7741312450017354e-05, "loss": 0.5317, "step": 46090 }, { "epoch": 2.78, "learning_rate": 2.773980353990313e-05, "loss": 0.5106, "step": 46092 }, { "epoch": 2.78, "learning_rate": 2.7738294629788904e-05, "loss": 0.4173, "step": 46094 }, { "epoch": 2.78, "learning_rate": 2.773678571967468e-05, "loss": 0.5298, "step": 46096 }, { "epoch": 2.78, "learning_rate": 2.7735276809560457e-05, "loss": 0.4399, "step": 46098 }, { "epoch": 2.78, "learning_rate": 2.7733767899446233e-05, "loss": 0.5167, "step": 46100 }, { "epoch": 2.78, "learning_rate": 2.7732258989332006e-05, "loss": 0.6381, "step": 46102 }, { "epoch": 2.78, "learning_rate": 2.7730750079217783e-05, "loss": 0.684, "step": 46104 }, { "epoch": 2.78, "learning_rate": 2.772924116910356e-05, "loss": 0.6378, "step": 46106 }, { "epoch": 2.78, "learning_rate": 2.7727732258989332e-05, "loss": 0.509, "step": 46108 }, { "epoch": 2.78, "learning_rate": 2.772622334887511e-05, "loss": 0.5663, "step": 46110 }, { "epoch": 2.78, "learning_rate": 2.7724714438760885e-05, "loss": 0.5187, "step": 46112 }, { "epoch": 2.78, "learning_rate": 2.7723205528646662e-05, "loss": 0.7184, "step": 46114 }, { "epoch": 2.78, "learning_rate": 2.7721696618532435e-05, "loss": 0.7582, "step": 46116 }, { "epoch": 2.78, "learning_rate": 2.772018770841821e-05, "loss": 0.7595, "step": 46118 }, { "epoch": 2.78, "learning_rate": 2.7718678798303988e-05, "loss": 0.5175, "step": 46120 }, { "epoch": 2.78, "learning_rate": 2.7717169888189765e-05, "loss": 0.5783, "step": 46122 }, { "epoch": 2.78, "learning_rate": 2.7715660978075538e-05, "loss": 0.5619, "step": 46124 }, { "epoch": 2.78, "learning_rate": 2.7714152067961314e-05, "loss": 0.6062, "step": 46126 }, { "epoch": 2.78, "learning_rate": 2.771264315784709e-05, "loss": 0.521, "step": 46128 }, { "epoch": 2.78, "learning_rate": 2.771113424773286e-05, "loss": 0.7259, "step": 46130 }, { "epoch": 2.78, "learning_rate": 2.770962533761864e-05, "loss": 0.6449, "step": 46132 }, { "epoch": 2.78, "learning_rate": 2.7708116427504417e-05, "loss": 0.5936, "step": 46134 }, { "epoch": 2.78, "learning_rate": 2.7706607517390193e-05, "loss": 0.5915, "step": 46136 }, { "epoch": 2.78, "learning_rate": 2.7705098607275963e-05, "loss": 0.5753, "step": 46138 }, { "epoch": 2.78, "learning_rate": 2.7703589697161743e-05, "loss": 0.5962, "step": 46140 }, { "epoch": 2.79, "learning_rate": 2.770208078704752e-05, "loss": 0.5001, "step": 46142 }, { "epoch": 2.79, "learning_rate": 2.770057187693329e-05, "loss": 0.5507, "step": 46144 }, { "epoch": 2.79, "learning_rate": 2.7699062966819066e-05, "loss": 0.4544, "step": 46146 }, { "epoch": 2.79, "learning_rate": 2.7697554056704846e-05, "loss": 0.6269, "step": 46148 }, { "epoch": 2.79, "learning_rate": 2.7696045146590622e-05, "loss": 0.4784, "step": 46150 }, { "epoch": 2.79, "learning_rate": 2.7694536236476392e-05, "loss": 0.4412, "step": 46152 }, { "epoch": 2.79, "learning_rate": 2.769302732636217e-05, "loss": 0.4426, "step": 46154 }, { "epoch": 2.79, "learning_rate": 2.769151841624795e-05, "loss": 0.5195, "step": 46156 }, { "epoch": 2.79, "learning_rate": 2.7690009506133718e-05, "loss": 0.6839, "step": 46158 }, { "epoch": 2.79, "learning_rate": 2.7688500596019495e-05, "loss": 0.7743, "step": 46160 }, { "epoch": 2.79, "learning_rate": 2.768699168590527e-05, "loss": 0.6033, "step": 46162 }, { "epoch": 2.79, "learning_rate": 2.768548277579105e-05, "loss": 0.4082, "step": 46164 }, { "epoch": 2.79, "learning_rate": 2.768397386567682e-05, "loss": 0.4701, "step": 46166 }, { "epoch": 2.79, "learning_rate": 2.7682464955562597e-05, "loss": 0.5089, "step": 46168 }, { "epoch": 2.79, "learning_rate": 2.7680956045448374e-05, "loss": 0.5026, "step": 46170 }, { "epoch": 2.79, "learning_rate": 2.7679447135334154e-05, "loss": 0.6615, "step": 46172 }, { "epoch": 2.79, "learning_rate": 2.7677938225219923e-05, "loss": 0.3751, "step": 46174 }, { "epoch": 2.79, "learning_rate": 2.76764293151057e-05, "loss": 0.7383, "step": 46176 }, { "epoch": 2.79, "learning_rate": 2.7674920404991476e-05, "loss": 0.5602, "step": 46178 }, { "epoch": 2.79, "learning_rate": 2.767341149487725e-05, "loss": 0.4768, "step": 46180 }, { "epoch": 2.79, "learning_rate": 2.7671902584763026e-05, "loss": 0.8692, "step": 46182 }, { "epoch": 2.79, "learning_rate": 2.7670393674648803e-05, "loss": 0.4822, "step": 46184 }, { "epoch": 2.79, "learning_rate": 2.766888476453458e-05, "loss": 0.4529, "step": 46186 }, { "epoch": 2.79, "learning_rate": 2.7667375854420352e-05, "loss": 0.4488, "step": 46188 }, { "epoch": 2.79, "learning_rate": 2.766586694430613e-05, "loss": 0.7697, "step": 46190 }, { "epoch": 2.79, "learning_rate": 2.7664358034191905e-05, "loss": 0.4359, "step": 46192 }, { "epoch": 2.79, "learning_rate": 2.766284912407768e-05, "loss": 0.52, "step": 46194 }, { "epoch": 2.79, "learning_rate": 2.7661340213963455e-05, "loss": 0.3891, "step": 46196 }, { "epoch": 2.79, "learning_rate": 2.765983130384923e-05, "loss": 0.6163, "step": 46198 }, { "epoch": 2.79, "learning_rate": 2.7658322393735008e-05, "loss": 0.5002, "step": 46200 }, { "epoch": 2.79, "learning_rate": 2.765681348362078e-05, "loss": 0.4805, "step": 46202 }, { "epoch": 2.79, "learning_rate": 2.7655304573506558e-05, "loss": 0.567, "step": 46204 }, { "epoch": 2.79, "learning_rate": 2.7653795663392334e-05, "loss": 0.6059, "step": 46206 }, { "epoch": 2.79, "learning_rate": 2.7652286753278107e-05, "loss": 0.6316, "step": 46208 }, { "epoch": 2.79, "learning_rate": 2.7650777843163884e-05, "loss": 0.7782, "step": 46210 }, { "epoch": 2.79, "learning_rate": 2.764926893304966e-05, "loss": 0.7445, "step": 46212 }, { "epoch": 2.79, "learning_rate": 2.7647760022935437e-05, "loss": 0.712, "step": 46214 }, { "epoch": 2.79, "learning_rate": 2.764625111282121e-05, "loss": 0.6653, "step": 46216 }, { "epoch": 2.79, "learning_rate": 2.7644742202706986e-05, "loss": 0.485, "step": 46218 }, { "epoch": 2.79, "learning_rate": 2.7643233292592763e-05, "loss": 0.5457, "step": 46220 }, { "epoch": 2.79, "learning_rate": 2.7641724382478536e-05, "loss": 0.5582, "step": 46222 }, { "epoch": 2.79, "learning_rate": 2.7640215472364313e-05, "loss": 0.5804, "step": 46224 }, { "epoch": 2.79, "learning_rate": 2.763870656225009e-05, "loss": 0.504, "step": 46226 }, { "epoch": 2.79, "learning_rate": 2.7637197652135866e-05, "loss": 0.4388, "step": 46228 }, { "epoch": 2.79, "learning_rate": 2.763568874202164e-05, "loss": 0.4969, "step": 46230 }, { "epoch": 2.79, "learning_rate": 2.7634179831907415e-05, "loss": 0.6605, "step": 46232 }, { "epoch": 2.79, "learning_rate": 2.7632670921793192e-05, "loss": 0.5772, "step": 46234 }, { "epoch": 2.79, "learning_rate": 2.7631162011678968e-05, "loss": 0.6068, "step": 46236 }, { "epoch": 2.79, "learning_rate": 2.762965310156474e-05, "loss": 0.5396, "step": 46238 }, { "epoch": 2.79, "learning_rate": 2.7628144191450518e-05, "loss": 0.4241, "step": 46240 }, { "epoch": 2.79, "learning_rate": 2.7626635281336294e-05, "loss": 0.5132, "step": 46242 }, { "epoch": 2.79, "learning_rate": 2.7625126371222064e-05, "loss": 0.5486, "step": 46244 }, { "epoch": 2.79, "learning_rate": 2.7623617461107844e-05, "loss": 0.3456, "step": 46246 }, { "epoch": 2.79, "learning_rate": 2.762210855099362e-05, "loss": 0.4528, "step": 46248 }, { "epoch": 2.79, "learning_rate": 2.7620599640879397e-05, "loss": 0.513, "step": 46250 }, { "epoch": 2.79, "learning_rate": 2.7619090730765167e-05, "loss": 0.4455, "step": 46252 }, { "epoch": 2.79, "learning_rate": 2.7617581820650947e-05, "loss": 0.3657, "step": 46254 }, { "epoch": 2.79, "learning_rate": 2.7616072910536723e-05, "loss": 0.6124, "step": 46256 }, { "epoch": 2.79, "learning_rate": 2.7614564000422493e-05, "loss": 0.6324, "step": 46258 }, { "epoch": 2.79, "learning_rate": 2.761305509030827e-05, "loss": 0.488, "step": 46260 }, { "epoch": 2.79, "learning_rate": 2.761154618019405e-05, "loss": 0.5952, "step": 46262 }, { "epoch": 2.79, "learning_rate": 2.7610037270079826e-05, "loss": 0.4585, "step": 46264 }, { "epoch": 2.79, "learning_rate": 2.7608528359965596e-05, "loss": 0.3097, "step": 46266 }, { "epoch": 2.79, "learning_rate": 2.7607019449851372e-05, "loss": 0.6328, "step": 46268 }, { "epoch": 2.79, "learning_rate": 2.7605510539737152e-05, "loss": 0.5193, "step": 46270 }, { "epoch": 2.79, "learning_rate": 2.7604001629622922e-05, "loss": 0.7254, "step": 46272 }, { "epoch": 2.79, "learning_rate": 2.7602492719508698e-05, "loss": 0.4887, "step": 46274 }, { "epoch": 2.79, "learning_rate": 2.7600983809394475e-05, "loss": 0.5717, "step": 46276 }, { "epoch": 2.79, "learning_rate": 2.7599474899280255e-05, "loss": 0.5424, "step": 46278 }, { "epoch": 2.79, "learning_rate": 2.7597965989166024e-05, "loss": 0.6651, "step": 46280 }, { "epoch": 2.79, "learning_rate": 2.75964570790518e-05, "loss": 0.7414, "step": 46282 }, { "epoch": 2.79, "learning_rate": 2.7594948168937577e-05, "loss": 0.7504, "step": 46284 }, { "epoch": 2.79, "learning_rate": 2.7593439258823357e-05, "loss": 0.3817, "step": 46286 }, { "epoch": 2.79, "learning_rate": 2.7591930348709127e-05, "loss": 0.6847, "step": 46288 }, { "epoch": 2.79, "learning_rate": 2.7590421438594904e-05, "loss": 0.414, "step": 46290 }, { "epoch": 2.79, "learning_rate": 2.758891252848068e-05, "loss": 0.6181, "step": 46292 }, { "epoch": 2.79, "learning_rate": 2.7587403618366453e-05, "loss": 0.5422, "step": 46294 }, { "epoch": 2.79, "learning_rate": 2.758589470825223e-05, "loss": 0.4677, "step": 46296 }, { "epoch": 2.79, "learning_rate": 2.7584385798138006e-05, "loss": 0.5742, "step": 46298 }, { "epoch": 2.79, "learning_rate": 2.7582876888023783e-05, "loss": 0.5069, "step": 46300 }, { "epoch": 2.79, "learning_rate": 2.7581367977909556e-05, "loss": 0.4973, "step": 46302 }, { "epoch": 2.79, "learning_rate": 2.7579859067795332e-05, "loss": 0.309, "step": 46304 }, { "epoch": 2.79, "learning_rate": 2.757835015768111e-05, "loss": 0.5813, "step": 46306 }, { "epoch": 2.8, "learning_rate": 2.7576841247566882e-05, "loss": 0.6088, "step": 46308 }, { "epoch": 2.8, "learning_rate": 2.757533233745266e-05, "loss": 0.6423, "step": 46310 }, { "epoch": 2.8, "learning_rate": 2.7573823427338435e-05, "loss": 0.4644, "step": 46312 }, { "epoch": 2.8, "learning_rate": 2.757231451722421e-05, "loss": 0.58, "step": 46314 }, { "epoch": 2.8, "learning_rate": 2.7570805607109985e-05, "loss": 0.51, "step": 46316 }, { "epoch": 2.8, "learning_rate": 2.756929669699576e-05, "loss": 0.661, "step": 46318 }, { "epoch": 2.8, "learning_rate": 2.7567787786881538e-05, "loss": 0.6863, "step": 46320 }, { "epoch": 2.8, "learning_rate": 2.756627887676731e-05, "loss": 0.4917, "step": 46322 }, { "epoch": 2.8, "learning_rate": 2.7564769966653087e-05, "loss": 0.6555, "step": 46324 }, { "epoch": 2.8, "learning_rate": 2.7563261056538864e-05, "loss": 0.6758, "step": 46326 }, { "epoch": 2.8, "learning_rate": 2.756175214642464e-05, "loss": 0.5934, "step": 46328 }, { "epoch": 2.8, "learning_rate": 2.7560243236310414e-05, "loss": 0.3925, "step": 46330 }, { "epoch": 2.8, "learning_rate": 2.755873432619619e-05, "loss": 0.5353, "step": 46332 }, { "epoch": 2.8, "learning_rate": 2.7557225416081967e-05, "loss": 0.4624, "step": 46334 }, { "epoch": 2.8, "learning_rate": 2.7555716505967743e-05, "loss": 0.4316, "step": 46336 }, { "epoch": 2.8, "learning_rate": 2.7554207595853516e-05, "loss": 0.4286, "step": 46338 }, { "epoch": 2.8, "learning_rate": 2.7552698685739293e-05, "loss": 0.5787, "step": 46340 }, { "epoch": 2.8, "learning_rate": 2.755118977562507e-05, "loss": 0.8315, "step": 46342 }, { "epoch": 2.8, "learning_rate": 2.7549680865510842e-05, "loss": 0.505, "step": 46344 }, { "epoch": 2.8, "learning_rate": 2.754817195539662e-05, "loss": 0.6059, "step": 46346 }, { "epoch": 2.8, "learning_rate": 2.7546663045282395e-05, "loss": 0.5469, "step": 46348 }, { "epoch": 2.8, "learning_rate": 2.7545154135168172e-05, "loss": 0.6492, "step": 46350 }, { "epoch": 2.8, "learning_rate": 2.7543645225053945e-05, "loss": 0.511, "step": 46352 }, { "epoch": 2.8, "learning_rate": 2.754213631493972e-05, "loss": 0.4943, "step": 46354 }, { "epoch": 2.8, "learning_rate": 2.7540627404825498e-05, "loss": 0.563, "step": 46356 }, { "epoch": 2.8, "learning_rate": 2.7539118494711268e-05, "loss": 0.6507, "step": 46358 }, { "epoch": 2.8, "learning_rate": 2.7537609584597048e-05, "loss": 0.4834, "step": 46360 }, { "epoch": 2.8, "learning_rate": 2.7536100674482824e-05, "loss": 0.8026, "step": 46362 }, { "epoch": 2.8, "learning_rate": 2.75345917643686e-05, "loss": 0.5485, "step": 46364 }, { "epoch": 2.8, "learning_rate": 2.753308285425437e-05, "loss": 0.2818, "step": 46366 }, { "epoch": 2.8, "learning_rate": 2.753157394414015e-05, "loss": 0.5091, "step": 46368 }, { "epoch": 2.8, "learning_rate": 2.7530065034025927e-05, "loss": 0.4467, "step": 46370 }, { "epoch": 2.8, "learning_rate": 2.7528556123911697e-05, "loss": 0.4179, "step": 46372 }, { "epoch": 2.8, "learning_rate": 2.7527047213797473e-05, "loss": 0.5363, "step": 46374 }, { "epoch": 2.8, "learning_rate": 2.7525538303683253e-05, "loss": 0.4984, "step": 46376 }, { "epoch": 2.8, "learning_rate": 2.752402939356903e-05, "loss": 0.3907, "step": 46378 }, { "epoch": 2.8, "learning_rate": 2.75225204834548e-05, "loss": 0.5173, "step": 46380 }, { "epoch": 2.8, "learning_rate": 2.7521011573340576e-05, "loss": 0.4781, "step": 46382 }, { "epoch": 2.8, "learning_rate": 2.7519502663226356e-05, "loss": 0.4923, "step": 46384 }, { "epoch": 2.8, "learning_rate": 2.7517993753112125e-05, "loss": 0.4833, "step": 46386 }, { "epoch": 2.8, "learning_rate": 2.7516484842997902e-05, "loss": 0.7141, "step": 46388 }, { "epoch": 2.8, "learning_rate": 2.751497593288368e-05, "loss": 0.4037, "step": 46390 }, { "epoch": 2.8, "learning_rate": 2.751346702276946e-05, "loss": 0.4185, "step": 46392 }, { "epoch": 2.8, "learning_rate": 2.7511958112655228e-05, "loss": 0.5197, "step": 46394 }, { "epoch": 2.8, "learning_rate": 2.7510449202541005e-05, "loss": 0.4724, "step": 46396 }, { "epoch": 2.8, "learning_rate": 2.750894029242678e-05, "loss": 0.4396, "step": 46398 }, { "epoch": 2.8, "learning_rate": 2.750743138231256e-05, "loss": 0.5681, "step": 46400 }, { "epoch": 2.8, "learning_rate": 2.750592247219833e-05, "loss": 0.4013, "step": 46402 }, { "epoch": 2.8, "learning_rate": 2.7504413562084107e-05, "loss": 0.8816, "step": 46404 }, { "epoch": 2.8, "learning_rate": 2.7502904651969884e-05, "loss": 0.6062, "step": 46406 }, { "epoch": 2.8, "learning_rate": 2.7501395741855657e-05, "loss": 0.4423, "step": 46408 }, { "epoch": 2.8, "learning_rate": 2.7499886831741433e-05, "loss": 0.6842, "step": 46410 }, { "epoch": 2.8, "learning_rate": 2.749837792162721e-05, "loss": 0.619, "step": 46412 }, { "epoch": 2.8, "learning_rate": 2.7496869011512986e-05, "loss": 0.324, "step": 46414 }, { "epoch": 2.8, "learning_rate": 2.749536010139876e-05, "loss": 0.5497, "step": 46416 }, { "epoch": 2.8, "learning_rate": 2.7493851191284536e-05, "loss": 0.5659, "step": 46418 }, { "epoch": 2.8, "learning_rate": 2.7492342281170313e-05, "loss": 0.4234, "step": 46420 }, { "epoch": 2.8, "learning_rate": 2.7490833371056086e-05, "loss": 0.4408, "step": 46422 }, { "epoch": 2.8, "learning_rate": 2.7489324460941862e-05, "loss": 0.5066, "step": 46424 }, { "epoch": 2.8, "learning_rate": 2.748781555082764e-05, "loss": 0.4189, "step": 46426 }, { "epoch": 2.8, "learning_rate": 2.7486306640713415e-05, "loss": 0.7273, "step": 46428 }, { "epoch": 2.8, "learning_rate": 2.748479773059919e-05, "loss": 0.5671, "step": 46430 }, { "epoch": 2.8, "learning_rate": 2.7483288820484965e-05, "loss": 0.5206, "step": 46432 }, { "epoch": 2.8, "learning_rate": 2.748177991037074e-05, "loss": 0.4249, "step": 46434 }, { "epoch": 2.8, "learning_rate": 2.7480271000256515e-05, "loss": 0.5762, "step": 46436 }, { "epoch": 2.8, "learning_rate": 2.747876209014229e-05, "loss": 0.3629, "step": 46438 }, { "epoch": 2.8, "learning_rate": 2.7477253180028068e-05, "loss": 0.5908, "step": 46440 }, { "epoch": 2.8, "learning_rate": 2.7475744269913844e-05, "loss": 0.5857, "step": 46442 }, { "epoch": 2.8, "learning_rate": 2.7474235359799617e-05, "loss": 0.516, "step": 46444 }, { "epoch": 2.8, "learning_rate": 2.7472726449685394e-05, "loss": 0.5222, "step": 46446 }, { "epoch": 2.8, "learning_rate": 2.747121753957117e-05, "loss": 0.603, "step": 46448 }, { "epoch": 2.8, "learning_rate": 2.7469708629456947e-05, "loss": 0.342, "step": 46450 }, { "epoch": 2.8, "learning_rate": 2.746819971934272e-05, "loss": 0.4171, "step": 46452 }, { "epoch": 2.8, "learning_rate": 2.7466690809228496e-05, "loss": 0.7594, "step": 46454 }, { "epoch": 2.8, "learning_rate": 2.7465181899114273e-05, "loss": 0.459, "step": 46456 }, { "epoch": 2.8, "learning_rate": 2.7463672989000046e-05, "loss": 0.5419, "step": 46458 }, { "epoch": 2.8, "learning_rate": 2.7462164078885823e-05, "loss": 0.5539, "step": 46460 }, { "epoch": 2.8, "learning_rate": 2.74606551687716e-05, "loss": 0.3667, "step": 46462 }, { "epoch": 2.8, "learning_rate": 2.7459146258657376e-05, "loss": 0.4857, "step": 46464 }, { "epoch": 2.8, "learning_rate": 2.745763734854315e-05, "loss": 0.6231, "step": 46466 }, { "epoch": 2.8, "learning_rate": 2.7456128438428925e-05, "loss": 0.6201, "step": 46468 }, { "epoch": 2.8, "learning_rate": 2.7454619528314702e-05, "loss": 0.448, "step": 46470 }, { "epoch": 2.8, "learning_rate": 2.745311061820047e-05, "loss": 0.6412, "step": 46472 }, { "epoch": 2.81, "learning_rate": 2.745160170808625e-05, "loss": 0.3357, "step": 46474 }, { "epoch": 2.81, "learning_rate": 2.7450092797972028e-05, "loss": 0.4694, "step": 46476 }, { "epoch": 2.81, "learning_rate": 2.7448583887857804e-05, "loss": 0.5413, "step": 46478 }, { "epoch": 2.81, "learning_rate": 2.7447074977743574e-05, "loss": 0.6644, "step": 46480 }, { "epoch": 2.81, "learning_rate": 2.7445566067629354e-05, "loss": 0.3556, "step": 46482 }, { "epoch": 2.81, "learning_rate": 2.744405715751513e-05, "loss": 0.4451, "step": 46484 }, { "epoch": 2.81, "learning_rate": 2.74425482474009e-05, "loss": 0.3931, "step": 46486 }, { "epoch": 2.81, "learning_rate": 2.7441039337286677e-05, "loss": 0.4611, "step": 46488 }, { "epoch": 2.81, "learning_rate": 2.7439530427172457e-05, "loss": 0.4245, "step": 46490 }, { "epoch": 2.81, "learning_rate": 2.7438021517058233e-05, "loss": 0.5563, "step": 46492 }, { "epoch": 2.81, "learning_rate": 2.7436512606944003e-05, "loss": 0.6166, "step": 46494 }, { "epoch": 2.81, "learning_rate": 2.743500369682978e-05, "loss": 0.6827, "step": 46496 }, { "epoch": 2.81, "learning_rate": 2.743349478671556e-05, "loss": 0.5808, "step": 46498 }, { "epoch": 2.81, "learning_rate": 2.7431985876601336e-05, "loss": 0.3667, "step": 46500 }, { "epoch": 2.81, "learning_rate": 2.7430476966487106e-05, "loss": 0.6393, "step": 46502 }, { "epoch": 2.81, "learning_rate": 2.7428968056372882e-05, "loss": 0.6296, "step": 46504 }, { "epoch": 2.81, "learning_rate": 2.7427459146258662e-05, "loss": 0.4343, "step": 46506 }, { "epoch": 2.81, "learning_rate": 2.7425950236144432e-05, "loss": 0.5857, "step": 46508 }, { "epoch": 2.81, "learning_rate": 2.742444132603021e-05, "loss": 0.3744, "step": 46510 }, { "epoch": 2.81, "learning_rate": 2.7422932415915985e-05, "loss": 0.7553, "step": 46512 }, { "epoch": 2.81, "learning_rate": 2.7421423505801765e-05, "loss": 0.4771, "step": 46514 }, { "epoch": 2.81, "learning_rate": 2.7419914595687534e-05, "loss": 0.608, "step": 46516 }, { "epoch": 2.81, "learning_rate": 2.741840568557331e-05, "loss": 0.6589, "step": 46518 }, { "epoch": 2.81, "learning_rate": 2.7416896775459087e-05, "loss": 0.4813, "step": 46520 }, { "epoch": 2.81, "learning_rate": 2.741538786534486e-05, "loss": 0.6308, "step": 46522 }, { "epoch": 2.81, "learning_rate": 2.7413878955230637e-05, "loss": 0.6707, "step": 46524 }, { "epoch": 2.81, "learning_rate": 2.7412370045116414e-05, "loss": 0.2967, "step": 46526 }, { "epoch": 2.81, "learning_rate": 2.741086113500219e-05, "loss": 0.47, "step": 46528 }, { "epoch": 2.81, "learning_rate": 2.7409352224887963e-05, "loss": 0.6336, "step": 46530 }, { "epoch": 2.81, "learning_rate": 2.740784331477374e-05, "loss": 0.6646, "step": 46532 }, { "epoch": 2.81, "learning_rate": 2.7406334404659516e-05, "loss": 0.6419, "step": 46534 }, { "epoch": 2.81, "learning_rate": 2.740482549454529e-05, "loss": 0.754, "step": 46536 }, { "epoch": 2.81, "learning_rate": 2.7403316584431066e-05, "loss": 0.5762, "step": 46538 }, { "epoch": 2.81, "learning_rate": 2.7401807674316842e-05, "loss": 0.5364, "step": 46540 }, { "epoch": 2.81, "learning_rate": 2.740029876420262e-05, "loss": 0.5274, "step": 46542 }, { "epoch": 2.81, "learning_rate": 2.7398789854088392e-05, "loss": 0.6543, "step": 46544 }, { "epoch": 2.81, "learning_rate": 2.739728094397417e-05, "loss": 0.5102, "step": 46546 }, { "epoch": 2.81, "learning_rate": 2.7395772033859945e-05, "loss": 0.615, "step": 46548 }, { "epoch": 2.81, "learning_rate": 2.7394263123745718e-05, "loss": 0.5089, "step": 46550 }, { "epoch": 2.81, "learning_rate": 2.7392754213631495e-05, "loss": 0.6547, "step": 46552 }, { "epoch": 2.81, "learning_rate": 2.739124530351727e-05, "loss": 0.5084, "step": 46554 }, { "epoch": 2.81, "learning_rate": 2.7389736393403048e-05, "loss": 0.4919, "step": 46556 }, { "epoch": 2.81, "learning_rate": 2.738822748328882e-05, "loss": 0.3719, "step": 46558 }, { "epoch": 2.81, "learning_rate": 2.7386718573174597e-05, "loss": 0.738, "step": 46560 }, { "epoch": 2.81, "learning_rate": 2.7385209663060374e-05, "loss": 0.4283, "step": 46562 }, { "epoch": 2.81, "learning_rate": 2.738370075294615e-05, "loss": 0.4754, "step": 46564 }, { "epoch": 2.81, "learning_rate": 2.7382191842831924e-05, "loss": 0.5441, "step": 46566 }, { "epoch": 2.81, "learning_rate": 2.73806829327177e-05, "loss": 0.6379, "step": 46568 }, { "epoch": 2.81, "learning_rate": 2.7379174022603477e-05, "loss": 0.4948, "step": 46570 }, { "epoch": 2.81, "learning_rate": 2.7377665112489246e-05, "loss": 0.4771, "step": 46572 }, { "epoch": 2.81, "learning_rate": 2.7376156202375026e-05, "loss": 0.4398, "step": 46574 }, { "epoch": 2.81, "learning_rate": 2.7374647292260803e-05, "loss": 0.4777, "step": 46576 }, { "epoch": 2.81, "learning_rate": 2.737313838214658e-05, "loss": 0.4527, "step": 46578 }, { "epoch": 2.81, "learning_rate": 2.7371629472032352e-05, "loss": 0.4933, "step": 46580 }, { "epoch": 2.81, "learning_rate": 2.737012056191813e-05, "loss": 0.6464, "step": 46582 }, { "epoch": 2.81, "learning_rate": 2.7368611651803905e-05, "loss": 0.4974, "step": 46584 }, { "epoch": 2.81, "learning_rate": 2.7367102741689675e-05, "loss": 0.6926, "step": 46586 }, { "epoch": 2.81, "learning_rate": 2.7365593831575455e-05, "loss": 0.7247, "step": 46588 }, { "epoch": 2.81, "learning_rate": 2.736408492146123e-05, "loss": 0.4821, "step": 46590 }, { "epoch": 2.81, "learning_rate": 2.7362576011347008e-05, "loss": 0.5487, "step": 46592 }, { "epoch": 2.81, "learning_rate": 2.7361067101232778e-05, "loss": 0.6007, "step": 46594 }, { "epoch": 2.81, "learning_rate": 2.7359558191118558e-05, "loss": 0.6654, "step": 46596 }, { "epoch": 2.81, "learning_rate": 2.7358049281004334e-05, "loss": 0.6953, "step": 46598 }, { "epoch": 2.81, "learning_rate": 2.7356540370890104e-05, "loss": 0.5062, "step": 46600 }, { "epoch": 2.81, "learning_rate": 2.735503146077588e-05, "loss": 0.6082, "step": 46602 }, { "epoch": 2.81, "learning_rate": 2.735352255066166e-05, "loss": 0.5452, "step": 46604 }, { "epoch": 2.81, "learning_rate": 2.7352013640547437e-05, "loss": 0.6141, "step": 46606 }, { "epoch": 2.81, "learning_rate": 2.7350504730433207e-05, "loss": 0.6292, "step": 46608 }, { "epoch": 2.81, "learning_rate": 2.7348995820318983e-05, "loss": 0.5161, "step": 46610 }, { "epoch": 2.81, "learning_rate": 2.7347486910204763e-05, "loss": 0.5467, "step": 46612 }, { "epoch": 2.81, "learning_rate": 2.734597800009054e-05, "loss": 0.575, "step": 46614 }, { "epoch": 2.81, "learning_rate": 2.734446908997631e-05, "loss": 0.516, "step": 46616 }, { "epoch": 2.81, "learning_rate": 2.7342960179862086e-05, "loss": 0.548, "step": 46618 }, { "epoch": 2.81, "learning_rate": 2.7341451269747866e-05, "loss": 0.4927, "step": 46620 }, { "epoch": 2.81, "learning_rate": 2.7339942359633635e-05, "loss": 0.7838, "step": 46622 }, { "epoch": 2.81, "learning_rate": 2.7338433449519412e-05, "loss": 0.3073, "step": 46624 }, { "epoch": 2.81, "learning_rate": 2.733692453940519e-05, "loss": 0.6226, "step": 46626 }, { "epoch": 2.81, "learning_rate": 2.733541562929097e-05, "loss": 0.3577, "step": 46628 }, { "epoch": 2.81, "learning_rate": 2.7333906719176738e-05, "loss": 0.7683, "step": 46630 }, { "epoch": 2.81, "learning_rate": 2.7332397809062515e-05, "loss": 0.6432, "step": 46632 }, { "epoch": 2.81, "learning_rate": 2.733088889894829e-05, "loss": 0.5014, "step": 46634 }, { "epoch": 2.81, "learning_rate": 2.7329379988834064e-05, "loss": 0.4892, "step": 46636 }, { "epoch": 2.81, "learning_rate": 2.732787107871984e-05, "loss": 0.6577, "step": 46638 }, { "epoch": 2.82, "learning_rate": 2.7326362168605617e-05, "loss": 0.4731, "step": 46640 }, { "epoch": 2.82, "learning_rate": 2.7324853258491394e-05, "loss": 0.7768, "step": 46642 }, { "epoch": 2.82, "learning_rate": 2.7323344348377167e-05, "loss": 0.5548, "step": 46644 }, { "epoch": 2.82, "learning_rate": 2.7321835438262943e-05, "loss": 0.4751, "step": 46646 }, { "epoch": 2.82, "learning_rate": 2.732032652814872e-05, "loss": 0.3524, "step": 46648 }, { "epoch": 2.82, "learning_rate": 2.7318817618034493e-05, "loss": 0.6346, "step": 46650 }, { "epoch": 2.82, "learning_rate": 2.731730870792027e-05, "loss": 0.3935, "step": 46652 }, { "epoch": 2.82, "learning_rate": 2.7315799797806046e-05, "loss": 0.5593, "step": 46654 }, { "epoch": 2.82, "learning_rate": 2.7314290887691823e-05, "loss": 0.4156, "step": 46656 }, { "epoch": 2.82, "learning_rate": 2.7312781977577596e-05, "loss": 0.6295, "step": 46658 }, { "epoch": 2.82, "learning_rate": 2.7311273067463372e-05, "loss": 0.5944, "step": 46660 }, { "epoch": 2.82, "learning_rate": 2.730976415734915e-05, "loss": 0.5229, "step": 46662 }, { "epoch": 2.82, "learning_rate": 2.7308255247234925e-05, "loss": 0.6967, "step": 46664 }, { "epoch": 2.82, "learning_rate": 2.73067463371207e-05, "loss": 0.6285, "step": 46666 }, { "epoch": 2.82, "learning_rate": 2.7305237427006475e-05, "loss": 0.6452, "step": 46668 }, { "epoch": 2.82, "learning_rate": 2.730372851689225e-05, "loss": 0.4911, "step": 46670 }, { "epoch": 2.82, "learning_rate": 2.7302219606778025e-05, "loss": 0.6316, "step": 46672 }, { "epoch": 2.82, "learning_rate": 2.73007106966638e-05, "loss": 0.4177, "step": 46674 }, { "epoch": 2.82, "learning_rate": 2.7299201786549578e-05, "loss": 0.4838, "step": 46676 }, { "epoch": 2.82, "learning_rate": 2.7297692876435354e-05, "loss": 0.6397, "step": 46678 }, { "epoch": 2.82, "learning_rate": 2.7296183966321127e-05, "loss": 0.3873, "step": 46680 }, { "epoch": 2.82, "learning_rate": 2.7294675056206904e-05, "loss": 0.4539, "step": 46682 }, { "epoch": 2.82, "learning_rate": 2.729316614609268e-05, "loss": 0.6393, "step": 46684 }, { "epoch": 2.82, "learning_rate": 2.729165723597845e-05, "loss": 0.5877, "step": 46686 }, { "epoch": 2.82, "learning_rate": 2.729014832586423e-05, "loss": 0.4783, "step": 46688 }, { "epoch": 2.82, "learning_rate": 2.7288639415750006e-05, "loss": 0.7899, "step": 46690 }, { "epoch": 2.82, "learning_rate": 2.7287130505635783e-05, "loss": 0.7239, "step": 46692 }, { "epoch": 2.82, "learning_rate": 2.7285621595521553e-05, "loss": 0.576, "step": 46694 }, { "epoch": 2.82, "learning_rate": 2.7284112685407333e-05, "loss": 0.7608, "step": 46696 }, { "epoch": 2.82, "learning_rate": 2.728260377529311e-05, "loss": 0.7882, "step": 46698 }, { "epoch": 2.82, "learning_rate": 2.728109486517888e-05, "loss": 0.4632, "step": 46700 }, { "epoch": 2.82, "learning_rate": 2.727958595506466e-05, "loss": 0.5077, "step": 46702 }, { "epoch": 2.82, "learning_rate": 2.7278077044950435e-05, "loss": 0.6347, "step": 46704 }, { "epoch": 2.82, "learning_rate": 2.7276568134836212e-05, "loss": 0.5321, "step": 46706 }, { "epoch": 2.82, "learning_rate": 2.727505922472198e-05, "loss": 0.6095, "step": 46708 }, { "epoch": 2.82, "learning_rate": 2.727355031460776e-05, "loss": 0.5425, "step": 46710 }, { "epoch": 2.82, "learning_rate": 2.7272041404493538e-05, "loss": 0.4651, "step": 46712 }, { "epoch": 2.82, "learning_rate": 2.7270532494379314e-05, "loss": 0.6781, "step": 46714 }, { "epoch": 2.82, "learning_rate": 2.7269023584265084e-05, "loss": 0.5699, "step": 46716 }, { "epoch": 2.82, "learning_rate": 2.7267514674150864e-05, "loss": 0.4683, "step": 46718 }, { "epoch": 2.82, "learning_rate": 2.726600576403664e-05, "loss": 0.5125, "step": 46720 }, { "epoch": 2.82, "learning_rate": 2.726449685392241e-05, "loss": 0.4046, "step": 46722 }, { "epoch": 2.82, "learning_rate": 2.7262987943808187e-05, "loss": 0.3972, "step": 46724 }, { "epoch": 2.82, "learning_rate": 2.7261479033693967e-05, "loss": 0.5951, "step": 46726 }, { "epoch": 2.82, "learning_rate": 2.7259970123579743e-05, "loss": 0.4629, "step": 46728 }, { "epoch": 2.82, "learning_rate": 2.7258461213465513e-05, "loss": 0.4923, "step": 46730 }, { "epoch": 2.82, "learning_rate": 2.725695230335129e-05, "loss": 0.6282, "step": 46732 }, { "epoch": 2.82, "learning_rate": 2.725544339323707e-05, "loss": 0.4326, "step": 46734 }, { "epoch": 2.82, "learning_rate": 2.725393448312284e-05, "loss": 0.5061, "step": 46736 }, { "epoch": 2.82, "learning_rate": 2.7252425573008616e-05, "loss": 0.3818, "step": 46738 }, { "epoch": 2.82, "learning_rate": 2.7250916662894392e-05, "loss": 0.4232, "step": 46740 }, { "epoch": 2.82, "learning_rate": 2.7249407752780172e-05, "loss": 0.5166, "step": 46742 }, { "epoch": 2.82, "learning_rate": 2.7247898842665942e-05, "loss": 0.4758, "step": 46744 }, { "epoch": 2.82, "learning_rate": 2.724638993255172e-05, "loss": 0.4705, "step": 46746 }, { "epoch": 2.82, "learning_rate": 2.7245635477494608e-05, "loss": 0.7012, "step": 46748 }, { "epoch": 2.82, "learning_rate": 2.7244126567380385e-05, "loss": 0.55, "step": 46750 }, { "epoch": 2.82, "learning_rate": 2.7242617657266155e-05, "loss": 0.3763, "step": 46752 }, { "epoch": 2.82, "learning_rate": 2.7241108747151934e-05, "loss": 0.5424, "step": 46754 }, { "epoch": 2.82, "learning_rate": 2.723959983703771e-05, "loss": 0.5278, "step": 46756 }, { "epoch": 2.82, "learning_rate": 2.723809092692348e-05, "loss": 0.5036, "step": 46758 }, { "epoch": 2.82, "learning_rate": 2.723658201680926e-05, "loss": 0.5526, "step": 46760 }, { "epoch": 2.82, "learning_rate": 2.7235073106695037e-05, "loss": 0.6763, "step": 46762 }, { "epoch": 2.82, "learning_rate": 2.7233564196580814e-05, "loss": 0.6397, "step": 46764 }, { "epoch": 2.82, "learning_rate": 2.7232055286466583e-05, "loss": 0.6374, "step": 46766 }, { "epoch": 2.82, "learning_rate": 2.7230546376352363e-05, "loss": 0.4768, "step": 46768 }, { "epoch": 2.82, "learning_rate": 2.722903746623814e-05, "loss": 0.5098, "step": 46770 }, { "epoch": 2.82, "learning_rate": 2.7227528556123916e-05, "loss": 0.5493, "step": 46772 }, { "epoch": 2.82, "learning_rate": 2.7226019646009686e-05, "loss": 0.794, "step": 46774 }, { "epoch": 2.82, "learning_rate": 2.7224510735895466e-05, "loss": 0.5908, "step": 46776 }, { "epoch": 2.82, "learning_rate": 2.7223001825781242e-05, "loss": 0.4544, "step": 46778 }, { "epoch": 2.82, "learning_rate": 2.7221492915667012e-05, "loss": 0.5422, "step": 46780 }, { "epoch": 2.82, "learning_rate": 2.721998400555279e-05, "loss": 0.8297, "step": 46782 }, { "epoch": 2.82, "learning_rate": 2.721847509543857e-05, "loss": 0.4389, "step": 46784 }, { "epoch": 2.82, "learning_rate": 2.7216966185324345e-05, "loss": 0.6673, "step": 46786 }, { "epoch": 2.82, "learning_rate": 2.7215457275210115e-05, "loss": 0.556, "step": 46788 }, { "epoch": 2.82, "learning_rate": 2.721394836509589e-05, "loss": 0.4482, "step": 46790 }, { "epoch": 2.82, "learning_rate": 2.721243945498167e-05, "loss": 0.5903, "step": 46792 }, { "epoch": 2.82, "learning_rate": 2.721093054486744e-05, "loss": 0.5917, "step": 46794 }, { "epoch": 2.82, "learning_rate": 2.7209421634753218e-05, "loss": 0.8926, "step": 46796 }, { "epoch": 2.82, "learning_rate": 2.7207912724638994e-05, "loss": 0.7725, "step": 46798 }, { "epoch": 2.82, "learning_rate": 2.7206403814524774e-05, "loss": 0.4774, "step": 46800 }, { "epoch": 2.82, "learning_rate": 2.7204894904410544e-05, "loss": 0.4729, "step": 46802 }, { "epoch": 2.82, "learning_rate": 2.720338599429632e-05, "loss": 0.6764, "step": 46804 }, { "epoch": 2.83, "learning_rate": 2.7201877084182097e-05, "loss": 0.5873, "step": 46806 }, { "epoch": 2.83, "learning_rate": 2.720036817406787e-05, "loss": 0.5924, "step": 46808 }, { "epoch": 2.83, "learning_rate": 2.7198859263953646e-05, "loss": 0.5678, "step": 46810 }, { "epoch": 2.83, "learning_rate": 2.7197350353839423e-05, "loss": 0.6085, "step": 46812 }, { "epoch": 2.83, "learning_rate": 2.71958414437252e-05, "loss": 0.7736, "step": 46814 }, { "epoch": 2.83, "learning_rate": 2.7194332533610972e-05, "loss": 0.3674, "step": 46816 }, { "epoch": 2.83, "learning_rate": 2.719282362349675e-05, "loss": 0.511, "step": 46818 }, { "epoch": 2.83, "learning_rate": 2.7191314713382526e-05, "loss": 0.5232, "step": 46820 }, { "epoch": 2.83, "learning_rate": 2.71898058032683e-05, "loss": 0.599, "step": 46822 }, { "epoch": 2.83, "learning_rate": 2.7188296893154075e-05, "loss": 0.5099, "step": 46824 }, { "epoch": 2.83, "learning_rate": 2.718678798303985e-05, "loss": 0.5042, "step": 46826 }, { "epoch": 2.83, "learning_rate": 2.7185279072925628e-05, "loss": 0.5228, "step": 46828 }, { "epoch": 2.83, "learning_rate": 2.71837701628114e-05, "loss": 0.5911, "step": 46830 }, { "epoch": 2.83, "learning_rate": 2.7182261252697178e-05, "loss": 0.5568, "step": 46832 }, { "epoch": 2.83, "learning_rate": 2.7180752342582954e-05, "loss": 0.5814, "step": 46834 }, { "epoch": 2.83, "learning_rate": 2.717924343246873e-05, "loss": 0.4446, "step": 46836 }, { "epoch": 2.83, "learning_rate": 2.7177734522354504e-05, "loss": 0.4966, "step": 46838 }, { "epoch": 2.83, "learning_rate": 2.717622561224028e-05, "loss": 0.477, "step": 46840 }, { "epoch": 2.83, "learning_rate": 2.7174716702126057e-05, "loss": 0.4428, "step": 46842 }, { "epoch": 2.83, "learning_rate": 2.717320779201183e-05, "loss": 0.4616, "step": 46844 }, { "epoch": 2.83, "learning_rate": 2.7171698881897607e-05, "loss": 0.5183, "step": 46846 }, { "epoch": 2.83, "learning_rate": 2.7170189971783383e-05, "loss": 0.4054, "step": 46848 }, { "epoch": 2.83, "learning_rate": 2.716868106166916e-05, "loss": 0.5683, "step": 46850 }, { "epoch": 2.83, "learning_rate": 2.7167172151554933e-05, "loss": 0.4468, "step": 46852 }, { "epoch": 2.83, "learning_rate": 2.716566324144071e-05, "loss": 0.6474, "step": 46854 }, { "epoch": 2.83, "learning_rate": 2.7164154331326486e-05, "loss": 0.6652, "step": 46856 }, { "epoch": 2.83, "learning_rate": 2.7162645421212256e-05, "loss": 0.3447, "step": 46858 }, { "epoch": 2.83, "learning_rate": 2.7161136511098035e-05, "loss": 0.3661, "step": 46860 }, { "epoch": 2.83, "learning_rate": 2.7159627600983812e-05, "loss": 0.4667, "step": 46862 }, { "epoch": 2.83, "learning_rate": 2.715811869086959e-05, "loss": 0.5488, "step": 46864 }, { "epoch": 2.83, "learning_rate": 2.7156609780755358e-05, "loss": 0.738, "step": 46866 }, { "epoch": 2.83, "learning_rate": 2.7155100870641138e-05, "loss": 0.4647, "step": 46868 }, { "epoch": 2.83, "learning_rate": 2.7153591960526915e-05, "loss": 0.5776, "step": 46870 }, { "epoch": 2.83, "learning_rate": 2.7152083050412684e-05, "loss": 0.6894, "step": 46872 }, { "epoch": 2.83, "learning_rate": 2.7150574140298464e-05, "loss": 0.6049, "step": 46874 }, { "epoch": 2.83, "learning_rate": 2.714906523018424e-05, "loss": 0.7194, "step": 46876 }, { "epoch": 2.83, "learning_rate": 2.7147556320070017e-05, "loss": 0.6232, "step": 46878 }, { "epoch": 2.83, "learning_rate": 2.7146047409955787e-05, "loss": 0.6111, "step": 46880 }, { "epoch": 2.83, "learning_rate": 2.7144538499841567e-05, "loss": 0.6489, "step": 46882 }, { "epoch": 2.83, "learning_rate": 2.7143029589727343e-05, "loss": 0.6464, "step": 46884 }, { "epoch": 2.83, "learning_rate": 2.714152067961312e-05, "loss": 0.4559, "step": 46886 }, { "epoch": 2.83, "learning_rate": 2.714001176949889e-05, "loss": 0.6386, "step": 46888 }, { "epoch": 2.83, "learning_rate": 2.713850285938467e-05, "loss": 0.6999, "step": 46890 }, { "epoch": 2.83, "learning_rate": 2.7136993949270446e-05, "loss": 0.484, "step": 46892 }, { "epoch": 2.83, "learning_rate": 2.7135485039156216e-05, "loss": 0.3408, "step": 46894 }, { "epoch": 2.83, "learning_rate": 2.7133976129041992e-05, "loss": 0.5629, "step": 46896 }, { "epoch": 2.83, "learning_rate": 2.7132467218927772e-05, "loss": 0.4612, "step": 46898 }, { "epoch": 2.83, "learning_rate": 2.713095830881355e-05, "loss": 0.5169, "step": 46900 }, { "epoch": 2.83, "learning_rate": 2.712944939869932e-05, "loss": 0.4514, "step": 46902 }, { "epoch": 2.83, "learning_rate": 2.7127940488585095e-05, "loss": 0.3799, "step": 46904 }, { "epoch": 2.83, "learning_rate": 2.7126431578470875e-05, "loss": 0.4987, "step": 46906 }, { "epoch": 2.83, "learning_rate": 2.7124922668356645e-05, "loss": 0.6691, "step": 46908 }, { "epoch": 2.83, "learning_rate": 2.712341375824242e-05, "loss": 0.923, "step": 46910 }, { "epoch": 2.83, "learning_rate": 2.7121904848128198e-05, "loss": 0.5796, "step": 46912 }, { "epoch": 2.83, "learning_rate": 2.7120395938013978e-05, "loss": 0.7123, "step": 46914 }, { "epoch": 2.83, "learning_rate": 2.7118887027899747e-05, "loss": 0.5432, "step": 46916 }, { "epoch": 2.83, "learning_rate": 2.7117378117785524e-05, "loss": 0.5344, "step": 46918 }, { "epoch": 2.83, "learning_rate": 2.71158692076713e-05, "loss": 0.788, "step": 46920 }, { "epoch": 2.83, "learning_rate": 2.7114360297557073e-05, "loss": 0.575, "step": 46922 }, { "epoch": 2.83, "learning_rate": 2.711285138744285e-05, "loss": 0.4836, "step": 46924 }, { "epoch": 2.83, "learning_rate": 2.7111342477328627e-05, "loss": 0.584, "step": 46926 }, { "epoch": 2.83, "learning_rate": 2.7109833567214403e-05, "loss": 0.3743, "step": 46928 }, { "epoch": 2.83, "learning_rate": 2.7108324657100176e-05, "loss": 0.4894, "step": 46930 }, { "epoch": 2.83, "learning_rate": 2.7106815746985953e-05, "loss": 0.5816, "step": 46932 }, { "epoch": 2.83, "learning_rate": 2.710530683687173e-05, "loss": 0.508, "step": 46934 }, { "epoch": 2.83, "learning_rate": 2.7103797926757506e-05, "loss": 0.5532, "step": 46936 }, { "epoch": 2.83, "learning_rate": 2.710228901664328e-05, "loss": 0.5886, "step": 46938 }, { "epoch": 2.83, "learning_rate": 2.7100780106529055e-05, "loss": 0.421, "step": 46940 }, { "epoch": 2.83, "learning_rate": 2.7099271196414832e-05, "loss": 0.5716, "step": 46942 }, { "epoch": 2.83, "learning_rate": 2.7097762286300605e-05, "loss": 0.5245, "step": 46944 }, { "epoch": 2.83, "learning_rate": 2.709625337618638e-05, "loss": 0.6423, "step": 46946 }, { "epoch": 2.83, "learning_rate": 2.7094744466072158e-05, "loss": 0.5981, "step": 46948 }, { "epoch": 2.83, "learning_rate": 2.7093235555957935e-05, "loss": 0.4684, "step": 46950 }, { "epoch": 2.83, "learning_rate": 2.7091726645843708e-05, "loss": 0.544, "step": 46952 }, { "epoch": 2.83, "learning_rate": 2.7090217735729484e-05, "loss": 0.3571, "step": 46954 }, { "epoch": 2.83, "learning_rate": 2.708870882561526e-05, "loss": 0.5754, "step": 46956 }, { "epoch": 2.83, "learning_rate": 2.7087199915501034e-05, "loss": 0.5716, "step": 46958 }, { "epoch": 2.83, "learning_rate": 2.708569100538681e-05, "loss": 0.5082, "step": 46960 }, { "epoch": 2.83, "learning_rate": 2.7084182095272587e-05, "loss": 0.3997, "step": 46962 }, { "epoch": 2.83, "learning_rate": 2.7082673185158363e-05, "loss": 0.6478, "step": 46964 }, { "epoch": 2.83, "learning_rate": 2.7081164275044136e-05, "loss": 0.6565, "step": 46966 }, { "epoch": 2.83, "learning_rate": 2.7079655364929913e-05, "loss": 0.4351, "step": 46968 }, { "epoch": 2.83, "learning_rate": 2.707814645481569e-05, "loss": 0.3719, "step": 46970 }, { "epoch": 2.84, "learning_rate": 2.707663754470146e-05, "loss": 0.5605, "step": 46972 }, { "epoch": 2.84, "learning_rate": 2.707512863458724e-05, "loss": 0.4563, "step": 46974 }, { "epoch": 2.84, "learning_rate": 2.7073619724473016e-05, "loss": 0.7852, "step": 46976 }, { "epoch": 2.84, "learning_rate": 2.7072110814358792e-05, "loss": 0.4285, "step": 46978 }, { "epoch": 2.84, "learning_rate": 2.7070601904244562e-05, "loss": 0.6647, "step": 46980 }, { "epoch": 2.84, "learning_rate": 2.7069092994130342e-05, "loss": 0.4864, "step": 46982 }, { "epoch": 2.84, "learning_rate": 2.706758408401612e-05, "loss": 0.4802, "step": 46984 }, { "epoch": 2.84, "learning_rate": 2.7066075173901895e-05, "loss": 0.5039, "step": 46986 }, { "epoch": 2.84, "learning_rate": 2.7064566263787665e-05, "loss": 0.5627, "step": 46988 }, { "epoch": 2.84, "learning_rate": 2.7063057353673444e-05, "loss": 0.5776, "step": 46990 }, { "epoch": 2.84, "learning_rate": 2.706154844355922e-05, "loss": 0.4503, "step": 46992 }, { "epoch": 2.84, "learning_rate": 2.706003953344499e-05, "loss": 0.6112, "step": 46994 }, { "epoch": 2.84, "learning_rate": 2.705853062333077e-05, "loss": 0.3638, "step": 46996 }, { "epoch": 2.84, "learning_rate": 2.7057021713216547e-05, "loss": 0.5259, "step": 46998 }, { "epoch": 2.84, "learning_rate": 2.7055512803102324e-05, "loss": 0.5476, "step": 47000 }, { "epoch": 2.84, "learning_rate": 2.7054003892988093e-05, "loss": 0.6509, "step": 47002 }, { "epoch": 2.84, "learning_rate": 2.7052494982873873e-05, "loss": 0.4418, "step": 47004 }, { "epoch": 2.84, "learning_rate": 2.705098607275965e-05, "loss": 0.4823, "step": 47006 }, { "epoch": 2.84, "learning_rate": 2.704947716264542e-05, "loss": 0.7877, "step": 47008 }, { "epoch": 2.84, "learning_rate": 2.7047968252531196e-05, "loss": 0.4699, "step": 47010 }, { "epoch": 2.84, "learning_rate": 2.7046459342416976e-05, "loss": 0.5412, "step": 47012 }, { "epoch": 2.84, "learning_rate": 2.7044950432302752e-05, "loss": 0.608, "step": 47014 }, { "epoch": 2.84, "learning_rate": 2.7043441522188522e-05, "loss": 0.6098, "step": 47016 }, { "epoch": 2.84, "learning_rate": 2.70419326120743e-05, "loss": 0.641, "step": 47018 }, { "epoch": 2.84, "learning_rate": 2.704042370196008e-05, "loss": 0.5527, "step": 47020 }, { "epoch": 2.84, "learning_rate": 2.703891479184585e-05, "loss": 0.6149, "step": 47022 }, { "epoch": 2.84, "learning_rate": 2.7037405881731625e-05, "loss": 0.452, "step": 47024 }, { "epoch": 2.84, "learning_rate": 2.70358969716174e-05, "loss": 0.5319, "step": 47026 }, { "epoch": 2.84, "learning_rate": 2.703438806150318e-05, "loss": 0.578, "step": 47028 }, { "epoch": 2.84, "learning_rate": 2.703287915138895e-05, "loss": 0.3347, "step": 47030 }, { "epoch": 2.84, "learning_rate": 2.7031370241274728e-05, "loss": 0.457, "step": 47032 }, { "epoch": 2.84, "learning_rate": 2.7029861331160504e-05, "loss": 0.5853, "step": 47034 }, { "epoch": 2.84, "learning_rate": 2.7028352421046277e-05, "loss": 0.4371, "step": 47036 }, { "epoch": 2.84, "learning_rate": 2.7026843510932054e-05, "loss": 0.271, "step": 47038 }, { "epoch": 2.84, "learning_rate": 2.702533460081783e-05, "loss": 0.5441, "step": 47040 }, { "epoch": 2.84, "learning_rate": 2.7023825690703607e-05, "loss": 0.4032, "step": 47042 }, { "epoch": 2.84, "learning_rate": 2.702231678058938e-05, "loss": 0.457, "step": 47044 }, { "epoch": 2.84, "learning_rate": 2.7020807870475156e-05, "loss": 0.433, "step": 47046 }, { "epoch": 2.84, "learning_rate": 2.7019298960360933e-05, "loss": 0.498, "step": 47048 }, { "epoch": 2.84, "learning_rate": 2.701779005024671e-05, "loss": 0.4802, "step": 47050 }, { "epoch": 2.84, "learning_rate": 2.7016281140132483e-05, "loss": 0.5539, "step": 47052 }, { "epoch": 2.84, "learning_rate": 2.701477223001826e-05, "loss": 0.3993, "step": 47054 }, { "epoch": 2.84, "learning_rate": 2.7013263319904036e-05, "loss": 0.5207, "step": 47056 }, { "epoch": 2.84, "learning_rate": 2.701175440978981e-05, "loss": 0.5463, "step": 47058 }, { "epoch": 2.84, "learning_rate": 2.7010245499675585e-05, "loss": 0.4249, "step": 47060 }, { "epoch": 2.84, "learning_rate": 2.700873658956136e-05, "loss": 0.4655, "step": 47062 }, { "epoch": 2.84, "learning_rate": 2.7007227679447138e-05, "loss": 0.6782, "step": 47064 }, { "epoch": 2.84, "learning_rate": 2.700571876933291e-05, "loss": 0.3263, "step": 47066 }, { "epoch": 2.84, "learning_rate": 2.7004209859218688e-05, "loss": 0.4545, "step": 47068 }, { "epoch": 2.84, "learning_rate": 2.7002700949104464e-05, "loss": 0.8107, "step": 47070 }, { "epoch": 2.84, "learning_rate": 2.7001192038990237e-05, "loss": 0.4652, "step": 47072 }, { "epoch": 2.84, "learning_rate": 2.6999683128876014e-05, "loss": 0.5456, "step": 47074 }, { "epoch": 2.84, "learning_rate": 2.699817421876179e-05, "loss": 0.4716, "step": 47076 }, { "epoch": 2.84, "learning_rate": 2.6996665308647567e-05, "loss": 0.5163, "step": 47078 }, { "epoch": 2.84, "learning_rate": 2.699515639853334e-05, "loss": 0.7291, "step": 47080 }, { "epoch": 2.84, "learning_rate": 2.6993647488419117e-05, "loss": 0.6081, "step": 47082 }, { "epoch": 2.84, "learning_rate": 2.6992138578304893e-05, "loss": 0.5522, "step": 47084 }, { "epoch": 2.84, "learning_rate": 2.6990629668190663e-05, "loss": 0.4881, "step": 47086 }, { "epoch": 2.84, "learning_rate": 2.6989120758076443e-05, "loss": 0.5866, "step": 47088 }, { "epoch": 2.84, "learning_rate": 2.698761184796222e-05, "loss": 0.5345, "step": 47090 }, { "epoch": 2.84, "learning_rate": 2.6986102937847996e-05, "loss": 0.4546, "step": 47092 }, { "epoch": 2.84, "learning_rate": 2.6984594027733766e-05, "loss": 0.397, "step": 47094 }, { "epoch": 2.84, "learning_rate": 2.6983085117619545e-05, "loss": 0.5057, "step": 47096 }, { "epoch": 2.84, "learning_rate": 2.6981576207505322e-05, "loss": 0.528, "step": 47098 }, { "epoch": 2.84, "learning_rate": 2.69800672973911e-05, "loss": 0.5698, "step": 47100 }, { "epoch": 2.84, "learning_rate": 2.6978558387276868e-05, "loss": 0.4166, "step": 47102 }, { "epoch": 2.84, "learning_rate": 2.6977049477162648e-05, "loss": 0.4938, "step": 47104 }, { "epoch": 2.84, "learning_rate": 2.6975540567048425e-05, "loss": 0.8166, "step": 47106 }, { "epoch": 2.84, "learning_rate": 2.6974031656934194e-05, "loss": 0.3669, "step": 47108 }, { "epoch": 2.84, "learning_rate": 2.697252274681997e-05, "loss": 0.7095, "step": 47110 }, { "epoch": 2.84, "learning_rate": 2.697101383670575e-05, "loss": 0.5227, "step": 47112 }, { "epoch": 2.84, "learning_rate": 2.6969504926591527e-05, "loss": 0.4795, "step": 47114 }, { "epoch": 2.84, "learning_rate": 2.6967996016477297e-05, "loss": 0.5821, "step": 47116 }, { "epoch": 2.84, "learning_rate": 2.6966487106363077e-05, "loss": 0.4478, "step": 47118 }, { "epoch": 2.84, "learning_rate": 2.6964978196248853e-05, "loss": 0.4943, "step": 47120 }, { "epoch": 2.84, "learning_rate": 2.6963469286134623e-05, "loss": 0.4648, "step": 47122 }, { "epoch": 2.84, "learning_rate": 2.69619603760204e-05, "loss": 0.4999, "step": 47124 }, { "epoch": 2.84, "learning_rate": 2.696045146590618e-05, "loss": 0.4749, "step": 47126 }, { "epoch": 2.84, "learning_rate": 2.6958942555791956e-05, "loss": 0.59, "step": 47128 }, { "epoch": 2.84, "learning_rate": 2.6957433645677726e-05, "loss": 0.4008, "step": 47130 }, { "epoch": 2.84, "learning_rate": 2.6955924735563502e-05, "loss": 0.4698, "step": 47132 }, { "epoch": 2.84, "learning_rate": 2.6954415825449282e-05, "loss": 0.4716, "step": 47134 }, { "epoch": 2.85, "learning_rate": 2.6952906915335052e-05, "loss": 0.558, "step": 47136 }, { "epoch": 2.85, "learning_rate": 2.695139800522083e-05, "loss": 0.6232, "step": 47138 }, { "epoch": 2.85, "learning_rate": 2.6949889095106605e-05, "loss": 0.4416, "step": 47140 }, { "epoch": 2.85, "learning_rate": 2.6948380184992385e-05, "loss": 0.8754, "step": 47142 }, { "epoch": 2.85, "learning_rate": 2.6946871274878155e-05, "loss": 0.3359, "step": 47144 }, { "epoch": 2.85, "learning_rate": 2.694536236476393e-05, "loss": 0.3801, "step": 47146 }, { "epoch": 2.85, "learning_rate": 2.6943853454649708e-05, "loss": 0.4449, "step": 47148 }, { "epoch": 2.85, "learning_rate": 2.6942344544535488e-05, "loss": 0.448, "step": 47150 }, { "epoch": 2.85, "learning_rate": 2.6940835634421257e-05, "loss": 0.4486, "step": 47152 }, { "epoch": 2.85, "learning_rate": 2.6939326724307034e-05, "loss": 0.3468, "step": 47154 }, { "epoch": 2.85, "learning_rate": 2.693781781419281e-05, "loss": 0.5159, "step": 47156 }, { "epoch": 2.85, "learning_rate": 2.6936308904078584e-05, "loss": 0.4303, "step": 47158 }, { "epoch": 2.85, "learning_rate": 2.693479999396436e-05, "loss": 0.6875, "step": 47160 }, { "epoch": 2.85, "learning_rate": 2.6933291083850137e-05, "loss": 0.661, "step": 47162 }, { "epoch": 2.85, "learning_rate": 2.6931782173735913e-05, "loss": 0.4314, "step": 47164 }, { "epoch": 2.85, "learning_rate": 2.6930273263621686e-05, "loss": 0.7392, "step": 47166 }, { "epoch": 2.85, "learning_rate": 2.6928764353507463e-05, "loss": 0.3996, "step": 47168 }, { "epoch": 2.85, "learning_rate": 2.692725544339324e-05, "loss": 0.6097, "step": 47170 }, { "epoch": 2.85, "learning_rate": 2.6925746533279012e-05, "loss": 0.5374, "step": 47172 }, { "epoch": 2.85, "learning_rate": 2.692423762316479e-05, "loss": 0.5842, "step": 47174 }, { "epoch": 2.85, "learning_rate": 2.6922728713050565e-05, "loss": 1.0432, "step": 47176 }, { "epoch": 2.85, "learning_rate": 2.6921219802936342e-05, "loss": 0.518, "step": 47178 }, { "epoch": 2.85, "learning_rate": 2.6919710892822115e-05, "loss": 0.5782, "step": 47180 }, { "epoch": 2.85, "learning_rate": 2.691820198270789e-05, "loss": 0.3847, "step": 47182 }, { "epoch": 2.85, "learning_rate": 2.6916693072593668e-05, "loss": 0.5802, "step": 47184 }, { "epoch": 2.85, "learning_rate": 2.691518416247944e-05, "loss": 0.5311, "step": 47186 }, { "epoch": 2.85, "learning_rate": 2.6913675252365218e-05, "loss": 0.5895, "step": 47188 }, { "epoch": 2.85, "learning_rate": 2.6912166342250994e-05, "loss": 0.3727, "step": 47190 }, { "epoch": 2.85, "learning_rate": 2.691065743213677e-05, "loss": 0.45, "step": 47192 }, { "epoch": 2.85, "learning_rate": 2.6909148522022544e-05, "loss": 0.492, "step": 47194 }, { "epoch": 2.85, "learning_rate": 2.690763961190832e-05, "loss": 0.5228, "step": 47196 }, { "epoch": 2.85, "learning_rate": 2.6906130701794097e-05, "loss": 0.3666, "step": 47198 }, { "epoch": 2.85, "learning_rate": 2.6904621791679867e-05, "loss": 0.6093, "step": 47200 }, { "epoch": 2.85, "learning_rate": 2.6903112881565646e-05, "loss": 0.4658, "step": 47202 }, { "epoch": 2.85, "learning_rate": 2.6901603971451423e-05, "loss": 0.5115, "step": 47204 }, { "epoch": 2.85, "learning_rate": 2.69000950613372e-05, "loss": 0.5794, "step": 47206 }, { "epoch": 2.85, "learning_rate": 2.689858615122297e-05, "loss": 0.6211, "step": 47208 }, { "epoch": 2.85, "learning_rate": 2.689707724110875e-05, "loss": 0.405, "step": 47210 }, { "epoch": 2.85, "learning_rate": 2.6895568330994526e-05, "loss": 0.4486, "step": 47212 }, { "epoch": 2.85, "learning_rate": 2.6894059420880302e-05, "loss": 0.5037, "step": 47214 }, { "epoch": 2.85, "learning_rate": 2.6892550510766072e-05, "loss": 0.4566, "step": 47216 }, { "epoch": 2.85, "learning_rate": 2.6891041600651852e-05, "loss": 0.4282, "step": 47218 }, { "epoch": 2.85, "learning_rate": 2.688953269053763e-05, "loss": 0.4359, "step": 47220 }, { "epoch": 2.85, "learning_rate": 2.6888023780423398e-05, "loss": 0.5044, "step": 47222 }, { "epoch": 2.85, "learning_rate": 2.6886514870309175e-05, "loss": 0.5325, "step": 47224 }, { "epoch": 2.85, "learning_rate": 2.6885005960194954e-05, "loss": 0.4472, "step": 47226 }, { "epoch": 2.85, "learning_rate": 2.688349705008073e-05, "loss": 0.3705, "step": 47228 }, { "epoch": 2.85, "learning_rate": 2.68819881399665e-05, "loss": 0.7823, "step": 47230 }, { "epoch": 2.85, "learning_rate": 2.688047922985228e-05, "loss": 0.3697, "step": 47232 }, { "epoch": 2.85, "learning_rate": 2.6878970319738057e-05, "loss": 0.4279, "step": 47234 }, { "epoch": 2.85, "learning_rate": 2.6877461409623827e-05, "loss": 0.4644, "step": 47236 }, { "epoch": 2.85, "learning_rate": 2.6875952499509603e-05, "loss": 0.4762, "step": 47238 }, { "epoch": 2.85, "learning_rate": 2.6874443589395383e-05, "loss": 0.7862, "step": 47240 }, { "epoch": 2.85, "learning_rate": 2.687293467928116e-05, "loss": 0.4532, "step": 47242 }, { "epoch": 2.85, "learning_rate": 2.687142576916693e-05, "loss": 0.562, "step": 47244 }, { "epoch": 2.85, "learning_rate": 2.6869916859052706e-05, "loss": 0.4497, "step": 47246 }, { "epoch": 2.85, "learning_rate": 2.6868407948938486e-05, "loss": 0.6667, "step": 47248 }, { "epoch": 2.85, "learning_rate": 2.6866899038824256e-05, "loss": 0.628, "step": 47250 }, { "epoch": 2.85, "learning_rate": 2.6865390128710032e-05, "loss": 0.6064, "step": 47252 }, { "epoch": 2.85, "learning_rate": 2.686388121859581e-05, "loss": 0.3576, "step": 47254 }, { "epoch": 2.85, "learning_rate": 2.686237230848159e-05, "loss": 0.5341, "step": 47256 }, { "epoch": 2.85, "learning_rate": 2.686086339836736e-05, "loss": 0.6729, "step": 47258 }, { "epoch": 2.85, "learning_rate": 2.6859354488253135e-05, "loss": 0.4641, "step": 47260 }, { "epoch": 2.85, "learning_rate": 2.685784557813891e-05, "loss": 0.5747, "step": 47262 }, { "epoch": 2.85, "learning_rate": 2.685633666802469e-05, "loss": 0.6388, "step": 47264 }, { "epoch": 2.85, "learning_rate": 2.685482775791046e-05, "loss": 0.4106, "step": 47266 }, { "epoch": 2.85, "learning_rate": 2.6853318847796238e-05, "loss": 0.4606, "step": 47268 }, { "epoch": 2.85, "learning_rate": 2.6851809937682014e-05, "loss": 0.4798, "step": 47270 }, { "epoch": 2.85, "learning_rate": 2.6850301027567787e-05, "loss": 0.5602, "step": 47272 }, { "epoch": 2.85, "learning_rate": 2.6848792117453564e-05, "loss": 0.6209, "step": 47274 }, { "epoch": 2.85, "learning_rate": 2.684728320733934e-05, "loss": 0.4709, "step": 47276 }, { "epoch": 2.85, "learning_rate": 2.6845774297225117e-05, "loss": 0.6792, "step": 47278 }, { "epoch": 2.85, "learning_rate": 2.684426538711089e-05, "loss": 0.4876, "step": 47280 }, { "epoch": 2.85, "learning_rate": 2.6842756476996666e-05, "loss": 0.5343, "step": 47282 }, { "epoch": 2.85, "learning_rate": 2.6841247566882443e-05, "loss": 0.6227, "step": 47284 }, { "epoch": 2.85, "learning_rate": 2.6839738656768216e-05, "loss": 0.6829, "step": 47286 }, { "epoch": 2.85, "learning_rate": 2.6838229746653993e-05, "loss": 0.4782, "step": 47288 }, { "epoch": 2.85, "learning_rate": 2.683672083653977e-05, "loss": 0.4116, "step": 47290 }, { "epoch": 2.85, "learning_rate": 2.6835211926425546e-05, "loss": 0.4931, "step": 47292 }, { "epoch": 2.85, "learning_rate": 2.683370301631132e-05, "loss": 0.6373, "step": 47294 }, { "epoch": 2.85, "learning_rate": 2.6832194106197095e-05, "loss": 0.4983, "step": 47296 }, { "epoch": 2.85, "learning_rate": 2.6830685196082872e-05, "loss": 0.7498, "step": 47298 }, { "epoch": 2.85, "learning_rate": 2.6829176285968645e-05, "loss": 0.4882, "step": 47300 }, { "epoch": 2.86, "learning_rate": 2.682766737585442e-05, "loss": 0.6467, "step": 47302 }, { "epoch": 2.86, "learning_rate": 2.6826158465740198e-05, "loss": 0.5814, "step": 47304 }, { "epoch": 2.86, "learning_rate": 2.6824649555625974e-05, "loss": 0.4568, "step": 47306 }, { "epoch": 2.86, "learning_rate": 2.6823140645511747e-05, "loss": 0.4635, "step": 47308 }, { "epoch": 2.86, "learning_rate": 2.6821631735397524e-05, "loss": 0.432, "step": 47310 }, { "epoch": 2.86, "learning_rate": 2.68201228252833e-05, "loss": 0.5164, "step": 47312 }, { "epoch": 2.86, "learning_rate": 2.6818613915169077e-05, "loss": 0.4743, "step": 47314 }, { "epoch": 2.86, "learning_rate": 2.681710500505485e-05, "loss": 0.3588, "step": 47316 }, { "epoch": 2.86, "learning_rate": 2.6815596094940627e-05, "loss": 0.5047, "step": 47318 }, { "epoch": 2.86, "learning_rate": 2.6814087184826403e-05, "loss": 0.7326, "step": 47320 }, { "epoch": 2.86, "learning_rate": 2.6812578274712173e-05, "loss": 0.3887, "step": 47322 }, { "epoch": 2.86, "learning_rate": 2.6811069364597953e-05, "loss": 0.4318, "step": 47324 }, { "epoch": 2.86, "learning_rate": 2.680956045448373e-05, "loss": 0.9708, "step": 47326 }, { "epoch": 2.86, "learning_rate": 2.6808051544369506e-05, "loss": 0.6288, "step": 47328 }, { "epoch": 2.86, "learning_rate": 2.6806542634255276e-05, "loss": 0.703, "step": 47330 }, { "epoch": 2.86, "learning_rate": 2.6805033724141055e-05, "loss": 0.5921, "step": 47332 }, { "epoch": 2.86, "learning_rate": 2.6803524814026832e-05, "loss": 0.4838, "step": 47334 }, { "epoch": 2.86, "learning_rate": 2.6802015903912602e-05, "loss": 0.483, "step": 47336 }, { "epoch": 2.86, "learning_rate": 2.6800506993798378e-05, "loss": 0.507, "step": 47338 }, { "epoch": 2.86, "learning_rate": 2.6798998083684158e-05, "loss": 0.5866, "step": 47340 }, { "epoch": 2.86, "learning_rate": 2.6797489173569935e-05, "loss": 0.4159, "step": 47342 }, { "epoch": 2.86, "learning_rate": 2.6795980263455704e-05, "loss": 0.5393, "step": 47344 }, { "epoch": 2.86, "learning_rate": 2.679447135334148e-05, "loss": 0.3968, "step": 47346 }, { "epoch": 2.86, "learning_rate": 2.679296244322726e-05, "loss": 0.5608, "step": 47348 }, { "epoch": 2.86, "learning_rate": 2.679145353311303e-05, "loss": 0.3877, "step": 47350 }, { "epoch": 2.86, "learning_rate": 2.6789944622998807e-05, "loss": 0.7266, "step": 47352 }, { "epoch": 2.86, "learning_rate": 2.6788435712884587e-05, "loss": 0.7044, "step": 47354 }, { "epoch": 2.86, "learning_rate": 2.6786926802770363e-05, "loss": 0.524, "step": 47356 }, { "epoch": 2.86, "learning_rate": 2.6785417892656133e-05, "loss": 0.4539, "step": 47358 }, { "epoch": 2.86, "learning_rate": 2.678390898254191e-05, "loss": 0.6026, "step": 47360 }, { "epoch": 2.86, "learning_rate": 2.678240007242769e-05, "loss": 0.6374, "step": 47362 }, { "epoch": 2.86, "learning_rate": 2.678089116231346e-05, "loss": 0.5759, "step": 47364 }, { "epoch": 2.86, "learning_rate": 2.6779382252199236e-05, "loss": 0.4456, "step": 47366 }, { "epoch": 2.86, "learning_rate": 2.6777873342085012e-05, "loss": 0.3668, "step": 47368 }, { "epoch": 2.86, "learning_rate": 2.6776364431970792e-05, "loss": 0.4789, "step": 47370 }, { "epoch": 2.86, "learning_rate": 2.6774855521856562e-05, "loss": 0.4617, "step": 47372 }, { "epoch": 2.86, "learning_rate": 2.677334661174234e-05, "loss": 0.5895, "step": 47374 }, { "epoch": 2.86, "learning_rate": 2.6771837701628115e-05, "loss": 0.6054, "step": 47376 }, { "epoch": 2.86, "learning_rate": 2.6770328791513895e-05, "loss": 0.3486, "step": 47378 }, { "epoch": 2.86, "learning_rate": 2.6768819881399665e-05, "loss": 0.4364, "step": 47380 }, { "epoch": 2.86, "learning_rate": 2.676731097128544e-05, "loss": 0.5859, "step": 47382 }, { "epoch": 2.86, "learning_rate": 2.6765802061171218e-05, "loss": 0.5064, "step": 47384 }, { "epoch": 2.86, "learning_rate": 2.676429315105699e-05, "loss": 0.4322, "step": 47386 }, { "epoch": 2.86, "learning_rate": 2.6762784240942767e-05, "loss": 0.4083, "step": 47388 }, { "epoch": 2.86, "learning_rate": 2.6761275330828544e-05, "loss": 0.395, "step": 47390 }, { "epoch": 2.86, "learning_rate": 2.675976642071432e-05, "loss": 0.6408, "step": 47392 }, { "epoch": 2.86, "learning_rate": 2.6758257510600094e-05, "loss": 0.746, "step": 47394 }, { "epoch": 2.86, "learning_rate": 2.675674860048587e-05, "loss": 0.5264, "step": 47396 }, { "epoch": 2.86, "learning_rate": 2.6755239690371647e-05, "loss": 0.6051, "step": 47398 }, { "epoch": 2.86, "learning_rate": 2.675373078025742e-05, "loss": 0.3889, "step": 47400 }, { "epoch": 2.86, "learning_rate": 2.6752221870143196e-05, "loss": 0.9319, "step": 47402 }, { "epoch": 2.86, "learning_rate": 2.6750712960028973e-05, "loss": 0.5256, "step": 47404 }, { "epoch": 2.86, "learning_rate": 2.674920404991475e-05, "loss": 0.4856, "step": 47406 }, { "epoch": 2.86, "learning_rate": 2.6747695139800522e-05, "loss": 0.5883, "step": 47408 }, { "epoch": 2.86, "learning_rate": 2.67461862296863e-05, "loss": 0.5154, "step": 47410 }, { "epoch": 2.86, "learning_rate": 2.6744677319572075e-05, "loss": 0.5148, "step": 47412 }, { "epoch": 2.86, "learning_rate": 2.674316840945785e-05, "loss": 0.5301, "step": 47414 }, { "epoch": 2.86, "learning_rate": 2.6741659499343625e-05, "loss": 0.4686, "step": 47416 }, { "epoch": 2.86, "learning_rate": 2.67401505892294e-05, "loss": 0.5236, "step": 47418 }, { "epoch": 2.86, "learning_rate": 2.6738641679115178e-05, "loss": 0.5498, "step": 47420 }, { "epoch": 2.86, "learning_rate": 2.673713276900095e-05, "loss": 0.4235, "step": 47422 }, { "epoch": 2.86, "learning_rate": 2.6735623858886728e-05, "loss": 0.5039, "step": 47424 }, { "epoch": 2.86, "learning_rate": 2.6734114948772504e-05, "loss": 0.5058, "step": 47426 }, { "epoch": 2.86, "learning_rate": 2.673260603865828e-05, "loss": 0.4796, "step": 47428 }, { "epoch": 2.86, "learning_rate": 2.6731097128544054e-05, "loss": 0.5312, "step": 47430 }, { "epoch": 2.86, "learning_rate": 2.672958821842983e-05, "loss": 0.6046, "step": 47432 }, { "epoch": 2.86, "learning_rate": 2.6728079308315607e-05, "loss": 0.5864, "step": 47434 }, { "epoch": 2.86, "learning_rate": 2.6726570398201377e-05, "loss": 0.505, "step": 47436 }, { "epoch": 2.86, "learning_rate": 2.6725061488087157e-05, "loss": 0.5565, "step": 47438 }, { "epoch": 2.86, "learning_rate": 2.6723552577972933e-05, "loss": 0.5124, "step": 47440 }, { "epoch": 2.86, "learning_rate": 2.672204366785871e-05, "loss": 0.5155, "step": 47442 }, { "epoch": 2.86, "learning_rate": 2.672053475774448e-05, "loss": 0.4715, "step": 47444 }, { "epoch": 2.86, "learning_rate": 2.671902584763026e-05, "loss": 0.496, "step": 47446 }, { "epoch": 2.86, "learning_rate": 2.6717516937516036e-05, "loss": 0.4307, "step": 47448 }, { "epoch": 2.86, "learning_rate": 2.6716008027401805e-05, "loss": 0.4075, "step": 47450 }, { "epoch": 2.86, "learning_rate": 2.6714499117287582e-05, "loss": 0.602, "step": 47452 }, { "epoch": 2.86, "learning_rate": 2.6712990207173362e-05, "loss": 0.6225, "step": 47454 }, { "epoch": 2.86, "learning_rate": 2.671148129705914e-05, "loss": 0.334, "step": 47456 }, { "epoch": 2.86, "learning_rate": 2.6709972386944908e-05, "loss": 0.3188, "step": 47458 }, { "epoch": 2.86, "learning_rate": 2.6708463476830685e-05, "loss": 0.5929, "step": 47460 }, { "epoch": 2.86, "learning_rate": 2.6706954566716465e-05, "loss": 0.3648, "step": 47462 }, { "epoch": 2.86, "learning_rate": 2.6705445656602234e-05, "loss": 0.6715, "step": 47464 }, { "epoch": 2.86, "learning_rate": 2.670393674648801e-05, "loss": 0.33, "step": 47466 }, { "epoch": 2.87, "learning_rate": 2.670242783637379e-05, "loss": 0.6977, "step": 47468 }, { "epoch": 2.87, "learning_rate": 2.6700918926259567e-05, "loss": 0.4553, "step": 47470 }, { "epoch": 2.87, "learning_rate": 2.6699410016145337e-05, "loss": 0.3739, "step": 47472 }, { "epoch": 2.87, "learning_rate": 2.6697901106031113e-05, "loss": 0.5582, "step": 47474 }, { "epoch": 2.87, "learning_rate": 2.6696392195916893e-05, "loss": 0.2929, "step": 47476 }, { "epoch": 2.87, "learning_rate": 2.669488328580267e-05, "loss": 0.5313, "step": 47478 }, { "epoch": 2.87, "learning_rate": 2.669337437568844e-05, "loss": 0.6505, "step": 47480 }, { "epoch": 2.87, "learning_rate": 2.6691865465574216e-05, "loss": 0.6675, "step": 47482 }, { "epoch": 2.87, "learning_rate": 2.6690356555459996e-05, "loss": 0.6909, "step": 47484 }, { "epoch": 2.87, "learning_rate": 2.6688847645345766e-05, "loss": 0.5429, "step": 47486 }, { "epoch": 2.87, "learning_rate": 2.6687338735231542e-05, "loss": 0.7654, "step": 47488 }, { "epoch": 2.87, "learning_rate": 2.668582982511732e-05, "loss": 0.3376, "step": 47490 }, { "epoch": 2.87, "learning_rate": 2.66843209150031e-05, "loss": 0.4631, "step": 47492 }, { "epoch": 2.87, "learning_rate": 2.668281200488887e-05, "loss": 0.6447, "step": 47494 }, { "epoch": 2.87, "learning_rate": 2.6681303094774645e-05, "loss": 0.4665, "step": 47496 }, { "epoch": 2.87, "learning_rate": 2.667979418466042e-05, "loss": 0.3726, "step": 47498 }, { "epoch": 2.87, "learning_rate": 2.6678285274546195e-05, "loss": 0.4123, "step": 47500 }, { "epoch": 2.87, "learning_rate": 2.667677636443197e-05, "loss": 0.6241, "step": 47502 }, { "epoch": 2.87, "learning_rate": 2.6675267454317748e-05, "loss": 0.6279, "step": 47504 }, { "epoch": 2.87, "learning_rate": 2.6673758544203524e-05, "loss": 0.4713, "step": 47506 }, { "epoch": 2.87, "learning_rate": 2.6672249634089297e-05, "loss": 0.5282, "step": 47508 }, { "epoch": 2.87, "learning_rate": 2.6670740723975074e-05, "loss": 0.3503, "step": 47510 }, { "epoch": 2.87, "learning_rate": 2.666923181386085e-05, "loss": 0.7474, "step": 47512 }, { "epoch": 2.87, "learning_rate": 2.6667722903746623e-05, "loss": 0.7451, "step": 47514 }, { "epoch": 2.87, "learning_rate": 2.66662139936324e-05, "loss": 0.5079, "step": 47516 }, { "epoch": 2.87, "learning_rate": 2.6664705083518176e-05, "loss": 0.5904, "step": 47518 }, { "epoch": 2.87, "learning_rate": 2.6663196173403953e-05, "loss": 0.5856, "step": 47520 }, { "epoch": 2.87, "learning_rate": 2.6661687263289726e-05, "loss": 0.5746, "step": 47522 }, { "epoch": 2.87, "learning_rate": 2.6660178353175503e-05, "loss": 0.3784, "step": 47524 }, { "epoch": 2.87, "learning_rate": 2.665866944306128e-05, "loss": 0.4116, "step": 47526 }, { "epoch": 2.87, "learning_rate": 2.6657160532947052e-05, "loss": 0.2989, "step": 47528 }, { "epoch": 2.87, "learning_rate": 2.665565162283283e-05, "loss": 0.5325, "step": 47530 }, { "epoch": 2.87, "learning_rate": 2.6654142712718605e-05, "loss": 0.4278, "step": 47532 }, { "epoch": 2.87, "learning_rate": 2.6652633802604382e-05, "loss": 0.6426, "step": 47534 }, { "epoch": 2.87, "learning_rate": 2.6651124892490155e-05, "loss": 0.4296, "step": 47536 }, { "epoch": 2.87, "learning_rate": 2.664961598237593e-05, "loss": 0.8079, "step": 47538 }, { "epoch": 2.87, "learning_rate": 2.6648107072261708e-05, "loss": 0.5088, "step": 47540 }, { "epoch": 2.87, "learning_rate": 2.6646598162147484e-05, "loss": 0.6403, "step": 47542 }, { "epoch": 2.87, "learning_rate": 2.6645089252033258e-05, "loss": 0.6415, "step": 47544 }, { "epoch": 2.87, "learning_rate": 2.6643580341919034e-05, "loss": 0.4017, "step": 47546 }, { "epoch": 2.87, "learning_rate": 2.664207143180481e-05, "loss": 0.658, "step": 47548 }, { "epoch": 2.87, "learning_rate": 2.664056252169058e-05, "loss": 0.4312, "step": 47550 }, { "epoch": 2.87, "learning_rate": 2.663905361157636e-05, "loss": 0.4344, "step": 47552 }, { "epoch": 2.87, "learning_rate": 2.6637544701462137e-05, "loss": 0.4196, "step": 47554 }, { "epoch": 2.87, "learning_rate": 2.6636035791347913e-05, "loss": 0.5047, "step": 47556 }, { "epoch": 2.87, "learning_rate": 2.6634526881233683e-05, "loss": 0.4714, "step": 47558 }, { "epoch": 2.87, "learning_rate": 2.6633017971119463e-05, "loss": 0.4298, "step": 47560 }, { "epoch": 2.87, "learning_rate": 2.663150906100524e-05, "loss": 0.5314, "step": 47562 }, { "epoch": 2.87, "learning_rate": 2.663000015089101e-05, "loss": 0.4856, "step": 47564 }, { "epoch": 2.87, "learning_rate": 2.6628491240776786e-05, "loss": 0.5328, "step": 47566 }, { "epoch": 2.87, "learning_rate": 2.6626982330662566e-05, "loss": 0.6015, "step": 47568 }, { "epoch": 2.87, "learning_rate": 2.6625473420548342e-05, "loss": 0.4609, "step": 47570 }, { "epoch": 2.87, "learning_rate": 2.6623964510434112e-05, "loss": 0.4552, "step": 47572 }, { "epoch": 2.87, "learning_rate": 2.6622455600319888e-05, "loss": 0.6243, "step": 47574 }, { "epoch": 2.87, "learning_rate": 2.6620946690205668e-05, "loss": 0.69, "step": 47576 }, { "epoch": 2.87, "learning_rate": 2.6619437780091438e-05, "loss": 0.6859, "step": 47578 }, { "epoch": 2.87, "learning_rate": 2.6617928869977214e-05, "loss": 0.5115, "step": 47580 }, { "epoch": 2.87, "learning_rate": 2.661641995986299e-05, "loss": 0.4455, "step": 47582 }, { "epoch": 2.87, "learning_rate": 2.661491104974877e-05, "loss": 0.4574, "step": 47584 }, { "epoch": 2.87, "learning_rate": 2.661340213963454e-05, "loss": 0.5339, "step": 47586 }, { "epoch": 2.87, "learning_rate": 2.6611893229520317e-05, "loss": 0.4468, "step": 47588 }, { "epoch": 2.87, "learning_rate": 2.6610384319406097e-05, "loss": 0.5541, "step": 47590 }, { "epoch": 2.87, "learning_rate": 2.6608875409291874e-05, "loss": 0.4678, "step": 47592 }, { "epoch": 2.87, "learning_rate": 2.6607366499177643e-05, "loss": 0.4988, "step": 47594 }, { "epoch": 2.87, "learning_rate": 2.660585758906342e-05, "loss": 0.5504, "step": 47596 }, { "epoch": 2.87, "learning_rate": 2.66043486789492e-05, "loss": 0.5058, "step": 47598 }, { "epoch": 2.87, "learning_rate": 2.660283976883497e-05, "loss": 0.6735, "step": 47600 }, { "epoch": 2.87, "learning_rate": 2.6601330858720746e-05, "loss": 0.6177, "step": 47602 }, { "epoch": 2.87, "learning_rate": 2.6599821948606522e-05, "loss": 0.5811, "step": 47604 }, { "epoch": 2.87, "learning_rate": 2.6598313038492302e-05, "loss": 0.502, "step": 47606 }, { "epoch": 2.87, "learning_rate": 2.6596804128378072e-05, "loss": 0.6486, "step": 47608 }, { "epoch": 2.87, "learning_rate": 2.659529521826385e-05, "loss": 0.5628, "step": 47610 }, { "epoch": 2.87, "learning_rate": 2.6593786308149625e-05, "loss": 0.3834, "step": 47612 }, { "epoch": 2.87, "learning_rate": 2.6592277398035398e-05, "loss": 0.4298, "step": 47614 }, { "epoch": 2.87, "learning_rate": 2.6590768487921175e-05, "loss": 0.6152, "step": 47616 }, { "epoch": 2.87, "learning_rate": 2.658925957780695e-05, "loss": 0.4409, "step": 47618 }, { "epoch": 2.87, "learning_rate": 2.6587750667692728e-05, "loss": 0.4746, "step": 47620 }, { "epoch": 2.87, "learning_rate": 2.65862417575785e-05, "loss": 0.6302, "step": 47622 }, { "epoch": 2.87, "learning_rate": 2.6584732847464277e-05, "loss": 0.4486, "step": 47624 }, { "epoch": 2.87, "learning_rate": 2.6583223937350054e-05, "loss": 0.5565, "step": 47626 }, { "epoch": 2.87, "learning_rate": 2.6581715027235827e-05, "loss": 0.5737, "step": 47628 }, { "epoch": 2.87, "learning_rate": 2.6580206117121604e-05, "loss": 0.6165, "step": 47630 }, { "epoch": 2.87, "learning_rate": 2.657869720700738e-05, "loss": 0.6121, "step": 47632 }, { "epoch": 2.88, "learning_rate": 2.6577188296893157e-05, "loss": 0.496, "step": 47634 }, { "epoch": 2.88, "learning_rate": 2.657567938677893e-05, "loss": 0.4553, "step": 47636 }, { "epoch": 2.88, "learning_rate": 2.6574170476664706e-05, "loss": 0.4146, "step": 47638 }, { "epoch": 2.88, "learning_rate": 2.6572661566550483e-05, "loss": 0.4568, "step": 47640 }, { "epoch": 2.88, "learning_rate": 2.657115265643626e-05, "loss": 0.3768, "step": 47642 }, { "epoch": 2.88, "learning_rate": 2.6569643746322032e-05, "loss": 0.4734, "step": 47644 }, { "epoch": 2.88, "learning_rate": 2.656813483620781e-05, "loss": 0.5964, "step": 47646 }, { "epoch": 2.88, "learning_rate": 2.6566625926093585e-05, "loss": 0.7975, "step": 47648 }, { "epoch": 2.88, "learning_rate": 2.656511701597936e-05, "loss": 0.4311, "step": 47650 }, { "epoch": 2.88, "learning_rate": 2.6563608105865135e-05, "loss": 0.7428, "step": 47652 }, { "epoch": 2.88, "learning_rate": 2.656209919575091e-05, "loss": 0.7399, "step": 47654 }, { "epoch": 2.88, "learning_rate": 2.6560590285636688e-05, "loss": 0.6056, "step": 47656 }, { "epoch": 2.88, "learning_rate": 2.655908137552246e-05, "loss": 0.5131, "step": 47658 }, { "epoch": 2.88, "learning_rate": 2.6557572465408238e-05, "loss": 0.514, "step": 47660 }, { "epoch": 2.88, "learning_rate": 2.6556063555294014e-05, "loss": 0.5676, "step": 47662 }, { "epoch": 2.88, "learning_rate": 2.6554554645179784e-05, "loss": 0.4031, "step": 47664 }, { "epoch": 2.88, "learning_rate": 2.6553045735065564e-05, "loss": 0.587, "step": 47666 }, { "epoch": 2.88, "learning_rate": 2.655153682495134e-05, "loss": 0.5514, "step": 47668 }, { "epoch": 2.88, "learning_rate": 2.6550027914837117e-05, "loss": 0.5205, "step": 47670 }, { "epoch": 2.88, "learning_rate": 2.6548519004722887e-05, "loss": 0.3906, "step": 47672 }, { "epoch": 2.88, "learning_rate": 2.6547010094608667e-05, "loss": 0.6378, "step": 47674 }, { "epoch": 2.88, "learning_rate": 2.6545501184494443e-05, "loss": 0.5216, "step": 47676 }, { "epoch": 2.88, "learning_rate": 2.6543992274380213e-05, "loss": 0.4626, "step": 47678 }, { "epoch": 2.88, "learning_rate": 2.654248336426599e-05, "loss": 0.4947, "step": 47680 }, { "epoch": 2.88, "learning_rate": 2.654097445415177e-05, "loss": 0.5209, "step": 47682 }, { "epoch": 2.88, "learning_rate": 2.6539465544037546e-05, "loss": 0.5014, "step": 47684 }, { "epoch": 2.88, "learning_rate": 2.6537956633923315e-05, "loss": 0.7074, "step": 47686 }, { "epoch": 2.88, "learning_rate": 2.6536447723809092e-05, "loss": 0.6682, "step": 47688 }, { "epoch": 2.88, "learning_rate": 2.6534938813694872e-05, "loss": 0.7069, "step": 47690 }, { "epoch": 2.88, "learning_rate": 2.653342990358065e-05, "loss": 0.6683, "step": 47692 }, { "epoch": 2.88, "learning_rate": 2.6531920993466418e-05, "loss": 0.6366, "step": 47694 }, { "epoch": 2.88, "learning_rate": 2.6530412083352195e-05, "loss": 0.3332, "step": 47696 }, { "epoch": 2.88, "learning_rate": 2.6528903173237975e-05, "loss": 0.5539, "step": 47698 }, { "epoch": 2.88, "learning_rate": 2.6527394263123744e-05, "loss": 0.5248, "step": 47700 }, { "epoch": 2.88, "learning_rate": 2.652588535300952e-05, "loss": 0.5967, "step": 47702 }, { "epoch": 2.88, "learning_rate": 2.65243764428953e-05, "loss": 0.6138, "step": 47704 }, { "epoch": 2.88, "learning_rate": 2.6522867532781077e-05, "loss": 0.5851, "step": 47706 }, { "epoch": 2.88, "learning_rate": 2.6521358622666847e-05, "loss": 0.4982, "step": 47708 }, { "epoch": 2.88, "learning_rate": 2.6519849712552623e-05, "loss": 0.491, "step": 47710 }, { "epoch": 2.88, "learning_rate": 2.6518340802438403e-05, "loss": 0.5602, "step": 47712 }, { "epoch": 2.88, "learning_rate": 2.6516831892324173e-05, "loss": 0.6061, "step": 47714 }, { "epoch": 2.88, "learning_rate": 2.651532298220995e-05, "loss": 0.5122, "step": 47716 }, { "epoch": 2.88, "learning_rate": 2.6513814072095726e-05, "loss": 0.5876, "step": 47718 }, { "epoch": 2.88, "learning_rate": 2.6512305161981506e-05, "loss": 0.7291, "step": 47720 }, { "epoch": 2.88, "learning_rate": 2.6510796251867276e-05, "loss": 0.7217, "step": 47722 }, { "epoch": 2.88, "learning_rate": 2.6509287341753052e-05, "loss": 0.3906, "step": 47724 }, { "epoch": 2.88, "learning_rate": 2.650777843163883e-05, "loss": 0.3117, "step": 47726 }, { "epoch": 2.88, "learning_rate": 2.6506269521524602e-05, "loss": 0.395, "step": 47728 }, { "epoch": 2.88, "learning_rate": 2.650476061141038e-05, "loss": 0.4291, "step": 47730 }, { "epoch": 2.88, "learning_rate": 2.6503251701296155e-05, "loss": 0.5322, "step": 47732 }, { "epoch": 2.88, "learning_rate": 2.650174279118193e-05, "loss": 0.5054, "step": 47734 }, { "epoch": 2.88, "learning_rate": 2.6500233881067705e-05, "loss": 0.5025, "step": 47736 }, { "epoch": 2.88, "learning_rate": 2.649872497095348e-05, "loss": 0.3735, "step": 47738 }, { "epoch": 2.88, "learning_rate": 2.6497216060839258e-05, "loss": 0.5814, "step": 47740 }, { "epoch": 2.88, "learning_rate": 2.649570715072503e-05, "loss": 0.5767, "step": 47742 }, { "epoch": 2.88, "learning_rate": 2.6494198240610807e-05, "loss": 0.5208, "step": 47744 }, { "epoch": 2.88, "learning_rate": 2.6492689330496584e-05, "loss": 0.4962, "step": 47746 }, { "epoch": 2.88, "learning_rate": 2.649118042038236e-05, "loss": 0.5715, "step": 47748 }, { "epoch": 2.88, "learning_rate": 2.6489671510268133e-05, "loss": 0.5779, "step": 47750 }, { "epoch": 2.88, "learning_rate": 2.648816260015391e-05, "loss": 0.3829, "step": 47752 }, { "epoch": 2.88, "learning_rate": 2.6486653690039686e-05, "loss": 0.537, "step": 47754 }, { "epoch": 2.88, "learning_rate": 2.6485144779925463e-05, "loss": 0.531, "step": 47756 }, { "epoch": 2.88, "learning_rate": 2.6483635869811236e-05, "loss": 0.4392, "step": 47758 }, { "epoch": 2.88, "learning_rate": 2.6482126959697013e-05, "loss": 0.5216, "step": 47760 }, { "epoch": 2.88, "learning_rate": 2.648061804958279e-05, "loss": 0.4682, "step": 47762 }, { "epoch": 2.88, "learning_rate": 2.6479109139468562e-05, "loss": 0.361, "step": 47764 }, { "epoch": 2.88, "learning_rate": 2.647760022935434e-05, "loss": 0.5545, "step": 47766 }, { "epoch": 2.88, "learning_rate": 2.6476091319240115e-05, "loss": 0.7356, "step": 47768 }, { "epoch": 2.88, "learning_rate": 2.6474582409125892e-05, "loss": 0.5507, "step": 47770 }, { "epoch": 2.88, "learning_rate": 2.6473073499011665e-05, "loss": 0.4286, "step": 47772 }, { "epoch": 2.88, "learning_rate": 2.647156458889744e-05, "loss": 0.5338, "step": 47774 }, { "epoch": 2.88, "learning_rate": 2.6470055678783218e-05, "loss": 0.4571, "step": 47776 }, { "epoch": 2.88, "learning_rate": 2.6468546768668988e-05, "loss": 0.6007, "step": 47778 }, { "epoch": 2.88, "learning_rate": 2.6467037858554768e-05, "loss": 0.6989, "step": 47780 }, { "epoch": 2.88, "learning_rate": 2.6465528948440544e-05, "loss": 0.5962, "step": 47782 }, { "epoch": 2.88, "learning_rate": 2.646402003832632e-05, "loss": 0.5869, "step": 47784 }, { "epoch": 2.88, "learning_rate": 2.646251112821209e-05, "loss": 0.5847, "step": 47786 }, { "epoch": 2.88, "learning_rate": 2.646100221809787e-05, "loss": 0.4873, "step": 47788 }, { "epoch": 2.88, "learning_rate": 2.6459493307983647e-05, "loss": 0.4872, "step": 47790 }, { "epoch": 2.88, "learning_rate": 2.6457984397869416e-05, "loss": 0.5669, "step": 47792 }, { "epoch": 2.88, "learning_rate": 2.6456475487755193e-05, "loss": 0.6969, "step": 47794 }, { "epoch": 2.88, "learning_rate": 2.6454966577640973e-05, "loss": 0.5094, "step": 47796 }, { "epoch": 2.88, "learning_rate": 2.645345766752675e-05, "loss": 0.3257, "step": 47798 }, { "epoch": 2.89, "learning_rate": 2.645194875741252e-05, "loss": 0.2862, "step": 47800 }, { "epoch": 2.89, "learning_rate": 2.6450439847298296e-05, "loss": 0.4233, "step": 47802 }, { "epoch": 2.89, "learning_rate": 2.6448930937184076e-05, "loss": 0.4717, "step": 47804 }, { "epoch": 2.89, "learning_rate": 2.6447422027069852e-05, "loss": 0.5574, "step": 47806 }, { "epoch": 2.89, "learning_rate": 2.6445913116955622e-05, "loss": 0.6403, "step": 47808 }, { "epoch": 2.89, "learning_rate": 2.64444042068414e-05, "loss": 0.7476, "step": 47810 }, { "epoch": 2.89, "learning_rate": 2.6442895296727178e-05, "loss": 0.6803, "step": 47812 }, { "epoch": 2.89, "learning_rate": 2.6441386386612948e-05, "loss": 0.5307, "step": 47814 }, { "epoch": 2.89, "learning_rate": 2.6439877476498724e-05, "loss": 0.6123, "step": 47816 }, { "epoch": 2.89, "learning_rate": 2.64383685663845e-05, "loss": 0.4218, "step": 47818 }, { "epoch": 2.89, "learning_rate": 2.643685965627028e-05, "loss": 0.5593, "step": 47820 }, { "epoch": 2.89, "learning_rate": 2.643535074615605e-05, "loss": 0.3748, "step": 47822 }, { "epoch": 2.89, "learning_rate": 2.6433841836041827e-05, "loss": 0.6068, "step": 47824 }, { "epoch": 2.89, "learning_rate": 2.6432332925927607e-05, "loss": 0.514, "step": 47826 }, { "epoch": 2.89, "learning_rate": 2.6430824015813377e-05, "loss": 0.4641, "step": 47828 }, { "epoch": 2.89, "learning_rate": 2.6429315105699153e-05, "loss": 0.4, "step": 47830 }, { "epoch": 2.89, "learning_rate": 2.642780619558493e-05, "loss": 0.7746, "step": 47832 }, { "epoch": 2.89, "learning_rate": 2.642629728547071e-05, "loss": 0.3878, "step": 47834 }, { "epoch": 2.89, "learning_rate": 2.642478837535648e-05, "loss": 0.5386, "step": 47836 }, { "epoch": 2.89, "learning_rate": 2.6423279465242256e-05, "loss": 0.6939, "step": 47838 }, { "epoch": 2.89, "learning_rate": 2.6421770555128032e-05, "loss": 0.9314, "step": 47840 }, { "epoch": 2.89, "learning_rate": 2.6420261645013806e-05, "loss": 0.3856, "step": 47842 }, { "epoch": 2.89, "learning_rate": 2.6418752734899582e-05, "loss": 0.3567, "step": 47844 }, { "epoch": 2.89, "learning_rate": 2.641724382478536e-05, "loss": 0.5337, "step": 47846 }, { "epoch": 2.89, "learning_rate": 2.6415734914671135e-05, "loss": 0.3499, "step": 47848 }, { "epoch": 2.89, "learning_rate": 2.6414226004556908e-05, "loss": 0.6053, "step": 47850 }, { "epoch": 2.89, "learning_rate": 2.6412717094442685e-05, "loss": 0.6865, "step": 47852 }, { "epoch": 2.89, "learning_rate": 2.641120818432846e-05, "loss": 0.5659, "step": 47854 }, { "epoch": 2.89, "learning_rate": 2.6409699274214238e-05, "loss": 0.5141, "step": 47856 }, { "epoch": 2.89, "learning_rate": 2.640819036410001e-05, "loss": 0.5115, "step": 47858 }, { "epoch": 2.89, "learning_rate": 2.6406681453985787e-05, "loss": 0.4329, "step": 47860 }, { "epoch": 2.89, "learning_rate": 2.6405172543871564e-05, "loss": 0.5648, "step": 47862 }, { "epoch": 2.89, "learning_rate": 2.6403663633757337e-05, "loss": 0.3881, "step": 47864 }, { "epoch": 2.89, "learning_rate": 2.6402154723643114e-05, "loss": 0.56, "step": 47866 }, { "epoch": 2.89, "learning_rate": 2.640064581352889e-05, "loss": 0.6533, "step": 47868 }, { "epoch": 2.89, "learning_rate": 2.6399136903414667e-05, "loss": 0.8295, "step": 47870 }, { "epoch": 2.89, "learning_rate": 2.639762799330044e-05, "loss": 0.4663, "step": 47872 }, { "epoch": 2.89, "learning_rate": 2.6396119083186216e-05, "loss": 0.5073, "step": 47874 }, { "epoch": 2.89, "learning_rate": 2.6394610173071993e-05, "loss": 0.5113, "step": 47876 }, { "epoch": 2.89, "learning_rate": 2.6393101262957766e-05, "loss": 0.5892, "step": 47878 }, { "epoch": 2.89, "learning_rate": 2.6391592352843542e-05, "loss": 0.5628, "step": 47880 }, { "epoch": 2.89, "learning_rate": 2.639008344272932e-05, "loss": 0.5397, "step": 47882 }, { "epoch": 2.89, "learning_rate": 2.6388574532615095e-05, "loss": 0.4022, "step": 47884 }, { "epoch": 2.89, "learning_rate": 2.638706562250087e-05, "loss": 0.3908, "step": 47886 }, { "epoch": 2.89, "learning_rate": 2.6385556712386645e-05, "loss": 0.4043, "step": 47888 }, { "epoch": 2.89, "learning_rate": 2.638404780227242e-05, "loss": 0.5496, "step": 47890 }, { "epoch": 2.89, "learning_rate": 2.638253889215819e-05, "loss": 0.6711, "step": 47892 }, { "epoch": 2.89, "learning_rate": 2.638102998204397e-05, "loss": 0.5498, "step": 47894 }, { "epoch": 2.89, "learning_rate": 2.6379521071929748e-05, "loss": 0.524, "step": 47896 }, { "epoch": 2.89, "learning_rate": 2.6378012161815524e-05, "loss": 0.3515, "step": 47898 }, { "epoch": 2.89, "learning_rate": 2.6376503251701294e-05, "loss": 0.6678, "step": 47900 }, { "epoch": 2.89, "learning_rate": 2.6374994341587074e-05, "loss": 0.425, "step": 47902 }, { "epoch": 2.89, "learning_rate": 2.637348543147285e-05, "loss": 0.5362, "step": 47904 }, { "epoch": 2.89, "learning_rate": 2.637197652135862e-05, "loss": 0.6306, "step": 47906 }, { "epoch": 2.89, "learning_rate": 2.6370467611244397e-05, "loss": 0.675, "step": 47908 }, { "epoch": 2.89, "learning_rate": 2.6368958701130177e-05, "loss": 0.5361, "step": 47910 }, { "epoch": 2.89, "learning_rate": 2.6367449791015953e-05, "loss": 0.6115, "step": 47912 }, { "epoch": 2.89, "learning_rate": 2.6365940880901723e-05, "loss": 0.3695, "step": 47914 }, { "epoch": 2.89, "learning_rate": 2.63644319707875e-05, "loss": 0.5963, "step": 47916 }, { "epoch": 2.89, "learning_rate": 2.636292306067328e-05, "loss": 0.618, "step": 47918 }, { "epoch": 2.89, "learning_rate": 2.6361414150559056e-05, "loss": 0.5115, "step": 47920 }, { "epoch": 2.89, "learning_rate": 2.6359905240444825e-05, "loss": 0.5167, "step": 47922 }, { "epoch": 2.89, "learning_rate": 2.6358396330330602e-05, "loss": 0.5149, "step": 47924 }, { "epoch": 2.89, "learning_rate": 2.6356887420216382e-05, "loss": 0.4939, "step": 47926 }, { "epoch": 2.89, "learning_rate": 2.635537851010215e-05, "loss": 0.7065, "step": 47928 }, { "epoch": 2.89, "learning_rate": 2.6353869599987928e-05, "loss": 0.4571, "step": 47930 }, { "epoch": 2.89, "learning_rate": 2.6352360689873705e-05, "loss": 0.6003, "step": 47932 }, { "epoch": 2.89, "learning_rate": 2.6350851779759485e-05, "loss": 0.3393, "step": 47934 }, { "epoch": 2.89, "learning_rate": 2.6349342869645254e-05, "loss": 0.4595, "step": 47936 }, { "epoch": 2.89, "learning_rate": 2.634783395953103e-05, "loss": 0.4781, "step": 47938 }, { "epoch": 2.89, "learning_rate": 2.634632504941681e-05, "loss": 0.628, "step": 47940 }, { "epoch": 2.89, "learning_rate": 2.634481613930258e-05, "loss": 0.4824, "step": 47942 }, { "epoch": 2.89, "learning_rate": 2.6343307229188357e-05, "loss": 0.5034, "step": 47944 }, { "epoch": 2.89, "learning_rate": 2.6341798319074133e-05, "loss": 0.3463, "step": 47946 }, { "epoch": 2.89, "learning_rate": 2.6340289408959913e-05, "loss": 0.5827, "step": 47948 }, { "epoch": 2.89, "learning_rate": 2.6338780498845683e-05, "loss": 0.5156, "step": 47950 }, { "epoch": 2.89, "learning_rate": 2.633727158873146e-05, "loss": 0.491, "step": 47952 }, { "epoch": 2.89, "learning_rate": 2.6335762678617236e-05, "loss": 0.4653, "step": 47954 }, { "epoch": 2.89, "learning_rate": 2.633425376850301e-05, "loss": 0.5197, "step": 47956 }, { "epoch": 2.89, "learning_rate": 2.6332744858388786e-05, "loss": 0.5593, "step": 47958 }, { "epoch": 2.89, "learning_rate": 2.6331235948274562e-05, "loss": 0.4067, "step": 47960 }, { "epoch": 2.89, "learning_rate": 2.632972703816034e-05, "loss": 0.4763, "step": 47962 }, { "epoch": 2.89, "learning_rate": 2.6328218128046112e-05, "loss": 0.5165, "step": 47964 }, { "epoch": 2.9, "learning_rate": 2.632670921793189e-05, "loss": 0.6079, "step": 47966 }, { "epoch": 2.9, "learning_rate": 2.6325200307817665e-05, "loss": 0.6105, "step": 47968 }, { "epoch": 2.9, "learning_rate": 2.632369139770344e-05, "loss": 0.544, "step": 47970 }, { "epoch": 2.9, "learning_rate": 2.6322182487589215e-05, "loss": 0.4997, "step": 47972 }, { "epoch": 2.9, "learning_rate": 2.632067357747499e-05, "loss": 0.4577, "step": 47974 }, { "epoch": 2.9, "learning_rate": 2.6319164667360768e-05, "loss": 0.6551, "step": 47976 }, { "epoch": 2.9, "learning_rate": 2.631765575724654e-05, "loss": 0.4735, "step": 47978 }, { "epoch": 2.9, "learning_rate": 2.6316146847132317e-05, "loss": 0.5373, "step": 47980 }, { "epoch": 2.9, "learning_rate": 2.6314637937018094e-05, "loss": 0.6084, "step": 47982 }, { "epoch": 2.9, "learning_rate": 2.631312902690387e-05, "loss": 0.5004, "step": 47984 }, { "epoch": 2.9, "learning_rate": 2.6311620116789643e-05, "loss": 0.3378, "step": 47986 }, { "epoch": 2.9, "learning_rate": 2.631011120667542e-05, "loss": 0.4239, "step": 47988 }, { "epoch": 2.9, "learning_rate": 2.6308602296561196e-05, "loss": 0.7007, "step": 47990 }, { "epoch": 2.9, "learning_rate": 2.630709338644697e-05, "loss": 0.5481, "step": 47992 }, { "epoch": 2.9, "learning_rate": 2.6305584476332746e-05, "loss": 0.4887, "step": 47994 }, { "epoch": 2.9, "learning_rate": 2.6304075566218523e-05, "loss": 0.3756, "step": 47996 }, { "epoch": 2.9, "learning_rate": 2.63025666561043e-05, "loss": 0.2598, "step": 47998 }, { "epoch": 2.9, "learning_rate": 2.6301057745990072e-05, "loss": 0.3882, "step": 48000 }, { "epoch": 2.9, "learning_rate": 2.629954883587585e-05, "loss": 0.5498, "step": 48002 }, { "epoch": 2.9, "learning_rate": 2.6298039925761625e-05, "loss": 0.4559, "step": 48004 }, { "epoch": 2.9, "learning_rate": 2.6296531015647395e-05, "loss": 0.4062, "step": 48006 }, { "epoch": 2.9, "learning_rate": 2.6295022105533175e-05, "loss": 0.7386, "step": 48008 }, { "epoch": 2.9, "learning_rate": 2.629351319541895e-05, "loss": 0.494, "step": 48010 }, { "epoch": 2.9, "learning_rate": 2.6292004285304728e-05, "loss": 0.5636, "step": 48012 }, { "epoch": 2.9, "learning_rate": 2.6290495375190498e-05, "loss": 0.5386, "step": 48014 }, { "epoch": 2.9, "learning_rate": 2.6288986465076278e-05, "loss": 0.2449, "step": 48016 }, { "epoch": 2.9, "learning_rate": 2.6287477554962054e-05, "loss": 0.3104, "step": 48018 }, { "epoch": 2.9, "learning_rate": 2.628596864484783e-05, "loss": 0.4131, "step": 48020 }, { "epoch": 2.9, "learning_rate": 2.62844597347336e-05, "loss": 0.3901, "step": 48022 }, { "epoch": 2.9, "learning_rate": 2.628295082461938e-05, "loss": 0.3898, "step": 48024 }, { "epoch": 2.9, "learning_rate": 2.6281441914505157e-05, "loss": 0.4168, "step": 48026 }, { "epoch": 2.9, "learning_rate": 2.6279933004390926e-05, "loss": 0.4418, "step": 48028 }, { "epoch": 2.9, "learning_rate": 2.6278424094276703e-05, "loss": 0.4769, "step": 48030 }, { "epoch": 2.9, "learning_rate": 2.6276915184162483e-05, "loss": 0.4519, "step": 48032 }, { "epoch": 2.9, "learning_rate": 2.627540627404826e-05, "loss": 0.4161, "step": 48034 }, { "epoch": 2.9, "learning_rate": 2.627389736393403e-05, "loss": 0.4153, "step": 48036 }, { "epoch": 2.9, "learning_rate": 2.6272388453819806e-05, "loss": 0.5855, "step": 48038 }, { "epoch": 2.9, "learning_rate": 2.6270879543705586e-05, "loss": 0.3957, "step": 48040 }, { "epoch": 2.9, "learning_rate": 2.6269370633591355e-05, "loss": 0.3233, "step": 48042 }, { "epoch": 2.9, "learning_rate": 2.6267861723477132e-05, "loss": 0.5576, "step": 48044 }, { "epoch": 2.9, "learning_rate": 2.626635281336291e-05, "loss": 0.5326, "step": 48046 }, { "epoch": 2.9, "learning_rate": 2.6264843903248688e-05, "loss": 0.466, "step": 48048 }, { "epoch": 2.9, "learning_rate": 2.6263334993134458e-05, "loss": 0.3793, "step": 48050 }, { "epoch": 2.9, "learning_rate": 2.6261826083020234e-05, "loss": 0.4778, "step": 48052 }, { "epoch": 2.9, "learning_rate": 2.626031717290601e-05, "loss": 0.6271, "step": 48054 }, { "epoch": 2.9, "learning_rate": 2.6258808262791784e-05, "loss": 0.415, "step": 48056 }, { "epoch": 2.9, "learning_rate": 2.625729935267756e-05, "loss": 0.4346, "step": 48058 }, { "epoch": 2.9, "learning_rate": 2.6255790442563337e-05, "loss": 0.2922, "step": 48060 }, { "epoch": 2.9, "learning_rate": 2.6254281532449117e-05, "loss": 0.4197, "step": 48062 }, { "epoch": 2.9, "learning_rate": 2.6252772622334887e-05, "loss": 0.3831, "step": 48064 }, { "epoch": 2.9, "learning_rate": 2.6251263712220663e-05, "loss": 0.4743, "step": 48066 }, { "epoch": 2.9, "learning_rate": 2.624975480210644e-05, "loss": 0.7421, "step": 48068 }, { "epoch": 2.9, "learning_rate": 2.6248245891992213e-05, "loss": 0.3632, "step": 48070 }, { "epoch": 2.9, "learning_rate": 2.624673698187799e-05, "loss": 0.5954, "step": 48072 }, { "epoch": 2.9, "learning_rate": 2.6245228071763766e-05, "loss": 0.4817, "step": 48074 }, { "epoch": 2.9, "learning_rate": 2.6243719161649542e-05, "loss": 0.4631, "step": 48076 }, { "epoch": 2.9, "learning_rate": 2.6242210251535316e-05, "loss": 0.6654, "step": 48078 }, { "epoch": 2.9, "learning_rate": 2.6240701341421092e-05, "loss": 0.5246, "step": 48080 }, { "epoch": 2.9, "learning_rate": 2.623919243130687e-05, "loss": 0.5594, "step": 48082 }, { "epoch": 2.9, "learning_rate": 2.6237683521192645e-05, "loss": 0.5217, "step": 48084 }, { "epoch": 2.9, "learning_rate": 2.6236174611078418e-05, "loss": 0.4767, "step": 48086 }, { "epoch": 2.9, "learning_rate": 2.6234665700964195e-05, "loss": 0.4918, "step": 48088 }, { "epoch": 2.9, "learning_rate": 2.623315679084997e-05, "loss": 0.4867, "step": 48090 }, { "epoch": 2.9, "learning_rate": 2.6231647880735744e-05, "loss": 0.4555, "step": 48092 }, { "epoch": 2.9, "learning_rate": 2.623013897062152e-05, "loss": 0.6936, "step": 48094 }, { "epoch": 2.9, "learning_rate": 2.6228630060507297e-05, "loss": 0.467, "step": 48096 }, { "epoch": 2.9, "learning_rate": 2.6227121150393074e-05, "loss": 0.6008, "step": 48098 }, { "epoch": 2.9, "learning_rate": 2.6225612240278847e-05, "loss": 0.4149, "step": 48100 }, { "epoch": 2.9, "learning_rate": 2.6224103330164624e-05, "loss": 0.7322, "step": 48102 }, { "epoch": 2.9, "learning_rate": 2.62225944200504e-05, "loss": 0.5601, "step": 48104 }, { "epoch": 2.9, "learning_rate": 2.6221085509936173e-05, "loss": 0.6726, "step": 48106 }, { "epoch": 2.9, "learning_rate": 2.621957659982195e-05, "loss": 0.4747, "step": 48108 }, { "epoch": 2.9, "learning_rate": 2.6218067689707726e-05, "loss": 0.5096, "step": 48110 }, { "epoch": 2.9, "learning_rate": 2.6216558779593503e-05, "loss": 0.4377, "step": 48112 }, { "epoch": 2.9, "learning_rate": 2.6215049869479276e-05, "loss": 0.6966, "step": 48114 }, { "epoch": 2.9, "learning_rate": 2.6213540959365052e-05, "loss": 0.4024, "step": 48116 }, { "epoch": 2.9, "learning_rate": 2.621203204925083e-05, "loss": 0.5643, "step": 48118 }, { "epoch": 2.9, "learning_rate": 2.62105231391366e-05, "loss": 0.4555, "step": 48120 }, { "epoch": 2.9, "learning_rate": 2.620901422902238e-05, "loss": 0.363, "step": 48122 }, { "epoch": 2.9, "learning_rate": 2.6207505318908155e-05, "loss": 0.3928, "step": 48124 }, { "epoch": 2.9, "learning_rate": 2.620599640879393e-05, "loss": 0.5341, "step": 48126 }, { "epoch": 2.9, "learning_rate": 2.62044874986797e-05, "loss": 0.703, "step": 48128 }, { "epoch": 2.9, "learning_rate": 2.620297858856548e-05, "loss": 0.5311, "step": 48130 }, { "epoch": 2.91, "learning_rate": 2.6201469678451258e-05, "loss": 0.4833, "step": 48132 }, { "epoch": 2.91, "learning_rate": 2.6199960768337034e-05, "loss": 0.4234, "step": 48134 }, { "epoch": 2.91, "learning_rate": 2.6198451858222804e-05, "loss": 0.3999, "step": 48136 }, { "epoch": 2.91, "learning_rate": 2.6196942948108584e-05, "loss": 0.6708, "step": 48138 }, { "epoch": 2.91, "learning_rate": 2.619543403799436e-05, "loss": 0.5573, "step": 48140 }, { "epoch": 2.91, "learning_rate": 2.619392512788013e-05, "loss": 0.5387, "step": 48142 }, { "epoch": 2.91, "learning_rate": 2.6192416217765907e-05, "loss": 0.4177, "step": 48144 }, { "epoch": 2.91, "learning_rate": 2.6190907307651687e-05, "loss": 0.5669, "step": 48146 }, { "epoch": 2.91, "learning_rate": 2.6189398397537463e-05, "loss": 0.5189, "step": 48148 }, { "epoch": 2.91, "learning_rate": 2.6187889487423233e-05, "loss": 0.9364, "step": 48150 }, { "epoch": 2.91, "learning_rate": 2.618638057730901e-05, "loss": 0.4933, "step": 48152 }, { "epoch": 2.91, "learning_rate": 2.618487166719479e-05, "loss": 0.6213, "step": 48154 }, { "epoch": 2.91, "learning_rate": 2.618336275708056e-05, "loss": 0.6341, "step": 48156 }, { "epoch": 2.91, "learning_rate": 2.6181853846966335e-05, "loss": 0.5121, "step": 48158 }, { "epoch": 2.91, "learning_rate": 2.6180344936852112e-05, "loss": 0.8137, "step": 48160 }, { "epoch": 2.91, "learning_rate": 2.6178836026737892e-05, "loss": 0.3713, "step": 48162 }, { "epoch": 2.91, "learning_rate": 2.617732711662366e-05, "loss": 0.5049, "step": 48164 }, { "epoch": 2.91, "learning_rate": 2.6175818206509438e-05, "loss": 0.3923, "step": 48166 }, { "epoch": 2.91, "learning_rate": 2.6174309296395215e-05, "loss": 0.636, "step": 48168 }, { "epoch": 2.91, "learning_rate": 2.6172800386280988e-05, "loss": 0.4629, "step": 48170 }, { "epoch": 2.91, "learning_rate": 2.6171291476166764e-05, "loss": 0.504, "step": 48172 }, { "epoch": 2.91, "learning_rate": 2.616978256605254e-05, "loss": 0.4306, "step": 48174 }, { "epoch": 2.91, "learning_rate": 2.6168273655938317e-05, "loss": 0.5349, "step": 48176 }, { "epoch": 2.91, "learning_rate": 2.616676474582409e-05, "loss": 0.478, "step": 48178 }, { "epoch": 2.91, "learning_rate": 2.6165255835709867e-05, "loss": 0.5112, "step": 48180 }, { "epoch": 2.91, "learning_rate": 2.6163746925595643e-05, "loss": 0.38, "step": 48182 }, { "epoch": 2.91, "learning_rate": 2.6162238015481423e-05, "loss": 0.4578, "step": 48184 }, { "epoch": 2.91, "learning_rate": 2.6160729105367193e-05, "loss": 0.6829, "step": 48186 }, { "epoch": 2.91, "learning_rate": 2.615922019525297e-05, "loss": 0.3485, "step": 48188 }, { "epoch": 2.91, "learning_rate": 2.6157711285138746e-05, "loss": 0.4949, "step": 48190 }, { "epoch": 2.91, "learning_rate": 2.615620237502452e-05, "loss": 0.5752, "step": 48192 }, { "epoch": 2.91, "learning_rate": 2.6154693464910296e-05, "loss": 0.4214, "step": 48194 }, { "epoch": 2.91, "learning_rate": 2.6153184554796072e-05, "loss": 0.5285, "step": 48196 }, { "epoch": 2.91, "learning_rate": 2.615167564468185e-05, "loss": 0.5887, "step": 48198 }, { "epoch": 2.91, "learning_rate": 2.6150166734567622e-05, "loss": 0.4412, "step": 48200 }, { "epoch": 2.91, "learning_rate": 2.61486578244534e-05, "loss": 0.6028, "step": 48202 }, { "epoch": 2.91, "learning_rate": 2.6147148914339175e-05, "loss": 0.3635, "step": 48204 }, { "epoch": 2.91, "learning_rate": 2.6145640004224948e-05, "loss": 0.3312, "step": 48206 }, { "epoch": 2.91, "learning_rate": 2.6144131094110725e-05, "loss": 0.5221, "step": 48208 }, { "epoch": 2.91, "learning_rate": 2.61426221839965e-05, "loss": 0.6441, "step": 48210 }, { "epoch": 2.91, "learning_rate": 2.6141113273882278e-05, "loss": 0.5651, "step": 48212 }, { "epoch": 2.91, "learning_rate": 2.613960436376805e-05, "loss": 0.5918, "step": 48214 }, { "epoch": 2.91, "learning_rate": 2.6138095453653827e-05, "loss": 0.4084, "step": 48216 }, { "epoch": 2.91, "learning_rate": 2.6136586543539604e-05, "loss": 0.4549, "step": 48218 }, { "epoch": 2.91, "learning_rate": 2.6135077633425377e-05, "loss": 0.4463, "step": 48220 }, { "epoch": 2.91, "learning_rate": 2.6133568723311153e-05, "loss": 0.4984, "step": 48222 }, { "epoch": 2.91, "learning_rate": 2.613205981319693e-05, "loss": 0.3385, "step": 48224 }, { "epoch": 2.91, "learning_rate": 2.6130550903082706e-05, "loss": 0.6128, "step": 48226 }, { "epoch": 2.91, "learning_rate": 2.612904199296848e-05, "loss": 0.5852, "step": 48228 }, { "epoch": 2.91, "learning_rate": 2.6127533082854256e-05, "loss": 0.4886, "step": 48230 }, { "epoch": 2.91, "learning_rate": 2.6126024172740033e-05, "loss": 0.3655, "step": 48232 }, { "epoch": 2.91, "learning_rate": 2.612451526262581e-05, "loss": 0.7374, "step": 48234 }, { "epoch": 2.91, "learning_rate": 2.6123006352511582e-05, "loss": 0.5564, "step": 48236 }, { "epoch": 2.91, "learning_rate": 2.612149744239736e-05, "loss": 0.4249, "step": 48238 }, { "epoch": 2.91, "learning_rate": 2.6119988532283135e-05, "loss": 0.5463, "step": 48240 }, { "epoch": 2.91, "learning_rate": 2.6118479622168905e-05, "loss": 0.4093, "step": 48242 }, { "epoch": 2.91, "learning_rate": 2.6116970712054685e-05, "loss": 0.6094, "step": 48244 }, { "epoch": 2.91, "learning_rate": 2.611546180194046e-05, "loss": 0.5143, "step": 48246 }, { "epoch": 2.91, "learning_rate": 2.6113952891826238e-05, "loss": 0.6042, "step": 48248 }, { "epoch": 2.91, "learning_rate": 2.6112443981712008e-05, "loss": 0.3942, "step": 48250 }, { "epoch": 2.91, "learning_rate": 2.6110935071597788e-05, "loss": 0.5726, "step": 48252 }, { "epoch": 2.91, "learning_rate": 2.6109426161483564e-05, "loss": 0.5075, "step": 48254 }, { "epoch": 2.91, "learning_rate": 2.6107917251369334e-05, "loss": 0.8318, "step": 48256 }, { "epoch": 2.91, "learning_rate": 2.610640834125511e-05, "loss": 0.6182, "step": 48258 }, { "epoch": 2.91, "learning_rate": 2.610489943114089e-05, "loss": 0.7695, "step": 48260 }, { "epoch": 2.91, "learning_rate": 2.6103390521026667e-05, "loss": 0.5446, "step": 48262 }, { "epoch": 2.91, "learning_rate": 2.6101881610912437e-05, "loss": 0.3727, "step": 48264 }, { "epoch": 2.91, "learning_rate": 2.6100372700798213e-05, "loss": 0.5916, "step": 48266 }, { "epoch": 2.91, "learning_rate": 2.6098863790683993e-05, "loss": 0.4251, "step": 48268 }, { "epoch": 2.91, "learning_rate": 2.6097354880569763e-05, "loss": 0.4996, "step": 48270 }, { "epoch": 2.91, "learning_rate": 2.609584597045554e-05, "loss": 0.6088, "step": 48272 }, { "epoch": 2.91, "learning_rate": 2.6094337060341316e-05, "loss": 0.3027, "step": 48274 }, { "epoch": 2.91, "learning_rate": 2.6092828150227096e-05, "loss": 0.7432, "step": 48276 }, { "epoch": 2.91, "learning_rate": 2.6091319240112865e-05, "loss": 0.4657, "step": 48278 }, { "epoch": 2.91, "learning_rate": 2.6089810329998642e-05, "loss": 0.4482, "step": 48280 }, { "epoch": 2.91, "learning_rate": 2.608830141988442e-05, "loss": 0.5293, "step": 48282 }, { "epoch": 2.91, "learning_rate": 2.608679250977019e-05, "loss": 0.498, "step": 48284 }, { "epoch": 2.91, "learning_rate": 2.6085283599655968e-05, "loss": 0.7025, "step": 48286 }, { "epoch": 2.91, "learning_rate": 2.6083774689541745e-05, "loss": 0.445, "step": 48288 }, { "epoch": 2.91, "learning_rate": 2.608226577942752e-05, "loss": 0.4288, "step": 48290 }, { "epoch": 2.91, "learning_rate": 2.6080756869313294e-05, "loss": 0.4712, "step": 48292 }, { "epoch": 2.91, "learning_rate": 2.607924795919907e-05, "loss": 0.5134, "step": 48294 }, { "epoch": 2.92, "learning_rate": 2.6077739049084847e-05, "loss": 0.3923, "step": 48296 }, { "epoch": 2.92, "learning_rate": 2.6076230138970627e-05, "loss": 0.4696, "step": 48298 }, { "epoch": 2.92, "learning_rate": 2.6074721228856397e-05, "loss": 0.5288, "step": 48300 }, { "epoch": 2.92, "learning_rate": 2.6073212318742173e-05, "loss": 0.5842, "step": 48302 }, { "epoch": 2.92, "learning_rate": 2.607170340862795e-05, "loss": 0.5099, "step": 48304 }, { "epoch": 2.92, "learning_rate": 2.6070194498513723e-05, "loss": 0.5383, "step": 48306 }, { "epoch": 2.92, "learning_rate": 2.60686855883995e-05, "loss": 0.4388, "step": 48308 }, { "epoch": 2.92, "learning_rate": 2.6067176678285276e-05, "loss": 0.5464, "step": 48310 }, { "epoch": 2.92, "learning_rate": 2.6065667768171053e-05, "loss": 0.5093, "step": 48312 }, { "epoch": 2.92, "learning_rate": 2.6064158858056826e-05, "loss": 0.6442, "step": 48314 }, { "epoch": 2.92, "learning_rate": 2.6062649947942602e-05, "loss": 0.5328, "step": 48316 }, { "epoch": 2.92, "learning_rate": 2.606114103782838e-05, "loss": 0.935, "step": 48318 }, { "epoch": 2.92, "learning_rate": 2.6059632127714152e-05, "loss": 0.4255, "step": 48320 }, { "epoch": 2.92, "learning_rate": 2.6058123217599928e-05, "loss": 0.755, "step": 48322 }, { "epoch": 2.92, "learning_rate": 2.6056614307485705e-05, "loss": 0.4774, "step": 48324 }, { "epoch": 2.92, "learning_rate": 2.605510539737148e-05, "loss": 0.6461, "step": 48326 }, { "epoch": 2.92, "learning_rate": 2.6053596487257254e-05, "loss": 0.557, "step": 48328 }, { "epoch": 2.92, "learning_rate": 2.605208757714303e-05, "loss": 0.6858, "step": 48330 }, { "epoch": 2.92, "learning_rate": 2.6050578667028807e-05, "loss": 0.5184, "step": 48332 }, { "epoch": 2.92, "learning_rate": 2.604906975691458e-05, "loss": 0.5421, "step": 48334 }, { "epoch": 2.92, "learning_rate": 2.6047560846800357e-05, "loss": 0.4414, "step": 48336 }, { "epoch": 2.92, "learning_rate": 2.6046051936686134e-05, "loss": 0.49, "step": 48338 }, { "epoch": 2.92, "learning_rate": 2.604454302657191e-05, "loss": 0.5756, "step": 48340 }, { "epoch": 2.92, "learning_rate": 2.6043034116457683e-05, "loss": 0.4317, "step": 48342 }, { "epoch": 2.92, "learning_rate": 2.604152520634346e-05, "loss": 0.6471, "step": 48344 }, { "epoch": 2.92, "learning_rate": 2.6040016296229236e-05, "loss": 0.4215, "step": 48346 }, { "epoch": 2.92, "learning_rate": 2.6038507386115013e-05, "loss": 0.6047, "step": 48348 }, { "epoch": 2.92, "learning_rate": 2.6036998476000786e-05, "loss": 0.7587, "step": 48350 }, { "epoch": 2.92, "learning_rate": 2.6035489565886562e-05, "loss": 0.6349, "step": 48352 }, { "epoch": 2.92, "learning_rate": 2.603398065577234e-05, "loss": 0.506, "step": 48354 }, { "epoch": 2.92, "learning_rate": 2.603247174565811e-05, "loss": 0.4259, "step": 48356 }, { "epoch": 2.92, "learning_rate": 2.603096283554389e-05, "loss": 0.4061, "step": 48358 }, { "epoch": 2.92, "learning_rate": 2.6029453925429665e-05, "loss": 0.513, "step": 48360 }, { "epoch": 2.92, "learning_rate": 2.602794501531544e-05, "loss": 0.7299, "step": 48362 }, { "epoch": 2.92, "learning_rate": 2.602643610520121e-05, "loss": 0.5527, "step": 48364 }, { "epoch": 2.92, "learning_rate": 2.602492719508699e-05, "loss": 0.4487, "step": 48366 }, { "epoch": 2.92, "learning_rate": 2.6023418284972768e-05, "loss": 0.5268, "step": 48368 }, { "epoch": 2.92, "learning_rate": 2.6021909374858538e-05, "loss": 0.4566, "step": 48370 }, { "epoch": 2.92, "learning_rate": 2.6020400464744314e-05, "loss": 0.4074, "step": 48372 }, { "epoch": 2.92, "learning_rate": 2.6018891554630094e-05, "loss": 0.9053, "step": 48374 }, { "epoch": 2.92, "learning_rate": 2.601738264451587e-05, "loss": 0.4452, "step": 48376 }, { "epoch": 2.92, "learning_rate": 2.601587373440164e-05, "loss": 0.5377, "step": 48378 }, { "epoch": 2.92, "learning_rate": 2.6014364824287417e-05, "loss": 0.4699, "step": 48380 }, { "epoch": 2.92, "learning_rate": 2.6012855914173197e-05, "loss": 0.5708, "step": 48382 }, { "epoch": 2.92, "learning_rate": 2.6011347004058966e-05, "loss": 0.6735, "step": 48384 }, { "epoch": 2.92, "learning_rate": 2.6009838093944743e-05, "loss": 0.7681, "step": 48386 }, { "epoch": 2.92, "learning_rate": 2.600832918383052e-05, "loss": 0.607, "step": 48388 }, { "epoch": 2.92, "learning_rate": 2.60068202737163e-05, "loss": 0.5567, "step": 48390 }, { "epoch": 2.92, "learning_rate": 2.600531136360207e-05, "loss": 0.4193, "step": 48392 }, { "epoch": 2.92, "learning_rate": 2.6003802453487846e-05, "loss": 0.2349, "step": 48394 }, { "epoch": 2.92, "learning_rate": 2.6002293543373622e-05, "loss": 0.5163, "step": 48396 }, { "epoch": 2.92, "learning_rate": 2.6000784633259402e-05, "loss": 0.5657, "step": 48398 }, { "epoch": 2.92, "learning_rate": 2.599927572314517e-05, "loss": 0.6177, "step": 48400 }, { "epoch": 2.92, "learning_rate": 2.5997766813030948e-05, "loss": 0.5671, "step": 48402 }, { "epoch": 2.92, "learning_rate": 2.5996257902916725e-05, "loss": 0.4825, "step": 48404 }, { "epoch": 2.92, "learning_rate": 2.5994748992802498e-05, "loss": 0.3968, "step": 48406 }, { "epoch": 2.92, "learning_rate": 2.5993240082688274e-05, "loss": 0.4826, "step": 48408 }, { "epoch": 2.92, "learning_rate": 2.599173117257405e-05, "loss": 0.5282, "step": 48410 }, { "epoch": 2.92, "learning_rate": 2.5990222262459827e-05, "loss": 0.4179, "step": 48412 }, { "epoch": 2.92, "learning_rate": 2.59887133523456e-05, "loss": 0.647, "step": 48414 }, { "epoch": 2.92, "learning_rate": 2.5987204442231377e-05, "loss": 0.6052, "step": 48416 }, { "epoch": 2.92, "learning_rate": 2.5985695532117154e-05, "loss": 0.4236, "step": 48418 }, { "epoch": 2.92, "learning_rate": 2.5984186622002927e-05, "loss": 0.4908, "step": 48420 }, { "epoch": 2.92, "learning_rate": 2.5982677711888703e-05, "loss": 0.5339, "step": 48422 }, { "epoch": 2.92, "learning_rate": 2.598116880177448e-05, "loss": 0.6445, "step": 48424 }, { "epoch": 2.92, "learning_rate": 2.5979659891660256e-05, "loss": 0.6107, "step": 48426 }, { "epoch": 2.92, "learning_rate": 2.597815098154603e-05, "loss": 0.5184, "step": 48428 }, { "epoch": 2.92, "learning_rate": 2.5976642071431806e-05, "loss": 0.3557, "step": 48430 }, { "epoch": 2.92, "learning_rate": 2.5975133161317582e-05, "loss": 0.4974, "step": 48432 }, { "epoch": 2.92, "learning_rate": 2.5973624251203355e-05, "loss": 0.3705, "step": 48434 }, { "epoch": 2.92, "learning_rate": 2.5972115341089132e-05, "loss": 0.4599, "step": 48436 }, { "epoch": 2.92, "learning_rate": 2.597060643097491e-05, "loss": 0.5445, "step": 48438 }, { "epoch": 2.92, "learning_rate": 2.5969097520860685e-05, "loss": 0.7676, "step": 48440 }, { "epoch": 2.92, "learning_rate": 2.5967588610746458e-05, "loss": 0.4469, "step": 48442 }, { "epoch": 2.92, "learning_rate": 2.5966079700632235e-05, "loss": 0.5754, "step": 48444 }, { "epoch": 2.92, "learning_rate": 2.596457079051801e-05, "loss": 0.4922, "step": 48446 }, { "epoch": 2.92, "learning_rate": 2.5963061880403784e-05, "loss": 0.6178, "step": 48448 }, { "epoch": 2.92, "learning_rate": 2.596155297028956e-05, "loss": 0.47, "step": 48450 }, { "epoch": 2.92, "learning_rate": 2.5960044060175337e-05, "loss": 0.5976, "step": 48452 }, { "epoch": 2.92, "learning_rate": 2.5958535150061114e-05, "loss": 0.4954, "step": 48454 }, { "epoch": 2.92, "learning_rate": 2.5957026239946887e-05, "loss": 0.5331, "step": 48456 }, { "epoch": 2.92, "learning_rate": 2.5955517329832663e-05, "loss": 0.5287, "step": 48458 }, { "epoch": 2.92, "learning_rate": 2.595400841971844e-05, "loss": 0.597, "step": 48460 }, { "epoch": 2.93, "learning_rate": 2.5952499509604216e-05, "loss": 0.7132, "step": 48462 }, { "epoch": 2.93, "learning_rate": 2.595099059948999e-05, "loss": 0.4718, "step": 48464 }, { "epoch": 2.93, "learning_rate": 2.5949481689375766e-05, "loss": 0.5143, "step": 48466 }, { "epoch": 2.93, "learning_rate": 2.5947972779261543e-05, "loss": 0.4578, "step": 48468 }, { "epoch": 2.93, "learning_rate": 2.5946463869147312e-05, "loss": 0.5877, "step": 48470 }, { "epoch": 2.93, "learning_rate": 2.5944954959033092e-05, "loss": 0.4558, "step": 48472 }, { "epoch": 2.93, "learning_rate": 2.594344604891887e-05, "loss": 0.4723, "step": 48474 }, { "epoch": 2.93, "learning_rate": 2.5941937138804645e-05, "loss": 0.5341, "step": 48476 }, { "epoch": 2.93, "learning_rate": 2.5940428228690415e-05, "loss": 0.486, "step": 48478 }, { "epoch": 2.93, "learning_rate": 2.5938919318576195e-05, "loss": 0.5393, "step": 48480 }, { "epoch": 2.93, "learning_rate": 2.593741040846197e-05, "loss": 0.4803, "step": 48482 }, { "epoch": 2.93, "learning_rate": 2.593590149834774e-05, "loss": 0.401, "step": 48484 }, { "epoch": 2.93, "learning_rate": 2.5934392588233518e-05, "loss": 0.4854, "step": 48486 }, { "epoch": 2.93, "learning_rate": 2.5932883678119298e-05, "loss": 0.7414, "step": 48488 }, { "epoch": 2.93, "learning_rate": 2.5931374768005074e-05, "loss": 0.5484, "step": 48490 }, { "epoch": 2.93, "learning_rate": 2.5929865857890844e-05, "loss": 0.6506, "step": 48492 }, { "epoch": 2.93, "learning_rate": 2.592835694777662e-05, "loss": 0.6664, "step": 48494 }, { "epoch": 2.93, "learning_rate": 2.59268480376624e-05, "loss": 0.3979, "step": 48496 }, { "epoch": 2.93, "learning_rate": 2.592533912754817e-05, "loss": 0.4018, "step": 48498 }, { "epoch": 2.93, "learning_rate": 2.5923830217433947e-05, "loss": 0.4825, "step": 48500 }, { "epoch": 2.93, "learning_rate": 2.5922321307319723e-05, "loss": 0.3915, "step": 48502 }, { "epoch": 2.93, "learning_rate": 2.5920812397205503e-05, "loss": 0.5238, "step": 48504 }, { "epoch": 2.93, "learning_rate": 2.5919303487091273e-05, "loss": 0.4691, "step": 48506 }, { "epoch": 2.93, "learning_rate": 2.591779457697705e-05, "loss": 0.4879, "step": 48508 }, { "epoch": 2.93, "learning_rate": 2.5916285666862826e-05, "loss": 0.5083, "step": 48510 }, { "epoch": 2.93, "learning_rate": 2.5914776756748606e-05, "loss": 0.4637, "step": 48512 }, { "epoch": 2.93, "learning_rate": 2.5913267846634375e-05, "loss": 0.6101, "step": 48514 }, { "epoch": 2.93, "learning_rate": 2.5911758936520152e-05, "loss": 0.4625, "step": 48516 }, { "epoch": 2.93, "learning_rate": 2.591025002640593e-05, "loss": 0.447, "step": 48518 }, { "epoch": 2.93, "learning_rate": 2.59087411162917e-05, "loss": 0.5579, "step": 48520 }, { "epoch": 2.93, "learning_rate": 2.5907232206177478e-05, "loss": 0.5078, "step": 48522 }, { "epoch": 2.93, "learning_rate": 2.5905723296063255e-05, "loss": 0.4413, "step": 48524 }, { "epoch": 2.93, "learning_rate": 2.590421438594903e-05, "loss": 0.347, "step": 48526 }, { "epoch": 2.93, "learning_rate": 2.5902705475834804e-05, "loss": 0.6767, "step": 48528 }, { "epoch": 2.93, "learning_rate": 2.590119656572058e-05, "loss": 0.4675, "step": 48530 }, { "epoch": 2.93, "learning_rate": 2.5899687655606357e-05, "loss": 0.4855, "step": 48532 }, { "epoch": 2.93, "learning_rate": 2.589817874549213e-05, "loss": 0.5298, "step": 48534 }, { "epoch": 2.93, "learning_rate": 2.5896669835377907e-05, "loss": 0.7107, "step": 48536 }, { "epoch": 2.93, "learning_rate": 2.5895160925263683e-05, "loss": 0.52, "step": 48538 }, { "epoch": 2.93, "learning_rate": 2.589365201514946e-05, "loss": 0.496, "step": 48540 }, { "epoch": 2.93, "learning_rate": 2.5892143105035233e-05, "loss": 0.5258, "step": 48542 }, { "epoch": 2.93, "learning_rate": 2.589063419492101e-05, "loss": 0.5714, "step": 48544 }, { "epoch": 2.93, "learning_rate": 2.5889125284806786e-05, "loss": 0.5243, "step": 48546 }, { "epoch": 2.93, "learning_rate": 2.588761637469256e-05, "loss": 0.5074, "step": 48548 }, { "epoch": 2.93, "learning_rate": 2.5886107464578336e-05, "loss": 0.3947, "step": 48550 }, { "epoch": 2.93, "learning_rate": 2.5884598554464112e-05, "loss": 0.6057, "step": 48552 }, { "epoch": 2.93, "learning_rate": 2.588308964434989e-05, "loss": 0.6413, "step": 48554 }, { "epoch": 2.93, "learning_rate": 2.5881580734235662e-05, "loss": 0.3732, "step": 48556 }, { "epoch": 2.93, "learning_rate": 2.588007182412144e-05, "loss": 0.6903, "step": 48558 }, { "epoch": 2.93, "learning_rate": 2.5878562914007215e-05, "loss": 0.6115, "step": 48560 }, { "epoch": 2.93, "learning_rate": 2.587705400389299e-05, "loss": 0.6491, "step": 48562 }, { "epoch": 2.93, "learning_rate": 2.5875545093778764e-05, "loss": 0.6791, "step": 48564 }, { "epoch": 2.93, "learning_rate": 2.587403618366454e-05, "loss": 0.519, "step": 48566 }, { "epoch": 2.93, "learning_rate": 2.5872527273550317e-05, "loss": 0.4938, "step": 48568 }, { "epoch": 2.93, "learning_rate": 2.587101836343609e-05, "loss": 0.477, "step": 48570 }, { "epoch": 2.93, "learning_rate": 2.5869509453321867e-05, "loss": 0.4831, "step": 48572 }, { "epoch": 2.93, "learning_rate": 2.5868000543207644e-05, "loss": 0.6015, "step": 48574 }, { "epoch": 2.93, "learning_rate": 2.586649163309342e-05, "loss": 0.4447, "step": 48576 }, { "epoch": 2.93, "learning_rate": 2.5864982722979193e-05, "loss": 0.5676, "step": 48578 }, { "epoch": 2.93, "learning_rate": 2.586347381286497e-05, "loss": 0.4085, "step": 48580 }, { "epoch": 2.93, "learning_rate": 2.5861964902750746e-05, "loss": 0.7633, "step": 48582 }, { "epoch": 2.93, "learning_rate": 2.5860455992636516e-05, "loss": 0.6433, "step": 48584 }, { "epoch": 2.93, "learning_rate": 2.5858947082522296e-05, "loss": 0.2792, "step": 48586 }, { "epoch": 2.93, "learning_rate": 2.5857438172408072e-05, "loss": 0.4508, "step": 48588 }, { "epoch": 2.93, "learning_rate": 2.585592926229385e-05, "loss": 0.304, "step": 48590 }, { "epoch": 2.93, "learning_rate": 2.585442035217962e-05, "loss": 0.5864, "step": 48592 }, { "epoch": 2.93, "learning_rate": 2.58529114420654e-05, "loss": 0.5147, "step": 48594 }, { "epoch": 2.93, "learning_rate": 2.5851402531951175e-05, "loss": 0.4022, "step": 48596 }, { "epoch": 2.93, "learning_rate": 2.5849893621836945e-05, "loss": 0.4487, "step": 48598 }, { "epoch": 2.93, "learning_rate": 2.584838471172272e-05, "loss": 0.6322, "step": 48600 }, { "epoch": 2.93, "learning_rate": 2.58468758016085e-05, "loss": 0.4113, "step": 48602 }, { "epoch": 2.93, "learning_rate": 2.5845366891494278e-05, "loss": 0.4267, "step": 48604 }, { "epoch": 2.93, "learning_rate": 2.5843857981380048e-05, "loss": 0.4185, "step": 48606 }, { "epoch": 2.93, "learning_rate": 2.5842349071265824e-05, "loss": 0.6724, "step": 48608 }, { "epoch": 2.93, "learning_rate": 2.5840840161151604e-05, "loss": 0.626, "step": 48610 }, { "epoch": 2.93, "learning_rate": 2.5839331251037374e-05, "loss": 0.4519, "step": 48612 }, { "epoch": 2.93, "learning_rate": 2.583782234092315e-05, "loss": 0.4323, "step": 48614 }, { "epoch": 2.93, "learning_rate": 2.5836313430808927e-05, "loss": 0.8057, "step": 48616 }, { "epoch": 2.93, "learning_rate": 2.5834804520694707e-05, "loss": 0.7965, "step": 48618 }, { "epoch": 2.93, "learning_rate": 2.5833295610580476e-05, "loss": 0.6992, "step": 48620 }, { "epoch": 2.93, "learning_rate": 2.5831786700466253e-05, "loss": 0.5787, "step": 48622 }, { "epoch": 2.93, "learning_rate": 2.583027779035203e-05, "loss": 0.5377, "step": 48624 }, { "epoch": 2.93, "learning_rate": 2.582876888023781e-05, "loss": 0.5042, "step": 48626 }, { "epoch": 2.94, "learning_rate": 2.582725997012358e-05, "loss": 0.3166, "step": 48628 }, { "epoch": 2.94, "learning_rate": 2.5825751060009356e-05, "loss": 0.5146, "step": 48630 }, { "epoch": 2.94, "learning_rate": 2.5824242149895132e-05, "loss": 0.5183, "step": 48632 }, { "epoch": 2.94, "learning_rate": 2.5822733239780905e-05, "loss": 0.4815, "step": 48634 }, { "epoch": 2.94, "learning_rate": 2.582122432966668e-05, "loss": 0.5395, "step": 48636 }, { "epoch": 2.94, "learning_rate": 2.5819715419552458e-05, "loss": 0.5019, "step": 48638 }, { "epoch": 2.94, "learning_rate": 2.5818206509438235e-05, "loss": 0.5931, "step": 48640 }, { "epoch": 2.94, "learning_rate": 2.5816697599324008e-05, "loss": 0.5186, "step": 48642 }, { "epoch": 2.94, "learning_rate": 2.5815188689209784e-05, "loss": 0.573, "step": 48644 }, { "epoch": 2.94, "learning_rate": 2.581367977909556e-05, "loss": 0.6236, "step": 48646 }, { "epoch": 2.94, "learning_rate": 2.5812170868981334e-05, "loss": 0.5889, "step": 48648 }, { "epoch": 2.94, "learning_rate": 2.581066195886711e-05, "loss": 0.4437, "step": 48650 }, { "epoch": 2.94, "learning_rate": 2.5809153048752887e-05, "loss": 0.5328, "step": 48652 }, { "epoch": 2.94, "learning_rate": 2.5807644138638664e-05, "loss": 0.5952, "step": 48654 }, { "epoch": 2.94, "learning_rate": 2.5806135228524437e-05, "loss": 0.6199, "step": 48656 }, { "epoch": 2.94, "learning_rate": 2.5804626318410213e-05, "loss": 0.6376, "step": 48658 }, { "epoch": 2.94, "learning_rate": 2.580311740829599e-05, "loss": 0.4752, "step": 48660 }, { "epoch": 2.94, "learning_rate": 2.5801608498181763e-05, "loss": 0.3918, "step": 48662 }, { "epoch": 2.94, "learning_rate": 2.580009958806754e-05, "loss": 0.5001, "step": 48664 }, { "epoch": 2.94, "learning_rate": 2.5798590677953316e-05, "loss": 0.654, "step": 48666 }, { "epoch": 2.94, "learning_rate": 2.5797081767839092e-05, "loss": 0.5266, "step": 48668 }, { "epoch": 2.94, "learning_rate": 2.5795572857724865e-05, "loss": 0.4354, "step": 48670 }, { "epoch": 2.94, "learning_rate": 2.5794063947610642e-05, "loss": 0.4392, "step": 48672 }, { "epoch": 2.94, "learning_rate": 2.579255503749642e-05, "loss": 0.5512, "step": 48674 }, { "epoch": 2.94, "learning_rate": 2.5791046127382195e-05, "loss": 0.6195, "step": 48676 }, { "epoch": 2.94, "learning_rate": 2.5789537217267968e-05, "loss": 0.5223, "step": 48678 }, { "epoch": 2.94, "learning_rate": 2.5788028307153745e-05, "loss": 0.4668, "step": 48680 }, { "epoch": 2.94, "learning_rate": 2.578651939703952e-05, "loss": 0.2874, "step": 48682 }, { "epoch": 2.94, "learning_rate": 2.5785010486925294e-05, "loss": 0.562, "step": 48684 }, { "epoch": 2.94, "learning_rate": 2.578350157681107e-05, "loss": 0.6746, "step": 48686 }, { "epoch": 2.94, "learning_rate": 2.5781992666696847e-05, "loss": 0.7007, "step": 48688 }, { "epoch": 2.94, "learning_rate": 2.5780483756582624e-05, "loss": 0.5906, "step": 48690 }, { "epoch": 2.94, "learning_rate": 2.5778974846468397e-05, "loss": 0.5712, "step": 48692 }, { "epoch": 2.94, "learning_rate": 2.5777465936354173e-05, "loss": 0.5459, "step": 48694 }, { "epoch": 2.94, "learning_rate": 2.577595702623995e-05, "loss": 0.4005, "step": 48696 }, { "epoch": 2.94, "learning_rate": 2.577444811612572e-05, "loss": 0.5483, "step": 48698 }, { "epoch": 2.94, "learning_rate": 2.57729392060115e-05, "loss": 0.544, "step": 48700 }, { "epoch": 2.94, "learning_rate": 2.5771430295897276e-05, "loss": 0.4877, "step": 48702 }, { "epoch": 2.94, "learning_rate": 2.5769921385783053e-05, "loss": 0.4529, "step": 48704 }, { "epoch": 2.94, "learning_rate": 2.5768412475668822e-05, "loss": 0.4509, "step": 48706 }, { "epoch": 2.94, "learning_rate": 2.5766903565554602e-05, "loss": 0.3513, "step": 48708 }, { "epoch": 2.94, "learning_rate": 2.576539465544038e-05, "loss": 0.4832, "step": 48710 }, { "epoch": 2.94, "learning_rate": 2.576388574532615e-05, "loss": 0.5166, "step": 48712 }, { "epoch": 2.94, "learning_rate": 2.5762376835211925e-05, "loss": 0.5606, "step": 48714 }, { "epoch": 2.94, "learning_rate": 2.5760867925097705e-05, "loss": 0.3837, "step": 48716 }, { "epoch": 2.94, "learning_rate": 2.575935901498348e-05, "loss": 0.6197, "step": 48718 }, { "epoch": 2.94, "learning_rate": 2.575785010486925e-05, "loss": 0.3233, "step": 48720 }, { "epoch": 2.94, "learning_rate": 2.5756341194755028e-05, "loss": 0.4311, "step": 48722 }, { "epoch": 2.94, "learning_rate": 2.5754832284640808e-05, "loss": 0.5156, "step": 48724 }, { "epoch": 2.94, "learning_rate": 2.5753323374526584e-05, "loss": 0.5235, "step": 48726 }, { "epoch": 2.94, "learning_rate": 2.5751814464412354e-05, "loss": 0.5839, "step": 48728 }, { "epoch": 2.94, "learning_rate": 2.575030555429813e-05, "loss": 0.5462, "step": 48730 }, { "epoch": 2.94, "learning_rate": 2.574879664418391e-05, "loss": 0.3401, "step": 48732 }, { "epoch": 2.94, "learning_rate": 2.574728773406968e-05, "loss": 0.4921, "step": 48734 }, { "epoch": 2.94, "learning_rate": 2.5745778823955457e-05, "loss": 0.5713, "step": 48736 }, { "epoch": 2.94, "learning_rate": 2.5744269913841233e-05, "loss": 0.7081, "step": 48738 }, { "epoch": 2.94, "learning_rate": 2.5742761003727013e-05, "loss": 0.4189, "step": 48740 }, { "epoch": 2.94, "learning_rate": 2.5741252093612783e-05, "loss": 0.4462, "step": 48742 }, { "epoch": 2.94, "learning_rate": 2.573974318349856e-05, "loss": 0.5349, "step": 48744 }, { "epoch": 2.94, "learning_rate": 2.5738234273384336e-05, "loss": 0.5444, "step": 48746 }, { "epoch": 2.94, "learning_rate": 2.573672536327011e-05, "loss": 0.6276, "step": 48748 }, { "epoch": 2.94, "learning_rate": 2.5735216453155885e-05, "loss": 0.5612, "step": 48750 }, { "epoch": 2.94, "learning_rate": 2.5733707543041662e-05, "loss": 0.6795, "step": 48752 }, { "epoch": 2.94, "learning_rate": 2.573219863292744e-05, "loss": 0.47, "step": 48754 }, { "epoch": 2.94, "learning_rate": 2.573068972281321e-05, "loss": 0.4466, "step": 48756 }, { "epoch": 2.94, "learning_rate": 2.5729180812698988e-05, "loss": 0.6941, "step": 48758 }, { "epoch": 2.94, "learning_rate": 2.5727671902584765e-05, "loss": 0.5382, "step": 48760 }, { "epoch": 2.94, "learning_rate": 2.5726162992470538e-05, "loss": 0.4634, "step": 48762 }, { "epoch": 2.94, "learning_rate": 2.5724654082356314e-05, "loss": 0.545, "step": 48764 }, { "epoch": 2.94, "learning_rate": 2.572314517224209e-05, "loss": 0.6251, "step": 48766 }, { "epoch": 2.94, "learning_rate": 2.5721636262127867e-05, "loss": 0.6775, "step": 48768 }, { "epoch": 2.94, "learning_rate": 2.572012735201364e-05, "loss": 0.48, "step": 48770 }, { "epoch": 2.94, "learning_rate": 2.5718618441899417e-05, "loss": 0.5947, "step": 48772 }, { "epoch": 2.94, "learning_rate": 2.5717109531785193e-05, "loss": 0.4757, "step": 48774 }, { "epoch": 2.94, "learning_rate": 2.571560062167097e-05, "loss": 0.3325, "step": 48776 }, { "epoch": 2.94, "learning_rate": 2.5714091711556743e-05, "loss": 0.4934, "step": 48778 }, { "epoch": 2.94, "learning_rate": 2.571258280144252e-05, "loss": 0.4587, "step": 48780 }, { "epoch": 2.94, "learning_rate": 2.5711073891328296e-05, "loss": 0.5775, "step": 48782 }, { "epoch": 2.94, "learning_rate": 2.570956498121407e-05, "loss": 0.6045, "step": 48784 }, { "epoch": 2.94, "learning_rate": 2.5708056071099846e-05, "loss": 0.4084, "step": 48786 }, { "epoch": 2.94, "learning_rate": 2.5706547160985622e-05, "loss": 0.6409, "step": 48788 }, { "epoch": 2.94, "learning_rate": 2.57050382508714e-05, "loss": 0.4425, "step": 48790 }, { "epoch": 2.94, "learning_rate": 2.5703529340757172e-05, "loss": 0.2967, "step": 48792 }, { "epoch": 2.95, "learning_rate": 2.570202043064295e-05, "loss": 0.5474, "step": 48794 }, { "epoch": 2.95, "learning_rate": 2.5700511520528725e-05, "loss": 0.6294, "step": 48796 }, { "epoch": 2.95, "learning_rate": 2.5699002610414498e-05, "loss": 0.5236, "step": 48798 }, { "epoch": 2.95, "learning_rate": 2.5697493700300274e-05, "loss": 0.5154, "step": 48800 }, { "epoch": 2.95, "learning_rate": 2.569598479018605e-05, "loss": 0.6422, "step": 48802 }, { "epoch": 2.95, "learning_rate": 2.5694475880071828e-05, "loss": 0.5115, "step": 48804 }, { "epoch": 2.95, "learning_rate": 2.56929669699576e-05, "loss": 0.3593, "step": 48806 }, { "epoch": 2.95, "learning_rate": 2.5691458059843377e-05, "loss": 0.3874, "step": 48808 }, { "epoch": 2.95, "learning_rate": 2.5689949149729154e-05, "loss": 0.4625, "step": 48810 }, { "epoch": 2.95, "learning_rate": 2.5688440239614923e-05, "loss": 0.7434, "step": 48812 }, { "epoch": 2.95, "learning_rate": 2.5686931329500703e-05, "loss": 0.5453, "step": 48814 }, { "epoch": 2.95, "learning_rate": 2.568542241938648e-05, "loss": 0.5689, "step": 48816 }, { "epoch": 2.95, "learning_rate": 2.5683913509272256e-05, "loss": 0.4619, "step": 48818 }, { "epoch": 2.95, "learning_rate": 2.5682404599158026e-05, "loss": 0.4118, "step": 48820 }, { "epoch": 2.95, "learning_rate": 2.5680895689043806e-05, "loss": 0.4973, "step": 48822 }, { "epoch": 2.95, "learning_rate": 2.5679386778929582e-05, "loss": 0.5745, "step": 48824 }, { "epoch": 2.95, "learning_rate": 2.5677877868815352e-05, "loss": 0.5718, "step": 48826 }, { "epoch": 2.95, "learning_rate": 2.567636895870113e-05, "loss": 0.6293, "step": 48828 }, { "epoch": 2.95, "learning_rate": 2.567486004858691e-05, "loss": 0.3839, "step": 48830 }, { "epoch": 2.95, "learning_rate": 2.5673351138472685e-05, "loss": 0.6116, "step": 48832 }, { "epoch": 2.95, "learning_rate": 2.5671842228358455e-05, "loss": 0.4837, "step": 48834 }, { "epoch": 2.95, "learning_rate": 2.567033331824423e-05, "loss": 0.3387, "step": 48836 }, { "epoch": 2.95, "learning_rate": 2.566882440813001e-05, "loss": 0.5342, "step": 48838 }, { "epoch": 2.95, "learning_rate": 2.5667315498015788e-05, "loss": 0.6673, "step": 48840 }, { "epoch": 2.95, "learning_rate": 2.5665806587901558e-05, "loss": 0.672, "step": 48842 }, { "epoch": 2.95, "learning_rate": 2.5664297677787334e-05, "loss": 0.4739, "step": 48844 }, { "epoch": 2.95, "learning_rate": 2.5662788767673114e-05, "loss": 0.6164, "step": 48846 }, { "epoch": 2.95, "learning_rate": 2.5661279857558884e-05, "loss": 0.2904, "step": 48848 }, { "epoch": 2.95, "learning_rate": 2.565977094744466e-05, "loss": 0.7013, "step": 48850 }, { "epoch": 2.95, "learning_rate": 2.5658262037330437e-05, "loss": 0.5406, "step": 48852 }, { "epoch": 2.95, "learning_rate": 2.5656753127216217e-05, "loss": 0.5211, "step": 48854 }, { "epoch": 2.95, "learning_rate": 2.5655244217101986e-05, "loss": 0.541, "step": 48856 }, { "epoch": 2.95, "learning_rate": 2.5653735306987763e-05, "loss": 0.4922, "step": 48858 }, { "epoch": 2.95, "learning_rate": 2.565222639687354e-05, "loss": 0.3699, "step": 48860 }, { "epoch": 2.95, "learning_rate": 2.5650717486759313e-05, "loss": 0.5947, "step": 48862 }, { "epoch": 2.95, "learning_rate": 2.564920857664509e-05, "loss": 0.4918, "step": 48864 }, { "epoch": 2.95, "learning_rate": 2.5647699666530866e-05, "loss": 0.517, "step": 48866 }, { "epoch": 2.95, "learning_rate": 2.5646190756416642e-05, "loss": 0.5535, "step": 48868 }, { "epoch": 2.95, "learning_rate": 2.5644681846302415e-05, "loss": 0.2825, "step": 48870 }, { "epoch": 2.95, "learning_rate": 2.5643172936188192e-05, "loss": 0.6794, "step": 48872 }, { "epoch": 2.95, "learning_rate": 2.5641664026073968e-05, "loss": 0.5298, "step": 48874 }, { "epoch": 2.95, "learning_rate": 2.564015511595974e-05, "loss": 0.4798, "step": 48876 }, { "epoch": 2.95, "learning_rate": 2.5638646205845518e-05, "loss": 0.7359, "step": 48878 }, { "epoch": 2.95, "learning_rate": 2.5637137295731294e-05, "loss": 0.4191, "step": 48880 }, { "epoch": 2.95, "learning_rate": 2.563562838561707e-05, "loss": 0.413, "step": 48882 }, { "epoch": 2.95, "learning_rate": 2.5634119475502844e-05, "loss": 0.5973, "step": 48884 }, { "epoch": 2.95, "learning_rate": 2.563261056538862e-05, "loss": 0.5451, "step": 48886 }, { "epoch": 2.95, "learning_rate": 2.5631101655274397e-05, "loss": 0.4712, "step": 48888 }, { "epoch": 2.95, "learning_rate": 2.5629592745160174e-05, "loss": 0.4158, "step": 48890 }, { "epoch": 2.95, "learning_rate": 2.5628083835045947e-05, "loss": 0.339, "step": 48892 }, { "epoch": 2.95, "learning_rate": 2.5626574924931723e-05, "loss": 0.4033, "step": 48894 }, { "epoch": 2.95, "learning_rate": 2.56250660148175e-05, "loss": 0.6345, "step": 48896 }, { "epoch": 2.95, "learning_rate": 2.5623557104703273e-05, "loss": 0.5637, "step": 48898 }, { "epoch": 2.95, "learning_rate": 2.562204819458905e-05, "loss": 0.4996, "step": 48900 }, { "epoch": 2.95, "learning_rate": 2.5620539284474826e-05, "loss": 0.5074, "step": 48902 }, { "epoch": 2.95, "learning_rate": 2.5619030374360602e-05, "loss": 0.5039, "step": 48904 }, { "epoch": 2.95, "learning_rate": 2.5617521464246376e-05, "loss": 0.664, "step": 48906 }, { "epoch": 2.95, "learning_rate": 2.5616012554132152e-05, "loss": 0.4392, "step": 48908 }, { "epoch": 2.95, "learning_rate": 2.561450364401793e-05, "loss": 0.5297, "step": 48910 }, { "epoch": 2.95, "learning_rate": 2.56129947339037e-05, "loss": 0.4656, "step": 48912 }, { "epoch": 2.95, "learning_rate": 2.5611485823789478e-05, "loss": 0.5696, "step": 48914 }, { "epoch": 2.95, "learning_rate": 2.5609976913675255e-05, "loss": 0.6463, "step": 48916 }, { "epoch": 2.95, "learning_rate": 2.560846800356103e-05, "loss": 0.5233, "step": 48918 }, { "epoch": 2.95, "learning_rate": 2.5606959093446804e-05, "loss": 0.7151, "step": 48920 }, { "epoch": 2.95, "learning_rate": 2.560545018333258e-05, "loss": 0.5574, "step": 48922 }, { "epoch": 2.95, "learning_rate": 2.5603941273218357e-05, "loss": 0.4844, "step": 48924 }, { "epoch": 2.95, "learning_rate": 2.5602432363104127e-05, "loss": 0.414, "step": 48926 }, { "epoch": 2.95, "learning_rate": 2.5600923452989907e-05, "loss": 0.5359, "step": 48928 }, { "epoch": 2.95, "learning_rate": 2.5599414542875684e-05, "loss": 0.5112, "step": 48930 }, { "epoch": 2.95, "learning_rate": 2.559790563276146e-05, "loss": 0.4848, "step": 48932 }, { "epoch": 2.95, "learning_rate": 2.559639672264723e-05, "loss": 0.5841, "step": 48934 }, { "epoch": 2.95, "learning_rate": 2.559488781253301e-05, "loss": 0.4069, "step": 48936 }, { "epoch": 2.95, "learning_rate": 2.5593378902418786e-05, "loss": 0.5675, "step": 48938 }, { "epoch": 2.95, "learning_rate": 2.5591869992304563e-05, "loss": 0.496, "step": 48940 }, { "epoch": 2.95, "learning_rate": 2.5590361082190332e-05, "loss": 0.5499, "step": 48942 }, { "epoch": 2.95, "learning_rate": 2.5588852172076112e-05, "loss": 0.6013, "step": 48944 }, { "epoch": 2.95, "learning_rate": 2.558734326196189e-05, "loss": 0.5024, "step": 48946 }, { "epoch": 2.95, "learning_rate": 2.558583435184766e-05, "loss": 0.294, "step": 48948 }, { "epoch": 2.95, "learning_rate": 2.5584325441733435e-05, "loss": 0.4848, "step": 48950 }, { "epoch": 2.95, "learning_rate": 2.5582816531619215e-05, "loss": 0.4687, "step": 48952 }, { "epoch": 2.95, "learning_rate": 2.558130762150499e-05, "loss": 0.5098, "step": 48954 }, { "epoch": 2.95, "learning_rate": 2.557979871139076e-05, "loss": 0.5052, "step": 48956 }, { "epoch": 2.95, "learning_rate": 2.5578289801276538e-05, "loss": 0.6322, "step": 48958 }, { "epoch": 2.96, "learning_rate": 2.5576780891162318e-05, "loss": 0.7113, "step": 48960 }, { "epoch": 2.96, "learning_rate": 2.5575271981048087e-05, "loss": 0.7142, "step": 48962 }, { "epoch": 2.96, "learning_rate": 2.5573763070933864e-05, "loss": 0.3243, "step": 48964 }, { "epoch": 2.96, "learning_rate": 2.557225416081964e-05, "loss": 0.2905, "step": 48966 }, { "epoch": 2.96, "learning_rate": 2.557074525070542e-05, "loss": 0.4501, "step": 48968 }, { "epoch": 2.96, "learning_rate": 2.556923634059119e-05, "loss": 0.5895, "step": 48970 }, { "epoch": 2.96, "learning_rate": 2.5567727430476967e-05, "loss": 0.5624, "step": 48972 }, { "epoch": 2.96, "learning_rate": 2.5566218520362743e-05, "loss": 0.6338, "step": 48974 }, { "epoch": 2.96, "learning_rate": 2.5564709610248516e-05, "loss": 0.4314, "step": 48976 }, { "epoch": 2.96, "learning_rate": 2.5563200700134293e-05, "loss": 0.6173, "step": 48978 }, { "epoch": 2.96, "learning_rate": 2.556169179002007e-05, "loss": 0.4919, "step": 48980 }, { "epoch": 2.96, "learning_rate": 2.5560182879905846e-05, "loss": 0.4869, "step": 48982 }, { "epoch": 2.96, "learning_rate": 2.555867396979162e-05, "loss": 0.4999, "step": 48984 }, { "epoch": 2.96, "learning_rate": 2.5557165059677395e-05, "loss": 0.5419, "step": 48986 }, { "epoch": 2.96, "learning_rate": 2.5555656149563172e-05, "loss": 0.7296, "step": 48988 }, { "epoch": 2.96, "learning_rate": 2.5554147239448945e-05, "loss": 0.4738, "step": 48990 }, { "epoch": 2.96, "learning_rate": 2.555263832933472e-05, "loss": 0.5119, "step": 48992 }, { "epoch": 2.96, "learning_rate": 2.5551129419220498e-05, "loss": 0.5161, "step": 48994 }, { "epoch": 2.96, "learning_rate": 2.5549620509106275e-05, "loss": 0.3855, "step": 48996 }, { "epoch": 2.96, "learning_rate": 2.5548111598992048e-05, "loss": 0.4164, "step": 48998 }, { "epoch": 2.96, "learning_rate": 2.5546602688877824e-05, "loss": 0.515, "step": 49000 }, { "epoch": 2.96, "learning_rate": 2.55450937787636e-05, "loss": 0.4356, "step": 49002 }, { "epoch": 2.96, "learning_rate": 2.5543584868649377e-05, "loss": 0.43, "step": 49004 }, { "epoch": 2.96, "learning_rate": 2.554207595853515e-05, "loss": 0.4954, "step": 49006 }, { "epoch": 2.96, "learning_rate": 2.5540567048420927e-05, "loss": 0.4248, "step": 49008 }, { "epoch": 2.96, "learning_rate": 2.5539058138306703e-05, "loss": 0.4806, "step": 49010 }, { "epoch": 2.96, "learning_rate": 2.5537549228192477e-05, "loss": 0.4486, "step": 49012 }, { "epoch": 2.96, "learning_rate": 2.5536040318078253e-05, "loss": 0.459, "step": 49014 }, { "epoch": 2.96, "learning_rate": 2.553453140796403e-05, "loss": 0.4836, "step": 49016 }, { "epoch": 2.96, "learning_rate": 2.5533022497849806e-05, "loss": 0.3549, "step": 49018 }, { "epoch": 2.96, "learning_rate": 2.553151358773558e-05, "loss": 0.3347, "step": 49020 }, { "epoch": 2.96, "learning_rate": 2.5530004677621356e-05, "loss": 0.4489, "step": 49022 }, { "epoch": 2.96, "learning_rate": 2.5528495767507132e-05, "loss": 0.61, "step": 49024 }, { "epoch": 2.96, "learning_rate": 2.5526986857392905e-05, "loss": 0.6143, "step": 49026 }, { "epoch": 2.96, "learning_rate": 2.5525477947278682e-05, "loss": 0.5174, "step": 49028 }, { "epoch": 2.96, "learning_rate": 2.552396903716446e-05, "loss": 0.3344, "step": 49030 }, { "epoch": 2.96, "learning_rate": 2.5522460127050235e-05, "loss": 0.5485, "step": 49032 }, { "epoch": 2.96, "learning_rate": 2.5520951216936008e-05, "loss": 0.4087, "step": 49034 }, { "epoch": 2.96, "learning_rate": 2.5519442306821785e-05, "loss": 0.5261, "step": 49036 }, { "epoch": 2.96, "learning_rate": 2.551793339670756e-05, "loss": 0.5206, "step": 49038 }, { "epoch": 2.96, "learning_rate": 2.551642448659333e-05, "loss": 0.5155, "step": 49040 }, { "epoch": 2.96, "learning_rate": 2.551491557647911e-05, "loss": 0.4674, "step": 49042 }, { "epoch": 2.96, "learning_rate": 2.5513406666364887e-05, "loss": 0.5072, "step": 49044 }, { "epoch": 2.96, "learning_rate": 2.5511897756250664e-05, "loss": 0.458, "step": 49046 }, { "epoch": 2.96, "learning_rate": 2.5510388846136433e-05, "loss": 0.59, "step": 49048 }, { "epoch": 2.96, "learning_rate": 2.5508879936022213e-05, "loss": 0.5312, "step": 49050 }, { "epoch": 2.96, "learning_rate": 2.550737102590799e-05, "loss": 0.5246, "step": 49052 }, { "epoch": 2.96, "learning_rate": 2.5505862115793766e-05, "loss": 0.3159, "step": 49054 }, { "epoch": 2.96, "learning_rate": 2.5504353205679536e-05, "loss": 0.54, "step": 49056 }, { "epoch": 2.96, "learning_rate": 2.5502844295565316e-05, "loss": 0.446, "step": 49058 }, { "epoch": 2.96, "learning_rate": 2.5501335385451093e-05, "loss": 0.3813, "step": 49060 }, { "epoch": 2.96, "learning_rate": 2.5499826475336862e-05, "loss": 0.4957, "step": 49062 }, { "epoch": 2.96, "learning_rate": 2.549831756522264e-05, "loss": 0.3606, "step": 49064 }, { "epoch": 2.96, "learning_rate": 2.549680865510842e-05, "loss": 0.4958, "step": 49066 }, { "epoch": 2.96, "learning_rate": 2.5495299744994195e-05, "loss": 0.7968, "step": 49068 }, { "epoch": 2.96, "learning_rate": 2.5493790834879965e-05, "loss": 0.5439, "step": 49070 }, { "epoch": 2.96, "learning_rate": 2.549228192476574e-05, "loss": 0.7457, "step": 49072 }, { "epoch": 2.96, "learning_rate": 2.549077301465152e-05, "loss": 0.4143, "step": 49074 }, { "epoch": 2.96, "learning_rate": 2.548926410453729e-05, "loss": 0.4417, "step": 49076 }, { "epoch": 2.96, "learning_rate": 2.5487755194423068e-05, "loss": 0.6254, "step": 49078 }, { "epoch": 2.96, "learning_rate": 2.5486246284308844e-05, "loss": 0.6016, "step": 49080 }, { "epoch": 2.96, "learning_rate": 2.5484737374194624e-05, "loss": 0.3129, "step": 49082 }, { "epoch": 2.96, "learning_rate": 2.5483228464080394e-05, "loss": 0.4817, "step": 49084 }, { "epoch": 2.96, "learning_rate": 2.548171955396617e-05, "loss": 0.4713, "step": 49086 }, { "epoch": 2.96, "learning_rate": 2.5480210643851947e-05, "loss": 0.5048, "step": 49088 }, { "epoch": 2.96, "learning_rate": 2.547870173373772e-05, "loss": 0.6865, "step": 49090 }, { "epoch": 2.96, "learning_rate": 2.5477192823623496e-05, "loss": 0.6915, "step": 49092 }, { "epoch": 2.96, "learning_rate": 2.5475683913509273e-05, "loss": 0.3871, "step": 49094 }, { "epoch": 2.96, "learning_rate": 2.547417500339505e-05, "loss": 0.4743, "step": 49096 }, { "epoch": 2.96, "learning_rate": 2.5472666093280823e-05, "loss": 0.6132, "step": 49098 }, { "epoch": 2.96, "learning_rate": 2.54711571831666e-05, "loss": 0.5733, "step": 49100 }, { "epoch": 2.96, "learning_rate": 2.5469648273052376e-05, "loss": 0.4669, "step": 49102 }, { "epoch": 2.96, "learning_rate": 2.5468139362938152e-05, "loss": 0.4979, "step": 49104 }, { "epoch": 2.96, "learning_rate": 2.5466630452823925e-05, "loss": 0.5845, "step": 49106 }, { "epoch": 2.96, "learning_rate": 2.5465121542709702e-05, "loss": 0.2801, "step": 49108 }, { "epoch": 2.96, "learning_rate": 2.5463612632595478e-05, "loss": 0.4674, "step": 49110 }, { "epoch": 2.96, "learning_rate": 2.546210372248125e-05, "loss": 0.6102, "step": 49112 }, { "epoch": 2.96, "learning_rate": 2.5460594812367028e-05, "loss": 0.5624, "step": 49114 }, { "epoch": 2.96, "learning_rate": 2.5459085902252804e-05, "loss": 0.3898, "step": 49116 }, { "epoch": 2.96, "learning_rate": 2.545757699213858e-05, "loss": 0.4752, "step": 49118 }, { "epoch": 2.96, "learning_rate": 2.5456068082024354e-05, "loss": 0.5402, "step": 49120 }, { "epoch": 2.96, "learning_rate": 2.545455917191013e-05, "loss": 0.4428, "step": 49122 }, { "epoch": 2.96, "learning_rate": 2.5453050261795907e-05, "loss": 0.5935, "step": 49124 }, { "epoch": 2.97, "learning_rate": 2.545154135168168e-05, "loss": 0.5368, "step": 49126 }, { "epoch": 2.97, "learning_rate": 2.5450032441567457e-05, "loss": 0.5975, "step": 49128 }, { "epoch": 2.97, "learning_rate": 2.5448523531453233e-05, "loss": 0.5428, "step": 49130 }, { "epoch": 2.97, "learning_rate": 2.544701462133901e-05, "loss": 0.6211, "step": 49132 }, { "epoch": 2.97, "learning_rate": 2.5445505711224783e-05, "loss": 0.6271, "step": 49134 }, { "epoch": 2.97, "learning_rate": 2.544399680111056e-05, "loss": 0.4554, "step": 49136 }, { "epoch": 2.97, "learning_rate": 2.5442487890996336e-05, "loss": 0.6726, "step": 49138 }, { "epoch": 2.97, "learning_rate": 2.544097898088211e-05, "loss": 0.3932, "step": 49140 }, { "epoch": 2.97, "learning_rate": 2.5439470070767886e-05, "loss": 0.4093, "step": 49142 }, { "epoch": 2.97, "learning_rate": 2.5437961160653662e-05, "loss": 0.4231, "step": 49144 }, { "epoch": 2.97, "learning_rate": 2.543645225053944e-05, "loss": 0.3097, "step": 49146 }, { "epoch": 2.97, "learning_rate": 2.543494334042521e-05, "loss": 0.3733, "step": 49148 }, { "epoch": 2.97, "learning_rate": 2.5433434430310988e-05, "loss": 0.3295, "step": 49150 }, { "epoch": 2.97, "learning_rate": 2.5431925520196765e-05, "loss": 0.4577, "step": 49152 }, { "epoch": 2.97, "learning_rate": 2.5430416610082534e-05, "loss": 0.3617, "step": 49154 }, { "epoch": 2.97, "learning_rate": 2.5428907699968314e-05, "loss": 0.382, "step": 49156 }, { "epoch": 2.97, "learning_rate": 2.542739878985409e-05, "loss": 0.4487, "step": 49158 }, { "epoch": 2.97, "learning_rate": 2.5425889879739867e-05, "loss": 0.4295, "step": 49160 }, { "epoch": 2.97, "learning_rate": 2.5424380969625637e-05, "loss": 0.5762, "step": 49162 }, { "epoch": 2.97, "learning_rate": 2.5422872059511417e-05, "loss": 0.3805, "step": 49164 }, { "epoch": 2.97, "learning_rate": 2.5421363149397194e-05, "loss": 0.4052, "step": 49166 }, { "epoch": 2.97, "learning_rate": 2.541985423928297e-05, "loss": 0.492, "step": 49168 }, { "epoch": 2.97, "learning_rate": 2.541834532916874e-05, "loss": 0.4199, "step": 49170 }, { "epoch": 2.97, "learning_rate": 2.541683641905452e-05, "loss": 0.4909, "step": 49172 }, { "epoch": 2.97, "learning_rate": 2.5415327508940296e-05, "loss": 0.4818, "step": 49174 }, { "epoch": 2.97, "learning_rate": 2.5413818598826066e-05, "loss": 0.3633, "step": 49176 }, { "epoch": 2.97, "learning_rate": 2.5412309688711842e-05, "loss": 0.569, "step": 49178 }, { "epoch": 2.97, "learning_rate": 2.5410800778597622e-05, "loss": 0.5625, "step": 49180 }, { "epoch": 2.97, "learning_rate": 2.54092918684834e-05, "loss": 0.6974, "step": 49182 }, { "epoch": 2.97, "learning_rate": 2.540778295836917e-05, "loss": 0.3907, "step": 49184 }, { "epoch": 2.97, "learning_rate": 2.5406274048254945e-05, "loss": 0.4826, "step": 49186 }, { "epoch": 2.97, "learning_rate": 2.5404765138140725e-05, "loss": 0.672, "step": 49188 }, { "epoch": 2.97, "learning_rate": 2.5403256228026495e-05, "loss": 0.5998, "step": 49190 }, { "epoch": 2.97, "learning_rate": 2.540174731791227e-05, "loss": 0.4439, "step": 49192 }, { "epoch": 2.97, "learning_rate": 2.5400238407798048e-05, "loss": 0.7391, "step": 49194 }, { "epoch": 2.97, "learning_rate": 2.5398729497683828e-05, "loss": 0.4848, "step": 49196 }, { "epoch": 2.97, "learning_rate": 2.5397220587569597e-05, "loss": 0.3129, "step": 49198 }, { "epoch": 2.97, "learning_rate": 2.5395711677455374e-05, "loss": 0.6422, "step": 49200 }, { "epoch": 2.97, "learning_rate": 2.539420276734115e-05, "loss": 0.3986, "step": 49202 }, { "epoch": 2.97, "learning_rate": 2.5392693857226924e-05, "loss": 0.4255, "step": 49204 }, { "epoch": 2.97, "learning_rate": 2.53911849471127e-05, "loss": 0.5112, "step": 49206 }, { "epoch": 2.97, "learning_rate": 2.5389676036998477e-05, "loss": 0.2893, "step": 49208 }, { "epoch": 2.97, "learning_rate": 2.5388167126884253e-05, "loss": 0.6706, "step": 49210 }, { "epoch": 2.97, "learning_rate": 2.5386658216770026e-05, "loss": 0.6946, "step": 49212 }, { "epoch": 2.97, "learning_rate": 2.5385149306655803e-05, "loss": 0.5156, "step": 49214 }, { "epoch": 2.97, "learning_rate": 2.538364039654158e-05, "loss": 0.4015, "step": 49216 }, { "epoch": 2.97, "learning_rate": 2.5382131486427356e-05, "loss": 0.7305, "step": 49218 }, { "epoch": 2.97, "learning_rate": 2.538062257631313e-05, "loss": 0.6098, "step": 49220 }, { "epoch": 2.97, "learning_rate": 2.5379113666198905e-05, "loss": 0.4124, "step": 49222 }, { "epoch": 2.97, "learning_rate": 2.5377604756084682e-05, "loss": 0.4999, "step": 49224 }, { "epoch": 2.97, "learning_rate": 2.5376095845970455e-05, "loss": 0.6338, "step": 49226 }, { "epoch": 2.97, "learning_rate": 2.537458693585623e-05, "loss": 0.5426, "step": 49228 }, { "epoch": 2.97, "learning_rate": 2.5373078025742008e-05, "loss": 0.366, "step": 49230 }, { "epoch": 2.97, "learning_rate": 2.5371569115627785e-05, "loss": 0.4213, "step": 49232 }, { "epoch": 2.97, "learning_rate": 2.5370060205513558e-05, "loss": 0.3146, "step": 49234 }, { "epoch": 2.97, "learning_rate": 2.5368551295399334e-05, "loss": 0.5122, "step": 49236 }, { "epoch": 2.97, "learning_rate": 2.536704238528511e-05, "loss": 0.5425, "step": 49238 }, { "epoch": 2.97, "learning_rate": 2.5365533475170884e-05, "loss": 0.5919, "step": 49240 }, { "epoch": 2.97, "learning_rate": 2.536402456505666e-05, "loss": 0.7251, "step": 49242 }, { "epoch": 2.97, "learning_rate": 2.5362515654942437e-05, "loss": 0.5504, "step": 49244 }, { "epoch": 2.97, "learning_rate": 2.5361006744828213e-05, "loss": 0.6093, "step": 49246 }, { "epoch": 2.97, "learning_rate": 2.5359497834713987e-05, "loss": 0.3635, "step": 49248 }, { "epoch": 2.97, "learning_rate": 2.5357988924599763e-05, "loss": 0.5455, "step": 49250 }, { "epoch": 2.97, "learning_rate": 2.535648001448554e-05, "loss": 0.3748, "step": 49252 }, { "epoch": 2.97, "learning_rate": 2.5354971104371313e-05, "loss": 0.3423, "step": 49254 }, { "epoch": 2.97, "learning_rate": 2.535346219425709e-05, "loss": 0.4561, "step": 49256 }, { "epoch": 2.97, "learning_rate": 2.5351953284142866e-05, "loss": 0.477, "step": 49258 }, { "epoch": 2.97, "learning_rate": 2.5350444374028642e-05, "loss": 0.3981, "step": 49260 }, { "epoch": 2.97, "learning_rate": 2.5348935463914415e-05, "loss": 0.6393, "step": 49262 }, { "epoch": 2.97, "learning_rate": 2.5347426553800192e-05, "loss": 0.7053, "step": 49264 }, { "epoch": 2.97, "learning_rate": 2.534591764368597e-05, "loss": 0.4883, "step": 49266 }, { "epoch": 2.97, "learning_rate": 2.5344408733571745e-05, "loss": 0.4822, "step": 49268 }, { "epoch": 2.97, "learning_rate": 2.5342899823457518e-05, "loss": 0.3825, "step": 49270 }, { "epoch": 2.97, "learning_rate": 2.5341390913343295e-05, "loss": 0.3415, "step": 49272 }, { "epoch": 2.97, "learning_rate": 2.533988200322907e-05, "loss": 0.5419, "step": 49274 }, { "epoch": 2.97, "learning_rate": 2.533837309311484e-05, "loss": 0.4893, "step": 49276 }, { "epoch": 2.97, "learning_rate": 2.533686418300062e-05, "loss": 0.7466, "step": 49278 }, { "epoch": 2.97, "learning_rate": 2.5335355272886397e-05, "loss": 0.4565, "step": 49280 }, { "epoch": 2.97, "learning_rate": 2.5333846362772174e-05, "loss": 0.3747, "step": 49282 }, { "epoch": 2.97, "learning_rate": 2.5332337452657943e-05, "loss": 0.5651, "step": 49284 }, { "epoch": 2.97, "learning_rate": 2.5330828542543723e-05, "loss": 0.7088, "step": 49286 }, { "epoch": 2.97, "learning_rate": 2.53293196324295e-05, "loss": 0.5469, "step": 49288 }, { "epoch": 2.98, "learning_rate": 2.532781072231527e-05, "loss": 0.802, "step": 49290 }, { "epoch": 2.98, "learning_rate": 2.5326301812201046e-05, "loss": 0.3713, "step": 49292 }, { "epoch": 2.98, "learning_rate": 2.5324792902086826e-05, "loss": 0.4676, "step": 49294 }, { "epoch": 2.98, "learning_rate": 2.5323283991972603e-05, "loss": 0.4892, "step": 49296 }, { "epoch": 2.98, "learning_rate": 2.5321775081858372e-05, "loss": 0.3892, "step": 49298 }, { "epoch": 2.98, "learning_rate": 2.532026617174415e-05, "loss": 0.2865, "step": 49300 }, { "epoch": 2.98, "learning_rate": 2.531875726162993e-05, "loss": 0.4265, "step": 49302 }, { "epoch": 2.98, "learning_rate": 2.53172483515157e-05, "loss": 0.6216, "step": 49304 }, { "epoch": 2.98, "learning_rate": 2.5315739441401475e-05, "loss": 0.4124, "step": 49306 }, { "epoch": 2.98, "learning_rate": 2.531423053128725e-05, "loss": 0.462, "step": 49308 }, { "epoch": 2.98, "learning_rate": 2.531272162117303e-05, "loss": 0.5872, "step": 49310 }, { "epoch": 2.98, "learning_rate": 2.53112127110588e-05, "loss": 0.4426, "step": 49312 }, { "epoch": 2.98, "learning_rate": 2.5309703800944578e-05, "loss": 0.5338, "step": 49314 }, { "epoch": 2.98, "learning_rate": 2.5308194890830354e-05, "loss": 0.4134, "step": 49316 }, { "epoch": 2.98, "learning_rate": 2.5306685980716134e-05, "loss": 0.5185, "step": 49318 }, { "epoch": 2.98, "learning_rate": 2.5305177070601904e-05, "loss": 0.4136, "step": 49320 }, { "epoch": 2.98, "learning_rate": 2.530366816048768e-05, "loss": 0.6026, "step": 49322 }, { "epoch": 2.98, "learning_rate": 2.5302159250373457e-05, "loss": 0.7432, "step": 49324 }, { "epoch": 2.98, "learning_rate": 2.530065034025923e-05, "loss": 0.598, "step": 49326 }, { "epoch": 2.98, "learning_rate": 2.5299141430145006e-05, "loss": 0.3794, "step": 49328 }, { "epoch": 2.98, "learning_rate": 2.5297632520030783e-05, "loss": 0.5865, "step": 49330 }, { "epoch": 2.98, "learning_rate": 2.529612360991656e-05, "loss": 0.4538, "step": 49332 }, { "epoch": 2.98, "learning_rate": 2.5294614699802333e-05, "loss": 0.75, "step": 49334 }, { "epoch": 2.98, "learning_rate": 2.529310578968811e-05, "loss": 0.502, "step": 49336 }, { "epoch": 2.98, "learning_rate": 2.5291596879573886e-05, "loss": 0.4918, "step": 49338 }, { "epoch": 2.98, "learning_rate": 2.529008796945966e-05, "loss": 0.6263, "step": 49340 }, { "epoch": 2.98, "learning_rate": 2.5288579059345435e-05, "loss": 0.6912, "step": 49342 }, { "epoch": 2.98, "learning_rate": 2.5287070149231212e-05, "loss": 0.3903, "step": 49344 }, { "epoch": 2.98, "learning_rate": 2.5285561239116988e-05, "loss": 0.3764, "step": 49346 }, { "epoch": 2.98, "learning_rate": 2.528405232900276e-05, "loss": 0.4198, "step": 49348 }, { "epoch": 2.98, "learning_rate": 2.5282543418888538e-05, "loss": 0.382, "step": 49350 }, { "epoch": 2.98, "learning_rate": 2.5281034508774314e-05, "loss": 0.4823, "step": 49352 }, { "epoch": 2.98, "learning_rate": 2.5279525598660088e-05, "loss": 0.4626, "step": 49354 }, { "epoch": 2.98, "learning_rate": 2.5278016688545864e-05, "loss": 0.6008, "step": 49356 }, { "epoch": 2.98, "learning_rate": 2.527650777843164e-05, "loss": 0.5865, "step": 49358 }, { "epoch": 2.98, "learning_rate": 2.5274998868317417e-05, "loss": 0.5368, "step": 49360 }, { "epoch": 2.98, "learning_rate": 2.527348995820319e-05, "loss": 0.5726, "step": 49362 }, { "epoch": 2.98, "learning_rate": 2.5271981048088967e-05, "loss": 0.3765, "step": 49364 }, { "epoch": 2.98, "learning_rate": 2.5270472137974743e-05, "loss": 0.4014, "step": 49366 }, { "epoch": 2.98, "learning_rate": 2.5268963227860516e-05, "loss": 0.3934, "step": 49368 }, { "epoch": 2.98, "learning_rate": 2.5267454317746293e-05, "loss": 0.3492, "step": 49370 }, { "epoch": 2.98, "learning_rate": 2.526594540763207e-05, "loss": 0.3579, "step": 49372 }, { "epoch": 2.98, "learning_rate": 2.5264436497517846e-05, "loss": 0.5598, "step": 49374 }, { "epoch": 2.98, "learning_rate": 2.526292758740362e-05, "loss": 0.43, "step": 49376 }, { "epoch": 2.98, "learning_rate": 2.5261418677289396e-05, "loss": 0.4714, "step": 49378 }, { "epoch": 2.98, "learning_rate": 2.5259909767175172e-05, "loss": 0.3524, "step": 49380 }, { "epoch": 2.98, "learning_rate": 2.525840085706095e-05, "loss": 0.6453, "step": 49382 }, { "epoch": 2.98, "learning_rate": 2.5256891946946722e-05, "loss": 0.6639, "step": 49384 }, { "epoch": 2.98, "learning_rate": 2.5255383036832498e-05, "loss": 0.4846, "step": 49386 }, { "epoch": 2.98, "learning_rate": 2.5253874126718275e-05, "loss": 0.475, "step": 49388 }, { "epoch": 2.98, "learning_rate": 2.5252365216604044e-05, "loss": 0.5257, "step": 49390 }, { "epoch": 2.98, "learning_rate": 2.5250856306489824e-05, "loss": 0.5683, "step": 49392 }, { "epoch": 2.98, "learning_rate": 2.52493473963756e-05, "loss": 0.5098, "step": 49394 }, { "epoch": 2.98, "learning_rate": 2.5247838486261377e-05, "loss": 0.4985, "step": 49396 }, { "epoch": 2.98, "learning_rate": 2.5246329576147147e-05, "loss": 0.6939, "step": 49398 }, { "epoch": 2.98, "learning_rate": 2.5244820666032927e-05, "loss": 0.6274, "step": 49400 }, { "epoch": 2.98, "learning_rate": 2.5243311755918704e-05, "loss": 0.3571, "step": 49402 }, { "epoch": 2.98, "learning_rate": 2.5241802845804473e-05, "loss": 0.5265, "step": 49404 }, { "epoch": 2.98, "learning_rate": 2.524029393569025e-05, "loss": 0.4502, "step": 49406 }, { "epoch": 2.98, "learning_rate": 2.523878502557603e-05, "loss": 0.3884, "step": 49408 }, { "epoch": 2.98, "learning_rate": 2.5237276115461806e-05, "loss": 0.4553, "step": 49410 }, { "epoch": 2.98, "learning_rate": 2.5235767205347576e-05, "loss": 0.5793, "step": 49412 }, { "epoch": 2.98, "learning_rate": 2.5234258295233352e-05, "loss": 0.5328, "step": 49414 }, { "epoch": 2.98, "learning_rate": 2.5232749385119132e-05, "loss": 0.5317, "step": 49416 }, { "epoch": 2.98, "learning_rate": 2.5231240475004902e-05, "loss": 0.546, "step": 49418 }, { "epoch": 2.98, "learning_rate": 2.522973156489068e-05, "loss": 0.4237, "step": 49420 }, { "epoch": 2.98, "learning_rate": 2.5228222654776455e-05, "loss": 0.5642, "step": 49422 }, { "epoch": 2.98, "learning_rate": 2.5226713744662235e-05, "loss": 0.4329, "step": 49424 }, { "epoch": 2.98, "learning_rate": 2.5225204834548005e-05, "loss": 0.5796, "step": 49426 }, { "epoch": 2.98, "learning_rate": 2.522369592443378e-05, "loss": 0.5905, "step": 49428 }, { "epoch": 2.98, "learning_rate": 2.5222187014319558e-05, "loss": 0.5903, "step": 49430 }, { "epoch": 2.98, "learning_rate": 2.5220678104205338e-05, "loss": 0.4686, "step": 49432 }, { "epoch": 2.98, "learning_rate": 2.5219169194091107e-05, "loss": 0.648, "step": 49434 }, { "epoch": 2.98, "learning_rate": 2.5217660283976884e-05, "loss": 0.576, "step": 49436 }, { "epoch": 2.98, "learning_rate": 2.521615137386266e-05, "loss": 0.6551, "step": 49438 }, { "epoch": 2.98, "learning_rate": 2.5214642463748434e-05, "loss": 0.4561, "step": 49440 }, { "epoch": 2.98, "learning_rate": 2.521313355363421e-05, "loss": 0.4356, "step": 49442 }, { "epoch": 2.98, "learning_rate": 2.5211624643519987e-05, "loss": 0.7224, "step": 49444 }, { "epoch": 2.98, "learning_rate": 2.5210115733405763e-05, "loss": 0.5951, "step": 49446 }, { "epoch": 2.98, "learning_rate": 2.5208606823291536e-05, "loss": 0.5792, "step": 49448 }, { "epoch": 2.98, "learning_rate": 2.5207097913177313e-05, "loss": 0.6655, "step": 49450 }, { "epoch": 2.98, "learning_rate": 2.520558900306309e-05, "loss": 0.4846, "step": 49452 }, { "epoch": 2.98, "learning_rate": 2.5204080092948862e-05, "loss": 0.3617, "step": 49454 }, { "epoch": 2.99, "learning_rate": 2.520257118283464e-05, "loss": 0.5147, "step": 49456 }, { "epoch": 2.99, "learning_rate": 2.5201062272720415e-05, "loss": 0.5077, "step": 49458 }, { "epoch": 2.99, "learning_rate": 2.5199553362606192e-05, "loss": 0.6386, "step": 49460 }, { "epoch": 2.99, "learning_rate": 2.5198044452491965e-05, "loss": 0.436, "step": 49462 }, { "epoch": 2.99, "learning_rate": 2.519653554237774e-05, "loss": 0.5649, "step": 49464 }, { "epoch": 2.99, "learning_rate": 2.5195026632263518e-05, "loss": 0.416, "step": 49466 }, { "epoch": 2.99, "learning_rate": 2.519351772214929e-05, "loss": 0.5957, "step": 49468 }, { "epoch": 2.99, "learning_rate": 2.5192008812035068e-05, "loss": 0.4355, "step": 49470 }, { "epoch": 2.99, "learning_rate": 2.5190499901920844e-05, "loss": 0.7126, "step": 49472 }, { "epoch": 2.99, "learning_rate": 2.518899099180662e-05, "loss": 0.4686, "step": 49474 }, { "epoch": 2.99, "learning_rate": 2.5187482081692394e-05, "loss": 0.3373, "step": 49476 }, { "epoch": 2.99, "learning_rate": 2.518597317157817e-05, "loss": 0.3712, "step": 49478 }, { "epoch": 2.99, "learning_rate": 2.5184464261463947e-05, "loss": 0.594, "step": 49480 }, { "epoch": 2.99, "learning_rate": 2.5182955351349723e-05, "loss": 0.4833, "step": 49482 }, { "epoch": 2.99, "learning_rate": 2.5181446441235497e-05, "loss": 0.3598, "step": 49484 }, { "epoch": 2.99, "learning_rate": 2.5179937531121273e-05, "loss": 0.4654, "step": 49486 }, { "epoch": 2.99, "learning_rate": 2.517842862100705e-05, "loss": 0.5037, "step": 49488 }, { "epoch": 2.99, "learning_rate": 2.5176919710892823e-05, "loss": 0.5742, "step": 49490 }, { "epoch": 2.99, "learning_rate": 2.51754108007786e-05, "loss": 0.4496, "step": 49492 }, { "epoch": 2.99, "learning_rate": 2.5173901890664376e-05, "loss": 0.5704, "step": 49494 }, { "epoch": 2.99, "learning_rate": 2.5172392980550152e-05, "loss": 0.5213, "step": 49496 }, { "epoch": 2.99, "learning_rate": 2.5170884070435925e-05, "loss": 0.7488, "step": 49498 }, { "epoch": 2.99, "learning_rate": 2.5169375160321702e-05, "loss": 0.5294, "step": 49500 }, { "epoch": 2.99, "learning_rate": 2.516786625020748e-05, "loss": 0.5917, "step": 49502 }, { "epoch": 2.99, "learning_rate": 2.5166357340093248e-05, "loss": 0.519, "step": 49504 }, { "epoch": 2.99, "learning_rate": 2.5164848429979028e-05, "loss": 0.5721, "step": 49506 }, { "epoch": 2.99, "learning_rate": 2.5163339519864805e-05, "loss": 0.5291, "step": 49508 }, { "epoch": 2.99, "learning_rate": 2.516183060975058e-05, "loss": 0.6078, "step": 49510 }, { "epoch": 2.99, "learning_rate": 2.516032169963635e-05, "loss": 0.5136, "step": 49512 }, { "epoch": 2.99, "learning_rate": 2.515881278952213e-05, "loss": 0.4272, "step": 49514 }, { "epoch": 2.99, "learning_rate": 2.5157303879407907e-05, "loss": 0.6324, "step": 49516 }, { "epoch": 2.99, "learning_rate": 2.5155794969293677e-05, "loss": 0.5156, "step": 49518 }, { "epoch": 2.99, "learning_rate": 2.5154286059179453e-05, "loss": 0.5844, "step": 49520 }, { "epoch": 2.99, "learning_rate": 2.5152777149065233e-05, "loss": 0.5536, "step": 49522 }, { "epoch": 2.99, "learning_rate": 2.515126823895101e-05, "loss": 0.5671, "step": 49524 }, { "epoch": 2.99, "learning_rate": 2.514975932883678e-05, "loss": 0.3444, "step": 49526 }, { "epoch": 2.99, "learning_rate": 2.5148250418722556e-05, "loss": 0.4927, "step": 49528 }, { "epoch": 2.99, "learning_rate": 2.5146741508608336e-05, "loss": 0.5227, "step": 49530 }, { "epoch": 2.99, "learning_rate": 2.5145232598494106e-05, "loss": 0.5441, "step": 49532 }, { "epoch": 2.99, "learning_rate": 2.5143723688379882e-05, "loss": 0.5244, "step": 49534 }, { "epoch": 2.99, "learning_rate": 2.514221477826566e-05, "loss": 0.5459, "step": 49536 }, { "epoch": 2.99, "learning_rate": 2.514070586815144e-05, "loss": 0.437, "step": 49538 }, { "epoch": 2.99, "learning_rate": 2.513919695803721e-05, "loss": 0.433, "step": 49540 }, { "epoch": 2.99, "learning_rate": 2.5137688047922985e-05, "loss": 0.4935, "step": 49542 }, { "epoch": 2.99, "learning_rate": 2.513617913780876e-05, "loss": 0.3138, "step": 49544 }, { "epoch": 2.99, "learning_rate": 2.513467022769454e-05, "loss": 0.4888, "step": 49546 }, { "epoch": 2.99, "learning_rate": 2.513316131758031e-05, "loss": 0.3702, "step": 49548 }, { "epoch": 2.99, "learning_rate": 2.5131652407466088e-05, "loss": 0.321, "step": 49550 }, { "epoch": 2.99, "learning_rate": 2.5130143497351864e-05, "loss": 0.3334, "step": 49552 }, { "epoch": 2.99, "learning_rate": 2.5128634587237637e-05, "loss": 0.4119, "step": 49554 }, { "epoch": 2.99, "learning_rate": 2.5127125677123414e-05, "loss": 0.485, "step": 49556 }, { "epoch": 2.99, "learning_rate": 2.512561676700919e-05, "loss": 0.5442, "step": 49558 }, { "epoch": 2.99, "learning_rate": 2.5124107856894967e-05, "loss": 0.3741, "step": 49560 }, { "epoch": 2.99, "learning_rate": 2.512259894678074e-05, "loss": 0.4706, "step": 49562 }, { "epoch": 2.99, "learning_rate": 2.5121090036666516e-05, "loss": 0.4594, "step": 49564 }, { "epoch": 2.99, "learning_rate": 2.5119581126552293e-05, "loss": 0.4869, "step": 49566 }, { "epoch": 2.99, "learning_rate": 2.5118072216438066e-05, "loss": 0.4947, "step": 49568 }, { "epoch": 2.99, "learning_rate": 2.5116563306323843e-05, "loss": 0.4289, "step": 49570 }, { "epoch": 2.99, "learning_rate": 2.511505439620962e-05, "loss": 0.4399, "step": 49572 }, { "epoch": 2.99, "learning_rate": 2.5113545486095396e-05, "loss": 0.5364, "step": 49574 }, { "epoch": 2.99, "learning_rate": 2.511203657598117e-05, "loss": 0.5726, "step": 49576 }, { "epoch": 2.99, "learning_rate": 2.5110527665866945e-05, "loss": 0.3036, "step": 49578 }, { "epoch": 2.99, "learning_rate": 2.5109018755752722e-05, "loss": 0.5728, "step": 49580 }, { "epoch": 2.99, "learning_rate": 2.5107509845638495e-05, "loss": 0.608, "step": 49582 }, { "epoch": 2.99, "learning_rate": 2.510600093552427e-05, "loss": 0.69, "step": 49584 }, { "epoch": 2.99, "learning_rate": 2.5104492025410048e-05, "loss": 0.4003, "step": 49586 }, { "epoch": 2.99, "learning_rate": 2.5102983115295824e-05, "loss": 0.4448, "step": 49588 }, { "epoch": 2.99, "learning_rate": 2.5101474205181598e-05, "loss": 0.4352, "step": 49590 }, { "epoch": 2.99, "learning_rate": 2.5099965295067374e-05, "loss": 0.7965, "step": 49592 }, { "epoch": 2.99, "learning_rate": 2.509845638495315e-05, "loss": 0.6557, "step": 49594 }, { "epoch": 2.99, "learning_rate": 2.5096947474838927e-05, "loss": 0.6341, "step": 49596 }, { "epoch": 2.99, "learning_rate": 2.50954385647247e-05, "loss": 0.2815, "step": 49598 }, { "epoch": 2.99, "learning_rate": 2.5093929654610477e-05, "loss": 0.5291, "step": 49600 }, { "epoch": 2.99, "learning_rate": 2.5092420744496253e-05, "loss": 0.7869, "step": 49602 }, { "epoch": 2.99, "learning_rate": 2.5090911834382026e-05, "loss": 0.6332, "step": 49604 }, { "epoch": 2.99, "learning_rate": 2.5089402924267803e-05, "loss": 0.6118, "step": 49606 }, { "epoch": 2.99, "learning_rate": 2.508789401415358e-05, "loss": 0.7012, "step": 49608 }, { "epoch": 2.99, "learning_rate": 2.5086385104039356e-05, "loss": 0.6774, "step": 49610 }, { "epoch": 2.99, "learning_rate": 2.508487619392513e-05, "loss": 0.5188, "step": 49612 }, { "epoch": 2.99, "learning_rate": 2.5083367283810906e-05, "loss": 0.3589, "step": 49614 }, { "epoch": 2.99, "learning_rate": 2.5081858373696682e-05, "loss": 0.3726, "step": 49616 }, { "epoch": 2.99, "learning_rate": 2.5080349463582452e-05, "loss": 0.6078, "step": 49618 }, { "epoch": 2.99, "learning_rate": 2.5078840553468232e-05, "loss": 0.5336, "step": 49620 }, { "epoch": 3.0, "learning_rate": 2.5077331643354008e-05, "loss": 0.3908, "step": 49622 }, { "epoch": 3.0, "learning_rate": 2.5075822733239785e-05, "loss": 0.5624, "step": 49624 }, { "epoch": 3.0, "learning_rate": 2.5074313823125554e-05, "loss": 0.5293, "step": 49626 }, { "epoch": 3.0, "learning_rate": 2.5072804913011334e-05, "loss": 0.6607, "step": 49628 }, { "epoch": 3.0, "learning_rate": 2.507129600289711e-05, "loss": 0.4839, "step": 49630 }, { "epoch": 3.0, "learning_rate": 2.506978709278288e-05, "loss": 0.5693, "step": 49632 }, { "epoch": 3.0, "learning_rate": 2.5068278182668657e-05, "loss": 0.6637, "step": 49634 }, { "epoch": 3.0, "learning_rate": 2.5066769272554437e-05, "loss": 0.4182, "step": 49636 }, { "epoch": 3.0, "learning_rate": 2.5065260362440214e-05, "loss": 0.4981, "step": 49638 }, { "epoch": 3.0, "learning_rate": 2.5063751452325983e-05, "loss": 0.2736, "step": 49640 }, { "epoch": 3.0, "learning_rate": 2.506224254221176e-05, "loss": 0.5256, "step": 49642 }, { "epoch": 3.0, "learning_rate": 2.506073363209754e-05, "loss": 0.4167, "step": 49644 }, { "epoch": 3.0, "learning_rate": 2.5059224721983316e-05, "loss": 0.5007, "step": 49646 }, { "epoch": 3.0, "learning_rate": 2.5057715811869086e-05, "loss": 0.5651, "step": 49648 }, { "epoch": 3.0, "learning_rate": 2.5056206901754862e-05, "loss": 0.3228, "step": 49650 }, { "epoch": 3.0, "learning_rate": 2.5054697991640642e-05, "loss": 0.6266, "step": 49652 }, { "epoch": 3.0, "learning_rate": 2.5053189081526412e-05, "loss": 0.4094, "step": 49654 }, { "epoch": 3.0, "learning_rate": 2.505168017141219e-05, "loss": 0.5082, "step": 49656 }, { "epoch": 3.0, "learning_rate": 2.5050171261297965e-05, "loss": 0.5318, "step": 49658 }, { "epoch": 3.0, "learning_rate": 2.5048662351183745e-05, "loss": 0.3978, "step": 49660 }, { "epoch": 3.0, "learning_rate": 2.5047153441069515e-05, "loss": 0.3897, "step": 49662 }, { "epoch": 3.0, "learning_rate": 2.504564453095529e-05, "loss": 0.4386, "step": 49664 }, { "epoch": 3.0, "learning_rate": 2.5044135620841068e-05, "loss": 0.5921, "step": 49666 }, { "epoch": 3.0, "learning_rate": 2.504262671072684e-05, "loss": 0.5441, "step": 49668 }, { "epoch": 3.0, "learning_rate": 2.5041117800612617e-05, "loss": 0.615, "step": 49670 }, { "epoch": 3.0, "learning_rate": 2.5039608890498394e-05, "loss": 0.4252, "step": 49672 }, { "epoch": 3.0, "learning_rate": 2.503809998038417e-05, "loss": 0.4504, "step": 49674 }, { "epoch": 3.0, "learning_rate": 2.5036591070269944e-05, "loss": 0.4843, "step": 49676 }, { "epoch": 3.0, "learning_rate": 2.503508216015572e-05, "loss": 0.5756, "step": 49678 }, { "epoch": 3.0, "learning_rate": 2.5033573250041497e-05, "loss": 0.568, "step": 49680 }, { "epoch": 3.0, "learning_rate": 2.503206433992727e-05, "loss": 0.6219, "step": 49682 }, { "epoch": 3.0, "learning_rate": 2.5030555429813046e-05, "loss": 0.5269, "step": 49684 }, { "epoch": 3.0, "learning_rate": 2.5029046519698823e-05, "loss": 0.6433, "step": 49686 }, { "epoch": 3.0, "learning_rate": 2.50275376095846e-05, "loss": 0.5146, "step": 49688 }, { "epoch": 3.0, "learning_rate": 2.5026028699470372e-05, "loss": 0.5653, "step": 49690 }, { "epoch": 3.0, "learning_rate": 2.502451978935615e-05, "loss": 0.4993, "step": 49692 }, { "epoch": 3.0, "learning_rate": 2.5023010879241925e-05, "loss": 0.4767, "step": 49694 }, { "epoch": 3.0, "learning_rate": 2.50215019691277e-05, "loss": 0.4231, "step": 49696 }, { "epoch": 3.0, "learning_rate": 2.5019993059013475e-05, "loss": 0.4477, "step": 49698 }, { "epoch": 3.0, "learning_rate": 2.501848414889925e-05, "loss": 0.5214, "step": 49700 }, { "epoch": 3.0, "eval_cer": 0.09257358867534708, "eval_loss": 0.573975145816803, "eval_runtime": 9373.7184, "eval_samples_per_second": 1.767, "eval_steps_per_second": 0.221, "step": 49701 }, { "epoch": 3.0, "learning_rate": 2.5016975238785028e-05, "loss": 0.3746, "step": 49702 }, { "epoch": 3.0, "learning_rate": 2.50154663286708e-05, "loss": 0.3476, "step": 49704 }, { "epoch": 3.0, "learning_rate": 2.5013957418556578e-05, "loss": 0.3868, "step": 49706 }, { "epoch": 3.0, "learning_rate": 2.5012448508442354e-05, "loss": 0.4208, "step": 49708 }, { "epoch": 3.0, "learning_rate": 2.501093959832813e-05, "loss": 0.2561, "step": 49710 }, { "epoch": 3.0, "learning_rate": 2.5009430688213904e-05, "loss": 0.3673, "step": 49712 }, { "epoch": 3.0, "learning_rate": 2.500792177809968e-05, "loss": 0.356, "step": 49714 }, { "epoch": 3.0, "learning_rate": 2.5006412867985457e-05, "loss": 0.3226, "step": 49716 }, { "epoch": 3.0, "learning_rate": 2.500490395787123e-05, "loss": 0.3307, "step": 49718 }, { "epoch": 3.0, "learning_rate": 2.5003395047757007e-05, "loss": 0.352, "step": 49720 }, { "epoch": 3.0, "learning_rate": 2.5001886137642783e-05, "loss": 0.3905, "step": 49722 }, { "epoch": 3.0, "learning_rate": 2.500037722752856e-05, "loss": 0.22, "step": 49724 }, { "epoch": 3.0, "learning_rate": 2.4998868317414333e-05, "loss": 0.3973, "step": 49726 }, { "epoch": 3.0, "learning_rate": 2.499735940730011e-05, "loss": 0.3772, "step": 49728 }, { "epoch": 3.0, "learning_rate": 2.4995850497185882e-05, "loss": 0.3054, "step": 49730 }, { "epoch": 3.0, "learning_rate": 2.499434158707166e-05, "loss": 0.438, "step": 49732 }, { "epoch": 3.0, "learning_rate": 2.4992832676957435e-05, "loss": 0.505, "step": 49734 }, { "epoch": 3.0, "learning_rate": 2.4991323766843212e-05, "loss": 0.2514, "step": 49736 }, { "epoch": 3.0, "learning_rate": 2.4989814856728985e-05, "loss": 0.394, "step": 49738 }, { "epoch": 3.0, "learning_rate": 2.498830594661476e-05, "loss": 0.376, "step": 49740 }, { "epoch": 3.0, "learning_rate": 2.4986797036500538e-05, "loss": 0.2397, "step": 49742 }, { "epoch": 3.0, "learning_rate": 2.4985288126386315e-05, "loss": 0.3801, "step": 49744 }, { "epoch": 3.0, "learning_rate": 2.4983779216272088e-05, "loss": 0.4768, "step": 49746 }, { "epoch": 3.0, "learning_rate": 2.498227030615786e-05, "loss": 0.2962, "step": 49748 }, { "epoch": 3.0, "learning_rate": 2.498076139604364e-05, "loss": 0.2608, "step": 49750 }, { "epoch": 3.0, "learning_rate": 2.4979252485929414e-05, "loss": 0.4183, "step": 49752 }, { "epoch": 3.0, "learning_rate": 2.497774357581519e-05, "loss": 0.3144, "step": 49754 }, { "epoch": 3.0, "learning_rate": 2.4976234665700963e-05, "loss": 0.4234, "step": 49756 }, { "epoch": 3.0, "learning_rate": 2.4974725755586743e-05, "loss": 0.4261, "step": 49758 }, { "epoch": 3.0, "learning_rate": 2.4973216845472517e-05, "loss": 0.4846, "step": 49760 }, { "epoch": 3.0, "learning_rate": 2.497170793535829e-05, "loss": 0.2861, "step": 49762 }, { "epoch": 3.0, "learning_rate": 2.4970199025244066e-05, "loss": 0.4512, "step": 49764 }, { "epoch": 3.0, "learning_rate": 2.4968690115129843e-05, "loss": 0.3189, "step": 49766 }, { "epoch": 3.0, "learning_rate": 2.496718120501562e-05, "loss": 0.5899, "step": 49768 }, { "epoch": 3.0, "learning_rate": 2.4965672294901392e-05, "loss": 0.2642, "step": 49770 }, { "epoch": 3.0, "learning_rate": 2.496416338478717e-05, "loss": 0.446, "step": 49772 }, { "epoch": 3.0, "learning_rate": 2.4962654474672945e-05, "loss": 0.2821, "step": 49774 }, { "epoch": 3.0, "learning_rate": 2.4961145564558722e-05, "loss": 0.4083, "step": 49776 }, { "epoch": 3.0, "learning_rate": 2.4959636654444495e-05, "loss": 0.2472, "step": 49778 }, { "epoch": 3.0, "learning_rate": 2.495812774433027e-05, "loss": 0.1963, "step": 49780 }, { "epoch": 3.0, "learning_rate": 2.4956618834216048e-05, "loss": 0.4303, "step": 49782 }, { "epoch": 3.0, "learning_rate": 2.495510992410182e-05, "loss": 0.3676, "step": 49784 }, { "epoch": 3.0, "learning_rate": 2.4953601013987598e-05, "loss": 0.2597, "step": 49786 }, { "epoch": 3.01, "learning_rate": 2.4952092103873374e-05, "loss": 0.3756, "step": 49788 }, { "epoch": 3.01, "learning_rate": 2.495058319375915e-05, "loss": 0.5536, "step": 49790 }, { "epoch": 3.01, "learning_rate": 2.4949074283644924e-05, "loss": 0.3031, "step": 49792 }, { "epoch": 3.01, "learning_rate": 2.49475653735307e-05, "loss": 0.3386, "step": 49794 }, { "epoch": 3.01, "learning_rate": 2.4946056463416477e-05, "loss": 0.2672, "step": 49796 }, { "epoch": 3.01, "learning_rate": 2.494454755330225e-05, "loss": 0.3092, "step": 49798 }, { "epoch": 3.01, "learning_rate": 2.4943038643188026e-05, "loss": 0.4498, "step": 49800 }, { "epoch": 3.01, "learning_rate": 2.49415297330738e-05, "loss": 0.3382, "step": 49802 }, { "epoch": 3.01, "learning_rate": 2.494002082295958e-05, "loss": 0.4764, "step": 49804 }, { "epoch": 3.01, "learning_rate": 2.4938511912845353e-05, "loss": 0.401, "step": 49806 }, { "epoch": 3.01, "learning_rate": 2.493700300273113e-05, "loss": 0.4196, "step": 49808 }, { "epoch": 3.01, "learning_rate": 2.4935494092616902e-05, "loss": 0.4676, "step": 49810 }, { "epoch": 3.01, "learning_rate": 2.493398518250268e-05, "loss": 0.5722, "step": 49812 }, { "epoch": 3.01, "learning_rate": 2.4932476272388455e-05, "loss": 0.3553, "step": 49814 }, { "epoch": 3.01, "learning_rate": 2.493096736227423e-05, "loss": 0.3972, "step": 49816 }, { "epoch": 3.01, "learning_rate": 2.492945845216001e-05, "loss": 0.3767, "step": 49818 }, { "epoch": 3.01, "learning_rate": 2.492794954204578e-05, "loss": 0.412, "step": 49820 }, { "epoch": 3.01, "learning_rate": 2.4926440631931558e-05, "loss": 0.3448, "step": 49822 }, { "epoch": 3.01, "learning_rate": 2.492493172181733e-05, "loss": 0.3875, "step": 49824 }, { "epoch": 3.01, "learning_rate": 2.492342281170311e-05, "loss": 0.2979, "step": 49826 }, { "epoch": 3.01, "learning_rate": 2.4921913901588884e-05, "loss": 0.4815, "step": 49828 }, { "epoch": 3.01, "learning_rate": 2.4920404991474657e-05, "loss": 0.4047, "step": 49830 }, { "epoch": 3.01, "learning_rate": 2.4918896081360434e-05, "loss": 0.4373, "step": 49832 }, { "epoch": 3.01, "learning_rate": 2.491738717124621e-05, "loss": 0.3742, "step": 49834 }, { "epoch": 3.01, "learning_rate": 2.4915878261131987e-05, "loss": 0.2996, "step": 49836 }, { "epoch": 3.01, "learning_rate": 2.491436935101776e-05, "loss": 0.4789, "step": 49838 }, { "epoch": 3.01, "learning_rate": 2.4912860440903536e-05, "loss": 0.5859, "step": 49840 }, { "epoch": 3.01, "learning_rate": 2.4911351530789313e-05, "loss": 0.3209, "step": 49842 }, { "epoch": 3.01, "learning_rate": 2.4909842620675086e-05, "loss": 0.3469, "step": 49844 }, { "epoch": 3.01, "learning_rate": 2.4908333710560863e-05, "loss": 0.4055, "step": 49846 }, { "epoch": 3.01, "learning_rate": 2.490682480044664e-05, "loss": 0.4237, "step": 49848 }, { "epoch": 3.01, "learning_rate": 2.4905315890332416e-05, "loss": 0.3781, "step": 49850 }, { "epoch": 3.01, "learning_rate": 2.490380698021819e-05, "loss": 0.3624, "step": 49852 }, { "epoch": 3.01, "learning_rate": 2.4902298070103965e-05, "loss": 0.3519, "step": 49854 }, { "epoch": 3.01, "learning_rate": 2.4900789159989742e-05, "loss": 0.4841, "step": 49856 }, { "epoch": 3.01, "learning_rate": 2.4899280249875518e-05, "loss": 0.2594, "step": 49858 }, { "epoch": 3.01, "learning_rate": 2.489777133976129e-05, "loss": 0.4156, "step": 49860 }, { "epoch": 3.01, "learning_rate": 2.4896262429647065e-05, "loss": 0.3335, "step": 49862 }, { "epoch": 3.01, "learning_rate": 2.4894753519532844e-05, "loss": 0.3664, "step": 49864 }, { "epoch": 3.01, "learning_rate": 2.4893244609418618e-05, "loss": 0.3019, "step": 49866 }, { "epoch": 3.01, "learning_rate": 2.4891735699304394e-05, "loss": 0.5174, "step": 49868 }, { "epoch": 3.01, "learning_rate": 2.4890226789190167e-05, "loss": 0.4683, "step": 49870 }, { "epoch": 3.01, "learning_rate": 2.4888717879075947e-05, "loss": 0.2765, "step": 49872 }, { "epoch": 3.01, "learning_rate": 2.488720896896172e-05, "loss": 0.3794, "step": 49874 }, { "epoch": 3.01, "learning_rate": 2.4885700058847497e-05, "loss": 0.3287, "step": 49876 }, { "epoch": 3.01, "learning_rate": 2.488419114873327e-05, "loss": 0.2629, "step": 49878 }, { "epoch": 3.01, "learning_rate": 2.4882682238619046e-05, "loss": 0.3976, "step": 49880 }, { "epoch": 3.01, "learning_rate": 2.4881173328504823e-05, "loss": 0.3279, "step": 49882 }, { "epoch": 3.01, "learning_rate": 2.4879664418390596e-05, "loss": 0.3603, "step": 49884 }, { "epoch": 3.01, "learning_rate": 2.4878155508276373e-05, "loss": 0.5643, "step": 49886 }, { "epoch": 3.01, "learning_rate": 2.487664659816215e-05, "loss": 0.4134, "step": 49888 }, { "epoch": 3.01, "learning_rate": 2.4875137688047926e-05, "loss": 0.3154, "step": 49890 }, { "epoch": 3.01, "learning_rate": 2.48736287779337e-05, "loss": 0.2452, "step": 49892 }, { "epoch": 3.01, "learning_rate": 2.4872119867819475e-05, "loss": 0.4417, "step": 49894 }, { "epoch": 3.01, "learning_rate": 2.487061095770525e-05, "loss": 0.3304, "step": 49896 }, { "epoch": 3.01, "learning_rate": 2.4869102047591025e-05, "loss": 0.4649, "step": 49898 }, { "epoch": 3.01, "learning_rate": 2.48675931374768e-05, "loss": 0.5944, "step": 49900 }, { "epoch": 3.01, "learning_rate": 2.4866084227362578e-05, "loss": 0.4351, "step": 49902 }, { "epoch": 3.01, "learning_rate": 2.4864575317248354e-05, "loss": 0.4167, "step": 49904 }, { "epoch": 3.01, "learning_rate": 2.4863066407134127e-05, "loss": 0.2356, "step": 49906 }, { "epoch": 3.01, "learning_rate": 2.4861557497019904e-05, "loss": 0.4617, "step": 49908 }, { "epoch": 3.01, "learning_rate": 2.486004858690568e-05, "loss": 0.2966, "step": 49910 }, { "epoch": 3.01, "learning_rate": 2.4858539676791454e-05, "loss": 0.5135, "step": 49912 }, { "epoch": 3.01, "learning_rate": 2.485703076667723e-05, "loss": 0.2889, "step": 49914 }, { "epoch": 3.01, "learning_rate": 2.4855521856563003e-05, "loss": 0.3875, "step": 49916 }, { "epoch": 3.01, "learning_rate": 2.4854012946448783e-05, "loss": 0.5133, "step": 49918 }, { "epoch": 3.01, "learning_rate": 2.4852504036334556e-05, "loss": 0.3788, "step": 49920 }, { "epoch": 3.01, "learning_rate": 2.4850995126220333e-05, "loss": 0.3809, "step": 49922 }, { "epoch": 3.01, "learning_rate": 2.4849486216106106e-05, "loss": 0.4703, "step": 49924 }, { "epoch": 3.01, "learning_rate": 2.4847977305991882e-05, "loss": 0.3684, "step": 49926 }, { "epoch": 3.01, "learning_rate": 2.484646839587766e-05, "loss": 0.3579, "step": 49928 }, { "epoch": 3.01, "learning_rate": 2.4844959485763432e-05, "loss": 0.6305, "step": 49930 }, { "epoch": 3.01, "learning_rate": 2.4843450575649212e-05, "loss": 0.3911, "step": 49932 }, { "epoch": 3.01, "learning_rate": 2.4841941665534985e-05, "loss": 0.4134, "step": 49934 }, { "epoch": 3.01, "learning_rate": 2.484043275542076e-05, "loss": 0.5034, "step": 49936 }, { "epoch": 3.01, "learning_rate": 2.4838923845306535e-05, "loss": 0.2538, "step": 49938 }, { "epoch": 3.01, "learning_rate": 2.4837414935192315e-05, "loss": 0.4863, "step": 49940 }, { "epoch": 3.01, "learning_rate": 2.4835906025078088e-05, "loss": 0.4448, "step": 49942 }, { "epoch": 3.01, "learning_rate": 2.483439711496386e-05, "loss": 0.3914, "step": 49944 }, { "epoch": 3.01, "learning_rate": 2.4832888204849637e-05, "loss": 0.3313, "step": 49946 }, { "epoch": 3.01, "learning_rate": 2.4831379294735414e-05, "loss": 0.2209, "step": 49948 }, { "epoch": 3.01, "learning_rate": 2.482987038462119e-05, "loss": 0.482, "step": 49950 }, { "epoch": 3.01, "learning_rate": 2.4828361474506964e-05, "loss": 0.3373, "step": 49952 }, { "epoch": 3.02, "learning_rate": 2.482685256439274e-05, "loss": 0.3401, "step": 49954 }, { "epoch": 3.02, "learning_rate": 2.4825343654278517e-05, "loss": 0.3348, "step": 49956 }, { "epoch": 3.02, "learning_rate": 2.4823834744164293e-05, "loss": 0.3873, "step": 49958 }, { "epoch": 3.02, "learning_rate": 2.4822325834050066e-05, "loss": 0.5219, "step": 49960 }, { "epoch": 3.02, "learning_rate": 2.4820816923935843e-05, "loss": 0.3222, "step": 49962 }, { "epoch": 3.02, "learning_rate": 2.481930801382162e-05, "loss": 0.3993, "step": 49964 }, { "epoch": 3.02, "learning_rate": 2.4817799103707392e-05, "loss": 0.3187, "step": 49966 }, { "epoch": 3.02, "learning_rate": 2.481629019359317e-05, "loss": 0.4368, "step": 49968 }, { "epoch": 3.02, "learning_rate": 2.4814781283478945e-05, "loss": 0.3758, "step": 49970 }, { "epoch": 3.02, "learning_rate": 2.4813272373364722e-05, "loss": 0.2393, "step": 49972 }, { "epoch": 3.02, "learning_rate": 2.4811763463250495e-05, "loss": 0.4332, "step": 49974 }, { "epoch": 3.02, "learning_rate": 2.4810254553136268e-05, "loss": 0.4105, "step": 49976 }, { "epoch": 3.02, "learning_rate": 2.4808745643022048e-05, "loss": 0.2898, "step": 49978 }, { "epoch": 3.02, "learning_rate": 2.480723673290782e-05, "loss": 0.6098, "step": 49980 }, { "epoch": 3.02, "learning_rate": 2.4805727822793598e-05, "loss": 0.4414, "step": 49982 }, { "epoch": 3.02, "learning_rate": 2.480421891267937e-05, "loss": 0.5416, "step": 49984 }, { "epoch": 3.02, "learning_rate": 2.480271000256515e-05, "loss": 0.3553, "step": 49986 }, { "epoch": 3.02, "learning_rate": 2.4801201092450924e-05, "loss": 0.3295, "step": 49988 }, { "epoch": 3.02, "learning_rate": 2.47996921823367e-05, "loss": 0.2753, "step": 49990 }, { "epoch": 3.02, "learning_rate": 2.4798183272222474e-05, "loss": 0.3836, "step": 49992 }, { "epoch": 3.02, "learning_rate": 2.479667436210825e-05, "loss": 0.5728, "step": 49994 }, { "epoch": 3.02, "learning_rate": 2.4795165451994027e-05, "loss": 0.4052, "step": 49996 }, { "epoch": 3.02, "learning_rate": 2.47936565418798e-05, "loss": 0.354, "step": 49998 }, { "epoch": 3.02, "learning_rate": 2.4792147631765576e-05, "loss": 0.4812, "step": 50000 }, { "epoch": 3.02, "learning_rate": 2.4790638721651353e-05, "loss": 0.4465, "step": 50002 }, { "epoch": 3.02, "learning_rate": 2.478912981153713e-05, "loss": 0.4473, "step": 50004 }, { "epoch": 3.02, "learning_rate": 2.4787620901422902e-05, "loss": 0.3296, "step": 50006 }, { "epoch": 3.02, "learning_rate": 2.478611199130868e-05, "loss": 0.3897, "step": 50008 }, { "epoch": 3.02, "learning_rate": 2.4784603081194455e-05, "loss": 0.4443, "step": 50010 }, { "epoch": 3.02, "learning_rate": 2.478309417108023e-05, "loss": 0.3881, "step": 50012 }, { "epoch": 3.02, "learning_rate": 2.4781585260966005e-05, "loss": 0.4313, "step": 50014 }, { "epoch": 3.02, "learning_rate": 2.478007635085178e-05, "loss": 0.2746, "step": 50016 }, { "epoch": 3.02, "learning_rate": 2.4778567440737558e-05, "loss": 0.3889, "step": 50018 }, { "epoch": 3.02, "learning_rate": 2.477705853062333e-05, "loss": 0.2888, "step": 50020 }, { "epoch": 3.02, "learning_rate": 2.4775549620509108e-05, "loss": 0.3834, "step": 50022 }, { "epoch": 3.02, "learning_rate": 2.4774040710394884e-05, "loss": 0.3108, "step": 50024 }, { "epoch": 3.02, "learning_rate": 2.4772531800280657e-05, "loss": 0.3782, "step": 50026 }, { "epoch": 3.02, "learning_rate": 2.4771022890166434e-05, "loss": 0.4364, "step": 50028 }, { "epoch": 3.02, "learning_rate": 2.4769513980052207e-05, "loss": 0.2251, "step": 50030 }, { "epoch": 3.02, "learning_rate": 2.4768005069937987e-05, "loss": 0.3151, "step": 50032 }, { "epoch": 3.02, "learning_rate": 2.476649615982376e-05, "loss": 0.311, "step": 50034 }, { "epoch": 3.02, "learning_rate": 2.4764987249709536e-05, "loss": 0.2873, "step": 50036 }, { "epoch": 3.02, "learning_rate": 2.476347833959531e-05, "loss": 0.5042, "step": 50038 }, { "epoch": 3.02, "learning_rate": 2.476196942948109e-05, "loss": 0.2945, "step": 50040 }, { "epoch": 3.02, "learning_rate": 2.4760460519366863e-05, "loss": 0.3708, "step": 50042 }, { "epoch": 3.02, "learning_rate": 2.4758951609252636e-05, "loss": 0.418, "step": 50044 }, { "epoch": 3.02, "learning_rate": 2.4757442699138412e-05, "loss": 0.3179, "step": 50046 }, { "epoch": 3.02, "learning_rate": 2.475593378902419e-05, "loss": 0.5301, "step": 50048 }, { "epoch": 3.02, "learning_rate": 2.4754424878909965e-05, "loss": 0.3198, "step": 50050 }, { "epoch": 3.02, "learning_rate": 2.475291596879574e-05, "loss": 0.3278, "step": 50052 }, { "epoch": 3.02, "learning_rate": 2.475140705868152e-05, "loss": 0.3719, "step": 50054 }, { "epoch": 3.02, "learning_rate": 2.474989814856729e-05, "loss": 0.4084, "step": 50056 }, { "epoch": 3.02, "learning_rate": 2.4748389238453065e-05, "loss": 0.4047, "step": 50058 }, { "epoch": 3.02, "learning_rate": 2.474688032833884e-05, "loss": 0.5002, "step": 50060 }, { "epoch": 3.02, "learning_rate": 2.4745371418224618e-05, "loss": 0.4295, "step": 50062 }, { "epoch": 3.02, "learning_rate": 2.4743862508110394e-05, "loss": 0.4071, "step": 50064 }, { "epoch": 3.02, "learning_rate": 2.4742353597996167e-05, "loss": 0.3942, "step": 50066 }, { "epoch": 3.02, "learning_rate": 2.4740844687881944e-05, "loss": 0.3481, "step": 50068 }, { "epoch": 3.02, "learning_rate": 2.473933577776772e-05, "loss": 0.4661, "step": 50070 }, { "epoch": 3.02, "learning_rate": 2.4737826867653497e-05, "loss": 0.3453, "step": 50072 }, { "epoch": 3.02, "learning_rate": 2.473631795753927e-05, "loss": 0.3415, "step": 50074 }, { "epoch": 3.02, "learning_rate": 2.4734809047425046e-05, "loss": 0.3851, "step": 50076 }, { "epoch": 3.02, "learning_rate": 2.4733300137310823e-05, "loss": 0.3657, "step": 50078 }, { "epoch": 3.02, "learning_rate": 2.4731791227196596e-05, "loss": 0.4635, "step": 50080 }, { "epoch": 3.02, "learning_rate": 2.4730282317082373e-05, "loss": 0.3494, "step": 50082 }, { "epoch": 3.02, "learning_rate": 2.472877340696815e-05, "loss": 0.2565, "step": 50084 }, { "epoch": 3.02, "learning_rate": 2.4727264496853926e-05, "loss": 0.2994, "step": 50086 }, { "epoch": 3.02, "learning_rate": 2.47257555867397e-05, "loss": 0.5244, "step": 50088 }, { "epoch": 3.02, "learning_rate": 2.4724246676625472e-05, "loss": 0.4231, "step": 50090 }, { "epoch": 3.02, "learning_rate": 2.4722737766511252e-05, "loss": 0.4041, "step": 50092 }, { "epoch": 3.02, "learning_rate": 2.4721228856397025e-05, "loss": 0.3864, "step": 50094 }, { "epoch": 3.02, "learning_rate": 2.47197199462828e-05, "loss": 0.3069, "step": 50096 }, { "epoch": 3.02, "learning_rate": 2.4718211036168575e-05, "loss": 0.3214, "step": 50098 }, { "epoch": 3.02, "learning_rate": 2.4716702126054354e-05, "loss": 0.4294, "step": 50100 }, { "epoch": 3.02, "learning_rate": 2.4715193215940128e-05, "loss": 0.3579, "step": 50102 }, { "epoch": 3.02, "learning_rate": 2.4713684305825904e-05, "loss": 0.4311, "step": 50104 }, { "epoch": 3.02, "learning_rate": 2.4712175395711677e-05, "loss": 0.4078, "step": 50106 }, { "epoch": 3.02, "learning_rate": 2.4710666485597454e-05, "loss": 0.3197, "step": 50108 }, { "epoch": 3.02, "learning_rate": 2.470915757548323e-05, "loss": 0.3994, "step": 50110 }, { "epoch": 3.02, "learning_rate": 2.4707648665369003e-05, "loss": 0.5264, "step": 50112 }, { "epoch": 3.02, "learning_rate": 2.470613975525478e-05, "loss": 0.3828, "step": 50114 }, { "epoch": 3.02, "learning_rate": 2.4704630845140556e-05, "loss": 0.2416, "step": 50116 }, { "epoch": 3.02, "learning_rate": 2.4703121935026333e-05, "loss": 0.3291, "step": 50118 }, { "epoch": 3.03, "learning_rate": 2.4701613024912106e-05, "loss": 0.3666, "step": 50120 }, { "epoch": 3.03, "learning_rate": 2.4700104114797883e-05, "loss": 0.2875, "step": 50122 }, { "epoch": 3.03, "learning_rate": 2.469859520468366e-05, "loss": 0.4742, "step": 50124 }, { "epoch": 3.03, "learning_rate": 2.4697086294569432e-05, "loss": 0.3272, "step": 50126 }, { "epoch": 3.03, "learning_rate": 2.469557738445521e-05, "loss": 0.3762, "step": 50128 }, { "epoch": 3.03, "learning_rate": 2.4694068474340985e-05, "loss": 0.4108, "step": 50130 }, { "epoch": 3.03, "learning_rate": 2.4692559564226762e-05, "loss": 0.2533, "step": 50132 }, { "epoch": 3.03, "learning_rate": 2.4691050654112535e-05, "loss": 0.5107, "step": 50134 }, { "epoch": 3.03, "learning_rate": 2.468954174399831e-05, "loss": 0.4027, "step": 50136 }, { "epoch": 3.03, "learning_rate": 2.4688032833884088e-05, "loss": 0.3498, "step": 50138 }, { "epoch": 3.03, "learning_rate": 2.468652392376986e-05, "loss": 0.3146, "step": 50140 }, { "epoch": 3.03, "learning_rate": 2.4685015013655638e-05, "loss": 0.3704, "step": 50142 }, { "epoch": 3.03, "learning_rate": 2.468350610354141e-05, "loss": 0.4446, "step": 50144 }, { "epoch": 3.03, "learning_rate": 2.468199719342719e-05, "loss": 0.3608, "step": 50146 }, { "epoch": 3.03, "learning_rate": 2.4680488283312964e-05, "loss": 0.4231, "step": 50148 }, { "epoch": 3.03, "learning_rate": 2.467897937319874e-05, "loss": 0.2959, "step": 50150 }, { "epoch": 3.03, "learning_rate": 2.4677470463084513e-05, "loss": 0.299, "step": 50152 }, { "epoch": 3.03, "learning_rate": 2.4675961552970293e-05, "loss": 0.3651, "step": 50154 }, { "epoch": 3.03, "learning_rate": 2.4674452642856066e-05, "loss": 0.5434, "step": 50156 }, { "epoch": 3.03, "learning_rate": 2.467294373274184e-05, "loss": 0.4699, "step": 50158 }, { "epoch": 3.03, "learning_rate": 2.4671434822627616e-05, "loss": 0.4278, "step": 50160 }, { "epoch": 3.03, "learning_rate": 2.4669925912513392e-05, "loss": 0.4464, "step": 50162 }, { "epoch": 3.03, "learning_rate": 2.466841700239917e-05, "loss": 0.4731, "step": 50164 }, { "epoch": 3.03, "learning_rate": 2.4666908092284942e-05, "loss": 0.3272, "step": 50166 }, { "epoch": 3.03, "learning_rate": 2.466539918217072e-05, "loss": 0.4023, "step": 50168 }, { "epoch": 3.03, "learning_rate": 2.4663890272056495e-05, "loss": 0.2514, "step": 50170 }, { "epoch": 3.03, "learning_rate": 2.466238136194227e-05, "loss": 0.4568, "step": 50172 }, { "epoch": 3.03, "learning_rate": 2.4660872451828045e-05, "loss": 0.2502, "step": 50174 }, { "epoch": 3.03, "learning_rate": 2.465936354171382e-05, "loss": 0.4759, "step": 50176 }, { "epoch": 3.03, "learning_rate": 2.4657854631599598e-05, "loss": 0.4461, "step": 50178 }, { "epoch": 3.03, "learning_rate": 2.465634572148537e-05, "loss": 0.4298, "step": 50180 }, { "epoch": 3.03, "learning_rate": 2.4654836811371147e-05, "loss": 0.413, "step": 50182 }, { "epoch": 3.03, "learning_rate": 2.4653327901256924e-05, "loss": 0.4829, "step": 50184 }, { "epoch": 3.03, "learning_rate": 2.46518189911427e-05, "loss": 0.3063, "step": 50186 }, { "epoch": 3.03, "learning_rate": 2.4650310081028474e-05, "loss": 0.4534, "step": 50188 }, { "epoch": 3.03, "learning_rate": 2.464880117091425e-05, "loss": 0.3799, "step": 50190 }, { "epoch": 3.03, "learning_rate": 2.4647292260800027e-05, "loss": 0.2709, "step": 50192 }, { "epoch": 3.03, "learning_rate": 2.46457833506858e-05, "loss": 0.3484, "step": 50194 }, { "epoch": 3.03, "learning_rate": 2.4644274440571576e-05, "loss": 0.4082, "step": 50196 }, { "epoch": 3.03, "learning_rate": 2.4642765530457353e-05, "loss": 0.4123, "step": 50198 }, { "epoch": 3.03, "learning_rate": 2.464125662034313e-05, "loss": 0.4338, "step": 50200 }, { "epoch": 3.03, "learning_rate": 2.4639747710228902e-05, "loss": 0.4735, "step": 50202 }, { "epoch": 3.03, "learning_rate": 2.463823880011468e-05, "loss": 0.4162, "step": 50204 }, { "epoch": 3.03, "learning_rate": 2.4636729890000455e-05, "loss": 0.5198, "step": 50206 }, { "epoch": 3.03, "learning_rate": 2.463522097988623e-05, "loss": 0.353, "step": 50208 }, { "epoch": 3.03, "learning_rate": 2.4633712069772005e-05, "loss": 0.2826, "step": 50210 }, { "epoch": 3.03, "learning_rate": 2.4632203159657778e-05, "loss": 0.3648, "step": 50212 }, { "epoch": 3.03, "learning_rate": 2.4630694249543558e-05, "loss": 0.4046, "step": 50214 }, { "epoch": 3.03, "learning_rate": 2.462918533942933e-05, "loss": 0.4274, "step": 50216 }, { "epoch": 3.03, "learning_rate": 2.4627676429315108e-05, "loss": 0.4502, "step": 50218 }, { "epoch": 3.03, "learning_rate": 2.462616751920088e-05, "loss": 0.407, "step": 50220 }, { "epoch": 3.03, "learning_rate": 2.4624658609086657e-05, "loss": 0.4038, "step": 50222 }, { "epoch": 3.03, "learning_rate": 2.4623149698972434e-05, "loss": 0.4083, "step": 50224 }, { "epoch": 3.03, "learning_rate": 2.4621640788858207e-05, "loss": 0.4644, "step": 50226 }, { "epoch": 3.03, "learning_rate": 2.4620131878743984e-05, "loss": 0.4451, "step": 50228 }, { "epoch": 3.03, "learning_rate": 2.461862296862976e-05, "loss": 0.5242, "step": 50230 }, { "epoch": 3.03, "learning_rate": 2.4617114058515537e-05, "loss": 0.359, "step": 50232 }, { "epoch": 3.03, "learning_rate": 2.461560514840131e-05, "loss": 0.333, "step": 50234 }, { "epoch": 3.03, "learning_rate": 2.4614096238287086e-05, "loss": 0.3595, "step": 50236 }, { "epoch": 3.03, "learning_rate": 2.4612587328172863e-05, "loss": 0.2722, "step": 50238 }, { "epoch": 3.03, "learning_rate": 2.4611078418058636e-05, "loss": 0.3743, "step": 50240 }, { "epoch": 3.03, "learning_rate": 2.4609569507944412e-05, "loss": 0.4731, "step": 50242 }, { "epoch": 3.03, "learning_rate": 2.460806059783019e-05, "loss": 0.4876, "step": 50244 }, { "epoch": 3.03, "learning_rate": 2.4606551687715965e-05, "loss": 0.4263, "step": 50246 }, { "epoch": 3.03, "learning_rate": 2.460504277760174e-05, "loss": 0.388, "step": 50248 }, { "epoch": 3.03, "learning_rate": 2.4603533867487515e-05, "loss": 0.3068, "step": 50250 }, { "epoch": 3.03, "learning_rate": 2.460202495737329e-05, "loss": 0.3044, "step": 50252 }, { "epoch": 3.03, "learning_rate": 2.4600516047259068e-05, "loss": 0.3244, "step": 50254 }, { "epoch": 3.03, "learning_rate": 2.459900713714484e-05, "loss": 0.4402, "step": 50256 }, { "epoch": 3.03, "learning_rate": 2.4597498227030614e-05, "loss": 0.5258, "step": 50258 }, { "epoch": 3.03, "learning_rate": 2.4595989316916394e-05, "loss": 0.324, "step": 50260 }, { "epoch": 3.03, "learning_rate": 2.4594480406802167e-05, "loss": 0.3548, "step": 50262 }, { "epoch": 3.03, "learning_rate": 2.4592971496687944e-05, "loss": 0.3686, "step": 50264 }, { "epoch": 3.03, "learning_rate": 2.4591462586573717e-05, "loss": 0.3832, "step": 50266 }, { "epoch": 3.03, "learning_rate": 2.4589953676459497e-05, "loss": 0.3775, "step": 50268 }, { "epoch": 3.03, "learning_rate": 2.458844476634527e-05, "loss": 0.4512, "step": 50270 }, { "epoch": 3.03, "learning_rate": 2.4586935856231043e-05, "loss": 0.2322, "step": 50272 }, { "epoch": 3.03, "learning_rate": 2.458542694611682e-05, "loss": 0.3851, "step": 50274 }, { "epoch": 3.03, "learning_rate": 2.4583918036002596e-05, "loss": 0.326, "step": 50276 }, { "epoch": 3.03, "learning_rate": 2.4582409125888373e-05, "loss": 0.2961, "step": 50278 }, { "epoch": 3.03, "learning_rate": 2.4580900215774146e-05, "loss": 0.4596, "step": 50280 }, { "epoch": 3.03, "learning_rate": 2.4579391305659922e-05, "loss": 0.3551, "step": 50282 }, { "epoch": 3.04, "learning_rate": 2.45778823955457e-05, "loss": 0.374, "step": 50284 }, { "epoch": 3.04, "learning_rate": 2.4576373485431475e-05, "loss": 0.3154, "step": 50286 }, { "epoch": 3.04, "learning_rate": 2.457486457531725e-05, "loss": 0.3996, "step": 50288 }, { "epoch": 3.04, "learning_rate": 2.4573355665203025e-05, "loss": 0.2858, "step": 50290 }, { "epoch": 3.04, "learning_rate": 2.45718467550888e-05, "loss": 0.51, "step": 50292 }, { "epoch": 3.04, "learning_rate": 2.4570337844974575e-05, "loss": 0.3811, "step": 50294 }, { "epoch": 3.04, "learning_rate": 2.456882893486035e-05, "loss": 0.2844, "step": 50296 }, { "epoch": 3.04, "learning_rate": 2.4567320024746128e-05, "loss": 0.3191, "step": 50298 }, { "epoch": 3.04, "learning_rate": 2.4565811114631904e-05, "loss": 0.4317, "step": 50300 }, { "epoch": 3.04, "learning_rate": 2.4564302204517677e-05, "loss": 0.2773, "step": 50302 }, { "epoch": 3.04, "learning_rate": 2.4562793294403454e-05, "loss": 0.3517, "step": 50304 }, { "epoch": 3.04, "learning_rate": 2.456128438428923e-05, "loss": 0.4101, "step": 50306 }, { "epoch": 3.04, "learning_rate": 2.4559775474175003e-05, "loss": 0.2551, "step": 50308 }, { "epoch": 3.04, "learning_rate": 2.455826656406078e-05, "loss": 0.416, "step": 50310 }, { "epoch": 3.04, "learning_rate": 2.4556757653946556e-05, "loss": 0.375, "step": 50312 }, { "epoch": 3.04, "learning_rate": 2.4555248743832333e-05, "loss": 0.5044, "step": 50314 }, { "epoch": 3.04, "learning_rate": 2.4553739833718106e-05, "loss": 0.3963, "step": 50316 }, { "epoch": 3.04, "learning_rate": 2.4552230923603883e-05, "loss": 0.3271, "step": 50318 }, { "epoch": 3.04, "learning_rate": 2.455072201348966e-05, "loss": 0.4032, "step": 50320 }, { "epoch": 3.04, "learning_rate": 2.4549213103375432e-05, "loss": 0.2995, "step": 50322 }, { "epoch": 3.04, "learning_rate": 2.454770419326121e-05, "loss": 0.3597, "step": 50324 }, { "epoch": 3.04, "learning_rate": 2.4546195283146982e-05, "loss": 0.4427, "step": 50326 }, { "epoch": 3.04, "learning_rate": 2.4544686373032762e-05, "loss": 0.3559, "step": 50328 }, { "epoch": 3.04, "learning_rate": 2.4543177462918535e-05, "loss": 0.4547, "step": 50330 }, { "epoch": 3.04, "learning_rate": 2.454166855280431e-05, "loss": 0.4595, "step": 50332 }, { "epoch": 3.04, "learning_rate": 2.4540159642690085e-05, "loss": 0.4249, "step": 50334 }, { "epoch": 3.04, "learning_rate": 2.4538650732575864e-05, "loss": 0.3436, "step": 50336 }, { "epoch": 3.04, "learning_rate": 2.4537141822461638e-05, "loss": 0.6516, "step": 50338 }, { "epoch": 3.04, "learning_rate": 2.453563291234741e-05, "loss": 0.3416, "step": 50340 }, { "epoch": 3.04, "learning_rate": 2.4534124002233187e-05, "loss": 0.494, "step": 50342 }, { "epoch": 3.04, "learning_rate": 2.4532615092118964e-05, "loss": 0.4416, "step": 50344 }, { "epoch": 3.04, "learning_rate": 2.453110618200474e-05, "loss": 0.2387, "step": 50346 }, { "epoch": 3.04, "learning_rate": 2.4529597271890513e-05, "loss": 0.2775, "step": 50348 }, { "epoch": 3.04, "learning_rate": 2.452808836177629e-05, "loss": 0.4572, "step": 50350 }, { "epoch": 3.04, "learning_rate": 2.4526579451662066e-05, "loss": 0.3946, "step": 50352 }, { "epoch": 3.04, "learning_rate": 2.452507054154784e-05, "loss": 0.2635, "step": 50354 }, { "epoch": 3.04, "learning_rate": 2.4523561631433616e-05, "loss": 0.3573, "step": 50356 }, { "epoch": 3.04, "learning_rate": 2.4522052721319393e-05, "loss": 0.5403, "step": 50358 }, { "epoch": 3.04, "learning_rate": 2.452054381120517e-05, "loss": 0.271, "step": 50360 }, { "epoch": 3.04, "learning_rate": 2.4519034901090942e-05, "loss": 0.3053, "step": 50362 }, { "epoch": 3.04, "learning_rate": 2.451752599097672e-05, "loss": 0.3305, "step": 50364 }, { "epoch": 3.04, "learning_rate": 2.4516017080862495e-05, "loss": 0.3575, "step": 50366 }, { "epoch": 3.04, "learning_rate": 2.4514508170748272e-05, "loss": 0.4962, "step": 50368 }, { "epoch": 3.04, "learning_rate": 2.4512999260634045e-05, "loss": 0.3806, "step": 50370 }, { "epoch": 3.04, "learning_rate": 2.4511490350519818e-05, "loss": 0.3808, "step": 50372 }, { "epoch": 3.04, "learning_rate": 2.4509981440405598e-05, "loss": 0.3063, "step": 50374 }, { "epoch": 3.04, "learning_rate": 2.450847253029137e-05, "loss": 0.5093, "step": 50376 }, { "epoch": 3.04, "learning_rate": 2.4506963620177148e-05, "loss": 0.3998, "step": 50378 }, { "epoch": 3.04, "learning_rate": 2.450545471006292e-05, "loss": 0.4546, "step": 50380 }, { "epoch": 3.04, "learning_rate": 2.45039457999487e-05, "loss": 0.3223, "step": 50382 }, { "epoch": 3.04, "learning_rate": 2.4502436889834474e-05, "loss": 0.3279, "step": 50384 }, { "epoch": 3.04, "learning_rate": 2.4500927979720247e-05, "loss": 0.5112, "step": 50386 }, { "epoch": 3.04, "learning_rate": 2.4499419069606023e-05, "loss": 0.3094, "step": 50388 }, { "epoch": 3.04, "learning_rate": 2.44979101594918e-05, "loss": 0.4065, "step": 50390 }, { "epoch": 3.04, "learning_rate": 2.4496401249377576e-05, "loss": 0.4752, "step": 50392 }, { "epoch": 3.04, "learning_rate": 2.449489233926335e-05, "loss": 0.3884, "step": 50394 }, { "epoch": 3.04, "learning_rate": 2.4493383429149126e-05, "loss": 0.297, "step": 50396 }, { "epoch": 3.04, "learning_rate": 2.4491874519034902e-05, "loss": 0.3312, "step": 50398 }, { "epoch": 3.04, "learning_rate": 2.449036560892068e-05, "loss": 0.3414, "step": 50400 }, { "epoch": 3.04, "learning_rate": 2.4488856698806452e-05, "loss": 0.357, "step": 50402 }, { "epoch": 3.04, "learning_rate": 2.448734778869223e-05, "loss": 0.316, "step": 50404 }, { "epoch": 3.04, "learning_rate": 2.4485838878578005e-05, "loss": 0.3293, "step": 50406 }, { "epoch": 3.04, "learning_rate": 2.4484329968463778e-05, "loss": 0.2923, "step": 50408 }, { "epoch": 3.04, "learning_rate": 2.4482821058349555e-05, "loss": 0.3594, "step": 50410 }, { "epoch": 3.04, "learning_rate": 2.448131214823533e-05, "loss": 0.5322, "step": 50412 }, { "epoch": 3.04, "learning_rate": 2.4479803238121108e-05, "loss": 0.2869, "step": 50414 }, { "epoch": 3.04, "learning_rate": 2.447829432800688e-05, "loss": 0.314, "step": 50416 }, { "epoch": 3.04, "learning_rate": 2.4476785417892657e-05, "loss": 0.3159, "step": 50418 }, { "epoch": 3.04, "learning_rate": 2.4475276507778434e-05, "loss": 0.3236, "step": 50420 }, { "epoch": 3.04, "learning_rate": 2.4473767597664207e-05, "loss": 0.3558, "step": 50422 }, { "epoch": 3.04, "learning_rate": 2.4472258687549984e-05, "loss": 0.2917, "step": 50424 }, { "epoch": 3.04, "learning_rate": 2.447074977743576e-05, "loss": 0.4756, "step": 50426 }, { "epoch": 3.04, "learning_rate": 2.4469240867321537e-05, "loss": 0.4462, "step": 50428 }, { "epoch": 3.04, "learning_rate": 2.446773195720731e-05, "loss": 0.3616, "step": 50430 }, { "epoch": 3.04, "learning_rate": 2.4466223047093086e-05, "loss": 0.4728, "step": 50432 }, { "epoch": 3.04, "learning_rate": 2.4464714136978863e-05, "loss": 0.4263, "step": 50434 }, { "epoch": 3.04, "learning_rate": 2.4463205226864636e-05, "loss": 0.3649, "step": 50436 }, { "epoch": 3.04, "learning_rate": 2.4461696316750412e-05, "loss": 0.3954, "step": 50438 }, { "epoch": 3.04, "learning_rate": 2.4460187406636186e-05, "loss": 0.5345, "step": 50440 }, { "epoch": 3.04, "learning_rate": 2.4458678496521965e-05, "loss": 0.4119, "step": 50442 }, { "epoch": 3.04, "learning_rate": 2.445716958640774e-05, "loss": 0.3737, "step": 50444 }, { "epoch": 3.04, "learning_rate": 2.4455660676293515e-05, "loss": 0.3687, "step": 50446 }, { "epoch": 3.04, "learning_rate": 2.4454151766179288e-05, "loss": 0.5036, "step": 50448 }, { "epoch": 3.05, "learning_rate": 2.4452642856065068e-05, "loss": 0.3255, "step": 50450 }, { "epoch": 3.05, "learning_rate": 2.445113394595084e-05, "loss": 0.3956, "step": 50452 }, { "epoch": 3.05, "learning_rate": 2.4449625035836614e-05, "loss": 0.3859, "step": 50454 }, { "epoch": 3.05, "learning_rate": 2.444811612572239e-05, "loss": 0.397, "step": 50456 }, { "epoch": 3.05, "learning_rate": 2.4446607215608167e-05, "loss": 0.3404, "step": 50458 }, { "epoch": 3.05, "learning_rate": 2.4445098305493944e-05, "loss": 0.4275, "step": 50460 }, { "epoch": 3.05, "learning_rate": 2.4443589395379717e-05, "loss": 0.4517, "step": 50462 }, { "epoch": 3.05, "learning_rate": 2.4442080485265494e-05, "loss": 0.4189, "step": 50464 }, { "epoch": 3.05, "learning_rate": 2.444057157515127e-05, "loss": 0.3292, "step": 50466 }, { "epoch": 3.05, "learning_rate": 2.4439062665037043e-05, "loss": 0.5478, "step": 50468 }, { "epoch": 3.05, "learning_rate": 2.443755375492282e-05, "loss": 0.4214, "step": 50470 }, { "epoch": 3.05, "learning_rate": 2.4436044844808596e-05, "loss": 0.4584, "step": 50472 }, { "epoch": 3.05, "learning_rate": 2.4434535934694373e-05, "loss": 0.365, "step": 50474 }, { "epoch": 3.05, "learning_rate": 2.4433027024580146e-05, "loss": 0.266, "step": 50476 }, { "epoch": 3.05, "learning_rate": 2.4431518114465922e-05, "loss": 0.424, "step": 50478 }, { "epoch": 3.05, "learning_rate": 2.44300092043517e-05, "loss": 0.4676, "step": 50480 }, { "epoch": 3.05, "learning_rate": 2.4428500294237475e-05, "loss": 0.4269, "step": 50482 }, { "epoch": 3.05, "learning_rate": 2.442699138412325e-05, "loss": 0.4371, "step": 50484 }, { "epoch": 3.05, "learning_rate": 2.442548247400902e-05, "loss": 0.4853, "step": 50486 }, { "epoch": 3.05, "learning_rate": 2.44239735638948e-05, "loss": 0.3928, "step": 50488 }, { "epoch": 3.05, "learning_rate": 2.4422464653780575e-05, "loss": 0.4131, "step": 50490 }, { "epoch": 3.05, "learning_rate": 2.442095574366635e-05, "loss": 0.2178, "step": 50492 }, { "epoch": 3.05, "learning_rate": 2.4419446833552124e-05, "loss": 0.4717, "step": 50494 }, { "epoch": 3.05, "learning_rate": 2.4417937923437904e-05, "loss": 0.2791, "step": 50496 }, { "epoch": 3.05, "learning_rate": 2.4416429013323677e-05, "loss": 0.293, "step": 50498 }, { "epoch": 3.05, "learning_rate": 2.4414920103209454e-05, "loss": 0.2591, "step": 50500 }, { "epoch": 3.05, "learning_rate": 2.4413411193095227e-05, "loss": 0.2878, "step": 50502 }, { "epoch": 3.05, "learning_rate": 2.4411902282981004e-05, "loss": 0.4582, "step": 50504 }, { "epoch": 3.05, "learning_rate": 2.441039337286678e-05, "loss": 0.4149, "step": 50506 }, { "epoch": 3.05, "learning_rate": 2.4408884462752553e-05, "loss": 0.3668, "step": 50508 }, { "epoch": 3.05, "learning_rate": 2.440737555263833e-05, "loss": 0.3862, "step": 50510 }, { "epoch": 3.05, "learning_rate": 2.4405866642524106e-05, "loss": 0.2945, "step": 50512 }, { "epoch": 3.05, "learning_rate": 2.4404357732409883e-05, "loss": 0.3811, "step": 50514 }, { "epoch": 3.05, "learning_rate": 2.4402848822295656e-05, "loss": 0.4035, "step": 50516 }, { "epoch": 3.05, "learning_rate": 2.4401339912181432e-05, "loss": 0.355, "step": 50518 }, { "epoch": 3.05, "learning_rate": 2.439983100206721e-05, "loss": 0.394, "step": 50520 }, { "epoch": 3.05, "learning_rate": 2.4398322091952982e-05, "loss": 0.3084, "step": 50522 }, { "epoch": 3.05, "learning_rate": 2.439681318183876e-05, "loss": 0.3671, "step": 50524 }, { "epoch": 3.05, "learning_rate": 2.4395304271724535e-05, "loss": 0.2918, "step": 50526 }, { "epoch": 3.05, "learning_rate": 2.439379536161031e-05, "loss": 0.4324, "step": 50528 }, { "epoch": 3.05, "learning_rate": 2.4392286451496085e-05, "loss": 0.3, "step": 50530 }, { "epoch": 3.05, "learning_rate": 2.439077754138186e-05, "loss": 0.317, "step": 50532 }, { "epoch": 3.05, "learning_rate": 2.4389268631267638e-05, "loss": 0.3972, "step": 50534 }, { "epoch": 3.05, "learning_rate": 2.438775972115341e-05, "loss": 0.3304, "step": 50536 }, { "epoch": 3.05, "learning_rate": 2.4386250811039187e-05, "loss": 0.5416, "step": 50538 }, { "epoch": 3.05, "learning_rate": 2.4384741900924964e-05, "loss": 0.3238, "step": 50540 }, { "epoch": 3.05, "learning_rate": 2.438323299081074e-05, "loss": 0.6123, "step": 50542 }, { "epoch": 3.05, "learning_rate": 2.4381724080696513e-05, "loss": 0.5109, "step": 50544 }, { "epoch": 3.05, "learning_rate": 2.438021517058229e-05, "loss": 0.3236, "step": 50546 }, { "epoch": 3.05, "learning_rate": 2.4378706260468066e-05, "loss": 0.4242, "step": 50548 }, { "epoch": 3.05, "learning_rate": 2.437719735035384e-05, "loss": 0.2972, "step": 50550 }, { "epoch": 3.05, "learning_rate": 2.4375688440239616e-05, "loss": 0.4143, "step": 50552 }, { "epoch": 3.05, "learning_rate": 2.437417953012539e-05, "loss": 0.3872, "step": 50554 }, { "epoch": 3.05, "learning_rate": 2.437267062001117e-05, "loss": 0.3286, "step": 50556 }, { "epoch": 3.05, "learning_rate": 2.4371161709896942e-05, "loss": 0.4695, "step": 50558 }, { "epoch": 3.05, "learning_rate": 2.436965279978272e-05, "loss": 0.2957, "step": 50560 }, { "epoch": 3.05, "learning_rate": 2.4368143889668492e-05, "loss": 0.4132, "step": 50562 }, { "epoch": 3.05, "learning_rate": 2.4366634979554272e-05, "loss": 0.4817, "step": 50564 }, { "epoch": 3.05, "learning_rate": 2.4365126069440045e-05, "loss": 0.3635, "step": 50566 }, { "epoch": 3.05, "learning_rate": 2.4363617159325818e-05, "loss": 0.6013, "step": 50568 }, { "epoch": 3.05, "learning_rate": 2.4362108249211595e-05, "loss": 0.2987, "step": 50570 }, { "epoch": 3.05, "learning_rate": 2.436059933909737e-05, "loss": 0.4534, "step": 50572 }, { "epoch": 3.05, "learning_rate": 2.4359090428983148e-05, "loss": 0.2773, "step": 50574 }, { "epoch": 3.05, "learning_rate": 2.435758151886892e-05, "loss": 0.4172, "step": 50576 }, { "epoch": 3.05, "learning_rate": 2.4356072608754697e-05, "loss": 0.2273, "step": 50578 }, { "epoch": 3.05, "learning_rate": 2.4354563698640474e-05, "loss": 0.2567, "step": 50580 }, { "epoch": 3.05, "learning_rate": 2.435305478852625e-05, "loss": 0.4778, "step": 50582 }, { "epoch": 3.05, "learning_rate": 2.4351545878412023e-05, "loss": 0.3457, "step": 50584 }, { "epoch": 3.05, "learning_rate": 2.43500369682978e-05, "loss": 0.3402, "step": 50586 }, { "epoch": 3.05, "learning_rate": 2.4348528058183576e-05, "loss": 0.4081, "step": 50588 }, { "epoch": 3.05, "learning_rate": 2.434701914806935e-05, "loss": 0.3564, "step": 50590 }, { "epoch": 3.05, "learning_rate": 2.4345510237955126e-05, "loss": 0.4205, "step": 50592 }, { "epoch": 3.05, "learning_rate": 2.4344001327840903e-05, "loss": 0.2484, "step": 50594 }, { "epoch": 3.05, "learning_rate": 2.434249241772668e-05, "loss": 0.3097, "step": 50596 }, { "epoch": 3.05, "learning_rate": 2.4340983507612452e-05, "loss": 0.2801, "step": 50598 }, { "epoch": 3.05, "learning_rate": 2.4339474597498225e-05, "loss": 0.3197, "step": 50600 }, { "epoch": 3.05, "learning_rate": 2.4337965687384005e-05, "loss": 0.428, "step": 50602 }, { "epoch": 3.05, "learning_rate": 2.433645677726978e-05, "loss": 0.4454, "step": 50604 }, { "epoch": 3.05, "learning_rate": 2.4334947867155555e-05, "loss": 0.4047, "step": 50606 }, { "epoch": 3.05, "learning_rate": 2.4333438957041328e-05, "loss": 0.4534, "step": 50608 }, { "epoch": 3.05, "learning_rate": 2.4331930046927108e-05, "loss": 0.3504, "step": 50610 }, { "epoch": 3.05, "learning_rate": 2.433042113681288e-05, "loss": 0.3322, "step": 50612 }, { "epoch": 3.05, "learning_rate": 2.4328912226698658e-05, "loss": 0.4426, "step": 50614 }, { "epoch": 3.06, "learning_rate": 2.432740331658443e-05, "loss": 0.4385, "step": 50616 }, { "epoch": 3.06, "learning_rate": 2.4325894406470207e-05, "loss": 0.5402, "step": 50618 }, { "epoch": 3.06, "learning_rate": 2.4324385496355984e-05, "loss": 0.3884, "step": 50620 }, { "epoch": 3.06, "learning_rate": 2.4322876586241757e-05, "loss": 0.3476, "step": 50622 }, { "epoch": 3.06, "learning_rate": 2.4321367676127533e-05, "loss": 0.425, "step": 50624 }, { "epoch": 3.06, "learning_rate": 2.431985876601331e-05, "loss": 0.5093, "step": 50626 }, { "epoch": 3.06, "learning_rate": 2.4318349855899086e-05, "loss": 0.4118, "step": 50628 }, { "epoch": 3.06, "learning_rate": 2.431684094578486e-05, "loss": 0.6802, "step": 50630 }, { "epoch": 3.06, "learning_rate": 2.4315332035670636e-05, "loss": 0.4171, "step": 50632 }, { "epoch": 3.06, "learning_rate": 2.4313823125556413e-05, "loss": 0.2935, "step": 50634 }, { "epoch": 3.06, "learning_rate": 2.4312314215442186e-05, "loss": 0.37, "step": 50636 }, { "epoch": 3.06, "learning_rate": 2.4310805305327962e-05, "loss": 0.5473, "step": 50638 }, { "epoch": 3.06, "learning_rate": 2.430929639521374e-05, "loss": 0.3973, "step": 50640 }, { "epoch": 3.06, "learning_rate": 2.4307787485099515e-05, "loss": 0.2537, "step": 50642 }, { "epoch": 3.06, "learning_rate": 2.430627857498529e-05, "loss": 0.5014, "step": 50644 }, { "epoch": 3.06, "learning_rate": 2.4304769664871065e-05, "loss": 0.4453, "step": 50646 }, { "epoch": 3.06, "learning_rate": 2.430326075475684e-05, "loss": 0.5292, "step": 50648 }, { "epoch": 3.06, "learning_rate": 2.4301751844642614e-05, "loss": 0.2965, "step": 50650 }, { "epoch": 3.06, "learning_rate": 2.430024293452839e-05, "loss": 0.2963, "step": 50652 }, { "epoch": 3.06, "learning_rate": 2.4298734024414167e-05, "loss": 0.4024, "step": 50654 }, { "epoch": 3.06, "learning_rate": 2.4297225114299944e-05, "loss": 0.5193, "step": 50656 }, { "epoch": 3.06, "learning_rate": 2.4295716204185717e-05, "loss": 0.3824, "step": 50658 }, { "epoch": 3.06, "learning_rate": 2.4294207294071494e-05, "loss": 0.435, "step": 50660 }, { "epoch": 3.06, "learning_rate": 2.429269838395727e-05, "loss": 0.366, "step": 50662 }, { "epoch": 3.06, "learning_rate": 2.4291189473843047e-05, "loss": 0.3049, "step": 50664 }, { "epoch": 3.06, "learning_rate": 2.428968056372882e-05, "loss": 0.3046, "step": 50666 }, { "epoch": 3.06, "learning_rate": 2.4288171653614593e-05, "loss": 0.3594, "step": 50668 }, { "epoch": 3.06, "learning_rate": 2.4286662743500373e-05, "loss": 0.3699, "step": 50670 }, { "epoch": 3.06, "learning_rate": 2.4285153833386146e-05, "loss": 0.3335, "step": 50672 }, { "epoch": 3.06, "learning_rate": 2.4283644923271922e-05, "loss": 0.2171, "step": 50674 }, { "epoch": 3.06, "learning_rate": 2.4282136013157696e-05, "loss": 0.3843, "step": 50676 }, { "epoch": 3.06, "learning_rate": 2.4280627103043475e-05, "loss": 0.2901, "step": 50678 }, { "epoch": 3.06, "learning_rate": 2.427911819292925e-05, "loss": 0.3708, "step": 50680 }, { "epoch": 3.06, "learning_rate": 2.4277609282815022e-05, "loss": 0.3506, "step": 50682 }, { "epoch": 3.06, "learning_rate": 2.4276100372700798e-05, "loss": 0.2264, "step": 50684 }, { "epoch": 3.06, "learning_rate": 2.4274591462586575e-05, "loss": 0.3839, "step": 50686 }, { "epoch": 3.06, "learning_rate": 2.427308255247235e-05, "loss": 0.4247, "step": 50688 }, { "epoch": 3.06, "learning_rate": 2.4271573642358124e-05, "loss": 0.4542, "step": 50690 }, { "epoch": 3.06, "learning_rate": 2.42700647322439e-05, "loss": 0.4199, "step": 50692 }, { "epoch": 3.06, "learning_rate": 2.4268555822129677e-05, "loss": 0.3494, "step": 50694 }, { "epoch": 3.06, "learning_rate": 2.4267046912015454e-05, "loss": 0.4082, "step": 50696 }, { "epoch": 3.06, "learning_rate": 2.4265538001901227e-05, "loss": 0.4248, "step": 50698 }, { "epoch": 3.06, "learning_rate": 2.4264029091787004e-05, "loss": 0.5437, "step": 50700 }, { "epoch": 3.06, "learning_rate": 2.426252018167278e-05, "loss": 0.3749, "step": 50702 }, { "epoch": 3.06, "learning_rate": 2.4261011271558553e-05, "loss": 0.4783, "step": 50704 }, { "epoch": 3.06, "learning_rate": 2.425950236144433e-05, "loss": 0.3676, "step": 50706 }, { "epoch": 3.06, "learning_rate": 2.4257993451330106e-05, "loss": 0.2737, "step": 50708 }, { "epoch": 3.06, "learning_rate": 2.4256484541215883e-05, "loss": 0.4385, "step": 50710 }, { "epoch": 3.06, "learning_rate": 2.4254975631101656e-05, "loss": 0.3458, "step": 50712 }, { "epoch": 3.06, "learning_rate": 2.425346672098743e-05, "loss": 0.3674, "step": 50714 }, { "epoch": 3.06, "learning_rate": 2.425195781087321e-05, "loss": 0.3623, "step": 50716 }, { "epoch": 3.06, "learning_rate": 2.4250448900758982e-05, "loss": 0.3565, "step": 50718 }, { "epoch": 3.06, "learning_rate": 2.424893999064476e-05, "loss": 0.3432, "step": 50720 }, { "epoch": 3.06, "learning_rate": 2.424743108053053e-05, "loss": 0.4775, "step": 50722 }, { "epoch": 3.06, "learning_rate": 2.424592217041631e-05, "loss": 0.4724, "step": 50724 }, { "epoch": 3.06, "learning_rate": 2.4244413260302085e-05, "loss": 0.4483, "step": 50726 }, { "epoch": 3.06, "learning_rate": 2.424290435018786e-05, "loss": 0.337, "step": 50728 }, { "epoch": 3.06, "learning_rate": 2.4241395440073634e-05, "loss": 0.3553, "step": 50730 }, { "epoch": 3.06, "learning_rate": 2.423988652995941e-05, "loss": 0.4382, "step": 50732 }, { "epoch": 3.06, "learning_rate": 2.4238377619845187e-05, "loss": 0.3657, "step": 50734 }, { "epoch": 3.06, "learning_rate": 2.423686870973096e-05, "loss": 0.3134, "step": 50736 }, { "epoch": 3.06, "learning_rate": 2.4235359799616737e-05, "loss": 0.3525, "step": 50738 }, { "epoch": 3.06, "learning_rate": 2.4233850889502514e-05, "loss": 0.3535, "step": 50740 }, { "epoch": 3.06, "learning_rate": 2.423234197938829e-05, "loss": 0.3092, "step": 50742 }, { "epoch": 3.06, "learning_rate": 2.4230833069274063e-05, "loss": 0.43, "step": 50744 }, { "epoch": 3.06, "learning_rate": 2.422932415915984e-05, "loss": 0.5119, "step": 50746 }, { "epoch": 3.06, "learning_rate": 2.4227815249045616e-05, "loss": 0.3555, "step": 50748 }, { "epoch": 3.06, "learning_rate": 2.422630633893139e-05, "loss": 0.3669, "step": 50750 }, { "epoch": 3.06, "learning_rate": 2.4224797428817166e-05, "loss": 0.3909, "step": 50752 }, { "epoch": 3.06, "learning_rate": 2.4223288518702942e-05, "loss": 0.3834, "step": 50754 }, { "epoch": 3.06, "learning_rate": 2.422177960858872e-05, "loss": 0.3368, "step": 50756 }, { "epoch": 3.06, "learning_rate": 2.4220270698474492e-05, "loss": 0.5397, "step": 50758 }, { "epoch": 3.06, "learning_rate": 2.421876178836027e-05, "loss": 0.3794, "step": 50760 }, { "epoch": 3.06, "learning_rate": 2.4217252878246045e-05, "loss": 0.352, "step": 50762 }, { "epoch": 3.06, "learning_rate": 2.4215743968131818e-05, "loss": 0.4047, "step": 50764 }, { "epoch": 3.06, "learning_rate": 2.4214235058017595e-05, "loss": 0.4617, "step": 50766 }, { "epoch": 3.06, "learning_rate": 2.421272614790337e-05, "loss": 0.3006, "step": 50768 }, { "epoch": 3.06, "learning_rate": 2.4211217237789148e-05, "loss": 0.4919, "step": 50770 }, { "epoch": 3.06, "learning_rate": 2.420970832767492e-05, "loss": 0.2326, "step": 50772 }, { "epoch": 3.06, "learning_rate": 2.4208199417560697e-05, "loss": 0.3381, "step": 50774 }, { "epoch": 3.06, "learning_rate": 2.4206690507446474e-05, "loss": 0.5026, "step": 50776 }, { "epoch": 3.06, "learning_rate": 2.420518159733225e-05, "loss": 0.3247, "step": 50778 }, { "epoch": 3.06, "learning_rate": 2.4203672687218023e-05, "loss": 0.2514, "step": 50780 }, { "epoch": 3.07, "learning_rate": 2.4202163777103797e-05, "loss": 0.3638, "step": 50782 }, { "epoch": 3.07, "learning_rate": 2.4200654866989577e-05, "loss": 0.2179, "step": 50784 }, { "epoch": 3.07, "learning_rate": 2.419914595687535e-05, "loss": 0.3842, "step": 50786 }, { "epoch": 3.07, "learning_rate": 2.4197637046761126e-05, "loss": 0.3459, "step": 50788 }, { "epoch": 3.07, "learning_rate": 2.41961281366469e-05, "loss": 0.5014, "step": 50790 }, { "epoch": 3.07, "learning_rate": 2.419461922653268e-05, "loss": 0.3338, "step": 50792 }, { "epoch": 3.07, "learning_rate": 2.4193110316418452e-05, "loss": 0.5231, "step": 50794 }, { "epoch": 3.07, "learning_rate": 2.419160140630423e-05, "loss": 0.2381, "step": 50796 }, { "epoch": 3.07, "learning_rate": 2.4190092496190002e-05, "loss": 0.3958, "step": 50798 }, { "epoch": 3.07, "learning_rate": 2.418858358607578e-05, "loss": 0.353, "step": 50800 }, { "epoch": 3.07, "learning_rate": 2.4187074675961555e-05, "loss": 0.3124, "step": 50802 }, { "epoch": 3.07, "learning_rate": 2.4185565765847328e-05, "loss": 0.3124, "step": 50804 }, { "epoch": 3.07, "learning_rate": 2.4184056855733105e-05, "loss": 0.3822, "step": 50806 }, { "epoch": 3.07, "learning_rate": 2.418254794561888e-05, "loss": 0.3484, "step": 50808 }, { "epoch": 3.07, "learning_rate": 2.4181039035504658e-05, "loss": 0.2941, "step": 50810 }, { "epoch": 3.07, "learning_rate": 2.417953012539043e-05, "loss": 0.3842, "step": 50812 }, { "epoch": 3.07, "learning_rate": 2.4178021215276207e-05, "loss": 0.4544, "step": 50814 }, { "epoch": 3.07, "learning_rate": 2.4176512305161984e-05, "loss": 0.4061, "step": 50816 }, { "epoch": 3.07, "learning_rate": 2.4175003395047757e-05, "loss": 0.3636, "step": 50818 }, { "epoch": 3.07, "learning_rate": 2.4173494484933533e-05, "loss": 0.3062, "step": 50820 }, { "epoch": 3.07, "learning_rate": 2.417198557481931e-05, "loss": 0.3617, "step": 50822 }, { "epoch": 3.07, "learning_rate": 2.4170476664705086e-05, "loss": 0.3552, "step": 50824 }, { "epoch": 3.07, "learning_rate": 2.416896775459086e-05, "loss": 0.3929, "step": 50826 }, { "epoch": 3.07, "learning_rate": 2.4167458844476636e-05, "loss": 0.3609, "step": 50828 }, { "epoch": 3.07, "learning_rate": 2.4165949934362413e-05, "loss": 0.4463, "step": 50830 }, { "epoch": 3.07, "learning_rate": 2.4164441024248186e-05, "loss": 0.4648, "step": 50832 }, { "epoch": 3.07, "learning_rate": 2.4162932114133962e-05, "loss": 0.3156, "step": 50834 }, { "epoch": 3.07, "learning_rate": 2.4161423204019735e-05, "loss": 0.3379, "step": 50836 }, { "epoch": 3.07, "learning_rate": 2.4159914293905515e-05, "loss": 0.3839, "step": 50838 }, { "epoch": 3.07, "learning_rate": 2.415840538379129e-05, "loss": 0.4909, "step": 50840 }, { "epoch": 3.07, "learning_rate": 2.4156896473677065e-05, "loss": 0.5832, "step": 50842 }, { "epoch": 3.07, "learning_rate": 2.4155387563562838e-05, "loss": 0.4275, "step": 50844 }, { "epoch": 3.07, "learning_rate": 2.4153878653448615e-05, "loss": 0.3551, "step": 50846 }, { "epoch": 3.07, "learning_rate": 2.415236974333439e-05, "loss": 0.4705, "step": 50848 }, { "epoch": 3.07, "learning_rate": 2.4150860833220164e-05, "loss": 0.4033, "step": 50850 }, { "epoch": 3.07, "learning_rate": 2.414935192310594e-05, "loss": 0.4437, "step": 50852 }, { "epoch": 3.07, "learning_rate": 2.4147843012991717e-05, "loss": 0.3504, "step": 50854 }, { "epoch": 3.07, "learning_rate": 2.4146334102877494e-05, "loss": 0.4399, "step": 50856 }, { "epoch": 3.07, "learning_rate": 2.4144825192763267e-05, "loss": 0.4412, "step": 50858 }, { "epoch": 3.07, "learning_rate": 2.4143316282649043e-05, "loss": 0.5294, "step": 50860 }, { "epoch": 3.07, "learning_rate": 2.414180737253482e-05, "loss": 0.347, "step": 50862 }, { "epoch": 3.07, "learning_rate": 2.4140298462420593e-05, "loss": 0.3867, "step": 50864 }, { "epoch": 3.07, "learning_rate": 2.413878955230637e-05, "loss": 0.5289, "step": 50866 }, { "epoch": 3.07, "learning_rate": 2.4137280642192146e-05, "loss": 0.432, "step": 50868 }, { "epoch": 3.07, "learning_rate": 2.4135771732077923e-05, "loss": 0.4831, "step": 50870 }, { "epoch": 3.07, "learning_rate": 2.4134262821963696e-05, "loss": 0.1672, "step": 50872 }, { "epoch": 3.07, "learning_rate": 2.4132753911849472e-05, "loss": 0.6167, "step": 50874 }, { "epoch": 3.07, "learning_rate": 2.413124500173525e-05, "loss": 0.3607, "step": 50876 }, { "epoch": 3.07, "learning_rate": 2.4129736091621025e-05, "loss": 0.31, "step": 50878 }, { "epoch": 3.07, "learning_rate": 2.41282271815068e-05, "loss": 0.4613, "step": 50880 }, { "epoch": 3.07, "learning_rate": 2.4126718271392575e-05, "loss": 0.3788, "step": 50882 }, { "epoch": 3.07, "learning_rate": 2.412520936127835e-05, "loss": 0.3931, "step": 50884 }, { "epoch": 3.07, "learning_rate": 2.4123700451164124e-05, "loss": 0.4282, "step": 50886 }, { "epoch": 3.07, "learning_rate": 2.41221915410499e-05, "loss": 0.4517, "step": 50888 }, { "epoch": 3.07, "learning_rate": 2.4120682630935678e-05, "loss": 0.3124, "step": 50890 }, { "epoch": 3.07, "learning_rate": 2.4119173720821454e-05, "loss": 0.3552, "step": 50892 }, { "epoch": 3.07, "learning_rate": 2.4117664810707227e-05, "loss": 0.3397, "step": 50894 }, { "epoch": 3.07, "learning_rate": 2.4116155900593e-05, "loss": 0.3797, "step": 50896 }, { "epoch": 3.07, "learning_rate": 2.411464699047878e-05, "loss": 0.3479, "step": 50898 }, { "epoch": 3.07, "learning_rate": 2.4113138080364553e-05, "loss": 0.3185, "step": 50900 }, { "epoch": 3.07, "learning_rate": 2.411162917025033e-05, "loss": 0.4176, "step": 50902 }, { "epoch": 3.07, "learning_rate": 2.4110120260136103e-05, "loss": 0.431, "step": 50904 }, { "epoch": 3.07, "learning_rate": 2.4108611350021883e-05, "loss": 0.4464, "step": 50906 }, { "epoch": 3.07, "learning_rate": 2.4107102439907656e-05, "loss": 0.3112, "step": 50908 }, { "epoch": 3.07, "learning_rate": 2.4105593529793432e-05, "loss": 0.3524, "step": 50910 }, { "epoch": 3.07, "learning_rate": 2.4104084619679206e-05, "loss": 0.3165, "step": 50912 }, { "epoch": 3.07, "learning_rate": 2.4102575709564982e-05, "loss": 0.2718, "step": 50914 }, { "epoch": 3.07, "learning_rate": 2.410106679945076e-05, "loss": 0.3948, "step": 50916 }, { "epoch": 3.07, "learning_rate": 2.4099557889336532e-05, "loss": 0.3866, "step": 50918 }, { "epoch": 3.07, "learning_rate": 2.4098048979222308e-05, "loss": 0.1825, "step": 50920 }, { "epoch": 3.07, "learning_rate": 2.4096540069108085e-05, "loss": 0.3575, "step": 50922 }, { "epoch": 3.07, "learning_rate": 2.409503115899386e-05, "loss": 0.3647, "step": 50924 }, { "epoch": 3.07, "learning_rate": 2.4093522248879634e-05, "loss": 0.433, "step": 50926 }, { "epoch": 3.07, "learning_rate": 2.409201333876541e-05, "loss": 0.4036, "step": 50928 }, { "epoch": 3.07, "learning_rate": 2.4090504428651187e-05, "loss": 0.4979, "step": 50930 }, { "epoch": 3.07, "learning_rate": 2.408899551853696e-05, "loss": 0.2941, "step": 50932 }, { "epoch": 3.07, "learning_rate": 2.4087486608422737e-05, "loss": 0.3672, "step": 50934 }, { "epoch": 3.07, "learning_rate": 2.4085977698308514e-05, "loss": 0.396, "step": 50936 }, { "epoch": 3.07, "learning_rate": 2.408446878819429e-05, "loss": 0.3468, "step": 50938 }, { "epoch": 3.07, "learning_rate": 2.4082959878080063e-05, "loss": 0.403, "step": 50940 }, { "epoch": 3.07, "learning_rate": 2.408145096796584e-05, "loss": 0.2441, "step": 50942 }, { "epoch": 3.07, "learning_rate": 2.4079942057851616e-05, "loss": 0.2134, "step": 50944 }, { "epoch": 3.07, "learning_rate": 2.407843314773739e-05, "loss": 0.3887, "step": 50946 }, { "epoch": 3.08, "learning_rate": 2.4076924237623166e-05, "loss": 0.3506, "step": 50948 }, { "epoch": 3.08, "learning_rate": 2.407541532750894e-05, "loss": 0.3907, "step": 50950 }, { "epoch": 3.08, "learning_rate": 2.407390641739472e-05, "loss": 0.5082, "step": 50952 }, { "epoch": 3.08, "learning_rate": 2.4072397507280492e-05, "loss": 0.2744, "step": 50954 }, { "epoch": 3.08, "learning_rate": 2.407088859716627e-05, "loss": 0.3936, "step": 50956 }, { "epoch": 3.08, "learning_rate": 2.4069379687052042e-05, "loss": 0.3671, "step": 50958 }, { "epoch": 3.08, "learning_rate": 2.406787077693782e-05, "loss": 0.4168, "step": 50960 }, { "epoch": 3.08, "learning_rate": 2.4066361866823595e-05, "loss": 0.3281, "step": 50962 }, { "epoch": 3.08, "learning_rate": 2.4064852956709368e-05, "loss": 0.3536, "step": 50964 }, { "epoch": 3.08, "learning_rate": 2.4063344046595144e-05, "loss": 0.2062, "step": 50966 }, { "epoch": 3.08, "learning_rate": 2.406183513648092e-05, "loss": 0.4715, "step": 50968 }, { "epoch": 3.08, "learning_rate": 2.4060326226366697e-05, "loss": 0.5849, "step": 50970 }, { "epoch": 3.08, "learning_rate": 2.405881731625247e-05, "loss": 0.2827, "step": 50972 }, { "epoch": 3.08, "learning_rate": 2.4057308406138247e-05, "loss": 0.3834, "step": 50974 }, { "epoch": 3.08, "learning_rate": 2.4055799496024024e-05, "loss": 0.3887, "step": 50976 }, { "epoch": 3.08, "learning_rate": 2.4054290585909797e-05, "loss": 0.3466, "step": 50978 }, { "epoch": 3.08, "learning_rate": 2.4052781675795573e-05, "loss": 0.4034, "step": 50980 }, { "epoch": 3.08, "learning_rate": 2.405127276568135e-05, "loss": 0.3218, "step": 50982 }, { "epoch": 3.08, "learning_rate": 2.4049763855567126e-05, "loss": 0.4427, "step": 50984 }, { "epoch": 3.08, "learning_rate": 2.40482549454529e-05, "loss": 0.3255, "step": 50986 }, { "epoch": 3.08, "learning_rate": 2.4046746035338676e-05, "loss": 0.275, "step": 50988 }, { "epoch": 3.08, "learning_rate": 2.4045237125224452e-05, "loss": 0.4167, "step": 50990 }, { "epoch": 3.08, "learning_rate": 2.404372821511023e-05, "loss": 0.3456, "step": 50992 }, { "epoch": 3.08, "learning_rate": 2.4042219304996002e-05, "loss": 0.345, "step": 50994 }, { "epoch": 3.08, "learning_rate": 2.404071039488178e-05, "loss": 0.2356, "step": 50996 }, { "epoch": 3.08, "learning_rate": 2.4039201484767555e-05, "loss": 0.5073, "step": 50998 }, { "epoch": 3.08, "learning_rate": 2.4037692574653328e-05, "loss": 0.4142, "step": 51000 }, { "epoch": 3.08, "learning_rate": 2.4036183664539105e-05, "loss": 0.3691, "step": 51002 }, { "epoch": 3.08, "learning_rate": 2.403467475442488e-05, "loss": 0.3669, "step": 51004 }, { "epoch": 3.08, "learning_rate": 2.4033165844310658e-05, "loss": 0.5149, "step": 51006 }, { "epoch": 3.08, "learning_rate": 2.403165693419643e-05, "loss": 0.3163, "step": 51008 }, { "epoch": 3.08, "learning_rate": 2.4030148024082204e-05, "loss": 0.3589, "step": 51010 }, { "epoch": 3.08, "learning_rate": 2.4028639113967984e-05, "loss": 0.298, "step": 51012 }, { "epoch": 3.08, "learning_rate": 2.4027130203853757e-05, "loss": 0.2682, "step": 51014 }, { "epoch": 3.08, "learning_rate": 2.4025621293739533e-05, "loss": 0.4804, "step": 51016 }, { "epoch": 3.08, "learning_rate": 2.4024112383625307e-05, "loss": 0.3791, "step": 51018 }, { "epoch": 3.08, "learning_rate": 2.4022603473511087e-05, "loss": 0.409, "step": 51020 }, { "epoch": 3.08, "learning_rate": 2.402109456339686e-05, "loss": 0.2711, "step": 51022 }, { "epoch": 3.08, "learning_rate": 2.4019585653282636e-05, "loss": 0.383, "step": 51024 }, { "epoch": 3.08, "learning_rate": 2.401807674316841e-05, "loss": 0.3465, "step": 51026 }, { "epoch": 3.08, "learning_rate": 2.4016567833054186e-05, "loss": 0.3179, "step": 51028 }, { "epoch": 3.08, "learning_rate": 2.4015058922939962e-05, "loss": 0.4383, "step": 51030 }, { "epoch": 3.08, "learning_rate": 2.4013550012825735e-05, "loss": 0.6163, "step": 51032 }, { "epoch": 3.08, "learning_rate": 2.4012041102711512e-05, "loss": 0.3132, "step": 51034 }, { "epoch": 3.08, "learning_rate": 2.401053219259729e-05, "loss": 0.3476, "step": 51036 }, { "epoch": 3.08, "learning_rate": 2.4009023282483065e-05, "loss": 0.3625, "step": 51038 }, { "epoch": 3.08, "learning_rate": 2.4007514372368838e-05, "loss": 0.3675, "step": 51040 }, { "epoch": 3.08, "learning_rate": 2.4006005462254615e-05, "loss": 0.36, "step": 51042 }, { "epoch": 3.08, "learning_rate": 2.400449655214039e-05, "loss": 0.3618, "step": 51044 }, { "epoch": 3.08, "learning_rate": 2.4002987642026164e-05, "loss": 0.3161, "step": 51046 }, { "epoch": 3.08, "learning_rate": 2.400147873191194e-05, "loss": 0.2786, "step": 51048 }, { "epoch": 3.08, "learning_rate": 2.3999969821797717e-05, "loss": 0.3189, "step": 51050 }, { "epoch": 3.08, "learning_rate": 2.3998460911683494e-05, "loss": 0.3718, "step": 51052 }, { "epoch": 3.08, "learning_rate": 2.3996952001569267e-05, "loss": 0.366, "step": 51054 }, { "epoch": 3.08, "learning_rate": 2.3995443091455043e-05, "loss": 0.3732, "step": 51056 }, { "epoch": 3.08, "learning_rate": 2.399393418134082e-05, "loss": 0.5234, "step": 51058 }, { "epoch": 3.08, "learning_rate": 2.3992425271226593e-05, "loss": 0.4339, "step": 51060 }, { "epoch": 3.08, "learning_rate": 2.399091636111237e-05, "loss": 0.4762, "step": 51062 }, { "epoch": 3.08, "learning_rate": 2.3989407450998143e-05, "loss": 0.3975, "step": 51064 }, { "epoch": 3.08, "learning_rate": 2.3987898540883923e-05, "loss": 0.458, "step": 51066 }, { "epoch": 3.08, "learning_rate": 2.3986389630769696e-05, "loss": 0.5266, "step": 51068 }, { "epoch": 3.08, "learning_rate": 2.3984880720655472e-05, "loss": 0.2978, "step": 51070 }, { "epoch": 3.08, "learning_rate": 2.3983371810541245e-05, "loss": 0.3787, "step": 51072 }, { "epoch": 3.08, "learning_rate": 2.3981862900427025e-05, "loss": 0.3546, "step": 51074 }, { "epoch": 3.08, "learning_rate": 2.39803539903128e-05, "loss": 0.3784, "step": 51076 }, { "epoch": 3.08, "learning_rate": 2.397884508019857e-05, "loss": 0.367, "step": 51078 }, { "epoch": 3.08, "learning_rate": 2.3977336170084348e-05, "loss": 0.3772, "step": 51080 }, { "epoch": 3.08, "learning_rate": 2.3975827259970125e-05, "loss": 0.392, "step": 51082 }, { "epoch": 3.08, "learning_rate": 2.39743183498559e-05, "loss": 0.4548, "step": 51084 }, { "epoch": 3.08, "learning_rate": 2.3972809439741674e-05, "loss": 0.3612, "step": 51086 }, { "epoch": 3.08, "learning_rate": 2.397130052962745e-05, "loss": 0.2669, "step": 51088 }, { "epoch": 3.08, "learning_rate": 2.3969791619513227e-05, "loss": 0.3186, "step": 51090 }, { "epoch": 3.08, "learning_rate": 2.3968282709399e-05, "loss": 0.3596, "step": 51092 }, { "epoch": 3.08, "learning_rate": 2.3966773799284777e-05, "loss": 0.3452, "step": 51094 }, { "epoch": 3.08, "learning_rate": 2.3965264889170553e-05, "loss": 0.2225, "step": 51096 }, { "epoch": 3.08, "learning_rate": 2.396375597905633e-05, "loss": 0.4134, "step": 51098 }, { "epoch": 3.08, "learning_rate": 2.3962247068942103e-05, "loss": 0.35, "step": 51100 }, { "epoch": 3.08, "learning_rate": 2.396073815882788e-05, "loss": 0.3653, "step": 51102 }, { "epoch": 3.08, "learning_rate": 2.3959229248713656e-05, "loss": 0.4011, "step": 51104 }, { "epoch": 3.08, "learning_rate": 2.3957720338599433e-05, "loss": 0.4963, "step": 51106 }, { "epoch": 3.08, "learning_rate": 2.3956211428485206e-05, "loss": 0.3024, "step": 51108 }, { "epoch": 3.08, "learning_rate": 2.3954702518370982e-05, "loss": 0.3974, "step": 51110 }, { "epoch": 3.08, "learning_rate": 2.395319360825676e-05, "loss": 0.2751, "step": 51112 }, { "epoch": 3.09, "learning_rate": 2.3951684698142532e-05, "loss": 0.2008, "step": 51114 }, { "epoch": 3.09, "learning_rate": 2.395017578802831e-05, "loss": 0.3317, "step": 51116 }, { "epoch": 3.09, "learning_rate": 2.3948666877914085e-05, "loss": 0.2798, "step": 51118 }, { "epoch": 3.09, "learning_rate": 2.394715796779986e-05, "loss": 0.3763, "step": 51120 }, { "epoch": 3.09, "learning_rate": 2.3945649057685635e-05, "loss": 0.4206, "step": 51122 }, { "epoch": 3.09, "learning_rate": 2.394414014757141e-05, "loss": 0.3989, "step": 51124 }, { "epoch": 3.09, "learning_rate": 2.3942631237457188e-05, "loss": 0.3561, "step": 51126 }, { "epoch": 3.09, "learning_rate": 2.394112232734296e-05, "loss": 0.3517, "step": 51128 }, { "epoch": 3.09, "learning_rate": 2.3939613417228737e-05, "loss": 0.3396, "step": 51130 }, { "epoch": 3.09, "learning_rate": 2.393810450711451e-05, "loss": 0.4135, "step": 51132 }, { "epoch": 3.09, "learning_rate": 2.393659559700029e-05, "loss": 0.2238, "step": 51134 }, { "epoch": 3.09, "learning_rate": 2.3935086686886063e-05, "loss": 0.4534, "step": 51136 }, { "epoch": 3.09, "learning_rate": 2.393357777677184e-05, "loss": 0.2606, "step": 51138 }, { "epoch": 3.09, "learning_rate": 2.3932068866657613e-05, "loss": 0.2432, "step": 51140 }, { "epoch": 3.09, "learning_rate": 2.393055995654339e-05, "loss": 0.4649, "step": 51142 }, { "epoch": 3.09, "learning_rate": 2.3929051046429166e-05, "loss": 0.4891, "step": 51144 }, { "epoch": 3.09, "learning_rate": 2.392754213631494e-05, "loss": 0.4771, "step": 51146 }, { "epoch": 3.09, "learning_rate": 2.3926033226200716e-05, "loss": 0.305, "step": 51148 }, { "epoch": 3.09, "learning_rate": 2.3924524316086492e-05, "loss": 0.2666, "step": 51150 }, { "epoch": 3.09, "learning_rate": 2.392301540597227e-05, "loss": 0.3134, "step": 51152 }, { "epoch": 3.09, "learning_rate": 2.3921506495858042e-05, "loss": 0.429, "step": 51154 }, { "epoch": 3.09, "learning_rate": 2.3919997585743818e-05, "loss": 0.4248, "step": 51156 }, { "epoch": 3.09, "learning_rate": 2.3918488675629595e-05, "loss": 0.3034, "step": 51158 }, { "epoch": 3.09, "learning_rate": 2.3916979765515368e-05, "loss": 0.5547, "step": 51160 }, { "epoch": 3.09, "learning_rate": 2.3915470855401144e-05, "loss": 0.4217, "step": 51162 }, { "epoch": 3.09, "learning_rate": 2.391396194528692e-05, "loss": 0.3023, "step": 51164 }, { "epoch": 3.09, "learning_rate": 2.3912453035172697e-05, "loss": 0.3181, "step": 51166 }, { "epoch": 3.09, "learning_rate": 2.391094412505847e-05, "loss": 0.4858, "step": 51168 }, { "epoch": 3.09, "learning_rate": 2.3909435214944247e-05, "loss": 0.4077, "step": 51170 }, { "epoch": 3.09, "learning_rate": 2.3907926304830024e-05, "loss": 0.2972, "step": 51172 }, { "epoch": 3.09, "learning_rate": 2.3906417394715797e-05, "loss": 0.4162, "step": 51174 }, { "epoch": 3.09, "learning_rate": 2.3904908484601573e-05, "loss": 0.4607, "step": 51176 }, { "epoch": 3.09, "learning_rate": 2.3903399574487346e-05, "loss": 0.2879, "step": 51178 }, { "epoch": 3.09, "learning_rate": 2.3901890664373126e-05, "loss": 0.2971, "step": 51180 }, { "epoch": 3.09, "learning_rate": 2.39003817542589e-05, "loss": 0.4503, "step": 51182 }, { "epoch": 3.09, "learning_rate": 2.3898872844144676e-05, "loss": 0.4656, "step": 51184 }, { "epoch": 3.09, "learning_rate": 2.389736393403045e-05, "loss": 0.3099, "step": 51186 }, { "epoch": 3.09, "learning_rate": 2.389585502391623e-05, "loss": 0.3382, "step": 51188 }, { "epoch": 3.09, "learning_rate": 2.3894346113802002e-05, "loss": 0.2332, "step": 51190 }, { "epoch": 3.09, "learning_rate": 2.3892837203687775e-05, "loss": 0.5285, "step": 51192 }, { "epoch": 3.09, "learning_rate": 2.3891328293573552e-05, "loss": 0.3614, "step": 51194 }, { "epoch": 3.09, "learning_rate": 2.3889819383459328e-05, "loss": 0.4658, "step": 51196 }, { "epoch": 3.09, "learning_rate": 2.3888310473345105e-05, "loss": 0.5019, "step": 51198 }, { "epoch": 3.09, "learning_rate": 2.3886801563230878e-05, "loss": 0.544, "step": 51200 }, { "epoch": 3.09, "learning_rate": 2.3885292653116654e-05, "loss": 0.2885, "step": 51202 }, { "epoch": 3.09, "learning_rate": 2.388378374300243e-05, "loss": 0.3533, "step": 51204 }, { "epoch": 3.09, "learning_rate": 2.3882274832888207e-05, "loss": 0.2852, "step": 51206 }, { "epoch": 3.09, "learning_rate": 2.388076592277398e-05, "loss": 0.5541, "step": 51208 }, { "epoch": 3.09, "learning_rate": 2.3879257012659757e-05, "loss": 0.3507, "step": 51210 }, { "epoch": 3.09, "learning_rate": 2.3877748102545534e-05, "loss": 0.3298, "step": 51212 }, { "epoch": 3.09, "learning_rate": 2.3876239192431307e-05, "loss": 0.413, "step": 51214 }, { "epoch": 3.09, "learning_rate": 2.3874730282317083e-05, "loss": 0.3471, "step": 51216 }, { "epoch": 3.09, "learning_rate": 2.387322137220286e-05, "loss": 0.4121, "step": 51218 }, { "epoch": 3.09, "learning_rate": 2.3871712462088636e-05, "loss": 0.5156, "step": 51220 }, { "epoch": 3.09, "learning_rate": 2.387020355197441e-05, "loss": 0.4149, "step": 51222 }, { "epoch": 3.09, "learning_rate": 2.3868694641860186e-05, "loss": 0.3478, "step": 51224 }, { "epoch": 3.09, "learning_rate": 2.3867185731745962e-05, "loss": 0.4692, "step": 51226 }, { "epoch": 3.09, "learning_rate": 2.3865676821631736e-05, "loss": 0.3288, "step": 51228 }, { "epoch": 3.09, "learning_rate": 2.3864167911517512e-05, "loss": 0.2456, "step": 51230 }, { "epoch": 3.09, "learning_rate": 2.386265900140329e-05, "loss": 0.4751, "step": 51232 }, { "epoch": 3.09, "learning_rate": 2.3861150091289065e-05, "loss": 0.4231, "step": 51234 }, { "epoch": 3.09, "learning_rate": 2.3859641181174838e-05, "loss": 0.5276, "step": 51236 }, { "epoch": 3.09, "learning_rate": 2.3858132271060615e-05, "loss": 0.315, "step": 51238 }, { "epoch": 3.09, "learning_rate": 2.385662336094639e-05, "loss": 0.3448, "step": 51240 }, { "epoch": 3.09, "learning_rate": 2.3855114450832164e-05, "loss": 0.3839, "step": 51242 }, { "epoch": 3.09, "learning_rate": 2.385360554071794e-05, "loss": 0.3727, "step": 51244 }, { "epoch": 3.09, "learning_rate": 2.3852096630603714e-05, "loss": 0.4782, "step": 51246 }, { "epoch": 3.09, "learning_rate": 2.3850587720489494e-05, "loss": 0.3866, "step": 51248 }, { "epoch": 3.09, "learning_rate": 2.3849078810375267e-05, "loss": 0.3493, "step": 51250 }, { "epoch": 3.09, "learning_rate": 2.3847569900261044e-05, "loss": 0.4272, "step": 51252 }, { "epoch": 3.09, "learning_rate": 2.3846060990146817e-05, "loss": 0.3533, "step": 51254 }, { "epoch": 3.09, "learning_rate": 2.3844552080032593e-05, "loss": 0.3889, "step": 51256 }, { "epoch": 3.09, "learning_rate": 2.384304316991837e-05, "loss": 0.535, "step": 51258 }, { "epoch": 3.09, "learning_rate": 2.3841534259804143e-05, "loss": 0.5658, "step": 51260 }, { "epoch": 3.09, "learning_rate": 2.384002534968992e-05, "loss": 0.3013, "step": 51262 }, { "epoch": 3.09, "learning_rate": 2.3838516439575696e-05, "loss": 0.4151, "step": 51264 }, { "epoch": 3.09, "learning_rate": 2.3837007529461472e-05, "loss": 0.3317, "step": 51266 }, { "epoch": 3.09, "learning_rate": 2.3835498619347245e-05, "loss": 0.3581, "step": 51268 }, { "epoch": 3.09, "learning_rate": 2.3833989709233022e-05, "loss": 0.4036, "step": 51270 }, { "epoch": 3.09, "learning_rate": 2.38324807991188e-05, "loss": 0.3787, "step": 51272 }, { "epoch": 3.09, "learning_rate": 2.383097188900457e-05, "loss": 0.396, "step": 51274 }, { "epoch": 3.09, "learning_rate": 2.3829462978890348e-05, "loss": 0.2792, "step": 51276 }, { "epoch": 3.1, "learning_rate": 2.3827954068776125e-05, "loss": 0.3903, "step": 51278 }, { "epoch": 3.1, "learning_rate": 2.38264451586619e-05, "loss": 0.2645, "step": 51280 }, { "epoch": 3.1, "learning_rate": 2.3824936248547674e-05, "loss": 0.4302, "step": 51282 }, { "epoch": 3.1, "learning_rate": 2.382342733843345e-05, "loss": 0.3398, "step": 51284 }, { "epoch": 3.1, "learning_rate": 2.3821918428319227e-05, "loss": 0.2968, "step": 51286 }, { "epoch": 3.1, "learning_rate": 2.3820409518205004e-05, "loss": 0.3871, "step": 51288 }, { "epoch": 3.1, "learning_rate": 2.3818900608090777e-05, "loss": 0.4835, "step": 51290 }, { "epoch": 3.1, "learning_rate": 2.381739169797655e-05, "loss": 0.4029, "step": 51292 }, { "epoch": 3.1, "learning_rate": 2.381588278786233e-05, "loss": 0.347, "step": 51294 }, { "epoch": 3.1, "learning_rate": 2.3814373877748103e-05, "loss": 0.5367, "step": 51296 }, { "epoch": 3.1, "learning_rate": 2.381286496763388e-05, "loss": 0.3195, "step": 51298 }, { "epoch": 3.1, "learning_rate": 2.3811356057519653e-05, "loss": 0.3675, "step": 51300 }, { "epoch": 3.1, "learning_rate": 2.3809847147405433e-05, "loss": 0.2781, "step": 51302 }, { "epoch": 3.1, "learning_rate": 2.3808338237291206e-05, "loss": 0.299, "step": 51304 }, { "epoch": 3.1, "learning_rate": 2.380682932717698e-05, "loss": 0.3271, "step": 51306 }, { "epoch": 3.1, "learning_rate": 2.3805320417062755e-05, "loss": 0.3748, "step": 51308 }, { "epoch": 3.1, "learning_rate": 2.3803811506948532e-05, "loss": 0.1948, "step": 51310 }, { "epoch": 3.1, "learning_rate": 2.380230259683431e-05, "loss": 0.3112, "step": 51312 }, { "epoch": 3.1, "learning_rate": 2.380079368672008e-05, "loss": 0.5192, "step": 51314 }, { "epoch": 3.1, "learning_rate": 2.3799284776605858e-05, "loss": 0.4518, "step": 51316 }, { "epoch": 3.1, "learning_rate": 2.3797775866491635e-05, "loss": 0.4109, "step": 51318 }, { "epoch": 3.1, "learning_rate": 2.379626695637741e-05, "loss": 0.2826, "step": 51320 }, { "epoch": 3.1, "learning_rate": 2.3794758046263184e-05, "loss": 0.3719, "step": 51322 }, { "epoch": 3.1, "learning_rate": 2.379324913614896e-05, "loss": 0.401, "step": 51324 }, { "epoch": 3.1, "learning_rate": 2.3791740226034737e-05, "loss": 0.4056, "step": 51326 }, { "epoch": 3.1, "learning_rate": 2.379023131592051e-05, "loss": 0.4444, "step": 51328 }, { "epoch": 3.1, "learning_rate": 2.3788722405806287e-05, "loss": 0.4263, "step": 51330 }, { "epoch": 3.1, "learning_rate": 2.3787213495692063e-05, "loss": 0.4096, "step": 51332 }, { "epoch": 3.1, "learning_rate": 2.378570458557784e-05, "loss": 0.3929, "step": 51334 }, { "epoch": 3.1, "learning_rate": 2.3784195675463613e-05, "loss": 0.3589, "step": 51336 }, { "epoch": 3.1, "learning_rate": 2.378268676534939e-05, "loss": 0.2818, "step": 51338 }, { "epoch": 3.1, "learning_rate": 2.3781177855235166e-05, "loss": 0.4, "step": 51340 }, { "epoch": 3.1, "learning_rate": 2.377966894512094e-05, "loss": 0.362, "step": 51342 }, { "epoch": 3.1, "learning_rate": 2.3778160035006716e-05, "loss": 0.373, "step": 51344 }, { "epoch": 3.1, "learning_rate": 2.3776651124892492e-05, "loss": 0.3059, "step": 51346 }, { "epoch": 3.1, "learning_rate": 2.377514221477827e-05, "loss": 0.2601, "step": 51348 }, { "epoch": 3.1, "learning_rate": 2.3773633304664042e-05, "loss": 0.4261, "step": 51350 }, { "epoch": 3.1, "learning_rate": 2.377212439454982e-05, "loss": 0.2728, "step": 51352 }, { "epoch": 3.1, "learning_rate": 2.3770615484435595e-05, "loss": 0.366, "step": 51354 }, { "epoch": 3.1, "learning_rate": 2.3769106574321368e-05, "loss": 0.3568, "step": 51356 }, { "epoch": 3.1, "learning_rate": 2.3767597664207145e-05, "loss": 0.3725, "step": 51358 }, { "epoch": 3.1, "learning_rate": 2.3766088754092918e-05, "loss": 0.5136, "step": 51360 }, { "epoch": 3.1, "learning_rate": 2.3764579843978698e-05, "loss": 0.4053, "step": 51362 }, { "epoch": 3.1, "learning_rate": 2.376307093386447e-05, "loss": 0.4162, "step": 51364 }, { "epoch": 3.1, "learning_rate": 2.3761562023750247e-05, "loss": 0.2919, "step": 51366 }, { "epoch": 3.1, "learning_rate": 2.376005311363602e-05, "loss": 0.4138, "step": 51368 }, { "epoch": 3.1, "learning_rate": 2.37585442035218e-05, "loss": 0.33, "step": 51370 }, { "epoch": 3.1, "learning_rate": 2.3757035293407573e-05, "loss": 0.4484, "step": 51372 }, { "epoch": 3.1, "learning_rate": 2.3755526383293346e-05, "loss": 0.5639, "step": 51374 }, { "epoch": 3.1, "learning_rate": 2.3754017473179123e-05, "loss": 0.1861, "step": 51376 }, { "epoch": 3.1, "learning_rate": 2.37525085630649e-05, "loss": 0.5078, "step": 51378 }, { "epoch": 3.1, "learning_rate": 2.3750999652950676e-05, "loss": 0.3457, "step": 51380 }, { "epoch": 3.1, "learning_rate": 2.374949074283645e-05, "loss": 0.3356, "step": 51382 }, { "epoch": 3.1, "learning_rate": 2.3747981832722226e-05, "loss": 0.4368, "step": 51384 }, { "epoch": 3.1, "learning_rate": 2.3746472922608002e-05, "loss": 0.257, "step": 51386 }, { "epoch": 3.1, "learning_rate": 2.3744964012493775e-05, "loss": 0.2176, "step": 51388 }, { "epoch": 3.1, "learning_rate": 2.3743455102379552e-05, "loss": 0.3898, "step": 51390 }, { "epoch": 3.1, "learning_rate": 2.374194619226533e-05, "loss": 0.3481, "step": 51392 }, { "epoch": 3.1, "learning_rate": 2.3740437282151105e-05, "loss": 0.4445, "step": 51394 }, { "epoch": 3.1, "learning_rate": 2.3738928372036878e-05, "loss": 0.3748, "step": 51396 }, { "epoch": 3.1, "learning_rate": 2.3737419461922654e-05, "loss": 0.2895, "step": 51398 }, { "epoch": 3.1, "learning_rate": 2.373591055180843e-05, "loss": 0.2034, "step": 51400 }, { "epoch": 3.1, "learning_rate": 2.3734401641694208e-05, "loss": 0.3842, "step": 51402 }, { "epoch": 3.1, "learning_rate": 2.373289273157998e-05, "loss": 0.3594, "step": 51404 }, { "epoch": 3.1, "learning_rate": 2.3731383821465754e-05, "loss": 0.4198, "step": 51406 }, { "epoch": 3.1, "learning_rate": 2.3729874911351534e-05, "loss": 0.3646, "step": 51408 }, { "epoch": 3.1, "learning_rate": 2.3728366001237307e-05, "loss": 0.3385, "step": 51410 }, { "epoch": 3.1, "learning_rate": 2.3726857091123083e-05, "loss": 0.3208, "step": 51412 }, { "epoch": 3.1, "learning_rate": 2.3725348181008856e-05, "loss": 0.3097, "step": 51414 }, { "epoch": 3.1, "learning_rate": 2.3723839270894636e-05, "loss": 0.2398, "step": 51416 }, { "epoch": 3.1, "learning_rate": 2.372233036078041e-05, "loss": 0.5221, "step": 51418 }, { "epoch": 3.1, "learning_rate": 2.3720821450666186e-05, "loss": 0.3057, "step": 51420 }, { "epoch": 3.1, "learning_rate": 2.371931254055196e-05, "loss": 0.3427, "step": 51422 }, { "epoch": 3.1, "learning_rate": 2.3717803630437736e-05, "loss": 0.2792, "step": 51424 }, { "epoch": 3.1, "learning_rate": 2.3716294720323512e-05, "loss": 0.4857, "step": 51426 }, { "epoch": 3.1, "learning_rate": 2.3714785810209285e-05, "loss": 0.2964, "step": 51428 }, { "epoch": 3.1, "learning_rate": 2.3713276900095062e-05, "loss": 0.5429, "step": 51430 }, { "epoch": 3.1, "learning_rate": 2.3711767989980838e-05, "loss": 0.4821, "step": 51432 }, { "epoch": 3.1, "learning_rate": 2.3710259079866615e-05, "loss": 0.3983, "step": 51434 }, { "epoch": 3.1, "learning_rate": 2.3708750169752388e-05, "loss": 0.4783, "step": 51436 }, { "epoch": 3.1, "learning_rate": 2.3707241259638164e-05, "loss": 0.3478, "step": 51438 }, { "epoch": 3.1, "learning_rate": 2.370573234952394e-05, "loss": 0.368, "step": 51440 }, { "epoch": 3.1, "learning_rate": 2.3704223439409714e-05, "loss": 0.4563, "step": 51442 }, { "epoch": 3.11, "learning_rate": 2.370271452929549e-05, "loss": 0.5454, "step": 51444 }, { "epoch": 3.11, "learning_rate": 2.3701205619181267e-05, "loss": 0.48, "step": 51446 }, { "epoch": 3.11, "learning_rate": 2.3699696709067044e-05, "loss": 0.2422, "step": 51448 }, { "epoch": 3.11, "learning_rate": 2.3698187798952817e-05, "loss": 0.3402, "step": 51450 }, { "epoch": 3.11, "learning_rate": 2.3696678888838593e-05, "loss": 0.3793, "step": 51452 }, { "epoch": 3.11, "learning_rate": 2.369516997872437e-05, "loss": 0.5348, "step": 51454 }, { "epoch": 3.11, "learning_rate": 2.3693661068610143e-05, "loss": 0.3542, "step": 51456 }, { "epoch": 3.11, "learning_rate": 2.369215215849592e-05, "loss": 0.3374, "step": 51458 }, { "epoch": 3.11, "learning_rate": 2.3690643248381696e-05, "loss": 0.4912, "step": 51460 }, { "epoch": 3.11, "learning_rate": 2.3689134338267472e-05, "loss": 0.2489, "step": 51462 }, { "epoch": 3.11, "learning_rate": 2.3687625428153246e-05, "loss": 0.2398, "step": 51464 }, { "epoch": 3.11, "learning_rate": 2.3686116518039022e-05, "loss": 0.321, "step": 51466 }, { "epoch": 3.11, "learning_rate": 2.36846076079248e-05, "loss": 0.4475, "step": 51468 }, { "epoch": 3.11, "learning_rate": 2.3683098697810572e-05, "loss": 0.5083, "step": 51470 }, { "epoch": 3.11, "learning_rate": 2.3681589787696348e-05, "loss": 0.412, "step": 51472 }, { "epoch": 3.11, "learning_rate": 2.368008087758212e-05, "loss": 0.4275, "step": 51474 }, { "epoch": 3.11, "learning_rate": 2.36785719674679e-05, "loss": 0.5005, "step": 51476 }, { "epoch": 3.11, "learning_rate": 2.3677063057353674e-05, "loss": 0.4411, "step": 51478 }, { "epoch": 3.11, "learning_rate": 2.367555414723945e-05, "loss": 0.2887, "step": 51480 }, { "epoch": 3.11, "learning_rate": 2.3674045237125224e-05, "loss": 0.3279, "step": 51482 }, { "epoch": 3.11, "learning_rate": 2.3672536327011004e-05, "loss": 0.466, "step": 51484 }, { "epoch": 3.11, "learning_rate": 2.3671027416896777e-05, "loss": 0.4014, "step": 51486 }, { "epoch": 3.11, "learning_rate": 2.366951850678255e-05, "loss": 0.4412, "step": 51488 }, { "epoch": 3.11, "learning_rate": 2.3668009596668327e-05, "loss": 0.397, "step": 51490 }, { "epoch": 3.11, "learning_rate": 2.3666500686554103e-05, "loss": 0.3642, "step": 51492 }, { "epoch": 3.11, "learning_rate": 2.366499177643988e-05, "loss": 0.319, "step": 51494 }, { "epoch": 3.11, "learning_rate": 2.3663482866325653e-05, "loss": 0.3161, "step": 51496 }, { "epoch": 3.11, "learning_rate": 2.366197395621143e-05, "loss": 0.4285, "step": 51498 }, { "epoch": 3.11, "learning_rate": 2.3660465046097206e-05, "loss": 0.345, "step": 51500 }, { "epoch": 3.11, "learning_rate": 2.3658956135982982e-05, "loss": 0.3663, "step": 51502 }, { "epoch": 3.11, "learning_rate": 2.3657447225868755e-05, "loss": 0.3231, "step": 51504 }, { "epoch": 3.11, "learning_rate": 2.3655938315754532e-05, "loss": 0.3742, "step": 51506 }, { "epoch": 3.11, "learning_rate": 2.365442940564031e-05, "loss": 0.3087, "step": 51508 }, { "epoch": 3.11, "learning_rate": 2.365292049552608e-05, "loss": 0.4544, "step": 51510 }, { "epoch": 3.11, "learning_rate": 2.3651411585411858e-05, "loss": 0.3615, "step": 51512 }, { "epoch": 3.11, "learning_rate": 2.3649902675297635e-05, "loss": 0.4588, "step": 51514 }, { "epoch": 3.11, "learning_rate": 2.364839376518341e-05, "loss": 0.3784, "step": 51516 }, { "epoch": 3.11, "learning_rate": 2.3646884855069184e-05, "loss": 0.3462, "step": 51518 }, { "epoch": 3.11, "learning_rate": 2.3645375944954957e-05, "loss": 0.3134, "step": 51520 }, { "epoch": 3.11, "learning_rate": 2.3643867034840737e-05, "loss": 0.2372, "step": 51522 }, { "epoch": 3.11, "learning_rate": 2.364235812472651e-05, "loss": 0.56, "step": 51524 }, { "epoch": 3.11, "learning_rate": 2.3640849214612287e-05, "loss": 0.5468, "step": 51526 }, { "epoch": 3.11, "learning_rate": 2.363934030449806e-05, "loss": 0.3631, "step": 51528 }, { "epoch": 3.11, "learning_rate": 2.363783139438384e-05, "loss": 0.4662, "step": 51530 }, { "epoch": 3.11, "learning_rate": 2.3636322484269613e-05, "loss": 0.4164, "step": 51532 }, { "epoch": 3.11, "learning_rate": 2.363481357415539e-05, "loss": 0.4289, "step": 51534 }, { "epoch": 3.11, "learning_rate": 2.3633304664041163e-05, "loss": 0.2667, "step": 51536 }, { "epoch": 3.11, "learning_rate": 2.363179575392694e-05, "loss": 0.4324, "step": 51538 }, { "epoch": 3.11, "learning_rate": 2.3630286843812716e-05, "loss": 0.3062, "step": 51540 }, { "epoch": 3.11, "learning_rate": 2.362877793369849e-05, "loss": 0.4377, "step": 51542 }, { "epoch": 3.11, "learning_rate": 2.3627269023584265e-05, "loss": 0.486, "step": 51544 }, { "epoch": 3.11, "learning_rate": 2.3625760113470042e-05, "loss": 0.4713, "step": 51546 }, { "epoch": 3.11, "learning_rate": 2.362425120335582e-05, "loss": 0.4704, "step": 51548 }, { "epoch": 3.11, "learning_rate": 2.362274229324159e-05, "loss": 0.4617, "step": 51550 }, { "epoch": 3.11, "learning_rate": 2.3621233383127368e-05, "loss": 0.36, "step": 51552 }, { "epoch": 3.11, "learning_rate": 2.3619724473013145e-05, "loss": 0.3789, "step": 51554 }, { "epoch": 3.11, "learning_rate": 2.3618215562898918e-05, "loss": 0.6045, "step": 51556 }, { "epoch": 3.11, "learning_rate": 2.3616706652784694e-05, "loss": 0.405, "step": 51558 }, { "epoch": 3.11, "learning_rate": 2.361519774267047e-05, "loss": 0.4443, "step": 51560 }, { "epoch": 3.11, "learning_rate": 2.3613688832556247e-05, "loss": 0.5081, "step": 51562 }, { "epoch": 3.11, "learning_rate": 2.361217992244202e-05, "loss": 0.3239, "step": 51564 }, { "epoch": 3.11, "learning_rate": 2.3610671012327797e-05, "loss": 0.4992, "step": 51566 }, { "epoch": 3.11, "learning_rate": 2.3609162102213573e-05, "loss": 0.2984, "step": 51568 }, { "epoch": 3.11, "learning_rate": 2.3607653192099347e-05, "loss": 0.5312, "step": 51570 }, { "epoch": 3.11, "learning_rate": 2.3606144281985123e-05, "loss": 0.3033, "step": 51572 }, { "epoch": 3.11, "learning_rate": 2.36046353718709e-05, "loss": 0.2223, "step": 51574 }, { "epoch": 3.11, "learning_rate": 2.3603126461756676e-05, "loss": 0.3754, "step": 51576 }, { "epoch": 3.11, "learning_rate": 2.360161755164245e-05, "loss": 0.4668, "step": 51578 }, { "epoch": 3.11, "learning_rate": 2.3600108641528226e-05, "loss": 0.5237, "step": 51580 }, { "epoch": 3.11, "learning_rate": 2.3598599731414002e-05, "loss": 0.4506, "step": 51582 }, { "epoch": 3.11, "learning_rate": 2.359709082129978e-05, "loss": 0.5095, "step": 51584 }, { "epoch": 3.11, "learning_rate": 2.3595581911185552e-05, "loss": 0.3084, "step": 51586 }, { "epoch": 3.11, "learning_rate": 2.3594073001071325e-05, "loss": 0.4414, "step": 51588 }, { "epoch": 3.11, "learning_rate": 2.3592564090957105e-05, "loss": 0.2893, "step": 51590 }, { "epoch": 3.11, "learning_rate": 2.3591055180842878e-05, "loss": 0.3842, "step": 51592 }, { "epoch": 3.11, "learning_rate": 2.3589546270728655e-05, "loss": 0.6294, "step": 51594 }, { "epoch": 3.11, "learning_rate": 2.3588037360614428e-05, "loss": 0.4152, "step": 51596 }, { "epoch": 3.11, "learning_rate": 2.3586528450500208e-05, "loss": 0.5397, "step": 51598 }, { "epoch": 3.11, "learning_rate": 2.358501954038598e-05, "loss": 0.3654, "step": 51600 }, { "epoch": 3.11, "learning_rate": 2.3583510630271754e-05, "loss": 0.3202, "step": 51602 }, { "epoch": 3.11, "learning_rate": 2.358200172015753e-05, "loss": 0.4059, "step": 51604 }, { "epoch": 3.11, "learning_rate": 2.3580492810043307e-05, "loss": 0.3829, "step": 51606 }, { "epoch": 3.11, "learning_rate": 2.3578983899929083e-05, "loss": 0.4018, "step": 51608 }, { "epoch": 3.12, "learning_rate": 2.3577474989814857e-05, "loss": 0.3744, "step": 51610 }, { "epoch": 3.12, "learning_rate": 2.3575966079700633e-05, "loss": 0.558, "step": 51612 }, { "epoch": 3.12, "learning_rate": 2.357445716958641e-05, "loss": 0.6904, "step": 51614 }, { "epoch": 3.12, "learning_rate": 2.3572948259472186e-05, "loss": 0.3609, "step": 51616 }, { "epoch": 3.12, "learning_rate": 2.357143934935796e-05, "loss": 0.3121, "step": 51618 }, { "epoch": 3.12, "learning_rate": 2.3569930439243736e-05, "loss": 0.4085, "step": 51620 }, { "epoch": 3.12, "learning_rate": 2.3568421529129512e-05, "loss": 0.3478, "step": 51622 }, { "epoch": 3.12, "learning_rate": 2.3566912619015285e-05, "loss": 0.3798, "step": 51624 }, { "epoch": 3.12, "learning_rate": 2.3565403708901062e-05, "loss": 0.4016, "step": 51626 }, { "epoch": 3.12, "learning_rate": 2.356464925384395e-05, "loss": 0.5488, "step": 51628 }, { "epoch": 3.12, "learning_rate": 2.3563140343729725e-05, "loss": 0.2055, "step": 51630 }, { "epoch": 3.12, "learning_rate": 2.35616314336155e-05, "loss": 0.662, "step": 51632 }, { "epoch": 3.12, "learning_rate": 2.3560122523501278e-05, "loss": 0.3929, "step": 51634 }, { "epoch": 3.12, "learning_rate": 2.355861361338705e-05, "loss": 0.291, "step": 51636 }, { "epoch": 3.12, "learning_rate": 2.3557104703272828e-05, "loss": 0.2945, "step": 51638 }, { "epoch": 3.12, "learning_rate": 2.3555595793158604e-05, "loss": 0.4123, "step": 51640 }, { "epoch": 3.12, "learning_rate": 2.355408688304438e-05, "loss": 0.3165, "step": 51642 }, { "epoch": 3.12, "learning_rate": 2.3552577972930154e-05, "loss": 0.5523, "step": 51644 }, { "epoch": 3.12, "learning_rate": 2.3551069062815927e-05, "loss": 0.3743, "step": 51646 }, { "epoch": 3.12, "learning_rate": 2.3549560152701707e-05, "loss": 0.2028, "step": 51648 }, { "epoch": 3.12, "learning_rate": 2.354805124258748e-05, "loss": 0.408, "step": 51650 }, { "epoch": 3.12, "learning_rate": 2.3546542332473256e-05, "loss": 0.3616, "step": 51652 }, { "epoch": 3.12, "learning_rate": 2.354503342235903e-05, "loss": 0.3245, "step": 51654 }, { "epoch": 3.12, "learning_rate": 2.354352451224481e-05, "loss": 0.3514, "step": 51656 }, { "epoch": 3.12, "learning_rate": 2.3542015602130583e-05, "loss": 0.3872, "step": 51658 }, { "epoch": 3.12, "learning_rate": 2.3540506692016356e-05, "loss": 0.2035, "step": 51660 }, { "epoch": 3.12, "learning_rate": 2.3538997781902132e-05, "loss": 0.5161, "step": 51662 }, { "epoch": 3.12, "learning_rate": 2.353748887178791e-05, "loss": 0.2883, "step": 51664 }, { "epoch": 3.12, "learning_rate": 2.3535979961673685e-05, "loss": 0.3016, "step": 51666 }, { "epoch": 3.12, "learning_rate": 2.353447105155946e-05, "loss": 0.4433, "step": 51668 }, { "epoch": 3.12, "learning_rate": 2.3532962141445235e-05, "loss": 0.3521, "step": 51670 }, { "epoch": 3.12, "learning_rate": 2.353145323133101e-05, "loss": 0.3292, "step": 51672 }, { "epoch": 3.12, "learning_rate": 2.3529944321216788e-05, "loss": 0.4671, "step": 51674 }, { "epoch": 3.12, "learning_rate": 2.352843541110256e-05, "loss": 0.2661, "step": 51676 }, { "epoch": 3.12, "learning_rate": 2.3526926500988338e-05, "loss": 0.4211, "step": 51678 }, { "epoch": 3.12, "learning_rate": 2.3525417590874114e-05, "loss": 0.284, "step": 51680 }, { "epoch": 3.12, "learning_rate": 2.3523908680759887e-05, "loss": 0.3516, "step": 51682 }, { "epoch": 3.12, "learning_rate": 2.3522399770645664e-05, "loss": 0.3836, "step": 51684 }, { "epoch": 3.12, "learning_rate": 2.352089086053144e-05, "loss": 0.3568, "step": 51686 }, { "epoch": 3.12, "learning_rate": 2.3519381950417217e-05, "loss": 0.4982, "step": 51688 }, { "epoch": 3.12, "learning_rate": 2.351787304030299e-05, "loss": 0.3513, "step": 51690 }, { "epoch": 3.12, "learning_rate": 2.3516364130188766e-05, "loss": 0.2775, "step": 51692 }, { "epoch": 3.12, "learning_rate": 2.3514855220074543e-05, "loss": 0.2413, "step": 51694 }, { "epoch": 3.12, "learning_rate": 2.3513346309960316e-05, "loss": 0.3367, "step": 51696 }, { "epoch": 3.12, "learning_rate": 2.3511837399846092e-05, "loss": 0.4746, "step": 51698 }, { "epoch": 3.12, "learning_rate": 2.3510328489731866e-05, "loss": 0.5383, "step": 51700 }, { "epoch": 3.12, "learning_rate": 2.3508819579617646e-05, "loss": 0.4414, "step": 51702 }, { "epoch": 3.12, "learning_rate": 2.350731066950342e-05, "loss": 0.3757, "step": 51704 }, { "epoch": 3.12, "learning_rate": 2.3505801759389195e-05, "loss": 0.4355, "step": 51706 }, { "epoch": 3.12, "learning_rate": 2.3504292849274968e-05, "loss": 0.3129, "step": 51708 }, { "epoch": 3.12, "learning_rate": 2.3502783939160745e-05, "loss": 0.2731, "step": 51710 }, { "epoch": 3.12, "learning_rate": 2.350127502904652e-05, "loss": 0.3092, "step": 51712 }, { "epoch": 3.12, "learning_rate": 2.3499766118932294e-05, "loss": 0.5332, "step": 51714 }, { "epoch": 3.12, "learning_rate": 2.349825720881807e-05, "loss": 0.3911, "step": 51716 }, { "epoch": 3.12, "learning_rate": 2.3496748298703847e-05, "loss": 0.3763, "step": 51718 }, { "epoch": 3.12, "learning_rate": 2.3495239388589624e-05, "loss": 0.2881, "step": 51720 }, { "epoch": 3.12, "learning_rate": 2.3493730478475397e-05, "loss": 0.3706, "step": 51722 }, { "epoch": 3.12, "learning_rate": 2.3492221568361174e-05, "loss": 0.5002, "step": 51724 }, { "epoch": 3.12, "learning_rate": 2.349071265824695e-05, "loss": 0.411, "step": 51726 }, { "epoch": 3.12, "learning_rate": 2.3489203748132723e-05, "loss": 0.3947, "step": 51728 }, { "epoch": 3.12, "learning_rate": 2.34876948380185e-05, "loss": 0.2931, "step": 51730 }, { "epoch": 3.12, "learning_rate": 2.3486185927904276e-05, "loss": 0.4137, "step": 51732 }, { "epoch": 3.12, "learning_rate": 2.3484677017790053e-05, "loss": 0.4976, "step": 51734 }, { "epoch": 3.12, "learning_rate": 2.3483168107675826e-05, "loss": 0.2867, "step": 51736 }, { "epoch": 3.12, "learning_rate": 2.3481659197561602e-05, "loss": 0.2813, "step": 51738 }, { "epoch": 3.12, "learning_rate": 2.348015028744738e-05, "loss": 0.4449, "step": 51740 }, { "epoch": 3.12, "learning_rate": 2.3478641377333152e-05, "loss": 0.3963, "step": 51742 }, { "epoch": 3.12, "learning_rate": 2.347713246721893e-05, "loss": 0.4315, "step": 51744 }, { "epoch": 3.12, "learning_rate": 2.3475623557104702e-05, "loss": 0.4115, "step": 51746 }, { "epoch": 3.12, "learning_rate": 2.347411464699048e-05, "loss": 0.3228, "step": 51748 }, { "epoch": 3.12, "learning_rate": 2.3472605736876255e-05, "loss": 0.3616, "step": 51750 }, { "epoch": 3.12, "learning_rate": 2.347109682676203e-05, "loss": 0.3761, "step": 51752 }, { "epoch": 3.12, "learning_rate": 2.3469587916647808e-05, "loss": 0.4152, "step": 51754 }, { "epoch": 3.12, "learning_rate": 2.3468079006533584e-05, "loss": 0.3127, "step": 51756 }, { "epoch": 3.12, "learning_rate": 2.3466570096419357e-05, "loss": 0.3328, "step": 51758 }, { "epoch": 3.12, "learning_rate": 2.346506118630513e-05, "loss": 0.3557, "step": 51760 }, { "epoch": 3.12, "learning_rate": 2.346355227619091e-05, "loss": 0.2762, "step": 51762 }, { "epoch": 3.12, "learning_rate": 2.3462043366076684e-05, "loss": 0.3926, "step": 51764 }, { "epoch": 3.12, "learning_rate": 2.346053445596246e-05, "loss": 0.4386, "step": 51766 }, { "epoch": 3.12, "learning_rate": 2.3459025545848233e-05, "loss": 0.2957, "step": 51768 }, { "epoch": 3.12, "learning_rate": 2.3457516635734013e-05, "loss": 0.3467, "step": 51770 }, { "epoch": 3.12, "learning_rate": 2.3456007725619786e-05, "loss": 0.4272, "step": 51772 }, { "epoch": 3.12, "learning_rate": 2.3454498815505563e-05, "loss": 0.406, "step": 51774 }, { "epoch": 3.13, "learning_rate": 2.3452989905391336e-05, "loss": 0.3909, "step": 51776 }, { "epoch": 3.13, "learning_rate": 2.3451480995277112e-05, "loss": 0.4543, "step": 51778 }, { "epoch": 3.13, "learning_rate": 2.344997208516289e-05, "loss": 0.4716, "step": 51780 }, { "epoch": 3.13, "learning_rate": 2.3448463175048662e-05, "loss": 0.3554, "step": 51782 }, { "epoch": 3.13, "learning_rate": 2.344695426493444e-05, "loss": 0.336, "step": 51784 }, { "epoch": 3.13, "learning_rate": 2.3445445354820215e-05, "loss": 0.4809, "step": 51786 }, { "epoch": 3.13, "learning_rate": 2.344393644470599e-05, "loss": 0.2486, "step": 51788 }, { "epoch": 3.13, "learning_rate": 2.3442427534591765e-05, "loss": 0.4236, "step": 51790 }, { "epoch": 3.13, "learning_rate": 2.344091862447754e-05, "loss": 0.2991, "step": 51792 }, { "epoch": 3.13, "learning_rate": 2.3439409714363318e-05, "loss": 0.4746, "step": 51794 }, { "epoch": 3.13, "learning_rate": 2.343790080424909e-05, "loss": 0.3119, "step": 51796 }, { "epoch": 3.13, "learning_rate": 2.3436391894134867e-05, "loss": 0.405, "step": 51798 }, { "epoch": 3.13, "learning_rate": 2.3434882984020644e-05, "loss": 0.4818, "step": 51800 }, { "epoch": 3.13, "learning_rate": 2.343337407390642e-05, "loss": 0.2371, "step": 51802 }, { "epoch": 3.13, "learning_rate": 2.3431865163792194e-05, "loss": 0.2891, "step": 51804 }, { "epoch": 3.13, "learning_rate": 2.343035625367797e-05, "loss": 0.4081, "step": 51806 }, { "epoch": 3.13, "learning_rate": 2.3428847343563747e-05, "loss": 0.3435, "step": 51808 }, { "epoch": 3.13, "learning_rate": 2.342733843344952e-05, "loss": 0.3517, "step": 51810 }, { "epoch": 3.13, "learning_rate": 2.3425829523335296e-05, "loss": 0.4167, "step": 51812 }, { "epoch": 3.13, "learning_rate": 2.342432061322107e-05, "loss": 0.3322, "step": 51814 }, { "epoch": 3.13, "learning_rate": 2.342281170310685e-05, "loss": 0.2574, "step": 51816 }, { "epoch": 3.13, "learning_rate": 2.3421302792992622e-05, "loss": 0.3432, "step": 51818 }, { "epoch": 3.13, "learning_rate": 2.34197938828784e-05, "loss": 0.3039, "step": 51820 }, { "epoch": 3.13, "learning_rate": 2.3418284972764172e-05, "loss": 0.3586, "step": 51822 }, { "epoch": 3.13, "learning_rate": 2.341677606264995e-05, "loss": 0.3738, "step": 51824 }, { "epoch": 3.13, "learning_rate": 2.3415267152535725e-05, "loss": 0.3505, "step": 51826 }, { "epoch": 3.13, "learning_rate": 2.3413758242421498e-05, "loss": 0.3268, "step": 51828 }, { "epoch": 3.13, "learning_rate": 2.3412249332307275e-05, "loss": 0.3715, "step": 51830 }, { "epoch": 3.13, "learning_rate": 2.341074042219305e-05, "loss": 0.468, "step": 51832 }, { "epoch": 3.13, "learning_rate": 2.3409231512078828e-05, "loss": 0.3795, "step": 51834 }, { "epoch": 3.13, "learning_rate": 2.34077226019646e-05, "loss": 0.3332, "step": 51836 }, { "epoch": 3.13, "learning_rate": 2.3406213691850377e-05, "loss": 0.4559, "step": 51838 }, { "epoch": 3.13, "learning_rate": 2.3404704781736154e-05, "loss": 0.3664, "step": 51840 }, { "epoch": 3.13, "learning_rate": 2.3403195871621927e-05, "loss": 0.301, "step": 51842 }, { "epoch": 3.13, "learning_rate": 2.3401686961507703e-05, "loss": 0.4111, "step": 51844 }, { "epoch": 3.13, "learning_rate": 2.340017805139348e-05, "loss": 0.2858, "step": 51846 }, { "epoch": 3.13, "learning_rate": 2.3398669141279256e-05, "loss": 0.3146, "step": 51848 }, { "epoch": 3.13, "learning_rate": 2.339716023116503e-05, "loss": 0.5769, "step": 51850 }, { "epoch": 3.13, "learning_rate": 2.3395651321050806e-05, "loss": 0.3929, "step": 51852 }, { "epoch": 3.13, "learning_rate": 2.3394142410936583e-05, "loss": 0.4148, "step": 51854 }, { "epoch": 3.13, "learning_rate": 2.339263350082236e-05, "loss": 0.3703, "step": 51856 }, { "epoch": 3.13, "learning_rate": 2.3391124590708132e-05, "loss": 0.3164, "step": 51858 }, { "epoch": 3.13, "learning_rate": 2.3389615680593905e-05, "loss": 0.374, "step": 51860 }, { "epoch": 3.13, "learning_rate": 2.3388106770479685e-05, "loss": 0.4578, "step": 51862 }, { "epoch": 3.13, "learning_rate": 2.338659786036546e-05, "loss": 0.449, "step": 51864 }, { "epoch": 3.13, "learning_rate": 2.3385088950251235e-05, "loss": 0.4042, "step": 51866 }, { "epoch": 3.13, "learning_rate": 2.338358004013701e-05, "loss": 0.4191, "step": 51868 }, { "epoch": 3.13, "learning_rate": 2.3382071130022788e-05, "loss": 0.441, "step": 51870 }, { "epoch": 3.13, "learning_rate": 2.338056221990856e-05, "loss": 0.4755, "step": 51872 }, { "epoch": 3.13, "learning_rate": 2.3379053309794334e-05, "loss": 0.432, "step": 51874 }, { "epoch": 3.13, "learning_rate": 2.3377544399680114e-05, "loss": 0.448, "step": 51876 }, { "epoch": 3.13, "learning_rate": 2.3376035489565887e-05, "loss": 0.3575, "step": 51878 }, { "epoch": 3.13, "learning_rate": 2.3374526579451664e-05, "loss": 0.3321, "step": 51880 }, { "epoch": 3.13, "learning_rate": 2.3373017669337437e-05, "loss": 0.3916, "step": 51882 }, { "epoch": 3.13, "learning_rate": 2.3371508759223217e-05, "loss": 0.4081, "step": 51884 }, { "epoch": 3.13, "learning_rate": 2.336999984910899e-05, "loss": 0.4147, "step": 51886 }, { "epoch": 3.13, "learning_rate": 2.3368490938994766e-05, "loss": 0.3113, "step": 51888 }, { "epoch": 3.13, "learning_rate": 2.336698202888054e-05, "loss": 0.3133, "step": 51890 }, { "epoch": 3.13, "learning_rate": 2.3365473118766316e-05, "loss": 0.4584, "step": 51892 }, { "epoch": 3.13, "learning_rate": 2.3363964208652093e-05, "loss": 0.271, "step": 51894 }, { "epoch": 3.13, "learning_rate": 2.3362455298537866e-05, "loss": 0.3669, "step": 51896 }, { "epoch": 3.13, "learning_rate": 2.3360946388423642e-05, "loss": 0.2835, "step": 51898 }, { "epoch": 3.13, "learning_rate": 2.335943747830942e-05, "loss": 0.6565, "step": 51900 }, { "epoch": 3.13, "learning_rate": 2.3357928568195195e-05, "loss": 0.4596, "step": 51902 }, { "epoch": 3.13, "learning_rate": 2.335641965808097e-05, "loss": 0.461, "step": 51904 }, { "epoch": 3.13, "learning_rate": 2.3354910747966745e-05, "loss": 0.4176, "step": 51906 }, { "epoch": 3.13, "learning_rate": 2.335340183785252e-05, "loss": 0.327, "step": 51908 }, { "epoch": 3.13, "learning_rate": 2.3351892927738295e-05, "loss": 0.3125, "step": 51910 }, { "epoch": 3.13, "learning_rate": 2.335038401762407e-05, "loss": 0.4111, "step": 51912 }, { "epoch": 3.13, "learning_rate": 2.3348875107509848e-05, "loss": 0.2581, "step": 51914 }, { "epoch": 3.13, "learning_rate": 2.3347366197395624e-05, "loss": 0.3909, "step": 51916 }, { "epoch": 3.13, "learning_rate": 2.3345857287281397e-05, "loss": 0.4879, "step": 51918 }, { "epoch": 3.13, "learning_rate": 2.3344348377167174e-05, "loss": 0.4025, "step": 51920 }, { "epoch": 3.13, "learning_rate": 2.334283946705295e-05, "loss": 0.5456, "step": 51922 }, { "epoch": 3.13, "learning_rate": 2.3341330556938723e-05, "loss": 0.4593, "step": 51924 }, { "epoch": 3.13, "learning_rate": 2.33398216468245e-05, "loss": 0.2354, "step": 51926 }, { "epoch": 3.13, "learning_rate": 2.3338312736710273e-05, "loss": 0.4829, "step": 51928 }, { "epoch": 3.13, "learning_rate": 2.3336803826596053e-05, "loss": 0.2919, "step": 51930 }, { "epoch": 3.13, "learning_rate": 2.3335294916481826e-05, "loss": 0.3738, "step": 51932 }, { "epoch": 3.13, "learning_rate": 2.3333786006367603e-05, "loss": 0.4455, "step": 51934 }, { "epoch": 3.13, "learning_rate": 2.3332277096253376e-05, "loss": 0.4124, "step": 51936 }, { "epoch": 3.13, "learning_rate": 2.3330768186139156e-05, "loss": 0.3218, "step": 51938 }, { "epoch": 3.13, "learning_rate": 2.332925927602493e-05, "loss": 0.3843, "step": 51940 }, { "epoch": 3.14, "learning_rate": 2.3327750365910702e-05, "loss": 0.3463, "step": 51942 }, { "epoch": 3.14, "learning_rate": 2.3326241455796478e-05, "loss": 0.3652, "step": 51944 }, { "epoch": 3.14, "learning_rate": 2.3324732545682255e-05, "loss": 0.5379, "step": 51946 }, { "epoch": 3.14, "learning_rate": 2.332322363556803e-05, "loss": 0.3296, "step": 51948 }, { "epoch": 3.14, "learning_rate": 2.3321714725453804e-05, "loss": 0.3586, "step": 51950 }, { "epoch": 3.14, "learning_rate": 2.332020581533958e-05, "loss": 0.5645, "step": 51952 }, { "epoch": 3.14, "learning_rate": 2.3318696905225357e-05, "loss": 0.3134, "step": 51954 }, { "epoch": 3.14, "learning_rate": 2.331718799511113e-05, "loss": 0.2152, "step": 51956 }, { "epoch": 3.14, "learning_rate": 2.3315679084996907e-05, "loss": 0.5004, "step": 51958 }, { "epoch": 3.14, "learning_rate": 2.3314170174882684e-05, "loss": 0.4136, "step": 51960 }, { "epoch": 3.14, "learning_rate": 2.331266126476846e-05, "loss": 0.5536, "step": 51962 }, { "epoch": 3.14, "learning_rate": 2.3311152354654233e-05, "loss": 0.4206, "step": 51964 }, { "epoch": 3.14, "learning_rate": 2.330964344454001e-05, "loss": 0.2655, "step": 51966 }, { "epoch": 3.14, "learning_rate": 2.3308134534425786e-05, "loss": 0.4905, "step": 51968 }, { "epoch": 3.14, "learning_rate": 2.3306625624311563e-05, "loss": 0.545, "step": 51970 }, { "epoch": 3.14, "learning_rate": 2.3305116714197336e-05, "loss": 0.3926, "step": 51972 }, { "epoch": 3.14, "learning_rate": 2.330360780408311e-05, "loss": 0.4872, "step": 51974 }, { "epoch": 3.14, "learning_rate": 2.330209889396889e-05, "loss": 0.2635, "step": 51976 }, { "epoch": 3.14, "learning_rate": 2.3300589983854662e-05, "loss": 0.5069, "step": 51978 }, { "epoch": 3.14, "learning_rate": 2.329908107374044e-05, "loss": 0.3558, "step": 51980 }, { "epoch": 3.14, "learning_rate": 2.3297572163626212e-05, "loss": 0.3136, "step": 51982 }, { "epoch": 3.14, "learning_rate": 2.329606325351199e-05, "loss": 0.3456, "step": 51984 }, { "epoch": 3.14, "learning_rate": 2.3294554343397765e-05, "loss": 0.3717, "step": 51986 }, { "epoch": 3.14, "learning_rate": 2.3293045433283538e-05, "loss": 0.2719, "step": 51988 }, { "epoch": 3.14, "learning_rate": 2.3291536523169318e-05, "loss": 0.3015, "step": 51990 }, { "epoch": 3.14, "learning_rate": 2.329002761305509e-05, "loss": 0.4982, "step": 51992 }, { "epoch": 3.14, "learning_rate": 2.3288518702940867e-05, "loss": 0.4071, "step": 51994 }, { "epoch": 3.14, "learning_rate": 2.328700979282664e-05, "loss": 0.3749, "step": 51996 }, { "epoch": 3.14, "learning_rate": 2.328550088271242e-05, "loss": 0.3025, "step": 51998 }, { "epoch": 3.14, "learning_rate": 2.3283991972598194e-05, "loss": 0.4409, "step": 52000 }, { "epoch": 3.14, "learning_rate": 2.328248306248397e-05, "loss": 0.2507, "step": 52002 }, { "epoch": 3.14, "learning_rate": 2.3280974152369743e-05, "loss": 0.5629, "step": 52004 }, { "epoch": 3.14, "learning_rate": 2.327946524225552e-05, "loss": 0.2878, "step": 52006 }, { "epoch": 3.14, "learning_rate": 2.3277956332141296e-05, "loss": 0.3574, "step": 52008 }, { "epoch": 3.14, "learning_rate": 2.327644742202707e-05, "loss": 0.4231, "step": 52010 }, { "epoch": 3.14, "learning_rate": 2.3274938511912846e-05, "loss": 0.2551, "step": 52012 }, { "epoch": 3.14, "learning_rate": 2.3273429601798622e-05, "loss": 0.3173, "step": 52014 }, { "epoch": 3.14, "learning_rate": 2.32719206916844e-05, "loss": 0.2194, "step": 52016 }, { "epoch": 3.14, "learning_rate": 2.3270411781570172e-05, "loss": 0.5009, "step": 52018 }, { "epoch": 3.14, "learning_rate": 2.326890287145595e-05, "loss": 0.3989, "step": 52020 }, { "epoch": 3.14, "learning_rate": 2.3267393961341725e-05, "loss": 0.51, "step": 52022 }, { "epoch": 3.14, "learning_rate": 2.3265885051227498e-05, "loss": 0.3606, "step": 52024 }, { "epoch": 3.14, "learning_rate": 2.3264376141113275e-05, "loss": 0.2652, "step": 52026 }, { "epoch": 3.14, "learning_rate": 2.326286723099905e-05, "loss": 0.4712, "step": 52028 }, { "epoch": 3.14, "learning_rate": 2.3261358320884828e-05, "loss": 0.2124, "step": 52030 }, { "epoch": 3.14, "learning_rate": 2.32598494107706e-05, "loss": 0.4383, "step": 52032 }, { "epoch": 3.14, "learning_rate": 2.3258340500656377e-05, "loss": 0.2783, "step": 52034 }, { "epoch": 3.14, "learning_rate": 2.3256831590542154e-05, "loss": 0.4204, "step": 52036 }, { "epoch": 3.14, "learning_rate": 2.3255322680427927e-05, "loss": 0.39, "step": 52038 }, { "epoch": 3.14, "learning_rate": 2.3253813770313704e-05, "loss": 0.3083, "step": 52040 }, { "epoch": 3.14, "learning_rate": 2.3252304860199477e-05, "loss": 0.5291, "step": 52042 }, { "epoch": 3.14, "learning_rate": 2.3250795950085257e-05, "loss": 0.3549, "step": 52044 }, { "epoch": 3.14, "learning_rate": 2.324928703997103e-05, "loss": 0.4509, "step": 52046 }, { "epoch": 3.14, "learning_rate": 2.3247778129856806e-05, "loss": 0.3889, "step": 52048 }, { "epoch": 3.14, "learning_rate": 2.324626921974258e-05, "loss": 0.5236, "step": 52050 }, { "epoch": 3.14, "learning_rate": 2.324476030962836e-05, "loss": 0.2857, "step": 52052 }, { "epoch": 3.14, "learning_rate": 2.3243251399514132e-05, "loss": 0.3538, "step": 52054 }, { "epoch": 3.14, "learning_rate": 2.3241742489399905e-05, "loss": 0.3342, "step": 52056 }, { "epoch": 3.14, "learning_rate": 2.3240233579285682e-05, "loss": 0.3917, "step": 52058 }, { "epoch": 3.14, "learning_rate": 2.323872466917146e-05, "loss": 0.3884, "step": 52060 }, { "epoch": 3.14, "learning_rate": 2.3237215759057235e-05, "loss": 0.1954, "step": 52062 }, { "epoch": 3.14, "learning_rate": 2.3235706848943008e-05, "loss": 0.4472, "step": 52064 }, { "epoch": 3.14, "learning_rate": 2.3234197938828785e-05, "loss": 0.3145, "step": 52066 }, { "epoch": 3.14, "learning_rate": 2.323268902871456e-05, "loss": 0.317, "step": 52068 }, { "epoch": 3.14, "learning_rate": 2.3231180118600334e-05, "loss": 0.5851, "step": 52070 }, { "epoch": 3.14, "learning_rate": 2.322967120848611e-05, "loss": 0.3555, "step": 52072 }, { "epoch": 3.14, "learning_rate": 2.3228162298371887e-05, "loss": 0.4907, "step": 52074 }, { "epoch": 3.14, "learning_rate": 2.3226653388257664e-05, "loss": 0.3287, "step": 52076 }, { "epoch": 3.14, "learning_rate": 2.3225144478143437e-05, "loss": 0.3785, "step": 52078 }, { "epoch": 3.14, "learning_rate": 2.3223635568029213e-05, "loss": 0.3762, "step": 52080 }, { "epoch": 3.14, "learning_rate": 2.322212665791499e-05, "loss": 0.2676, "step": 52082 }, { "epoch": 3.14, "learning_rate": 2.3220617747800766e-05, "loss": 0.3446, "step": 52084 }, { "epoch": 3.14, "learning_rate": 2.321910883768654e-05, "loss": 0.2645, "step": 52086 }, { "epoch": 3.14, "learning_rate": 2.3217599927572313e-05, "loss": 0.2947, "step": 52088 }, { "epoch": 3.14, "learning_rate": 2.3216091017458093e-05, "loss": 0.3176, "step": 52090 }, { "epoch": 3.14, "learning_rate": 2.3214582107343866e-05, "loss": 0.4524, "step": 52092 }, { "epoch": 3.14, "learning_rate": 2.3213073197229642e-05, "loss": 0.4069, "step": 52094 }, { "epoch": 3.14, "learning_rate": 2.3211564287115415e-05, "loss": 0.4353, "step": 52096 }, { "epoch": 3.14, "learning_rate": 2.3210055377001195e-05, "loss": 0.3132, "step": 52098 }, { "epoch": 3.14, "learning_rate": 2.320854646688697e-05, "loss": 0.4132, "step": 52100 }, { "epoch": 3.14, "learning_rate": 2.3207037556772745e-05, "loss": 0.35, "step": 52102 }, { "epoch": 3.14, "learning_rate": 2.320552864665852e-05, "loss": 0.3046, "step": 52104 }, { "epoch": 3.14, "learning_rate": 2.3204019736544295e-05, "loss": 0.3965, "step": 52106 }, { "epoch": 3.15, "learning_rate": 2.320251082643007e-05, "loss": 0.5275, "step": 52108 }, { "epoch": 3.15, "learning_rate": 2.3201001916315844e-05, "loss": 0.4806, "step": 52110 }, { "epoch": 3.15, "learning_rate": 2.3199493006201624e-05, "loss": 0.2654, "step": 52112 }, { "epoch": 3.15, "learning_rate": 2.3197984096087397e-05, "loss": 0.3503, "step": 52114 }, { "epoch": 3.15, "learning_rate": 2.3196475185973174e-05, "loss": 0.4446, "step": 52116 }, { "epoch": 3.15, "learning_rate": 2.3194966275858947e-05, "loss": 0.5991, "step": 52118 }, { "epoch": 3.15, "learning_rate": 2.3193457365744723e-05, "loss": 0.1946, "step": 52120 }, { "epoch": 3.15, "learning_rate": 2.31919484556305e-05, "loss": 0.3738, "step": 52122 }, { "epoch": 3.15, "learning_rate": 2.3190439545516273e-05, "loss": 0.4916, "step": 52124 }, { "epoch": 3.15, "learning_rate": 2.318893063540205e-05, "loss": 0.3144, "step": 52126 }, { "epoch": 3.15, "learning_rate": 2.3187421725287826e-05, "loss": 0.5573, "step": 52128 }, { "epoch": 3.15, "learning_rate": 2.3185912815173603e-05, "loss": 0.2622, "step": 52130 }, { "epoch": 3.15, "learning_rate": 2.3184403905059376e-05, "loss": 0.3128, "step": 52132 }, { "epoch": 3.15, "learning_rate": 2.3182894994945152e-05, "loss": 0.3118, "step": 52134 }, { "epoch": 3.15, "learning_rate": 2.318138608483093e-05, "loss": 0.3715, "step": 52136 }, { "epoch": 3.15, "learning_rate": 2.3179877174716702e-05, "loss": 0.3443, "step": 52138 }, { "epoch": 3.15, "learning_rate": 2.317836826460248e-05, "loss": 0.4492, "step": 52140 }, { "epoch": 3.15, "learning_rate": 2.3176859354488255e-05, "loss": 0.5068, "step": 52142 }, { "epoch": 3.15, "learning_rate": 2.317535044437403e-05, "loss": 0.3874, "step": 52144 }, { "epoch": 3.15, "learning_rate": 2.3173841534259805e-05, "loss": 0.5304, "step": 52146 }, { "epoch": 3.15, "learning_rate": 2.317233262414558e-05, "loss": 0.3593, "step": 52148 }, { "epoch": 3.15, "learning_rate": 2.3170823714031358e-05, "loss": 0.4877, "step": 52150 }, { "epoch": 3.15, "learning_rate": 2.316931480391713e-05, "loss": 0.4522, "step": 52152 }, { "epoch": 3.15, "learning_rate": 2.3167805893802907e-05, "loss": 0.4915, "step": 52154 }, { "epoch": 3.15, "learning_rate": 2.316629698368868e-05, "loss": 0.2664, "step": 52156 }, { "epoch": 3.15, "learning_rate": 2.316478807357446e-05, "loss": 0.4763, "step": 52158 }, { "epoch": 3.15, "learning_rate": 2.3163279163460233e-05, "loss": 0.4729, "step": 52160 }, { "epoch": 3.15, "learning_rate": 2.316177025334601e-05, "loss": 0.3891, "step": 52162 }, { "epoch": 3.15, "learning_rate": 2.3160261343231783e-05, "loss": 0.3953, "step": 52164 }, { "epoch": 3.15, "learning_rate": 2.3158752433117563e-05, "loss": 0.3852, "step": 52166 }, { "epoch": 3.15, "learning_rate": 2.3157243523003336e-05, "loss": 0.4259, "step": 52168 }, { "epoch": 3.15, "learning_rate": 2.315573461288911e-05, "loss": 0.413, "step": 52170 }, { "epoch": 3.15, "learning_rate": 2.3154225702774886e-05, "loss": 0.4256, "step": 52172 }, { "epoch": 3.15, "learning_rate": 2.3152716792660662e-05, "loss": 0.3963, "step": 52174 }, { "epoch": 3.15, "learning_rate": 2.315120788254644e-05, "loss": 0.3806, "step": 52176 }, { "epoch": 3.15, "learning_rate": 2.3149698972432212e-05, "loss": 0.2935, "step": 52178 }, { "epoch": 3.15, "learning_rate": 2.314819006231799e-05, "loss": 0.3099, "step": 52180 }, { "epoch": 3.15, "learning_rate": 2.3146681152203765e-05, "loss": 0.3836, "step": 52182 }, { "epoch": 3.15, "learning_rate": 2.314517224208954e-05, "loss": 0.3724, "step": 52184 }, { "epoch": 3.15, "learning_rate": 2.3143663331975314e-05, "loss": 0.4067, "step": 52186 }, { "epoch": 3.15, "learning_rate": 2.314215442186109e-05, "loss": 0.3793, "step": 52188 }, { "epoch": 3.15, "learning_rate": 2.3140645511746868e-05, "loss": 0.4073, "step": 52190 }, { "epoch": 3.15, "learning_rate": 2.313913660163264e-05, "loss": 0.3832, "step": 52192 }, { "epoch": 3.15, "learning_rate": 2.3137627691518417e-05, "loss": 0.4148, "step": 52194 }, { "epoch": 3.15, "learning_rate": 2.3136118781404194e-05, "loss": 0.2366, "step": 52196 }, { "epoch": 3.15, "learning_rate": 2.313460987128997e-05, "loss": 0.3913, "step": 52198 }, { "epoch": 3.15, "learning_rate": 2.3133100961175743e-05, "loss": 0.3045, "step": 52200 }, { "epoch": 3.15, "learning_rate": 2.3131592051061516e-05, "loss": 0.4984, "step": 52202 }, { "epoch": 3.15, "learning_rate": 2.3130083140947296e-05, "loss": 0.4257, "step": 52204 }, { "epoch": 3.15, "learning_rate": 2.312857423083307e-05, "loss": 0.3147, "step": 52206 }, { "epoch": 3.15, "learning_rate": 2.3127065320718846e-05, "loss": 0.3702, "step": 52208 }, { "epoch": 3.15, "learning_rate": 2.312555641060462e-05, "loss": 0.3944, "step": 52210 }, { "epoch": 3.15, "learning_rate": 2.31240475004904e-05, "loss": 0.4427, "step": 52212 }, { "epoch": 3.15, "learning_rate": 2.3122538590376172e-05, "loss": 0.3311, "step": 52214 }, { "epoch": 3.15, "learning_rate": 2.312102968026195e-05, "loss": 0.3602, "step": 52216 }, { "epoch": 3.15, "learning_rate": 2.3119520770147722e-05, "loss": 0.4757, "step": 52218 }, { "epoch": 3.15, "learning_rate": 2.3118011860033498e-05, "loss": 0.382, "step": 52220 }, { "epoch": 3.15, "learning_rate": 2.3116502949919275e-05, "loss": 0.2849, "step": 52222 }, { "epoch": 3.15, "learning_rate": 2.3114994039805048e-05, "loss": 0.5032, "step": 52224 }, { "epoch": 3.15, "learning_rate": 2.3113485129690828e-05, "loss": 0.4068, "step": 52226 }, { "epoch": 3.15, "learning_rate": 2.31119762195766e-05, "loss": 0.3215, "step": 52228 }, { "epoch": 3.15, "learning_rate": 2.3110467309462377e-05, "loss": 0.6589, "step": 52230 }, { "epoch": 3.15, "learning_rate": 2.310895839934815e-05, "loss": 0.4933, "step": 52232 }, { "epoch": 3.15, "learning_rate": 2.310744948923393e-05, "loss": 0.4392, "step": 52234 }, { "epoch": 3.15, "learning_rate": 2.3105940579119704e-05, "loss": 0.6681, "step": 52236 }, { "epoch": 3.15, "learning_rate": 2.3104431669005477e-05, "loss": 0.3022, "step": 52238 }, { "epoch": 3.15, "learning_rate": 2.3102922758891253e-05, "loss": 0.3981, "step": 52240 }, { "epoch": 3.15, "learning_rate": 2.310141384877703e-05, "loss": 0.3568, "step": 52242 }, { "epoch": 3.15, "learning_rate": 2.3099904938662806e-05, "loss": 0.316, "step": 52244 }, { "epoch": 3.15, "learning_rate": 2.309839602854858e-05, "loss": 0.2926, "step": 52246 }, { "epoch": 3.15, "learning_rate": 2.3096887118434356e-05, "loss": 0.3413, "step": 52248 }, { "epoch": 3.15, "learning_rate": 2.3095378208320132e-05, "loss": 0.3991, "step": 52250 }, { "epoch": 3.15, "learning_rate": 2.3093869298205906e-05, "loss": 0.5733, "step": 52252 }, { "epoch": 3.15, "learning_rate": 2.3092360388091682e-05, "loss": 0.3669, "step": 52254 }, { "epoch": 3.15, "learning_rate": 2.309085147797746e-05, "loss": 0.4434, "step": 52256 }, { "epoch": 3.15, "learning_rate": 2.3089342567863235e-05, "loss": 0.3648, "step": 52258 }, { "epoch": 3.15, "learning_rate": 2.3087833657749008e-05, "loss": 0.3642, "step": 52260 }, { "epoch": 3.15, "learning_rate": 2.3086324747634785e-05, "loss": 0.2802, "step": 52262 }, { "epoch": 3.15, "learning_rate": 2.308481583752056e-05, "loss": 0.3596, "step": 52264 }, { "epoch": 3.15, "learning_rate": 2.3083306927406338e-05, "loss": 0.2987, "step": 52266 }, { "epoch": 3.15, "learning_rate": 2.308179801729211e-05, "loss": 0.4124, "step": 52268 }, { "epoch": 3.15, "learning_rate": 2.3080289107177884e-05, "loss": 0.3255, "step": 52270 }, { "epoch": 3.15, "learning_rate": 2.3078780197063664e-05, "loss": 0.2864, "step": 52272 }, { "epoch": 3.16, "learning_rate": 2.3077271286949437e-05, "loss": 0.3175, "step": 52274 }, { "epoch": 3.16, "learning_rate": 2.3075762376835214e-05, "loss": 0.4535, "step": 52276 }, { "epoch": 3.16, "learning_rate": 2.3074253466720987e-05, "loss": 0.3734, "step": 52278 }, { "epoch": 3.16, "learning_rate": 2.3072744556606767e-05, "loss": 0.4357, "step": 52280 }, { "epoch": 3.16, "learning_rate": 2.307123564649254e-05, "loss": 0.3352, "step": 52282 }, { "epoch": 3.16, "learning_rate": 2.3069726736378313e-05, "loss": 0.4204, "step": 52284 }, { "epoch": 3.16, "learning_rate": 2.306821782626409e-05, "loss": 0.2441, "step": 52286 }, { "epoch": 3.16, "learning_rate": 2.3066708916149866e-05, "loss": 0.5636, "step": 52288 }, { "epoch": 3.16, "learning_rate": 2.3065200006035642e-05, "loss": 0.4604, "step": 52290 }, { "epoch": 3.16, "learning_rate": 2.3063691095921415e-05, "loss": 0.4429, "step": 52292 }, { "epoch": 3.16, "learning_rate": 2.3062182185807192e-05, "loss": 0.282, "step": 52294 }, { "epoch": 3.16, "learning_rate": 2.306067327569297e-05, "loss": 0.3758, "step": 52296 }, { "epoch": 3.16, "learning_rate": 2.3059164365578745e-05, "loss": 0.4465, "step": 52298 }, { "epoch": 3.16, "learning_rate": 2.3057655455464518e-05, "loss": 0.3552, "step": 52300 }, { "epoch": 3.16, "learning_rate": 2.3056146545350295e-05, "loss": 0.328, "step": 52302 }, { "epoch": 3.16, "learning_rate": 2.305463763523607e-05, "loss": 0.3589, "step": 52304 }, { "epoch": 3.16, "learning_rate": 2.3053128725121844e-05, "loss": 0.4218, "step": 52306 }, { "epoch": 3.16, "learning_rate": 2.305161981500762e-05, "loss": 0.3891, "step": 52308 }, { "epoch": 3.16, "learning_rate": 2.3050110904893397e-05, "loss": 0.5248, "step": 52310 }, { "epoch": 3.16, "learning_rate": 2.3048601994779174e-05, "loss": 0.3743, "step": 52312 }, { "epoch": 3.16, "learning_rate": 2.3047093084664947e-05, "loss": 0.3346, "step": 52314 }, { "epoch": 3.16, "learning_rate": 2.3045584174550723e-05, "loss": 0.286, "step": 52316 }, { "epoch": 3.16, "learning_rate": 2.30440752644365e-05, "loss": 0.4185, "step": 52318 }, { "epoch": 3.16, "learning_rate": 2.3042566354322273e-05, "loss": 0.2881, "step": 52320 }, { "epoch": 3.16, "learning_rate": 2.304105744420805e-05, "loss": 0.3093, "step": 52322 }, { "epoch": 3.16, "learning_rate": 2.3039548534093823e-05, "loss": 0.5386, "step": 52324 }, { "epoch": 3.16, "learning_rate": 2.3038039623979603e-05, "loss": 0.4644, "step": 52326 }, { "epoch": 3.16, "learning_rate": 2.3036530713865376e-05, "loss": 0.4718, "step": 52328 }, { "epoch": 3.16, "learning_rate": 2.3035021803751152e-05, "loss": 0.3017, "step": 52330 }, { "epoch": 3.16, "learning_rate": 2.3033512893636925e-05, "loss": 0.5408, "step": 52332 }, { "epoch": 3.16, "learning_rate": 2.3032003983522702e-05, "loss": 0.3176, "step": 52334 }, { "epoch": 3.16, "learning_rate": 2.303049507340848e-05, "loss": 0.4611, "step": 52336 }, { "epoch": 3.16, "learning_rate": 2.302898616329425e-05, "loss": 0.4064, "step": 52338 }, { "epoch": 3.16, "learning_rate": 2.302747725318003e-05, "loss": 0.5493, "step": 52340 }, { "epoch": 3.16, "learning_rate": 2.3025968343065805e-05, "loss": 0.3625, "step": 52342 }, { "epoch": 3.16, "learning_rate": 2.302445943295158e-05, "loss": 0.2609, "step": 52344 }, { "epoch": 3.16, "learning_rate": 2.3022950522837354e-05, "loss": 0.3798, "step": 52346 }, { "epoch": 3.16, "learning_rate": 2.3021441612723134e-05, "loss": 0.3859, "step": 52348 }, { "epoch": 3.16, "learning_rate": 2.3019932702608907e-05, "loss": 0.3864, "step": 52350 }, { "epoch": 3.16, "learning_rate": 2.301842379249468e-05, "loss": 0.3106, "step": 52352 }, { "epoch": 3.16, "learning_rate": 2.3016914882380457e-05, "loss": 0.3559, "step": 52354 }, { "epoch": 3.16, "learning_rate": 2.3015405972266233e-05, "loss": 0.3404, "step": 52356 }, { "epoch": 3.16, "learning_rate": 2.301389706215201e-05, "loss": 0.3763, "step": 52358 }, { "epoch": 3.16, "learning_rate": 2.3012388152037783e-05, "loss": 0.3238, "step": 52360 }, { "epoch": 3.16, "learning_rate": 2.301087924192356e-05, "loss": 0.5492, "step": 52362 }, { "epoch": 3.16, "learning_rate": 2.3009370331809336e-05, "loss": 0.4286, "step": 52364 }, { "epoch": 3.16, "learning_rate": 2.300786142169511e-05, "loss": 0.3211, "step": 52366 }, { "epoch": 3.16, "learning_rate": 2.3006352511580886e-05, "loss": 0.4803, "step": 52368 }, { "epoch": 3.16, "learning_rate": 2.3004843601466662e-05, "loss": 0.4491, "step": 52370 }, { "epoch": 3.16, "learning_rate": 2.300333469135244e-05, "loss": 0.2334, "step": 52372 }, { "epoch": 3.16, "learning_rate": 2.3001825781238212e-05, "loss": 0.3821, "step": 52374 }, { "epoch": 3.16, "learning_rate": 2.300031687112399e-05, "loss": 0.3881, "step": 52376 }, { "epoch": 3.16, "learning_rate": 2.2998807961009765e-05, "loss": 0.2574, "step": 52378 }, { "epoch": 3.16, "learning_rate": 2.299729905089554e-05, "loss": 0.3589, "step": 52380 }, { "epoch": 3.16, "learning_rate": 2.2995790140781315e-05, "loss": 0.4406, "step": 52382 }, { "epoch": 3.16, "learning_rate": 2.2994281230667088e-05, "loss": 0.3076, "step": 52384 }, { "epoch": 3.16, "learning_rate": 2.2992772320552868e-05, "loss": 0.3686, "step": 52386 }, { "epoch": 3.16, "learning_rate": 2.299126341043864e-05, "loss": 0.3049, "step": 52388 }, { "epoch": 3.16, "learning_rate": 2.2989754500324417e-05, "loss": 0.3985, "step": 52390 }, { "epoch": 3.16, "learning_rate": 2.298824559021019e-05, "loss": 0.4816, "step": 52392 }, { "epoch": 3.16, "learning_rate": 2.298673668009597e-05, "loss": 0.5487, "step": 52394 }, { "epoch": 3.16, "learning_rate": 2.2985227769981743e-05, "loss": 0.4193, "step": 52396 }, { "epoch": 3.16, "learning_rate": 2.298371885986752e-05, "loss": 0.3808, "step": 52398 }, { "epoch": 3.16, "learning_rate": 2.2982209949753293e-05, "loss": 0.4219, "step": 52400 }, { "epoch": 3.16, "learning_rate": 2.298070103963907e-05, "loss": 0.3045, "step": 52402 }, { "epoch": 3.16, "learning_rate": 2.2979192129524846e-05, "loss": 0.4455, "step": 52404 }, { "epoch": 3.16, "learning_rate": 2.297768321941062e-05, "loss": 0.5242, "step": 52406 }, { "epoch": 3.16, "learning_rate": 2.2976174309296396e-05, "loss": 0.4367, "step": 52408 }, { "epoch": 3.16, "learning_rate": 2.2974665399182172e-05, "loss": 0.3581, "step": 52410 }, { "epoch": 3.16, "learning_rate": 2.297315648906795e-05, "loss": 0.4872, "step": 52412 }, { "epoch": 3.16, "learning_rate": 2.2971647578953722e-05, "loss": 0.4589, "step": 52414 }, { "epoch": 3.16, "learning_rate": 2.29701386688395e-05, "loss": 0.4134, "step": 52416 }, { "epoch": 3.16, "learning_rate": 2.2968629758725275e-05, "loss": 0.2913, "step": 52418 }, { "epoch": 3.16, "learning_rate": 2.2967120848611048e-05, "loss": 0.4083, "step": 52420 }, { "epoch": 3.16, "learning_rate": 2.2965611938496825e-05, "loss": 0.3487, "step": 52422 }, { "epoch": 3.16, "learning_rate": 2.29641030283826e-05, "loss": 0.3356, "step": 52424 }, { "epoch": 3.16, "learning_rate": 2.2962594118268378e-05, "loss": 0.4396, "step": 52426 }, { "epoch": 3.16, "learning_rate": 2.296108520815415e-05, "loss": 0.3184, "step": 52428 }, { "epoch": 3.16, "learning_rate": 2.2959576298039927e-05, "loss": 0.4215, "step": 52430 }, { "epoch": 3.16, "learning_rate": 2.2958067387925704e-05, "loss": 0.452, "step": 52432 }, { "epoch": 3.16, "learning_rate": 2.2956558477811477e-05, "loss": 0.2042, "step": 52434 }, { "epoch": 3.16, "learning_rate": 2.2955049567697253e-05, "loss": 0.4637, "step": 52436 }, { "epoch": 3.17, "learning_rate": 2.2953540657583026e-05, "loss": 0.2749, "step": 52438 }, { "epoch": 3.17, "learning_rate": 2.2952031747468806e-05, "loss": 0.5111, "step": 52440 }, { "epoch": 3.17, "learning_rate": 2.295052283735458e-05, "loss": 0.4533, "step": 52442 }, { "epoch": 3.17, "learning_rate": 2.2949013927240356e-05, "loss": 0.3698, "step": 52444 }, { "epoch": 3.17, "learning_rate": 2.294750501712613e-05, "loss": 0.548, "step": 52446 }, { "epoch": 3.17, "learning_rate": 2.2945996107011906e-05, "loss": 0.2885, "step": 52448 }, { "epoch": 3.17, "learning_rate": 2.2944487196897682e-05, "loss": 0.3532, "step": 52450 }, { "epoch": 3.17, "learning_rate": 2.2942978286783455e-05, "loss": 0.4595, "step": 52452 }, { "epoch": 3.17, "learning_rate": 2.2941469376669232e-05, "loss": 0.2058, "step": 52454 }, { "epoch": 3.17, "learning_rate": 2.2939960466555008e-05, "loss": 0.3994, "step": 52456 }, { "epoch": 3.17, "learning_rate": 2.2938451556440785e-05, "loss": 0.428, "step": 52458 }, { "epoch": 3.17, "learning_rate": 2.2936942646326558e-05, "loss": 0.4162, "step": 52460 }, { "epoch": 3.17, "learning_rate": 2.2935433736212338e-05, "loss": 0.3699, "step": 52462 }, { "epoch": 3.17, "learning_rate": 2.293392482609811e-05, "loss": 0.3308, "step": 52464 }, { "epoch": 3.17, "learning_rate": 2.2932415915983884e-05, "loss": 0.459, "step": 52466 }, { "epoch": 3.17, "learning_rate": 2.293090700586966e-05, "loss": 0.3967, "step": 52468 }, { "epoch": 3.17, "learning_rate": 2.2929398095755437e-05, "loss": 0.3762, "step": 52470 }, { "epoch": 3.17, "learning_rate": 2.2927889185641214e-05, "loss": 0.4426, "step": 52472 }, { "epoch": 3.17, "learning_rate": 2.2926380275526987e-05, "loss": 0.4037, "step": 52474 }, { "epoch": 3.17, "learning_rate": 2.2924871365412763e-05, "loss": 0.4646, "step": 52476 }, { "epoch": 3.17, "learning_rate": 2.292336245529854e-05, "loss": 0.3044, "step": 52478 }, { "epoch": 3.17, "learning_rate": 2.2921853545184316e-05, "loss": 0.3444, "step": 52480 }, { "epoch": 3.17, "learning_rate": 2.292034463507009e-05, "loss": 0.4066, "step": 52482 }, { "epoch": 3.17, "learning_rate": 2.2918835724955866e-05, "loss": 0.3772, "step": 52484 }, { "epoch": 3.17, "learning_rate": 2.2917326814841642e-05, "loss": 0.5127, "step": 52486 }, { "epoch": 3.17, "learning_rate": 2.2915817904727416e-05, "loss": 0.3852, "step": 52488 }, { "epoch": 3.17, "learning_rate": 2.2914308994613192e-05, "loss": 0.3526, "step": 52490 }, { "epoch": 3.17, "learning_rate": 2.291280008449897e-05, "loss": 0.3999, "step": 52492 }, { "epoch": 3.17, "learning_rate": 2.2911291174384745e-05, "loss": 0.3661, "step": 52494 }, { "epoch": 3.17, "learning_rate": 2.2909782264270518e-05, "loss": 0.3213, "step": 52496 }, { "epoch": 3.17, "learning_rate": 2.290827335415629e-05, "loss": 0.2539, "step": 52498 }, { "epoch": 3.17, "learning_rate": 2.290676444404207e-05, "loss": 0.6272, "step": 52500 }, { "epoch": 3.17, "learning_rate": 2.2905255533927844e-05, "loss": 0.4284, "step": 52502 }, { "epoch": 3.17, "learning_rate": 2.290374662381362e-05, "loss": 0.3826, "step": 52504 }, { "epoch": 3.17, "learning_rate": 2.2902237713699394e-05, "loss": 0.4945, "step": 52506 }, { "epoch": 3.17, "learning_rate": 2.2900728803585174e-05, "loss": 0.4641, "step": 52508 }, { "epoch": 3.17, "learning_rate": 2.2899219893470947e-05, "loss": 0.2688, "step": 52510 }, { "epoch": 3.17, "learning_rate": 2.2897710983356724e-05, "loss": 0.3509, "step": 52512 }, { "epoch": 3.17, "learning_rate": 2.2896202073242497e-05, "loss": 0.3826, "step": 52514 }, { "epoch": 3.17, "learning_rate": 2.2894693163128273e-05, "loss": 0.3349, "step": 52516 }, { "epoch": 3.17, "learning_rate": 2.289318425301405e-05, "loss": 0.2256, "step": 52518 }, { "epoch": 3.17, "learning_rate": 2.2891675342899823e-05, "loss": 0.4081, "step": 52520 }, { "epoch": 3.17, "learning_rate": 2.28901664327856e-05, "loss": 0.4476, "step": 52522 }, { "epoch": 3.17, "learning_rate": 2.2888657522671376e-05, "loss": 0.3796, "step": 52524 }, { "epoch": 3.17, "learning_rate": 2.2887148612557152e-05, "loss": 0.4379, "step": 52526 }, { "epoch": 3.17, "learning_rate": 2.2885639702442926e-05, "loss": 0.4138, "step": 52528 }, { "epoch": 3.17, "learning_rate": 2.2884130792328702e-05, "loss": 0.295, "step": 52530 }, { "epoch": 3.17, "learning_rate": 2.288262188221448e-05, "loss": 0.2551, "step": 52532 }, { "epoch": 3.17, "learning_rate": 2.288111297210025e-05, "loss": 0.3141, "step": 52534 }, { "epoch": 3.17, "learning_rate": 2.2879604061986028e-05, "loss": 0.3983, "step": 52536 }, { "epoch": 3.17, "learning_rate": 2.2878095151871805e-05, "loss": 0.541, "step": 52538 }, { "epoch": 3.17, "learning_rate": 2.287658624175758e-05, "loss": 0.318, "step": 52540 }, { "epoch": 3.17, "learning_rate": 2.2875077331643354e-05, "loss": 0.3755, "step": 52542 }, { "epoch": 3.17, "learning_rate": 2.287356842152913e-05, "loss": 0.381, "step": 52544 }, { "epoch": 3.17, "learning_rate": 2.2872059511414907e-05, "loss": 0.3707, "step": 52546 }, { "epoch": 3.17, "learning_rate": 2.287055060130068e-05, "loss": 0.5475, "step": 52548 }, { "epoch": 3.17, "learning_rate": 2.2869041691186457e-05, "loss": 0.3139, "step": 52550 }, { "epoch": 3.17, "learning_rate": 2.286753278107223e-05, "loss": 0.5396, "step": 52552 }, { "epoch": 3.17, "learning_rate": 2.286602387095801e-05, "loss": 0.2799, "step": 52554 }, { "epoch": 3.17, "learning_rate": 2.2864514960843783e-05, "loss": 0.4953, "step": 52556 }, { "epoch": 3.17, "learning_rate": 2.286300605072956e-05, "loss": 0.3723, "step": 52558 }, { "epoch": 3.17, "learning_rate": 2.2861497140615333e-05, "loss": 0.2903, "step": 52560 }, { "epoch": 3.17, "learning_rate": 2.2859988230501113e-05, "loss": 0.3457, "step": 52562 }, { "epoch": 3.17, "learning_rate": 2.2858479320386886e-05, "loss": 0.4789, "step": 52564 }, { "epoch": 3.17, "learning_rate": 2.285697041027266e-05, "loss": 0.3853, "step": 52566 }, { "epoch": 3.17, "learning_rate": 2.2855461500158435e-05, "loss": 0.4999, "step": 52568 }, { "epoch": 3.17, "learning_rate": 2.2853952590044212e-05, "loss": 0.4218, "step": 52570 }, { "epoch": 3.17, "learning_rate": 2.285244367992999e-05, "loss": 0.3707, "step": 52572 }, { "epoch": 3.17, "learning_rate": 2.285093476981576e-05, "loss": 0.2928, "step": 52574 }, { "epoch": 3.17, "learning_rate": 2.2849425859701538e-05, "loss": 0.4084, "step": 52576 }, { "epoch": 3.17, "learning_rate": 2.2847916949587315e-05, "loss": 0.3124, "step": 52578 }, { "epoch": 3.17, "learning_rate": 2.2846408039473088e-05, "loss": 0.4638, "step": 52580 }, { "epoch": 3.17, "learning_rate": 2.2844899129358864e-05, "loss": 0.5987, "step": 52582 }, { "epoch": 3.17, "learning_rate": 2.284339021924464e-05, "loss": 0.3972, "step": 52584 }, { "epoch": 3.17, "learning_rate": 2.2841881309130417e-05, "loss": 0.347, "step": 52586 }, { "epoch": 3.17, "learning_rate": 2.284037239901619e-05, "loss": 0.4065, "step": 52588 }, { "epoch": 3.17, "learning_rate": 2.2838863488901967e-05, "loss": 0.3948, "step": 52590 }, { "epoch": 3.17, "learning_rate": 2.2837354578787743e-05, "loss": 0.5157, "step": 52592 }, { "epoch": 3.17, "learning_rate": 2.283584566867352e-05, "loss": 0.3526, "step": 52594 }, { "epoch": 3.17, "learning_rate": 2.2834336758559293e-05, "loss": 0.5077, "step": 52596 }, { "epoch": 3.17, "learning_rate": 2.283282784844507e-05, "loss": 0.3662, "step": 52598 }, { "epoch": 3.17, "learning_rate": 2.2831318938330846e-05, "loss": 0.4052, "step": 52600 }, { "epoch": 3.17, "learning_rate": 2.282981002821662e-05, "loss": 0.3646, "step": 52602 }, { "epoch": 3.18, "learning_rate": 2.2828301118102396e-05, "loss": 0.3858, "step": 52604 }, { "epoch": 3.18, "learning_rate": 2.2826792207988172e-05, "loss": 0.3801, "step": 52606 }, { "epoch": 3.18, "learning_rate": 2.282528329787395e-05, "loss": 0.3539, "step": 52608 }, { "epoch": 3.18, "learning_rate": 2.2823774387759722e-05, "loss": 0.4541, "step": 52610 }, { "epoch": 3.18, "learning_rate": 2.2822265477645495e-05, "loss": 0.2432, "step": 52612 }, { "epoch": 3.18, "learning_rate": 2.2820756567531275e-05, "loss": 0.3271, "step": 52614 }, { "epoch": 3.18, "learning_rate": 2.2819247657417048e-05, "loss": 0.4063, "step": 52616 }, { "epoch": 3.18, "learning_rate": 2.2817738747302825e-05, "loss": 0.3594, "step": 52618 }, { "epoch": 3.18, "learning_rate": 2.2816229837188598e-05, "loss": 0.3692, "step": 52620 }, { "epoch": 3.18, "learning_rate": 2.2814720927074378e-05, "loss": 0.2978, "step": 52622 }, { "epoch": 3.18, "learning_rate": 2.281321201696015e-05, "loss": 0.2725, "step": 52624 }, { "epoch": 3.18, "learning_rate": 2.2811703106845927e-05, "loss": 0.4307, "step": 52626 }, { "epoch": 3.18, "learning_rate": 2.28101941967317e-05, "loss": 0.4423, "step": 52628 }, { "epoch": 3.18, "learning_rate": 2.2808685286617477e-05, "loss": 0.4285, "step": 52630 }, { "epoch": 3.18, "learning_rate": 2.2807176376503253e-05, "loss": 0.3465, "step": 52632 }, { "epoch": 3.18, "learning_rate": 2.2805667466389027e-05, "loss": 0.7823, "step": 52634 }, { "epoch": 3.18, "learning_rate": 2.2804158556274803e-05, "loss": 0.5991, "step": 52636 }, { "epoch": 3.18, "learning_rate": 2.280264964616058e-05, "loss": 0.293, "step": 52638 }, { "epoch": 3.18, "learning_rate": 2.2801140736046356e-05, "loss": 0.3414, "step": 52640 }, { "epoch": 3.18, "learning_rate": 2.279963182593213e-05, "loss": 0.424, "step": 52642 }, { "epoch": 3.18, "learning_rate": 2.2798122915817906e-05, "loss": 0.4201, "step": 52644 }, { "epoch": 3.18, "learning_rate": 2.2796614005703682e-05, "loss": 0.486, "step": 52646 }, { "epoch": 3.18, "learning_rate": 2.2795105095589455e-05, "loss": 0.5231, "step": 52648 }, { "epoch": 3.18, "learning_rate": 2.2793596185475232e-05, "loss": 0.3439, "step": 52650 }, { "epoch": 3.18, "learning_rate": 2.279208727536101e-05, "loss": 0.399, "step": 52652 }, { "epoch": 3.18, "learning_rate": 2.2790578365246785e-05, "loss": 0.3693, "step": 52654 }, { "epoch": 3.18, "learning_rate": 2.2789069455132558e-05, "loss": 0.3216, "step": 52656 }, { "epoch": 3.18, "learning_rate": 2.2787560545018335e-05, "loss": 0.3137, "step": 52658 }, { "epoch": 3.18, "learning_rate": 2.278605163490411e-05, "loss": 0.4139, "step": 52660 }, { "epoch": 3.18, "learning_rate": 2.2784542724789884e-05, "loss": 0.4996, "step": 52662 }, { "epoch": 3.18, "learning_rate": 2.278303381467566e-05, "loss": 0.363, "step": 52664 }, { "epoch": 3.18, "learning_rate": 2.2781524904561434e-05, "loss": 0.2785, "step": 52666 }, { "epoch": 3.18, "learning_rate": 2.2780015994447214e-05, "loss": 0.3108, "step": 52668 }, { "epoch": 3.18, "learning_rate": 2.2778507084332987e-05, "loss": 0.2816, "step": 52670 }, { "epoch": 3.18, "learning_rate": 2.2776998174218763e-05, "loss": 0.4694, "step": 52672 }, { "epoch": 3.18, "learning_rate": 2.2775489264104536e-05, "loss": 0.4843, "step": 52674 }, { "epoch": 3.18, "learning_rate": 2.2773980353990316e-05, "loss": 0.4602, "step": 52676 }, { "epoch": 3.18, "learning_rate": 2.277247144387609e-05, "loss": 0.3897, "step": 52678 }, { "epoch": 3.18, "learning_rate": 2.2770962533761863e-05, "loss": 0.2965, "step": 52680 }, { "epoch": 3.18, "learning_rate": 2.276945362364764e-05, "loss": 0.6715, "step": 52682 }, { "epoch": 3.18, "learning_rate": 2.2767944713533416e-05, "loss": 0.3275, "step": 52684 }, { "epoch": 3.18, "learning_rate": 2.2766435803419192e-05, "loss": 0.3121, "step": 52686 }, { "epoch": 3.18, "learning_rate": 2.2764926893304965e-05, "loss": 0.4291, "step": 52688 }, { "epoch": 3.18, "learning_rate": 2.2763417983190742e-05, "loss": 0.3626, "step": 52690 }, { "epoch": 3.18, "learning_rate": 2.276190907307652e-05, "loss": 0.3826, "step": 52692 }, { "epoch": 3.18, "learning_rate": 2.276040016296229e-05, "loss": 0.3711, "step": 52694 }, { "epoch": 3.18, "learning_rate": 2.2758891252848068e-05, "loss": 0.3665, "step": 52696 }, { "epoch": 3.18, "learning_rate": 2.2757382342733844e-05, "loss": 0.3859, "step": 52698 }, { "epoch": 3.18, "learning_rate": 2.275587343261962e-05, "loss": 0.3498, "step": 52700 }, { "epoch": 3.18, "learning_rate": 2.2754364522505394e-05, "loss": 0.3396, "step": 52702 }, { "epoch": 3.18, "learning_rate": 2.275285561239117e-05, "loss": 0.375, "step": 52704 }, { "epoch": 3.18, "learning_rate": 2.2751346702276947e-05, "loss": 0.3384, "step": 52706 }, { "epoch": 3.18, "learning_rate": 2.2749837792162724e-05, "loss": 0.3385, "step": 52708 }, { "epoch": 3.18, "learning_rate": 2.2748328882048497e-05, "loss": 0.3921, "step": 52710 }, { "epoch": 3.18, "learning_rate": 2.2746819971934273e-05, "loss": 0.5645, "step": 52712 }, { "epoch": 3.18, "learning_rate": 2.274531106182005e-05, "loss": 0.3479, "step": 52714 }, { "epoch": 3.18, "learning_rate": 2.2743802151705823e-05, "loss": 0.3202, "step": 52716 }, { "epoch": 3.18, "learning_rate": 2.27422932415916e-05, "loss": 0.4538, "step": 52718 }, { "epoch": 3.18, "learning_rate": 2.2740784331477376e-05, "loss": 0.5044, "step": 52720 }, { "epoch": 3.18, "learning_rate": 2.2739275421363152e-05, "loss": 0.2735, "step": 52722 }, { "epoch": 3.18, "learning_rate": 2.2737766511248926e-05, "loss": 0.424, "step": 52724 }, { "epoch": 3.18, "learning_rate": 2.2736257601134702e-05, "loss": 0.3312, "step": 52726 }, { "epoch": 3.18, "learning_rate": 2.273474869102048e-05, "loss": 0.4359, "step": 52728 }, { "epoch": 3.18, "learning_rate": 2.2733239780906252e-05, "loss": 0.4487, "step": 52730 }, { "epoch": 3.18, "learning_rate": 2.2731730870792028e-05, "loss": 0.3792, "step": 52732 }, { "epoch": 3.18, "learning_rate": 2.27302219606778e-05, "loss": 0.2663, "step": 52734 }, { "epoch": 3.18, "learning_rate": 2.272871305056358e-05, "loss": 0.4174, "step": 52736 }, { "epoch": 3.18, "learning_rate": 2.2727204140449354e-05, "loss": 0.3889, "step": 52738 }, { "epoch": 3.18, "learning_rate": 2.272569523033513e-05, "loss": 0.2249, "step": 52740 }, { "epoch": 3.18, "learning_rate": 2.2724186320220904e-05, "loss": 0.2482, "step": 52742 }, { "epoch": 3.18, "learning_rate": 2.272267741010668e-05, "loss": 0.5004, "step": 52744 }, { "epoch": 3.18, "learning_rate": 2.2721168499992457e-05, "loss": 0.2194, "step": 52746 }, { "epoch": 3.18, "learning_rate": 2.271965958987823e-05, "loss": 0.3892, "step": 52748 }, { "epoch": 3.18, "learning_rate": 2.2718150679764007e-05, "loss": 0.4718, "step": 52750 }, { "epoch": 3.18, "learning_rate": 2.2716641769649783e-05, "loss": 0.3293, "step": 52752 }, { "epoch": 3.18, "learning_rate": 2.271513285953556e-05, "loss": 0.4253, "step": 52754 }, { "epoch": 3.18, "learning_rate": 2.2713623949421333e-05, "loss": 0.3536, "step": 52756 }, { "epoch": 3.18, "learning_rate": 2.271211503930711e-05, "loss": 0.3182, "step": 52758 }, { "epoch": 3.18, "learning_rate": 2.2710606129192886e-05, "loss": 0.2183, "step": 52760 }, { "epoch": 3.18, "learning_rate": 2.270909721907866e-05, "loss": 0.3346, "step": 52762 }, { "epoch": 3.18, "learning_rate": 2.2707588308964436e-05, "loss": 0.3968, "step": 52764 }, { "epoch": 3.18, "learning_rate": 2.2706079398850212e-05, "loss": 0.4408, "step": 52766 }, { "epoch": 3.18, "learning_rate": 2.270457048873599e-05, "loss": 0.3349, "step": 52768 }, { "epoch": 3.19, "learning_rate": 2.270306157862176e-05, "loss": 0.2986, "step": 52770 }, { "epoch": 3.19, "learning_rate": 2.2701552668507538e-05, "loss": 0.3852, "step": 52772 }, { "epoch": 3.19, "learning_rate": 2.2700043758393315e-05, "loss": 0.3288, "step": 52774 }, { "epoch": 3.19, "learning_rate": 2.2698534848279088e-05, "loss": 0.401, "step": 52776 }, { "epoch": 3.19, "learning_rate": 2.2697025938164864e-05, "loss": 0.343, "step": 52778 }, { "epoch": 3.19, "learning_rate": 2.2695517028050637e-05, "loss": 0.5146, "step": 52780 }, { "epoch": 3.19, "learning_rate": 2.2694008117936417e-05, "loss": 0.5867, "step": 52782 }, { "epoch": 3.19, "learning_rate": 2.269249920782219e-05, "loss": 0.3063, "step": 52784 }, { "epoch": 3.19, "learning_rate": 2.2690990297707967e-05, "loss": 0.3556, "step": 52786 }, { "epoch": 3.19, "learning_rate": 2.268948138759374e-05, "loss": 0.3388, "step": 52788 }, { "epoch": 3.19, "learning_rate": 2.268797247747952e-05, "loss": 0.2357, "step": 52790 }, { "epoch": 3.19, "learning_rate": 2.2686463567365293e-05, "loss": 0.4173, "step": 52792 }, { "epoch": 3.19, "learning_rate": 2.2684954657251066e-05, "loss": 0.4939, "step": 52794 }, { "epoch": 3.19, "learning_rate": 2.2683445747136843e-05, "loss": 0.3926, "step": 52796 }, { "epoch": 3.19, "learning_rate": 2.268193683702262e-05, "loss": 0.3194, "step": 52798 }, { "epoch": 3.19, "learning_rate": 2.2680427926908396e-05, "loss": 0.378, "step": 52800 }, { "epoch": 3.19, "learning_rate": 2.267891901679417e-05, "loss": 0.4249, "step": 52802 }, { "epoch": 3.19, "learning_rate": 2.2677410106679945e-05, "loss": 0.306, "step": 52804 }, { "epoch": 3.19, "learning_rate": 2.2675901196565722e-05, "loss": 0.3736, "step": 52806 }, { "epoch": 3.19, "learning_rate": 2.26743922864515e-05, "loss": 0.4299, "step": 52808 }, { "epoch": 3.19, "learning_rate": 2.267288337633727e-05, "loss": 0.5048, "step": 52810 }, { "epoch": 3.19, "learning_rate": 2.2671374466223048e-05, "loss": 0.461, "step": 52812 }, { "epoch": 3.19, "learning_rate": 2.2669865556108825e-05, "loss": 0.3711, "step": 52814 }, { "epoch": 3.19, "learning_rate": 2.2668356645994598e-05, "loss": 0.3431, "step": 52816 }, { "epoch": 3.19, "learning_rate": 2.2666847735880374e-05, "loss": 0.2677, "step": 52818 }, { "epoch": 3.19, "learning_rate": 2.266533882576615e-05, "loss": 0.4205, "step": 52820 }, { "epoch": 3.19, "learning_rate": 2.2663829915651927e-05, "loss": 0.3695, "step": 52822 }, { "epoch": 3.19, "learning_rate": 2.26623210055377e-05, "loss": 0.3846, "step": 52824 }, { "epoch": 3.19, "learning_rate": 2.2660812095423477e-05, "loss": 0.3068, "step": 52826 }, { "epoch": 3.19, "learning_rate": 2.2659303185309253e-05, "loss": 0.5026, "step": 52828 }, { "epoch": 3.19, "learning_rate": 2.2657794275195027e-05, "loss": 0.3429, "step": 52830 }, { "epoch": 3.19, "learning_rate": 2.2656285365080803e-05, "loss": 0.322, "step": 52832 }, { "epoch": 3.19, "learning_rate": 2.265477645496658e-05, "loss": 0.4108, "step": 52834 }, { "epoch": 3.19, "learning_rate": 2.2653267544852356e-05, "loss": 0.4931, "step": 52836 }, { "epoch": 3.19, "learning_rate": 2.265175863473813e-05, "loss": 0.4154, "step": 52838 }, { "epoch": 3.19, "learning_rate": 2.2650249724623906e-05, "loss": 0.442, "step": 52840 }, { "epoch": 3.19, "learning_rate": 2.2648740814509682e-05, "loss": 0.4237, "step": 52842 }, { "epoch": 3.19, "learning_rate": 2.2647231904395455e-05, "loss": 0.3445, "step": 52844 }, { "epoch": 3.19, "learning_rate": 2.2645722994281232e-05, "loss": 0.3593, "step": 52846 }, { "epoch": 3.19, "learning_rate": 2.2644214084167005e-05, "loss": 0.2979, "step": 52848 }, { "epoch": 3.19, "learning_rate": 2.2642705174052785e-05, "loss": 0.47, "step": 52850 }, { "epoch": 3.19, "learning_rate": 2.2641196263938558e-05, "loss": 0.3811, "step": 52852 }, { "epoch": 3.19, "learning_rate": 2.2639687353824335e-05, "loss": 0.3892, "step": 52854 }, { "epoch": 3.19, "learning_rate": 2.2638178443710108e-05, "loss": 0.3754, "step": 52856 }, { "epoch": 3.19, "learning_rate": 2.2636669533595888e-05, "loss": 0.3722, "step": 52858 }, { "epoch": 3.19, "learning_rate": 2.263516062348166e-05, "loss": 0.2734, "step": 52860 }, { "epoch": 3.19, "learning_rate": 2.2633651713367434e-05, "loss": 0.6066, "step": 52862 }, { "epoch": 3.19, "learning_rate": 2.263214280325321e-05, "loss": 0.3584, "step": 52864 }, { "epoch": 3.19, "learning_rate": 2.2630633893138987e-05, "loss": 0.3159, "step": 52866 }, { "epoch": 3.19, "learning_rate": 2.2629124983024763e-05, "loss": 0.4297, "step": 52868 }, { "epoch": 3.19, "learning_rate": 2.2627616072910537e-05, "loss": 0.3389, "step": 52870 }, { "epoch": 3.19, "learning_rate": 2.2626107162796313e-05, "loss": 0.4145, "step": 52872 }, { "epoch": 3.19, "learning_rate": 2.262459825268209e-05, "loss": 0.3457, "step": 52874 }, { "epoch": 3.19, "learning_rate": 2.2623089342567863e-05, "loss": 0.4116, "step": 52876 }, { "epoch": 3.19, "learning_rate": 2.262158043245364e-05, "loss": 0.4319, "step": 52878 }, { "epoch": 3.19, "learning_rate": 2.2620071522339416e-05, "loss": 0.4394, "step": 52880 }, { "epoch": 3.19, "learning_rate": 2.2618562612225192e-05, "loss": 0.4049, "step": 52882 }, { "epoch": 3.19, "learning_rate": 2.2617053702110965e-05, "loss": 0.5325, "step": 52884 }, { "epoch": 3.19, "learning_rate": 2.2615544791996742e-05, "loss": 0.381, "step": 52886 }, { "epoch": 3.19, "learning_rate": 2.261403588188252e-05, "loss": 0.2286, "step": 52888 }, { "epoch": 3.19, "learning_rate": 2.2612526971768295e-05, "loss": 0.4795, "step": 52890 }, { "epoch": 3.19, "learning_rate": 2.2611018061654068e-05, "loss": 0.4305, "step": 52892 }, { "epoch": 3.19, "learning_rate": 2.260950915153984e-05, "loss": 0.435, "step": 52894 }, { "epoch": 3.19, "learning_rate": 2.260800024142562e-05, "loss": 0.3747, "step": 52896 }, { "epoch": 3.19, "learning_rate": 2.2606491331311394e-05, "loss": 0.6135, "step": 52898 }, { "epoch": 3.19, "learning_rate": 2.260498242119717e-05, "loss": 0.3117, "step": 52900 }, { "epoch": 3.19, "learning_rate": 2.2603473511082944e-05, "loss": 0.4495, "step": 52902 }, { "epoch": 3.19, "learning_rate": 2.2601964600968724e-05, "loss": 0.4071, "step": 52904 }, { "epoch": 3.19, "learning_rate": 2.2600455690854497e-05, "loss": 0.3561, "step": 52906 }, { "epoch": 3.19, "learning_rate": 2.259894678074027e-05, "loss": 0.4495, "step": 52908 }, { "epoch": 3.19, "learning_rate": 2.2597437870626046e-05, "loss": 0.3512, "step": 52910 }, { "epoch": 3.19, "learning_rate": 2.2595928960511823e-05, "loss": 0.4441, "step": 52912 }, { "epoch": 3.19, "learning_rate": 2.25944200503976e-05, "loss": 0.3689, "step": 52914 }, { "epoch": 3.19, "learning_rate": 2.2592911140283373e-05, "loss": 0.5529, "step": 52916 }, { "epoch": 3.19, "learning_rate": 2.259140223016915e-05, "loss": 0.5235, "step": 52918 }, { "epoch": 3.19, "learning_rate": 2.2589893320054926e-05, "loss": 0.2339, "step": 52920 }, { "epoch": 3.19, "learning_rate": 2.2588384409940702e-05, "loss": 0.3924, "step": 52922 }, { "epoch": 3.19, "learning_rate": 2.2586875499826475e-05, "loss": 0.5744, "step": 52924 }, { "epoch": 3.19, "learning_rate": 2.2585366589712252e-05, "loss": 0.4266, "step": 52926 }, { "epoch": 3.19, "learning_rate": 2.258385767959803e-05, "loss": 0.2979, "step": 52928 }, { "epoch": 3.19, "learning_rate": 2.25823487694838e-05, "loss": 0.3623, "step": 52930 }, { "epoch": 3.19, "learning_rate": 2.2580839859369578e-05, "loss": 0.4737, "step": 52932 }, { "epoch": 3.19, "learning_rate": 2.2579330949255354e-05, "loss": 0.3379, "step": 52934 }, { "epoch": 3.2, "learning_rate": 2.257782203914113e-05, "loss": 0.2386, "step": 52936 }, { "epoch": 3.2, "learning_rate": 2.2576313129026904e-05, "loss": 0.319, "step": 52938 }, { "epoch": 3.2, "learning_rate": 2.257480421891268e-05, "loss": 0.3978, "step": 52940 }, { "epoch": 3.2, "learning_rate": 2.2573295308798457e-05, "loss": 0.3943, "step": 52942 }, { "epoch": 3.2, "learning_rate": 2.257178639868423e-05, "loss": 0.3677, "step": 52944 }, { "epoch": 3.2, "learning_rate": 2.2570277488570007e-05, "loss": 0.4226, "step": 52946 }, { "epoch": 3.2, "learning_rate": 2.2568768578455783e-05, "loss": 0.3231, "step": 52948 }, { "epoch": 3.2, "learning_rate": 2.256725966834156e-05, "loss": 0.427, "step": 52950 }, { "epoch": 3.2, "learning_rate": 2.2565750758227333e-05, "loss": 0.3321, "step": 52952 }, { "epoch": 3.2, "learning_rate": 2.256424184811311e-05, "loss": 0.4634, "step": 52954 }, { "epoch": 3.2, "learning_rate": 2.2562732937998886e-05, "loss": 0.3794, "step": 52956 }, { "epoch": 3.2, "learning_rate": 2.256122402788466e-05, "loss": 0.4102, "step": 52958 }, { "epoch": 3.2, "learning_rate": 2.2559715117770436e-05, "loss": 0.4551, "step": 52960 }, { "epoch": 3.2, "learning_rate": 2.255820620765621e-05, "loss": 0.3968, "step": 52962 }, { "epoch": 3.2, "learning_rate": 2.255669729754199e-05, "loss": 0.454, "step": 52964 }, { "epoch": 3.2, "learning_rate": 2.2555188387427762e-05, "loss": 0.4901, "step": 52966 }, { "epoch": 3.2, "learning_rate": 2.2553679477313538e-05, "loss": 0.3544, "step": 52968 }, { "epoch": 3.2, "learning_rate": 2.255217056719931e-05, "loss": 0.5475, "step": 52970 }, { "epoch": 3.2, "learning_rate": 2.255066165708509e-05, "loss": 0.4134, "step": 52972 }, { "epoch": 3.2, "learning_rate": 2.2549152746970864e-05, "loss": 0.2842, "step": 52974 }, { "epoch": 3.2, "learning_rate": 2.2547643836856638e-05, "loss": 0.4233, "step": 52976 }, { "epoch": 3.2, "learning_rate": 2.2546134926742414e-05, "loss": 0.4171, "step": 52978 }, { "epoch": 3.2, "learning_rate": 2.254462601662819e-05, "loss": 0.4517, "step": 52980 }, { "epoch": 3.2, "learning_rate": 2.2543117106513967e-05, "loss": 0.3431, "step": 52982 }, { "epoch": 3.2, "learning_rate": 2.254160819639974e-05, "loss": 0.3187, "step": 52984 }, { "epoch": 3.2, "learning_rate": 2.2540099286285517e-05, "loss": 0.2975, "step": 52986 }, { "epoch": 3.2, "learning_rate": 2.2538590376171293e-05, "loss": 0.2554, "step": 52988 }, { "epoch": 3.2, "learning_rate": 2.2537081466057066e-05, "loss": 0.3452, "step": 52990 }, { "epoch": 3.2, "learning_rate": 2.2535572555942843e-05, "loss": 0.4945, "step": 52992 }, { "epoch": 3.2, "learning_rate": 2.253406364582862e-05, "loss": 0.4049, "step": 52994 }, { "epoch": 3.2, "learning_rate": 2.2532554735714396e-05, "loss": 0.485, "step": 52996 }, { "epoch": 3.2, "learning_rate": 2.253104582560017e-05, "loss": 0.4502, "step": 52998 }, { "epoch": 3.2, "learning_rate": 2.2529536915485946e-05, "loss": 0.3176, "step": 53000 }, { "epoch": 3.2, "learning_rate": 2.2528028005371722e-05, "loss": 0.2903, "step": 53002 }, { "epoch": 3.2, "learning_rate": 2.25265190952575e-05, "loss": 0.4263, "step": 53004 }, { "epoch": 3.2, "learning_rate": 2.2525010185143272e-05, "loss": 0.4663, "step": 53006 }, { "epoch": 3.2, "learning_rate": 2.2523501275029045e-05, "loss": 0.3963, "step": 53008 }, { "epoch": 3.2, "learning_rate": 2.2521992364914825e-05, "loss": 0.441, "step": 53010 }, { "epoch": 3.2, "learning_rate": 2.2520483454800598e-05, "loss": 0.3174, "step": 53012 }, { "epoch": 3.2, "learning_rate": 2.2518974544686374e-05, "loss": 0.2612, "step": 53014 }, { "epoch": 3.2, "learning_rate": 2.2517465634572148e-05, "loss": 0.3507, "step": 53016 }, { "epoch": 3.2, "learning_rate": 2.2515956724457927e-05, "loss": 0.3781, "step": 53018 }, { "epoch": 3.2, "learning_rate": 2.25144478143437e-05, "loss": 0.4305, "step": 53020 }, { "epoch": 3.2, "learning_rate": 2.2512938904229477e-05, "loss": 0.4338, "step": 53022 }, { "epoch": 3.2, "learning_rate": 2.251142999411525e-05, "loss": 0.3634, "step": 53024 }, { "epoch": 3.2, "learning_rate": 2.2509921084001027e-05, "loss": 0.4203, "step": 53026 }, { "epoch": 3.2, "learning_rate": 2.2508412173886803e-05, "loss": 0.3599, "step": 53028 }, { "epoch": 3.2, "learning_rate": 2.2506903263772576e-05, "loss": 0.4436, "step": 53030 }, { "epoch": 3.2, "learning_rate": 2.2505394353658353e-05, "loss": 0.4329, "step": 53032 }, { "epoch": 3.2, "learning_rate": 2.250388544354413e-05, "loss": 0.4293, "step": 53034 }, { "epoch": 3.2, "learning_rate": 2.2502376533429906e-05, "loss": 0.281, "step": 53036 }, { "epoch": 3.2, "learning_rate": 2.250086762331568e-05, "loss": 0.423, "step": 53038 }, { "epoch": 3.2, "learning_rate": 2.2499358713201456e-05, "loss": 0.3467, "step": 53040 }, { "epoch": 3.2, "learning_rate": 2.2497849803087232e-05, "loss": 0.4457, "step": 53042 }, { "epoch": 3.2, "learning_rate": 2.2496340892973005e-05, "loss": 0.3242, "step": 53044 }, { "epoch": 3.2, "learning_rate": 2.249483198285878e-05, "loss": 0.3056, "step": 53046 }, { "epoch": 3.2, "learning_rate": 2.2493323072744558e-05, "loss": 0.3732, "step": 53048 }, { "epoch": 3.2, "learning_rate": 2.2491814162630335e-05, "loss": 0.2616, "step": 53050 }, { "epoch": 3.2, "learning_rate": 2.2490305252516108e-05, "loss": 0.3963, "step": 53052 }, { "epoch": 3.2, "learning_rate": 2.2488796342401884e-05, "loss": 0.4139, "step": 53054 }, { "epoch": 3.2, "learning_rate": 2.248728743228766e-05, "loss": 0.5071, "step": 53056 }, { "epoch": 3.2, "learning_rate": 2.2485778522173434e-05, "loss": 0.2528, "step": 53058 }, { "epoch": 3.2, "learning_rate": 2.248426961205921e-05, "loss": 0.2799, "step": 53060 }, { "epoch": 3.2, "learning_rate": 2.2482760701944987e-05, "loss": 0.4312, "step": 53062 }, { "epoch": 3.2, "learning_rate": 2.2481251791830764e-05, "loss": 0.4811, "step": 53064 }, { "epoch": 3.2, "learning_rate": 2.2479742881716537e-05, "loss": 0.4225, "step": 53066 }, { "epoch": 3.2, "learning_rate": 2.2478233971602313e-05, "loss": 0.5006, "step": 53068 }, { "epoch": 3.2, "learning_rate": 2.247672506148809e-05, "loss": 0.3292, "step": 53070 }, { "epoch": 3.2, "learning_rate": 2.2475216151373863e-05, "loss": 0.3226, "step": 53072 }, { "epoch": 3.2, "learning_rate": 2.247370724125964e-05, "loss": 0.397, "step": 53074 }, { "epoch": 3.2, "learning_rate": 2.2472198331145412e-05, "loss": 0.3994, "step": 53076 }, { "epoch": 3.2, "learning_rate": 2.2470689421031192e-05, "loss": 0.3738, "step": 53078 }, { "epoch": 3.2, "learning_rate": 2.2469180510916965e-05, "loss": 0.3627, "step": 53080 }, { "epoch": 3.2, "learning_rate": 2.2467671600802742e-05, "loss": 0.304, "step": 53082 }, { "epoch": 3.2, "learning_rate": 2.2466162690688515e-05, "loss": 0.3909, "step": 53084 }, { "epoch": 3.2, "learning_rate": 2.2464653780574295e-05, "loss": 0.4385, "step": 53086 }, { "epoch": 3.2, "learning_rate": 2.2463144870460068e-05, "loss": 0.4621, "step": 53088 }, { "epoch": 3.2, "learning_rate": 2.246163596034584e-05, "loss": 0.526, "step": 53090 }, { "epoch": 3.2, "learning_rate": 2.2460127050231618e-05, "loss": 0.3716, "step": 53092 }, { "epoch": 3.2, "learning_rate": 2.2458618140117394e-05, "loss": 0.421, "step": 53094 }, { "epoch": 3.2, "learning_rate": 2.245710923000317e-05, "loss": 0.5112, "step": 53096 }, { "epoch": 3.2, "learning_rate": 2.2455600319888944e-05, "loss": 0.3363, "step": 53098 }, { "epoch": 3.2, "learning_rate": 2.245409140977472e-05, "loss": 0.3118, "step": 53100 }, { "epoch": 3.21, "learning_rate": 2.2452582499660497e-05, "loss": 0.3798, "step": 53102 }, { "epoch": 3.21, "learning_rate": 2.2451073589546273e-05, "loss": 0.4907, "step": 53104 }, { "epoch": 3.21, "learning_rate": 2.2449564679432047e-05, "loss": 0.3896, "step": 53106 }, { "epoch": 3.21, "learning_rate": 2.2448055769317823e-05, "loss": 0.5911, "step": 53108 }, { "epoch": 3.21, "learning_rate": 2.24465468592036e-05, "loss": 0.3174, "step": 53110 }, { "epoch": 3.21, "learning_rate": 2.2445037949089373e-05, "loss": 0.4407, "step": 53112 }, { "epoch": 3.21, "learning_rate": 2.244352903897515e-05, "loss": 0.241, "step": 53114 }, { "epoch": 3.21, "learning_rate": 2.2442020128860926e-05, "loss": 0.4243, "step": 53116 }, { "epoch": 3.21, "learning_rate": 2.2440511218746702e-05, "loss": 0.4209, "step": 53118 }, { "epoch": 3.21, "learning_rate": 2.2439002308632475e-05, "loss": 0.3947, "step": 53120 }, { "epoch": 3.21, "learning_rate": 2.243749339851825e-05, "loss": 0.3034, "step": 53122 }, { "epoch": 3.21, "learning_rate": 2.243598448840403e-05, "loss": 0.4457, "step": 53124 }, { "epoch": 3.21, "learning_rate": 2.24344755782898e-05, "loss": 0.3563, "step": 53126 }, { "epoch": 3.21, "learning_rate": 2.2432966668175578e-05, "loss": 0.644, "step": 53128 }, { "epoch": 3.21, "learning_rate": 2.243145775806135e-05, "loss": 0.497, "step": 53130 }, { "epoch": 3.21, "learning_rate": 2.242994884794713e-05, "loss": 0.4702, "step": 53132 }, { "epoch": 3.21, "learning_rate": 2.2428439937832904e-05, "loss": 0.5101, "step": 53134 }, { "epoch": 3.21, "learning_rate": 2.242693102771868e-05, "loss": 0.3597, "step": 53136 }, { "epoch": 3.21, "learning_rate": 2.2425422117604454e-05, "loss": 0.3553, "step": 53138 }, { "epoch": 3.21, "learning_rate": 2.242391320749023e-05, "loss": 0.3884, "step": 53140 }, { "epoch": 3.21, "learning_rate": 2.2422404297376007e-05, "loss": 0.3193, "step": 53142 }, { "epoch": 3.21, "learning_rate": 2.242089538726178e-05, "loss": 0.27, "step": 53144 }, { "epoch": 3.21, "learning_rate": 2.2419386477147557e-05, "loss": 0.4124, "step": 53146 }, { "epoch": 3.21, "learning_rate": 2.2417877567033333e-05, "loss": 0.5294, "step": 53148 }, { "epoch": 3.21, "learning_rate": 2.241636865691911e-05, "loss": 0.3177, "step": 53150 }, { "epoch": 3.21, "learning_rate": 2.2414859746804883e-05, "loss": 0.2863, "step": 53152 }, { "epoch": 3.21, "learning_rate": 2.241335083669066e-05, "loss": 0.3888, "step": 53154 }, { "epoch": 3.21, "learning_rate": 2.2411841926576436e-05, "loss": 0.4325, "step": 53156 }, { "epoch": 3.21, "learning_rate": 2.241033301646221e-05, "loss": 0.4767, "step": 53158 }, { "epoch": 3.21, "learning_rate": 2.2408824106347985e-05, "loss": 0.3356, "step": 53160 }, { "epoch": 3.21, "learning_rate": 2.2407315196233762e-05, "loss": 0.3943, "step": 53162 }, { "epoch": 3.21, "learning_rate": 2.240580628611954e-05, "loss": 0.3379, "step": 53164 }, { "epoch": 3.21, "learning_rate": 2.240429737600531e-05, "loss": 0.3218, "step": 53166 }, { "epoch": 3.21, "learning_rate": 2.2402788465891088e-05, "loss": 0.2747, "step": 53168 }, { "epoch": 3.21, "learning_rate": 2.2401279555776865e-05, "loss": 0.3703, "step": 53170 }, { "epoch": 3.21, "learning_rate": 2.2399770645662638e-05, "loss": 0.5088, "step": 53172 }, { "epoch": 3.21, "learning_rate": 2.2398261735548414e-05, "loss": 0.3081, "step": 53174 }, { "epoch": 3.21, "learning_rate": 2.239675282543419e-05, "loss": 0.3099, "step": 53176 }, { "epoch": 3.21, "learning_rate": 2.2395243915319967e-05, "loss": 0.3722, "step": 53178 }, { "epoch": 3.21, "learning_rate": 2.239373500520574e-05, "loss": 0.4147, "step": 53180 }, { "epoch": 3.21, "learning_rate": 2.2392226095091517e-05, "loss": 0.3404, "step": 53182 }, { "epoch": 3.21, "learning_rate": 2.2390717184977293e-05, "loss": 0.4659, "step": 53184 }, { "epoch": 3.21, "learning_rate": 2.238920827486307e-05, "loss": 0.3488, "step": 53186 }, { "epoch": 3.21, "learning_rate": 2.2387699364748843e-05, "loss": 0.2393, "step": 53188 }, { "epoch": 3.21, "learning_rate": 2.2386190454634616e-05, "loss": 0.3577, "step": 53190 }, { "epoch": 3.21, "learning_rate": 2.2384681544520396e-05, "loss": 0.448, "step": 53192 }, { "epoch": 3.21, "learning_rate": 2.238317263440617e-05, "loss": 0.3684, "step": 53194 }, { "epoch": 3.21, "learning_rate": 2.2381663724291946e-05, "loss": 0.3164, "step": 53196 }, { "epoch": 3.21, "learning_rate": 2.238015481417772e-05, "loss": 0.3464, "step": 53198 }, { "epoch": 3.21, "learning_rate": 2.23786459040635e-05, "loss": 0.4621, "step": 53200 }, { "epoch": 3.21, "learning_rate": 2.2377136993949272e-05, "loss": 0.2745, "step": 53202 }, { "epoch": 3.21, "learning_rate": 2.2375628083835045e-05, "loss": 0.3714, "step": 53204 }, { "epoch": 3.21, "learning_rate": 2.237411917372082e-05, "loss": 0.374, "step": 53206 }, { "epoch": 3.21, "learning_rate": 2.2372610263606598e-05, "loss": 0.315, "step": 53208 }, { "epoch": 3.21, "learning_rate": 2.2371101353492374e-05, "loss": 0.3741, "step": 53210 }, { "epoch": 3.21, "learning_rate": 2.2369592443378148e-05, "loss": 0.427, "step": 53212 }, { "epoch": 3.21, "learning_rate": 2.2368083533263924e-05, "loss": 0.5239, "step": 53214 }, { "epoch": 3.21, "learning_rate": 2.23665746231497e-05, "loss": 0.4939, "step": 53216 }, { "epoch": 3.21, "learning_rate": 2.2365065713035477e-05, "loss": 0.243, "step": 53218 }, { "epoch": 3.21, "learning_rate": 2.236355680292125e-05, "loss": 0.3336, "step": 53220 }, { "epoch": 3.21, "learning_rate": 2.2362047892807027e-05, "loss": 0.3664, "step": 53222 }, { "epoch": 3.21, "learning_rate": 2.2360538982692803e-05, "loss": 0.3018, "step": 53224 }, { "epoch": 3.21, "learning_rate": 2.2359030072578576e-05, "loss": 0.3348, "step": 53226 }, { "epoch": 3.21, "learning_rate": 2.2357521162464353e-05, "loss": 0.3295, "step": 53228 }, { "epoch": 3.21, "learning_rate": 2.235601225235013e-05, "loss": 0.4277, "step": 53230 }, { "epoch": 3.21, "learning_rate": 2.2354503342235906e-05, "loss": 0.4364, "step": 53232 }, { "epoch": 3.21, "learning_rate": 2.235299443212168e-05, "loss": 0.2937, "step": 53234 }, { "epoch": 3.21, "learning_rate": 2.2351485522007452e-05, "loss": 0.4107, "step": 53236 }, { "epoch": 3.21, "learning_rate": 2.2349976611893232e-05, "loss": 0.3862, "step": 53238 }, { "epoch": 3.21, "learning_rate": 2.2348467701779005e-05, "loss": 0.3207, "step": 53240 }, { "epoch": 3.21, "learning_rate": 2.2346958791664782e-05, "loss": 0.3448, "step": 53242 }, { "epoch": 3.21, "learning_rate": 2.2345449881550555e-05, "loss": 0.4932, "step": 53244 }, { "epoch": 3.21, "learning_rate": 2.2343940971436335e-05, "loss": 0.4407, "step": 53246 }, { "epoch": 3.21, "learning_rate": 2.2342432061322108e-05, "loss": 0.468, "step": 53248 }, { "epoch": 3.21, "learning_rate": 2.2340923151207884e-05, "loss": 0.3755, "step": 53250 }, { "epoch": 3.21, "learning_rate": 2.2339414241093658e-05, "loss": 0.3612, "step": 53252 }, { "epoch": 3.21, "learning_rate": 2.2337905330979434e-05, "loss": 0.3178, "step": 53254 }, { "epoch": 3.21, "learning_rate": 2.233639642086521e-05, "loss": 0.3265, "step": 53256 }, { "epoch": 3.21, "learning_rate": 2.2334887510750984e-05, "loss": 0.5222, "step": 53258 }, { "epoch": 3.21, "learning_rate": 2.233337860063676e-05, "loss": 0.3085, "step": 53260 }, { "epoch": 3.21, "learning_rate": 2.2331869690522537e-05, "loss": 0.313, "step": 53262 }, { "epoch": 3.21, "learning_rate": 2.2330360780408313e-05, "loss": 0.4659, "step": 53264 }, { "epoch": 3.21, "learning_rate": 2.2328851870294086e-05, "loss": 0.393, "step": 53266 }, { "epoch": 3.22, "learning_rate": 2.2327342960179863e-05, "loss": 0.4159, "step": 53268 }, { "epoch": 3.22, "learning_rate": 2.232583405006564e-05, "loss": 0.3617, "step": 53270 }, { "epoch": 3.22, "learning_rate": 2.2324325139951412e-05, "loss": 0.3888, "step": 53272 }, { "epoch": 3.22, "learning_rate": 2.232281622983719e-05, "loss": 0.3654, "step": 53274 }, { "epoch": 3.22, "learning_rate": 2.2321307319722966e-05, "loss": 0.5141, "step": 53276 }, { "epoch": 3.22, "learning_rate": 2.2319798409608742e-05, "loss": 0.3798, "step": 53278 }, { "epoch": 3.22, "learning_rate": 2.2318289499494515e-05, "loss": 0.3667, "step": 53280 }, { "epoch": 3.22, "learning_rate": 2.231678058938029e-05, "loss": 0.3667, "step": 53282 }, { "epoch": 3.22, "learning_rate": 2.2315271679266068e-05, "loss": 0.3655, "step": 53284 }, { "epoch": 3.22, "learning_rate": 2.231376276915184e-05, "loss": 0.4478, "step": 53286 }, { "epoch": 3.22, "learning_rate": 2.2312253859037618e-05, "loss": 0.3062, "step": 53288 }, { "epoch": 3.22, "learning_rate": 2.2310744948923394e-05, "loss": 0.4433, "step": 53290 }, { "epoch": 3.22, "learning_rate": 2.230923603880917e-05, "loss": 0.3299, "step": 53292 }, { "epoch": 3.22, "learning_rate": 2.2307727128694944e-05, "loss": 0.3889, "step": 53294 }, { "epoch": 3.22, "learning_rate": 2.230621821858072e-05, "loss": 0.314, "step": 53296 }, { "epoch": 3.22, "learning_rate": 2.2304709308466497e-05, "loss": 0.4214, "step": 53298 }, { "epoch": 3.22, "learning_rate": 2.2303200398352274e-05, "loss": 0.338, "step": 53300 }, { "epoch": 3.22, "learning_rate": 2.2301691488238047e-05, "loss": 0.6267, "step": 53302 }, { "epoch": 3.22, "learning_rate": 2.230018257812382e-05, "loss": 0.4336, "step": 53304 }, { "epoch": 3.22, "learning_rate": 2.22986736680096e-05, "loss": 0.264, "step": 53306 }, { "epoch": 3.22, "learning_rate": 2.2297164757895373e-05, "loss": 0.3564, "step": 53308 }, { "epoch": 3.22, "learning_rate": 2.229565584778115e-05, "loss": 0.2459, "step": 53310 }, { "epoch": 3.22, "learning_rate": 2.2294146937666922e-05, "loss": 0.4761, "step": 53312 }, { "epoch": 3.22, "learning_rate": 2.2292638027552702e-05, "loss": 0.4703, "step": 53314 }, { "epoch": 3.22, "learning_rate": 2.2291129117438475e-05, "loss": 0.4967, "step": 53316 }, { "epoch": 3.22, "learning_rate": 2.228962020732425e-05, "loss": 0.3144, "step": 53318 }, { "epoch": 3.22, "learning_rate": 2.2288111297210025e-05, "loss": 0.4354, "step": 53320 }, { "epoch": 3.22, "learning_rate": 2.22866023870958e-05, "loss": 0.2532, "step": 53322 }, { "epoch": 3.22, "learning_rate": 2.2285093476981578e-05, "loss": 0.3265, "step": 53324 }, { "epoch": 3.22, "learning_rate": 2.228358456686735e-05, "loss": 0.3264, "step": 53326 }, { "epoch": 3.22, "learning_rate": 2.2282075656753128e-05, "loss": 0.4947, "step": 53328 }, { "epoch": 3.22, "learning_rate": 2.2280566746638904e-05, "loss": 0.4349, "step": 53330 }, { "epoch": 3.22, "learning_rate": 2.227905783652468e-05, "loss": 0.4268, "step": 53332 }, { "epoch": 3.22, "learning_rate": 2.2277548926410454e-05, "loss": 0.3193, "step": 53334 }, { "epoch": 3.22, "learning_rate": 2.227604001629623e-05, "loss": 0.3449, "step": 53336 }, { "epoch": 3.22, "learning_rate": 2.2274531106182007e-05, "loss": 0.4243, "step": 53338 }, { "epoch": 3.22, "learning_rate": 2.227302219606778e-05, "loss": 0.2829, "step": 53340 }, { "epoch": 3.22, "learning_rate": 2.2271513285953557e-05, "loss": 0.5441, "step": 53342 }, { "epoch": 3.22, "learning_rate": 2.2270004375839333e-05, "loss": 0.4106, "step": 53344 }, { "epoch": 3.22, "learning_rate": 2.226849546572511e-05, "loss": 0.3762, "step": 53346 }, { "epoch": 3.22, "learning_rate": 2.2266986555610883e-05, "loss": 0.4013, "step": 53348 }, { "epoch": 3.22, "learning_rate": 2.226547764549666e-05, "loss": 0.3854, "step": 53350 }, { "epoch": 3.22, "learning_rate": 2.2263968735382436e-05, "loss": 0.4237, "step": 53352 }, { "epoch": 3.22, "learning_rate": 2.226245982526821e-05, "loss": 0.2716, "step": 53354 }, { "epoch": 3.22, "learning_rate": 2.2260950915153985e-05, "loss": 0.3838, "step": 53356 }, { "epoch": 3.22, "learning_rate": 2.225944200503976e-05, "loss": 0.5049, "step": 53358 }, { "epoch": 3.22, "learning_rate": 2.225793309492554e-05, "loss": 0.3674, "step": 53360 }, { "epoch": 3.22, "learning_rate": 2.225642418481131e-05, "loss": 0.3479, "step": 53362 }, { "epoch": 3.22, "learning_rate": 2.2254915274697088e-05, "loss": 0.4979, "step": 53364 }, { "epoch": 3.22, "learning_rate": 2.225340636458286e-05, "loss": 0.3562, "step": 53366 }, { "epoch": 3.22, "learning_rate": 2.2251897454468638e-05, "loss": 0.3854, "step": 53368 }, { "epoch": 3.22, "learning_rate": 2.2250388544354414e-05, "loss": 0.362, "step": 53370 }, { "epoch": 3.22, "learning_rate": 2.2248879634240187e-05, "loss": 0.2529, "step": 53372 }, { "epoch": 3.22, "learning_rate": 2.2247370724125964e-05, "loss": 0.3869, "step": 53374 }, { "epoch": 3.22, "learning_rate": 2.224586181401174e-05, "loss": 0.5034, "step": 53376 }, { "epoch": 3.22, "learning_rate": 2.2244352903897517e-05, "loss": 0.4181, "step": 53378 }, { "epoch": 3.22, "learning_rate": 2.224284399378329e-05, "loss": 0.4402, "step": 53380 }, { "epoch": 3.22, "learning_rate": 2.2241335083669067e-05, "loss": 0.4521, "step": 53382 }, { "epoch": 3.22, "learning_rate": 2.2239826173554843e-05, "loss": 0.3985, "step": 53384 }, { "epoch": 3.22, "learning_rate": 2.2238317263440616e-05, "loss": 0.3363, "step": 53386 }, { "epoch": 3.22, "learning_rate": 2.2236808353326393e-05, "loss": 0.4696, "step": 53388 }, { "epoch": 3.22, "learning_rate": 2.223529944321217e-05, "loss": 0.4313, "step": 53390 }, { "epoch": 3.22, "learning_rate": 2.2233790533097946e-05, "loss": 0.3223, "step": 53392 }, { "epoch": 3.22, "learning_rate": 2.223228162298372e-05, "loss": 0.4017, "step": 53394 }, { "epoch": 3.22, "learning_rate": 2.2230772712869495e-05, "loss": 0.3829, "step": 53396 }, { "epoch": 3.22, "learning_rate": 2.2229263802755272e-05, "loss": 0.3774, "step": 53398 }, { "epoch": 3.22, "learning_rate": 2.222775489264105e-05, "loss": 0.3901, "step": 53400 }, { "epoch": 3.22, "learning_rate": 2.222624598252682e-05, "loss": 0.3976, "step": 53402 }, { "epoch": 3.22, "learning_rate": 2.2224737072412598e-05, "loss": 0.4023, "step": 53404 }, { "epoch": 3.22, "learning_rate": 2.2223228162298375e-05, "loss": 0.427, "step": 53406 }, { "epoch": 3.22, "learning_rate": 2.2221719252184148e-05, "loss": 0.3839, "step": 53408 }, { "epoch": 3.22, "learning_rate": 2.2220210342069924e-05, "loss": 0.3736, "step": 53410 }, { "epoch": 3.22, "learning_rate": 2.22187014319557e-05, "loss": 0.2405, "step": 53412 }, { "epoch": 3.22, "learning_rate": 2.2217192521841477e-05, "loss": 0.349, "step": 53414 }, { "epoch": 3.22, "learning_rate": 2.221568361172725e-05, "loss": 0.6134, "step": 53416 }, { "epoch": 3.22, "learning_rate": 2.2214174701613023e-05, "loss": 0.4519, "step": 53418 }, { "epoch": 3.22, "learning_rate": 2.2212665791498803e-05, "loss": 0.3718, "step": 53420 }, { "epoch": 3.22, "learning_rate": 2.2211156881384576e-05, "loss": 0.3793, "step": 53422 }, { "epoch": 3.22, "learning_rate": 2.2209647971270353e-05, "loss": 0.442, "step": 53424 }, { "epoch": 3.22, "learning_rate": 2.2208139061156126e-05, "loss": 0.2826, "step": 53426 }, { "epoch": 3.22, "learning_rate": 2.2206630151041906e-05, "loss": 0.3644, "step": 53428 }, { "epoch": 3.22, "learning_rate": 2.220512124092768e-05, "loss": 0.2587, "step": 53430 }, { "epoch": 3.23, "learning_rate": 2.2203612330813456e-05, "loss": 0.3132, "step": 53432 }, { "epoch": 3.23, "learning_rate": 2.220210342069923e-05, "loss": 0.2587, "step": 53434 }, { "epoch": 3.23, "learning_rate": 2.2200594510585005e-05, "loss": 0.4305, "step": 53436 }, { "epoch": 3.23, "learning_rate": 2.2199085600470782e-05, "loss": 0.4507, "step": 53438 }, { "epoch": 3.23, "learning_rate": 2.2197576690356555e-05, "loss": 0.3287, "step": 53440 }, { "epoch": 3.23, "learning_rate": 2.219606778024233e-05, "loss": 0.5632, "step": 53442 }, { "epoch": 3.23, "learning_rate": 2.2194558870128108e-05, "loss": 0.3044, "step": 53444 }, { "epoch": 3.23, "learning_rate": 2.2193049960013884e-05, "loss": 0.2783, "step": 53446 }, { "epoch": 3.23, "learning_rate": 2.2191541049899658e-05, "loss": 0.3829, "step": 53448 }, { "epoch": 3.23, "learning_rate": 2.2190032139785434e-05, "loss": 0.3633, "step": 53450 }, { "epoch": 3.23, "learning_rate": 2.218852322967121e-05, "loss": 0.3746, "step": 53452 }, { "epoch": 3.23, "learning_rate": 2.2187014319556984e-05, "loss": 0.5159, "step": 53454 }, { "epoch": 3.23, "learning_rate": 2.218550540944276e-05, "loss": 0.3243, "step": 53456 }, { "epoch": 3.23, "learning_rate": 2.2183996499328537e-05, "loss": 0.3079, "step": 53458 }, { "epoch": 3.23, "learning_rate": 2.2182487589214313e-05, "loss": 0.4519, "step": 53460 }, { "epoch": 3.23, "learning_rate": 2.2180978679100086e-05, "loss": 0.4982, "step": 53462 }, { "epoch": 3.23, "learning_rate": 2.2179469768985863e-05, "loss": 0.3455, "step": 53464 }, { "epoch": 3.23, "learning_rate": 2.217796085887164e-05, "loss": 0.4217, "step": 53466 }, { "epoch": 3.23, "learning_rate": 2.2176451948757413e-05, "loss": 0.4757, "step": 53468 }, { "epoch": 3.23, "learning_rate": 2.217494303864319e-05, "loss": 0.3536, "step": 53470 }, { "epoch": 3.23, "learning_rate": 2.2173434128528962e-05, "loss": 0.2753, "step": 53472 }, { "epoch": 3.23, "learning_rate": 2.2171925218414742e-05, "loss": 0.4554, "step": 53474 }, { "epoch": 3.23, "learning_rate": 2.2170416308300515e-05, "loss": 0.4353, "step": 53476 }, { "epoch": 3.23, "learning_rate": 2.2168907398186292e-05, "loss": 0.343, "step": 53478 }, { "epoch": 3.23, "learning_rate": 2.2167398488072065e-05, "loss": 0.3819, "step": 53480 }, { "epoch": 3.23, "learning_rate": 2.2165889577957845e-05, "loss": 0.2842, "step": 53482 }, { "epoch": 3.23, "learning_rate": 2.2164380667843618e-05, "loss": 0.4594, "step": 53484 }, { "epoch": 3.23, "learning_rate": 2.216287175772939e-05, "loss": 0.295, "step": 53486 }, { "epoch": 3.23, "learning_rate": 2.2161362847615168e-05, "loss": 0.4277, "step": 53488 }, { "epoch": 3.23, "learning_rate": 2.2159853937500944e-05, "loss": 0.3569, "step": 53490 }, { "epoch": 3.23, "learning_rate": 2.215834502738672e-05, "loss": 0.3197, "step": 53492 }, { "epoch": 3.23, "learning_rate": 2.2156836117272494e-05, "loss": 0.4662, "step": 53494 }, { "epoch": 3.23, "learning_rate": 2.215532720715827e-05, "loss": 0.3812, "step": 53496 }, { "epoch": 3.23, "learning_rate": 2.2153818297044047e-05, "loss": 0.4988, "step": 53498 }, { "epoch": 3.23, "learning_rate": 2.215230938692982e-05, "loss": 0.298, "step": 53500 }, { "epoch": 3.23, "learning_rate": 2.2150800476815596e-05, "loss": 0.4483, "step": 53502 }, { "epoch": 3.23, "learning_rate": 2.2149291566701373e-05, "loss": 0.4151, "step": 53504 }, { "epoch": 3.23, "learning_rate": 2.214778265658715e-05, "loss": 0.4353, "step": 53506 }, { "epoch": 3.23, "learning_rate": 2.2146273746472923e-05, "loss": 0.4399, "step": 53508 }, { "epoch": 3.23, "learning_rate": 2.21447648363587e-05, "loss": 0.238, "step": 53510 }, { "epoch": 3.23, "learning_rate": 2.2143255926244476e-05, "loss": 0.5934, "step": 53512 }, { "epoch": 3.23, "learning_rate": 2.2141747016130252e-05, "loss": 0.394, "step": 53514 }, { "epoch": 3.23, "learning_rate": 2.2140238106016025e-05, "loss": 0.6499, "step": 53516 }, { "epoch": 3.23, "learning_rate": 2.2138729195901802e-05, "loss": 0.4315, "step": 53518 }, { "epoch": 3.23, "learning_rate": 2.2137220285787578e-05, "loss": 0.2198, "step": 53520 }, { "epoch": 3.23, "learning_rate": 2.213571137567335e-05, "loss": 0.3558, "step": 53522 }, { "epoch": 3.23, "learning_rate": 2.2134202465559128e-05, "loss": 0.3989, "step": 53524 }, { "epoch": 3.23, "learning_rate": 2.2132693555444904e-05, "loss": 0.3377, "step": 53526 }, { "epoch": 3.23, "learning_rate": 2.213118464533068e-05, "loss": 0.3318, "step": 53528 }, { "epoch": 3.23, "learning_rate": 2.2129675735216454e-05, "loss": 0.4574, "step": 53530 }, { "epoch": 3.23, "learning_rate": 2.2128166825102227e-05, "loss": 0.2685, "step": 53532 }, { "epoch": 3.23, "learning_rate": 2.2126657914988007e-05, "loss": 0.4094, "step": 53534 }, { "epoch": 3.23, "learning_rate": 2.212514900487378e-05, "loss": 0.3849, "step": 53536 }, { "epoch": 3.23, "learning_rate": 2.2123640094759557e-05, "loss": 0.3086, "step": 53538 }, { "epoch": 3.23, "learning_rate": 2.212213118464533e-05, "loss": 0.3986, "step": 53540 }, { "epoch": 3.23, "learning_rate": 2.212062227453111e-05, "loss": 0.4723, "step": 53542 }, { "epoch": 3.23, "learning_rate": 2.2119113364416883e-05, "loss": 0.3447, "step": 53544 }, { "epoch": 3.23, "learning_rate": 2.211760445430266e-05, "loss": 0.3518, "step": 53546 }, { "epoch": 3.23, "learning_rate": 2.2116095544188432e-05, "loss": 0.454, "step": 53548 }, { "epoch": 3.23, "learning_rate": 2.211458663407421e-05, "loss": 0.4261, "step": 53550 }, { "epoch": 3.23, "learning_rate": 2.2113077723959985e-05, "loss": 0.2725, "step": 53552 }, { "epoch": 3.23, "learning_rate": 2.211156881384576e-05, "loss": 0.4853, "step": 53554 }, { "epoch": 3.23, "learning_rate": 2.2110059903731535e-05, "loss": 0.4271, "step": 53556 }, { "epoch": 3.23, "learning_rate": 2.210855099361731e-05, "loss": 0.4621, "step": 53558 }, { "epoch": 3.23, "learning_rate": 2.2107042083503088e-05, "loss": 0.4019, "step": 53560 }, { "epoch": 3.23, "learning_rate": 2.210553317338886e-05, "loss": 0.4283, "step": 53562 }, { "epoch": 3.23, "learning_rate": 2.2104024263274638e-05, "loss": 0.3963, "step": 53564 }, { "epoch": 3.23, "learning_rate": 2.2102515353160414e-05, "loss": 0.413, "step": 53566 }, { "epoch": 3.23, "learning_rate": 2.2101006443046187e-05, "loss": 0.328, "step": 53568 }, { "epoch": 3.23, "learning_rate": 2.2099497532931964e-05, "loss": 0.4322, "step": 53570 }, { "epoch": 3.23, "learning_rate": 2.209798862281774e-05, "loss": 0.2541, "step": 53572 }, { "epoch": 3.23, "learning_rate": 2.2096479712703517e-05, "loss": 0.5886, "step": 53574 }, { "epoch": 3.23, "learning_rate": 2.209497080258929e-05, "loss": 0.3657, "step": 53576 }, { "epoch": 3.23, "learning_rate": 2.2093461892475067e-05, "loss": 0.2941, "step": 53578 }, { "epoch": 3.23, "learning_rate": 2.2091952982360843e-05, "loss": 0.441, "step": 53580 }, { "epoch": 3.23, "learning_rate": 2.2090444072246616e-05, "loss": 0.5135, "step": 53582 }, { "epoch": 3.23, "learning_rate": 2.2088935162132393e-05, "loss": 0.4412, "step": 53584 }, { "epoch": 3.23, "learning_rate": 2.2087426252018166e-05, "loss": 0.4034, "step": 53586 }, { "epoch": 3.23, "learning_rate": 2.2085917341903946e-05, "loss": 0.3735, "step": 53588 }, { "epoch": 3.23, "learning_rate": 2.208440843178972e-05, "loss": 0.3648, "step": 53590 }, { "epoch": 3.23, "learning_rate": 2.2082899521675495e-05, "loss": 0.3671, "step": 53592 }, { "epoch": 3.23, "learning_rate": 2.208139061156127e-05, "loss": 0.4519, "step": 53594 }, { "epoch": 3.23, "learning_rate": 2.207988170144705e-05, "loss": 0.5946, "step": 53596 }, { "epoch": 3.24, "learning_rate": 2.207837279133282e-05, "loss": 0.7582, "step": 53598 }, { "epoch": 3.24, "learning_rate": 2.2076863881218595e-05, "loss": 0.3165, "step": 53600 }, { "epoch": 3.24, "learning_rate": 2.207535497110437e-05, "loss": 0.3544, "step": 53602 }, { "epoch": 3.24, "learning_rate": 2.2073846060990148e-05, "loss": 0.4281, "step": 53604 }, { "epoch": 3.24, "learning_rate": 2.2072337150875924e-05, "loss": 0.4357, "step": 53606 }, { "epoch": 3.24, "learning_rate": 2.2070828240761697e-05, "loss": 0.5014, "step": 53608 }, { "epoch": 3.24, "learning_rate": 2.2069319330647474e-05, "loss": 0.4018, "step": 53610 }, { "epoch": 3.24, "learning_rate": 2.206781042053325e-05, "loss": 0.335, "step": 53612 }, { "epoch": 3.24, "learning_rate": 2.2066301510419024e-05, "loss": 0.5718, "step": 53614 }, { "epoch": 3.24, "learning_rate": 2.20647926003048e-05, "loss": 0.394, "step": 53616 }, { "epoch": 3.24, "learning_rate": 2.2063283690190577e-05, "loss": 0.3947, "step": 53618 }, { "epoch": 3.24, "learning_rate": 2.2061774780076353e-05, "loss": 0.4073, "step": 53620 }, { "epoch": 3.24, "learning_rate": 2.2060265869962126e-05, "loss": 0.2852, "step": 53622 }, { "epoch": 3.24, "learning_rate": 2.2058756959847903e-05, "loss": 0.5736, "step": 53624 }, { "epoch": 3.24, "learning_rate": 2.205724804973368e-05, "loss": 0.3351, "step": 53626 }, { "epoch": 3.24, "learning_rate": 2.2055739139619456e-05, "loss": 0.4186, "step": 53628 }, { "epoch": 3.24, "learning_rate": 2.205423022950523e-05, "loss": 0.3038, "step": 53630 }, { "epoch": 3.24, "learning_rate": 2.2052721319391005e-05, "loss": 0.2502, "step": 53632 }, { "epoch": 3.24, "learning_rate": 2.2051212409276782e-05, "loss": 0.4547, "step": 53634 }, { "epoch": 3.24, "learning_rate": 2.2049703499162555e-05, "loss": 0.3641, "step": 53636 }, { "epoch": 3.24, "learning_rate": 2.204819458904833e-05, "loss": 0.4089, "step": 53638 }, { "epoch": 3.24, "learning_rate": 2.2046685678934108e-05, "loss": 0.2786, "step": 53640 }, { "epoch": 3.24, "learning_rate": 2.2045176768819885e-05, "loss": 0.3496, "step": 53642 }, { "epoch": 3.24, "learning_rate": 2.2043667858705658e-05, "loss": 0.4668, "step": 53644 }, { "epoch": 3.24, "learning_rate": 2.2042158948591434e-05, "loss": 0.4631, "step": 53646 }, { "epoch": 3.24, "learning_rate": 2.204065003847721e-05, "loss": 0.3159, "step": 53648 }, { "epoch": 3.24, "learning_rate": 2.2039141128362984e-05, "loss": 0.2843, "step": 53650 }, { "epoch": 3.24, "learning_rate": 2.203763221824876e-05, "loss": 0.3567, "step": 53652 }, { "epoch": 3.24, "learning_rate": 2.2036123308134533e-05, "loss": 0.3555, "step": 53654 }, { "epoch": 3.24, "learning_rate": 2.2034614398020313e-05, "loss": 0.2783, "step": 53656 }, { "epoch": 3.24, "learning_rate": 2.2033105487906087e-05, "loss": 0.3175, "step": 53658 }, { "epoch": 3.24, "learning_rate": 2.2031596577791863e-05, "loss": 0.4138, "step": 53660 }, { "epoch": 3.24, "learning_rate": 2.2030087667677636e-05, "loss": 0.3374, "step": 53662 }, { "epoch": 3.24, "learning_rate": 2.2028578757563413e-05, "loss": 0.5157, "step": 53664 }, { "epoch": 3.24, "learning_rate": 2.202706984744919e-05, "loss": 0.3738, "step": 53666 }, { "epoch": 3.24, "learning_rate": 2.2025560937334962e-05, "loss": 0.3423, "step": 53668 }, { "epoch": 3.24, "learning_rate": 2.202405202722074e-05, "loss": 0.495, "step": 53670 }, { "epoch": 3.24, "learning_rate": 2.2022543117106515e-05, "loss": 0.5068, "step": 53672 }, { "epoch": 3.24, "learning_rate": 2.2021034206992292e-05, "loss": 0.6284, "step": 53674 }, { "epoch": 3.24, "learning_rate": 2.2019525296878065e-05, "loss": 0.3973, "step": 53676 }, { "epoch": 3.24, "learning_rate": 2.201801638676384e-05, "loss": 0.3561, "step": 53678 }, { "epoch": 3.24, "learning_rate": 2.2016507476649618e-05, "loss": 0.4493, "step": 53680 }, { "epoch": 3.24, "learning_rate": 2.201499856653539e-05, "loss": 0.3696, "step": 53682 }, { "epoch": 3.24, "learning_rate": 2.2013489656421168e-05, "loss": 0.3209, "step": 53684 }, { "epoch": 3.24, "learning_rate": 2.2011980746306944e-05, "loss": 0.3935, "step": 53686 }, { "epoch": 3.24, "learning_rate": 2.201047183619272e-05, "loss": 0.3554, "step": 53688 }, { "epoch": 3.24, "learning_rate": 2.2008962926078494e-05, "loss": 0.4489, "step": 53690 }, { "epoch": 3.24, "learning_rate": 2.200745401596427e-05, "loss": 0.3414, "step": 53692 }, { "epoch": 3.24, "learning_rate": 2.2005945105850047e-05, "loss": 0.3592, "step": 53694 }, { "epoch": 3.24, "learning_rate": 2.200443619573582e-05, "loss": 0.3163, "step": 53696 }, { "epoch": 3.24, "learning_rate": 2.2002927285621596e-05, "loss": 0.4868, "step": 53698 }, { "epoch": 3.24, "learning_rate": 2.200141837550737e-05, "loss": 0.3059, "step": 53700 }, { "epoch": 3.24, "learning_rate": 2.199990946539315e-05, "loss": 0.321, "step": 53702 }, { "epoch": 3.24, "learning_rate": 2.1998400555278923e-05, "loss": 0.2999, "step": 53704 }, { "epoch": 3.24, "learning_rate": 2.19968916451647e-05, "loss": 0.5457, "step": 53706 }, { "epoch": 3.24, "learning_rate": 2.1995382735050472e-05, "loss": 0.4134, "step": 53708 }, { "epoch": 3.24, "learning_rate": 2.1993873824936252e-05, "loss": 0.4052, "step": 53710 }, { "epoch": 3.24, "learning_rate": 2.1992364914822025e-05, "loss": 0.446, "step": 53712 }, { "epoch": 3.24, "learning_rate": 2.19908560047078e-05, "loss": 0.4151, "step": 53714 }, { "epoch": 3.24, "learning_rate": 2.1989347094593575e-05, "loss": 0.3093, "step": 53716 }, { "epoch": 3.24, "learning_rate": 2.198783818447935e-05, "loss": 0.3291, "step": 53718 }, { "epoch": 3.24, "learning_rate": 2.1986329274365128e-05, "loss": 0.6149, "step": 53720 }, { "epoch": 3.24, "learning_rate": 2.19848203642509e-05, "loss": 0.3168, "step": 53722 }, { "epoch": 3.24, "learning_rate": 2.1983311454136678e-05, "loss": 0.5147, "step": 53724 }, { "epoch": 3.24, "learning_rate": 2.1981802544022454e-05, "loss": 0.3792, "step": 53726 }, { "epoch": 3.24, "learning_rate": 2.198029363390823e-05, "loss": 0.3834, "step": 53728 }, { "epoch": 3.24, "learning_rate": 2.1978784723794004e-05, "loss": 0.4366, "step": 53730 }, { "epoch": 3.24, "learning_rate": 2.197727581367978e-05, "loss": 0.3361, "step": 53732 }, { "epoch": 3.24, "learning_rate": 2.1975766903565557e-05, "loss": 0.2844, "step": 53734 }, { "epoch": 3.24, "learning_rate": 2.197425799345133e-05, "loss": 0.4028, "step": 53736 }, { "epoch": 3.24, "learning_rate": 2.1972749083337106e-05, "loss": 0.3764, "step": 53738 }, { "epoch": 3.24, "learning_rate": 2.1971240173222883e-05, "loss": 0.4787, "step": 53740 }, { "epoch": 3.24, "learning_rate": 2.196973126310866e-05, "loss": 0.4643, "step": 53742 }, { "epoch": 3.24, "learning_rate": 2.1968222352994433e-05, "loss": 0.4159, "step": 53744 }, { "epoch": 3.24, "learning_rate": 2.196671344288021e-05, "loss": 0.4173, "step": 53746 }, { "epoch": 3.24, "learning_rate": 2.1965204532765986e-05, "loss": 0.401, "step": 53748 }, { "epoch": 3.24, "learning_rate": 2.196369562265176e-05, "loss": 0.299, "step": 53750 }, { "epoch": 3.24, "learning_rate": 2.1962186712537535e-05, "loss": 0.4122, "step": 53752 }, { "epoch": 3.24, "learning_rate": 2.1960677802423312e-05, "loss": 0.5044, "step": 53754 }, { "epoch": 3.24, "learning_rate": 2.1959168892309088e-05, "loss": 0.3569, "step": 53756 }, { "epoch": 3.24, "learning_rate": 2.195765998219486e-05, "loss": 0.3795, "step": 53758 }, { "epoch": 3.24, "learning_rate": 2.1956151072080638e-05, "loss": 0.3618, "step": 53760 }, { "epoch": 3.24, "learning_rate": 2.1954642161966414e-05, "loss": 0.2857, "step": 53762 }, { "epoch": 3.25, "learning_rate": 2.1953133251852188e-05, "loss": 0.4351, "step": 53764 }, { "epoch": 3.25, "learning_rate": 2.1951624341737964e-05, "loss": 0.3448, "step": 53766 }, { "epoch": 3.25, "learning_rate": 2.1950115431623737e-05, "loss": 0.3634, "step": 53768 }, { "epoch": 3.25, "learning_rate": 2.1948606521509517e-05, "loss": 0.4175, "step": 53770 }, { "epoch": 3.25, "learning_rate": 2.194709761139529e-05, "loss": 0.4596, "step": 53772 }, { "epoch": 3.25, "learning_rate": 2.1945588701281067e-05, "loss": 0.3082, "step": 53774 }, { "epoch": 3.25, "learning_rate": 2.194407979116684e-05, "loss": 0.303, "step": 53776 }, { "epoch": 3.25, "learning_rate": 2.1942570881052616e-05, "loss": 0.3103, "step": 53778 }, { "epoch": 3.25, "learning_rate": 2.1941061970938393e-05, "loss": 0.4941, "step": 53780 }, { "epoch": 3.25, "learning_rate": 2.1939553060824166e-05, "loss": 0.4465, "step": 53782 }, { "epoch": 3.25, "learning_rate": 2.1938044150709942e-05, "loss": 0.2929, "step": 53784 }, { "epoch": 3.25, "learning_rate": 2.193653524059572e-05, "loss": 0.3741, "step": 53786 }, { "epoch": 3.25, "learning_rate": 2.1935026330481496e-05, "loss": 0.3682, "step": 53788 }, { "epoch": 3.25, "learning_rate": 2.193351742036727e-05, "loss": 0.56, "step": 53790 }, { "epoch": 3.25, "learning_rate": 2.1932008510253045e-05, "loss": 0.3239, "step": 53792 }, { "epoch": 3.25, "learning_rate": 2.193049960013882e-05, "loss": 0.4599, "step": 53794 }, { "epoch": 3.25, "learning_rate": 2.1928990690024595e-05, "loss": 0.3737, "step": 53796 }, { "epoch": 3.25, "learning_rate": 2.192748177991037e-05, "loss": 0.5106, "step": 53798 }, { "epoch": 3.25, "learning_rate": 2.1925972869796148e-05, "loss": 0.3162, "step": 53800 }, { "epoch": 3.25, "learning_rate": 2.1924463959681924e-05, "loss": 0.3377, "step": 53802 }, { "epoch": 3.25, "learning_rate": 2.1922955049567697e-05, "loss": 0.3258, "step": 53804 }, { "epoch": 3.25, "learning_rate": 2.1921446139453474e-05, "loss": 0.4187, "step": 53806 }, { "epoch": 3.25, "learning_rate": 2.191993722933925e-05, "loss": 0.2162, "step": 53808 }, { "epoch": 3.25, "learning_rate": 2.1918428319225027e-05, "loss": 0.4397, "step": 53810 }, { "epoch": 3.25, "learning_rate": 2.19169194091108e-05, "loss": 0.3032, "step": 53812 }, { "epoch": 3.25, "learning_rate": 2.1915410498996573e-05, "loss": 0.3955, "step": 53814 }, { "epoch": 3.25, "learning_rate": 2.1913901588882353e-05, "loss": 0.2973, "step": 53816 }, { "epoch": 3.25, "learning_rate": 2.1912392678768126e-05, "loss": 0.3325, "step": 53818 }, { "epoch": 3.25, "learning_rate": 2.1910883768653903e-05, "loss": 0.4157, "step": 53820 }, { "epoch": 3.25, "learning_rate": 2.1909374858539676e-05, "loss": 0.3275, "step": 53822 }, { "epoch": 3.25, "learning_rate": 2.1907865948425456e-05, "loss": 0.4141, "step": 53824 }, { "epoch": 3.25, "learning_rate": 2.190635703831123e-05, "loss": 0.3468, "step": 53826 }, { "epoch": 3.25, "learning_rate": 2.1904848128197002e-05, "loss": 0.422, "step": 53828 }, { "epoch": 3.25, "learning_rate": 2.190333921808278e-05, "loss": 0.3356, "step": 53830 }, { "epoch": 3.25, "learning_rate": 2.1901830307968555e-05, "loss": 0.4102, "step": 53832 }, { "epoch": 3.25, "learning_rate": 2.190032139785433e-05, "loss": 0.3023, "step": 53834 }, { "epoch": 3.25, "learning_rate": 2.1898812487740105e-05, "loss": 0.3272, "step": 53836 }, { "epoch": 3.25, "learning_rate": 2.189730357762588e-05, "loss": 0.2662, "step": 53838 }, { "epoch": 3.25, "learning_rate": 2.1895794667511658e-05, "loss": 0.3462, "step": 53840 }, { "epoch": 3.25, "learning_rate": 2.1894285757397434e-05, "loss": 0.4279, "step": 53842 }, { "epoch": 3.25, "learning_rate": 2.1892776847283207e-05, "loss": 0.6422, "step": 53844 }, { "epoch": 3.25, "learning_rate": 2.1891267937168984e-05, "loss": 0.3444, "step": 53846 }, { "epoch": 3.25, "learning_rate": 2.188975902705476e-05, "loss": 0.2647, "step": 53848 }, { "epoch": 3.25, "learning_rate": 2.1888250116940534e-05, "loss": 0.3041, "step": 53850 }, { "epoch": 3.25, "learning_rate": 2.188674120682631e-05, "loss": 0.2919, "step": 53852 }, { "epoch": 3.25, "learning_rate": 2.1885232296712087e-05, "loss": 0.4396, "step": 53854 }, { "epoch": 3.25, "learning_rate": 2.1883723386597863e-05, "loss": 0.4383, "step": 53856 }, { "epoch": 3.25, "learning_rate": 2.1882214476483636e-05, "loss": 0.3463, "step": 53858 }, { "epoch": 3.25, "learning_rate": 2.1880705566369413e-05, "loss": 0.5181, "step": 53860 }, { "epoch": 3.25, "learning_rate": 2.187919665625519e-05, "loss": 0.3336, "step": 53862 }, { "epoch": 3.25, "learning_rate": 2.1877687746140962e-05, "loss": 0.3445, "step": 53864 }, { "epoch": 3.25, "learning_rate": 2.187617883602674e-05, "loss": 0.2578, "step": 53866 }, { "epoch": 3.25, "learning_rate": 2.1874669925912515e-05, "loss": 0.4099, "step": 53868 }, { "epoch": 3.25, "learning_rate": 2.1873161015798292e-05, "loss": 0.3437, "step": 53870 }, { "epoch": 3.25, "learning_rate": 2.1871652105684065e-05, "loss": 0.4934, "step": 53872 }, { "epoch": 3.25, "learning_rate": 2.187014319556984e-05, "loss": 0.3983, "step": 53874 }, { "epoch": 3.25, "learning_rate": 2.1868634285455618e-05, "loss": 0.3274, "step": 53876 }, { "epoch": 3.25, "learning_rate": 2.186712537534139e-05, "loss": 0.4711, "step": 53878 }, { "epoch": 3.25, "learning_rate": 2.1865616465227168e-05, "loss": 0.2707, "step": 53880 }, { "epoch": 3.25, "learning_rate": 2.186410755511294e-05, "loss": 0.3486, "step": 53882 }, { "epoch": 3.25, "learning_rate": 2.186259864499872e-05, "loss": 0.2765, "step": 53884 }, { "epoch": 3.25, "learning_rate": 2.1861089734884494e-05, "loss": 0.3581, "step": 53886 }, { "epoch": 3.25, "learning_rate": 2.185958082477027e-05, "loss": 0.2599, "step": 53888 }, { "epoch": 3.25, "learning_rate": 2.1858071914656043e-05, "loss": 0.3517, "step": 53890 }, { "epoch": 3.25, "learning_rate": 2.1856563004541823e-05, "loss": 0.3526, "step": 53892 }, { "epoch": 3.25, "learning_rate": 2.1855054094427597e-05, "loss": 0.3747, "step": 53894 }, { "epoch": 3.25, "learning_rate": 2.185354518431337e-05, "loss": 0.3749, "step": 53896 }, { "epoch": 3.25, "learning_rate": 2.1852036274199146e-05, "loss": 0.3272, "step": 53898 }, { "epoch": 3.25, "learning_rate": 2.1850527364084923e-05, "loss": 0.4244, "step": 53900 }, { "epoch": 3.25, "learning_rate": 2.18490184539707e-05, "loss": 0.3826, "step": 53902 }, { "epoch": 3.25, "learning_rate": 2.1847509543856472e-05, "loss": 0.3313, "step": 53904 }, { "epoch": 3.25, "learning_rate": 2.184600063374225e-05, "loss": 0.3967, "step": 53906 }, { "epoch": 3.25, "learning_rate": 2.1844491723628025e-05, "loss": 0.3995, "step": 53908 }, { "epoch": 3.25, "learning_rate": 2.18429828135138e-05, "loss": 0.3748, "step": 53910 }, { "epoch": 3.25, "learning_rate": 2.1841473903399575e-05, "loss": 0.3249, "step": 53912 }, { "epoch": 3.25, "learning_rate": 2.183996499328535e-05, "loss": 0.4079, "step": 53914 }, { "epoch": 3.25, "learning_rate": 2.1838456083171128e-05, "loss": 0.4282, "step": 53916 }, { "epoch": 3.25, "learning_rate": 2.18369471730569e-05, "loss": 0.2622, "step": 53918 }, { "epoch": 3.25, "learning_rate": 2.1835438262942678e-05, "loss": 0.413, "step": 53920 }, { "epoch": 3.25, "learning_rate": 2.1833929352828454e-05, "loss": 0.4061, "step": 53922 }, { "epoch": 3.25, "learning_rate": 2.183242044271423e-05, "loss": 0.4867, "step": 53924 }, { "epoch": 3.25, "learning_rate": 2.1830911532600004e-05, "loss": 0.2821, "step": 53926 }, { "epoch": 3.25, "learning_rate": 2.1829402622485777e-05, "loss": 0.4475, "step": 53928 }, { "epoch": 3.26, "learning_rate": 2.1827893712371557e-05, "loss": 0.4265, "step": 53930 }, { "epoch": 3.26, "learning_rate": 2.182638480225733e-05, "loss": 0.391, "step": 53932 }, { "epoch": 3.26, "learning_rate": 2.1824875892143106e-05, "loss": 0.3962, "step": 53934 }, { "epoch": 3.26, "learning_rate": 2.182336698202888e-05, "loss": 0.3985, "step": 53936 }, { "epoch": 3.26, "learning_rate": 2.182185807191466e-05, "loss": 0.4488, "step": 53938 }, { "epoch": 3.26, "learning_rate": 2.1820349161800433e-05, "loss": 0.403, "step": 53940 }, { "epoch": 3.26, "learning_rate": 2.1818840251686206e-05, "loss": 0.3508, "step": 53942 }, { "epoch": 3.26, "learning_rate": 2.1817331341571982e-05, "loss": 0.3286, "step": 53944 }, { "epoch": 3.26, "learning_rate": 2.181582243145776e-05, "loss": 0.2415, "step": 53946 }, { "epoch": 3.26, "learning_rate": 2.1814313521343535e-05, "loss": 0.4336, "step": 53948 }, { "epoch": 3.26, "learning_rate": 2.181280461122931e-05, "loss": 0.4672, "step": 53950 }, { "epoch": 3.26, "learning_rate": 2.1811295701115085e-05, "loss": 0.431, "step": 53952 }, { "epoch": 3.26, "learning_rate": 2.180978679100086e-05, "loss": 0.5422, "step": 53954 }, { "epoch": 3.26, "learning_rate": 2.1808277880886638e-05, "loss": 0.3077, "step": 53956 }, { "epoch": 3.26, "learning_rate": 2.180676897077241e-05, "loss": 0.3013, "step": 53958 }, { "epoch": 3.26, "learning_rate": 2.1805260060658188e-05, "loss": 0.31, "step": 53960 }, { "epoch": 3.26, "learning_rate": 2.1803751150543964e-05, "loss": 0.4042, "step": 53962 }, { "epoch": 3.26, "learning_rate": 2.1802242240429737e-05, "loss": 0.5188, "step": 53964 }, { "epoch": 3.26, "learning_rate": 2.1800733330315514e-05, "loss": 0.4388, "step": 53966 }, { "epoch": 3.26, "learning_rate": 2.179922442020129e-05, "loss": 0.2812, "step": 53968 }, { "epoch": 3.26, "learning_rate": 2.1797715510087067e-05, "loss": 0.2805, "step": 53970 }, { "epoch": 3.26, "learning_rate": 2.179620659997284e-05, "loss": 0.3267, "step": 53972 }, { "epoch": 3.26, "learning_rate": 2.1794697689858616e-05, "loss": 0.5365, "step": 53974 }, { "epoch": 3.26, "learning_rate": 2.1793188779744393e-05, "loss": 0.5283, "step": 53976 }, { "epoch": 3.26, "learning_rate": 2.1791679869630166e-05, "loss": 0.4463, "step": 53978 }, { "epoch": 3.26, "learning_rate": 2.1790170959515943e-05, "loss": 0.2529, "step": 53980 }, { "epoch": 3.26, "learning_rate": 2.178866204940172e-05, "loss": 0.3907, "step": 53982 }, { "epoch": 3.26, "learning_rate": 2.1787153139287496e-05, "loss": 0.3874, "step": 53984 }, { "epoch": 3.26, "learning_rate": 2.178564422917327e-05, "loss": 0.3446, "step": 53986 }, { "epoch": 3.26, "learning_rate": 2.1784135319059045e-05, "loss": 0.3402, "step": 53988 }, { "epoch": 3.26, "learning_rate": 2.1782626408944822e-05, "loss": 0.3053, "step": 53990 }, { "epoch": 3.26, "learning_rate": 2.1781117498830595e-05, "loss": 0.2434, "step": 53992 }, { "epoch": 3.26, "learning_rate": 2.177960858871637e-05, "loss": 0.4883, "step": 53994 }, { "epoch": 3.26, "learning_rate": 2.1778099678602145e-05, "loss": 0.3481, "step": 53996 }, { "epoch": 3.26, "learning_rate": 2.1776590768487924e-05, "loss": 0.4586, "step": 53998 }, { "epoch": 3.26, "learning_rate": 2.1775081858373698e-05, "loss": 0.3459, "step": 54000 }, { "epoch": 3.26, "learning_rate": 2.1773572948259474e-05, "loss": 0.2303, "step": 54002 }, { "epoch": 3.26, "learning_rate": 2.1772064038145247e-05, "loss": 0.4437, "step": 54004 }, { "epoch": 3.26, "learning_rate": 2.1770555128031027e-05, "loss": 0.37, "step": 54006 }, { "epoch": 3.26, "learning_rate": 2.17690462179168e-05, "loss": 0.1908, "step": 54008 }, { "epoch": 3.26, "learning_rate": 2.1767537307802573e-05, "loss": 0.2623, "step": 54010 }, { "epoch": 3.26, "learning_rate": 2.176602839768835e-05, "loss": 0.2675, "step": 54012 }, { "epoch": 3.26, "learning_rate": 2.1764519487574126e-05, "loss": 0.4834, "step": 54014 }, { "epoch": 3.26, "learning_rate": 2.1763010577459903e-05, "loss": 0.3631, "step": 54016 }, { "epoch": 3.26, "learning_rate": 2.1761501667345676e-05, "loss": 0.3476, "step": 54018 }, { "epoch": 3.26, "learning_rate": 2.1759992757231453e-05, "loss": 0.5998, "step": 54020 }, { "epoch": 3.26, "learning_rate": 2.175848384711723e-05, "loss": 0.358, "step": 54022 }, { "epoch": 3.26, "learning_rate": 2.1756974937003006e-05, "loss": 0.2603, "step": 54024 }, { "epoch": 3.26, "learning_rate": 2.175546602688878e-05, "loss": 0.3364, "step": 54026 }, { "epoch": 3.26, "learning_rate": 2.1753957116774555e-05, "loss": 0.3594, "step": 54028 }, { "epoch": 3.26, "learning_rate": 2.175244820666033e-05, "loss": 0.4916, "step": 54030 }, { "epoch": 3.26, "learning_rate": 2.1750939296546105e-05, "loss": 0.2221, "step": 54032 }, { "epoch": 3.26, "learning_rate": 2.174943038643188e-05, "loss": 0.4635, "step": 54034 }, { "epoch": 3.26, "learning_rate": 2.1747921476317658e-05, "loss": 0.3061, "step": 54036 }, { "epoch": 3.26, "learning_rate": 2.1746412566203434e-05, "loss": 0.5703, "step": 54038 }, { "epoch": 3.26, "learning_rate": 2.1744903656089207e-05, "loss": 0.3169, "step": 54040 }, { "epoch": 3.26, "learning_rate": 2.174339474597498e-05, "loss": 0.4845, "step": 54042 }, { "epoch": 3.26, "learning_rate": 2.174188583586076e-05, "loss": 0.3913, "step": 54044 }, { "epoch": 3.26, "learning_rate": 2.1740376925746534e-05, "loss": 0.3583, "step": 54046 }, { "epoch": 3.26, "learning_rate": 2.173886801563231e-05, "loss": 0.2993, "step": 54048 }, { "epoch": 3.26, "learning_rate": 2.1737359105518083e-05, "loss": 0.4837, "step": 54050 }, { "epoch": 3.26, "learning_rate": 2.1735850195403863e-05, "loss": 0.3994, "step": 54052 }, { "epoch": 3.26, "learning_rate": 2.1734341285289636e-05, "loss": 0.3106, "step": 54054 }, { "epoch": 3.26, "learning_rate": 2.1732832375175413e-05, "loss": 0.4383, "step": 54056 }, { "epoch": 3.26, "learning_rate": 2.1731323465061186e-05, "loss": 0.3401, "step": 54058 }, { "epoch": 3.26, "learning_rate": 2.1729814554946962e-05, "loss": 0.5326, "step": 54060 }, { "epoch": 3.26, "learning_rate": 2.172830564483274e-05, "loss": 0.2837, "step": 54062 }, { "epoch": 3.26, "learning_rate": 2.1726796734718512e-05, "loss": 0.3535, "step": 54064 }, { "epoch": 3.26, "learning_rate": 2.172528782460429e-05, "loss": 0.6046, "step": 54066 }, { "epoch": 3.26, "learning_rate": 2.1723778914490065e-05, "loss": 0.5577, "step": 54068 }, { "epoch": 3.26, "learning_rate": 2.172227000437584e-05, "loss": 0.3477, "step": 54070 }, { "epoch": 3.26, "learning_rate": 2.1720761094261615e-05, "loss": 0.5235, "step": 54072 }, { "epoch": 3.26, "learning_rate": 2.171925218414739e-05, "loss": 0.4316, "step": 54074 }, { "epoch": 3.26, "learning_rate": 2.1717743274033168e-05, "loss": 0.3614, "step": 54076 }, { "epoch": 3.26, "learning_rate": 2.171623436391894e-05, "loss": 0.4222, "step": 54078 }, { "epoch": 3.26, "learning_rate": 2.1714725453804717e-05, "loss": 0.3553, "step": 54080 }, { "epoch": 3.26, "learning_rate": 2.1713216543690494e-05, "loss": 0.5059, "step": 54082 }, { "epoch": 3.26, "learning_rate": 2.171170763357627e-05, "loss": 0.287, "step": 54084 }, { "epoch": 3.26, "learning_rate": 2.1710198723462044e-05, "loss": 0.3817, "step": 54086 }, { "epoch": 3.26, "learning_rate": 2.170868981334782e-05, "loss": 0.37, "step": 54088 }, { "epoch": 3.26, "learning_rate": 2.1707180903233597e-05, "loss": 0.2749, "step": 54090 }, { "epoch": 3.26, "learning_rate": 2.170567199311937e-05, "loss": 0.4186, "step": 54092 }, { "epoch": 3.26, "learning_rate": 2.1704163083005146e-05, "loss": 0.4187, "step": 54094 }, { "epoch": 3.27, "learning_rate": 2.1702654172890923e-05, "loss": 0.2768, "step": 54096 }, { "epoch": 3.27, "learning_rate": 2.17011452627767e-05, "loss": 0.5078, "step": 54098 }, { "epoch": 3.27, "learning_rate": 2.1699636352662472e-05, "loss": 0.3779, "step": 54100 }, { "epoch": 3.27, "learning_rate": 2.169812744254825e-05, "loss": 0.3938, "step": 54102 }, { "epoch": 3.27, "learning_rate": 2.1696618532434025e-05, "loss": 0.331, "step": 54104 }, { "epoch": 3.27, "learning_rate": 2.1695109622319802e-05, "loss": 0.4777, "step": 54106 }, { "epoch": 3.27, "learning_rate": 2.1693600712205575e-05, "loss": 0.487, "step": 54108 }, { "epoch": 3.27, "learning_rate": 2.1692091802091348e-05, "loss": 0.3707, "step": 54110 }, { "epoch": 3.27, "learning_rate": 2.1690582891977128e-05, "loss": 0.2874, "step": 54112 }, { "epoch": 3.27, "learning_rate": 2.16890739818629e-05, "loss": 0.3116, "step": 54114 }, { "epoch": 3.27, "learning_rate": 2.1687565071748678e-05, "loss": 0.4366, "step": 54116 }, { "epoch": 3.27, "learning_rate": 2.168605616163445e-05, "loss": 0.3003, "step": 54118 }, { "epoch": 3.27, "learning_rate": 2.168454725152023e-05, "loss": 0.4332, "step": 54120 }, { "epoch": 3.27, "learning_rate": 2.1683038341406004e-05, "loss": 0.4531, "step": 54122 }, { "epoch": 3.27, "learning_rate": 2.1681529431291777e-05, "loss": 0.4295, "step": 54124 }, { "epoch": 3.27, "learning_rate": 2.1680020521177554e-05, "loss": 0.6586, "step": 54126 }, { "epoch": 3.27, "learning_rate": 2.167851161106333e-05, "loss": 0.553, "step": 54128 }, { "epoch": 3.27, "learning_rate": 2.1677002700949107e-05, "loss": 0.3899, "step": 54130 }, { "epoch": 3.27, "learning_rate": 2.167549379083488e-05, "loss": 0.4872, "step": 54132 }, { "epoch": 3.27, "learning_rate": 2.1673984880720656e-05, "loss": 0.325, "step": 54134 }, { "epoch": 3.27, "learning_rate": 2.1672475970606433e-05, "loss": 0.3455, "step": 54136 }, { "epoch": 3.27, "learning_rate": 2.167096706049221e-05, "loss": 0.3461, "step": 54138 }, { "epoch": 3.27, "learning_rate": 2.1669458150377982e-05, "loss": 0.4202, "step": 54140 }, { "epoch": 3.27, "learning_rate": 2.166794924026376e-05, "loss": 0.3559, "step": 54142 }, { "epoch": 3.27, "learning_rate": 2.1666440330149535e-05, "loss": 0.4353, "step": 54144 }, { "epoch": 3.27, "learning_rate": 2.166493142003531e-05, "loss": 0.407, "step": 54146 }, { "epoch": 3.27, "learning_rate": 2.1663422509921085e-05, "loss": 0.4103, "step": 54148 }, { "epoch": 3.27, "learning_rate": 2.166191359980686e-05, "loss": 0.3798, "step": 54150 }, { "epoch": 3.27, "learning_rate": 2.1660404689692638e-05, "loss": 0.4487, "step": 54152 }, { "epoch": 3.27, "learning_rate": 2.165889577957841e-05, "loss": 0.3911, "step": 54154 }, { "epoch": 3.27, "learning_rate": 2.1657386869464184e-05, "loss": 0.4141, "step": 54156 }, { "epoch": 3.27, "learning_rate": 2.1655877959349964e-05, "loss": 0.5355, "step": 54158 }, { "epoch": 3.27, "learning_rate": 2.1654369049235737e-05, "loss": 0.4448, "step": 54160 }, { "epoch": 3.27, "learning_rate": 2.1652860139121514e-05, "loss": 0.5804, "step": 54162 }, { "epoch": 3.27, "learning_rate": 2.1651351229007287e-05, "loss": 0.3511, "step": 54164 }, { "epoch": 3.27, "learning_rate": 2.1649842318893067e-05, "loss": 0.3089, "step": 54166 }, { "epoch": 3.27, "learning_rate": 2.164833340877884e-05, "loss": 0.5546, "step": 54168 }, { "epoch": 3.27, "learning_rate": 2.1646824498664616e-05, "loss": 0.3317, "step": 54170 }, { "epoch": 3.27, "learning_rate": 2.164531558855039e-05, "loss": 0.5598, "step": 54172 }, { "epoch": 3.27, "learning_rate": 2.1643806678436166e-05, "loss": 0.3553, "step": 54174 }, { "epoch": 3.27, "learning_rate": 2.1642297768321943e-05, "loss": 0.2939, "step": 54176 }, { "epoch": 3.27, "learning_rate": 2.1640788858207716e-05, "loss": 0.3305, "step": 54178 }, { "epoch": 3.27, "learning_rate": 2.1639279948093492e-05, "loss": 0.3793, "step": 54180 }, { "epoch": 3.27, "learning_rate": 2.163777103797927e-05, "loss": 0.4451, "step": 54182 }, { "epoch": 3.27, "learning_rate": 2.1636262127865045e-05, "loss": 0.4087, "step": 54184 }, { "epoch": 3.27, "learning_rate": 2.163475321775082e-05, "loss": 0.3017, "step": 54186 }, { "epoch": 3.27, "learning_rate": 2.1633244307636595e-05, "loss": 0.3264, "step": 54188 }, { "epoch": 3.27, "learning_rate": 2.163173539752237e-05, "loss": 0.3241, "step": 54190 }, { "epoch": 3.27, "learning_rate": 2.1630226487408145e-05, "loss": 0.4452, "step": 54192 }, { "epoch": 3.27, "learning_rate": 2.162871757729392e-05, "loss": 0.383, "step": 54194 }, { "epoch": 3.27, "learning_rate": 2.1627208667179698e-05, "loss": 0.5205, "step": 54196 }, { "epoch": 3.27, "learning_rate": 2.1625699757065474e-05, "loss": 0.4527, "step": 54198 }, { "epoch": 3.27, "learning_rate": 2.1624190846951247e-05, "loss": 0.2937, "step": 54200 }, { "epoch": 3.27, "learning_rate": 2.1622681936837024e-05, "loss": 0.4936, "step": 54202 }, { "epoch": 3.27, "learning_rate": 2.16211730267228e-05, "loss": 0.4591, "step": 54204 }, { "epoch": 3.27, "learning_rate": 2.1619664116608573e-05, "loss": 0.33, "step": 54206 }, { "epoch": 3.27, "learning_rate": 2.161815520649435e-05, "loss": 0.3776, "step": 54208 }, { "epoch": 3.27, "learning_rate": 2.1616646296380123e-05, "loss": 0.3128, "step": 54210 }, { "epoch": 3.27, "learning_rate": 2.1615137386265903e-05, "loss": 0.2761, "step": 54212 }, { "epoch": 3.27, "learning_rate": 2.1613628476151676e-05, "loss": 0.5481, "step": 54214 }, { "epoch": 3.27, "learning_rate": 2.1612119566037453e-05, "loss": 0.3315, "step": 54216 }, { "epoch": 3.27, "learning_rate": 2.161061065592323e-05, "loss": 0.3805, "step": 54218 }, { "epoch": 3.27, "learning_rate": 2.1609101745809006e-05, "loss": 0.4381, "step": 54220 }, { "epoch": 3.27, "learning_rate": 2.160759283569478e-05, "loss": 0.3438, "step": 54222 }, { "epoch": 3.27, "learning_rate": 2.1606083925580552e-05, "loss": 0.308, "step": 54224 }, { "epoch": 3.27, "learning_rate": 2.1604575015466332e-05, "loss": 0.3236, "step": 54226 }, { "epoch": 3.27, "learning_rate": 2.1603066105352105e-05, "loss": 0.3576, "step": 54228 }, { "epoch": 3.27, "learning_rate": 2.160155719523788e-05, "loss": 0.537, "step": 54230 }, { "epoch": 3.27, "learning_rate": 2.1600048285123655e-05, "loss": 0.5138, "step": 54232 }, { "epoch": 3.27, "learning_rate": 2.1598539375009434e-05, "loss": 0.4382, "step": 54234 }, { "epoch": 3.27, "learning_rate": 2.1597030464895208e-05, "loss": 0.3216, "step": 54236 }, { "epoch": 3.27, "learning_rate": 2.159552155478098e-05, "loss": 0.355, "step": 54238 }, { "epoch": 3.27, "learning_rate": 2.1594012644666757e-05, "loss": 0.4362, "step": 54240 }, { "epoch": 3.27, "learning_rate": 2.1592503734552534e-05, "loss": 0.2902, "step": 54242 }, { "epoch": 3.27, "learning_rate": 2.159099482443831e-05, "loss": 0.5041, "step": 54244 }, { "epoch": 3.27, "learning_rate": 2.1589485914324083e-05, "loss": 0.4267, "step": 54246 }, { "epoch": 3.27, "learning_rate": 2.158797700420986e-05, "loss": 0.4009, "step": 54248 }, { "epoch": 3.27, "learning_rate": 2.1586468094095636e-05, "loss": 1.1015, "step": 54250 }, { "epoch": 3.27, "learning_rate": 2.1584959183981413e-05, "loss": 0.4007, "step": 54252 }, { "epoch": 3.27, "learning_rate": 2.1583450273867186e-05, "loss": 0.287, "step": 54254 }, { "epoch": 3.27, "learning_rate": 2.1581941363752963e-05, "loss": 0.4529, "step": 54256 }, { "epoch": 3.27, "learning_rate": 2.158043245363874e-05, "loss": 0.3249, "step": 54258 }, { "epoch": 3.27, "learning_rate": 2.1578923543524512e-05, "loss": 0.3009, "step": 54260 }, { "epoch": 3.28, "learning_rate": 2.157741463341029e-05, "loss": 0.2204, "step": 54262 }, { "epoch": 3.28, "learning_rate": 2.1575905723296065e-05, "loss": 0.4819, "step": 54264 }, { "epoch": 3.28, "learning_rate": 2.1574396813181842e-05, "loss": 0.4412, "step": 54266 }, { "epoch": 3.28, "learning_rate": 2.1572887903067615e-05, "loss": 0.3096, "step": 54268 }, { "epoch": 3.28, "learning_rate": 2.157137899295339e-05, "loss": 0.3475, "step": 54270 }, { "epoch": 3.28, "learning_rate": 2.1569870082839168e-05, "loss": 0.4329, "step": 54272 }, { "epoch": 3.28, "learning_rate": 2.156836117272494e-05, "loss": 0.372, "step": 54274 }, { "epoch": 3.28, "learning_rate": 2.1566852262610718e-05, "loss": 0.3711, "step": 54276 }, { "epoch": 3.28, "learning_rate": 2.156534335249649e-05, "loss": 0.2732, "step": 54278 }, { "epoch": 3.28, "learning_rate": 2.156383444238227e-05, "loss": 0.3323, "step": 54280 }, { "epoch": 3.28, "learning_rate": 2.1562325532268044e-05, "loss": 0.3458, "step": 54282 }, { "epoch": 3.28, "learning_rate": 2.156081662215382e-05, "loss": 0.3796, "step": 54284 }, { "epoch": 3.28, "learning_rate": 2.1559307712039593e-05, "loss": 0.4568, "step": 54286 }, { "epoch": 3.28, "learning_rate": 2.155779880192537e-05, "loss": 0.3496, "step": 54288 }, { "epoch": 3.28, "learning_rate": 2.1556289891811146e-05, "loss": 0.4883, "step": 54290 }, { "epoch": 3.28, "learning_rate": 2.155478098169692e-05, "loss": 0.5576, "step": 54292 }, { "epoch": 3.28, "learning_rate": 2.1553272071582696e-05, "loss": 0.2737, "step": 54294 }, { "epoch": 3.28, "learning_rate": 2.1551763161468472e-05, "loss": 0.2833, "step": 54296 }, { "epoch": 3.28, "learning_rate": 2.155025425135425e-05, "loss": 0.3583, "step": 54298 }, { "epoch": 3.28, "learning_rate": 2.1548745341240022e-05, "loss": 0.2669, "step": 54300 }, { "epoch": 3.28, "learning_rate": 2.15472364311258e-05, "loss": 0.425, "step": 54302 }, { "epoch": 3.28, "learning_rate": 2.1545727521011575e-05, "loss": 0.4374, "step": 54304 }, { "epoch": 3.28, "learning_rate": 2.1544218610897348e-05, "loss": 0.4557, "step": 54306 }, { "epoch": 3.28, "learning_rate": 2.1542709700783125e-05, "loss": 0.4639, "step": 54308 }, { "epoch": 3.28, "learning_rate": 2.15412007906689e-05, "loss": 0.3505, "step": 54310 }, { "epoch": 3.28, "learning_rate": 2.1539691880554678e-05, "loss": 0.2981, "step": 54312 }, { "epoch": 3.28, "learning_rate": 2.153818297044045e-05, "loss": 0.4487, "step": 54314 }, { "epoch": 3.28, "learning_rate": 2.1536674060326227e-05, "loss": 0.4002, "step": 54316 }, { "epoch": 3.28, "learning_rate": 2.1535165150212004e-05, "loss": 0.4445, "step": 54318 }, { "epoch": 3.28, "learning_rate": 2.1533656240097777e-05, "loss": 0.3107, "step": 54320 }, { "epoch": 3.28, "learning_rate": 2.1532147329983554e-05, "loss": 0.4204, "step": 54322 }, { "epoch": 3.28, "learning_rate": 2.1530638419869327e-05, "loss": 0.3138, "step": 54324 }, { "epoch": 3.28, "learning_rate": 2.1529129509755107e-05, "loss": 0.3335, "step": 54326 }, { "epoch": 3.28, "learning_rate": 2.152762059964088e-05, "loss": 0.2731, "step": 54328 }, { "epoch": 3.28, "learning_rate": 2.1526111689526656e-05, "loss": 0.3795, "step": 54330 }, { "epoch": 3.28, "learning_rate": 2.152460277941243e-05, "loss": 0.3616, "step": 54332 }, { "epoch": 3.28, "learning_rate": 2.152309386929821e-05, "loss": 0.497, "step": 54334 }, { "epoch": 3.28, "learning_rate": 2.1521584959183982e-05, "loss": 0.3554, "step": 54336 }, { "epoch": 3.28, "learning_rate": 2.1520076049069756e-05, "loss": 0.2599, "step": 54338 }, { "epoch": 3.28, "learning_rate": 2.1518567138955535e-05, "loss": 0.4404, "step": 54340 }, { "epoch": 3.28, "learning_rate": 2.151705822884131e-05, "loss": 0.2291, "step": 54342 }, { "epoch": 3.28, "learning_rate": 2.1515549318727085e-05, "loss": 0.2871, "step": 54344 }, { "epoch": 3.28, "learning_rate": 2.1514040408612858e-05, "loss": 0.5695, "step": 54346 }, { "epoch": 3.28, "learning_rate": 2.1512531498498638e-05, "loss": 0.2707, "step": 54348 }, { "epoch": 3.28, "learning_rate": 2.151102258838441e-05, "loss": 0.2608, "step": 54350 }, { "epoch": 3.28, "learning_rate": 2.1509513678270188e-05, "loss": 0.288, "step": 54352 }, { "epoch": 3.28, "learning_rate": 2.150800476815596e-05, "loss": 0.3396, "step": 54354 }, { "epoch": 3.28, "learning_rate": 2.1506495858041737e-05, "loss": 0.354, "step": 54356 }, { "epoch": 3.28, "learning_rate": 2.1504986947927514e-05, "loss": 0.3538, "step": 54358 }, { "epoch": 3.28, "learning_rate": 2.1503478037813287e-05, "loss": 0.3643, "step": 54360 }, { "epoch": 3.28, "learning_rate": 2.1501969127699064e-05, "loss": 0.4907, "step": 54362 }, { "epoch": 3.28, "learning_rate": 2.150046021758484e-05, "loss": 0.4324, "step": 54364 }, { "epoch": 3.28, "learning_rate": 2.1498951307470617e-05, "loss": 0.5955, "step": 54366 }, { "epoch": 3.28, "learning_rate": 2.149744239735639e-05, "loss": 0.4107, "step": 54368 }, { "epoch": 3.28, "learning_rate": 2.1495933487242166e-05, "loss": 0.2903, "step": 54370 }, { "epoch": 3.28, "learning_rate": 2.1494424577127943e-05, "loss": 0.4846, "step": 54372 }, { "epoch": 3.28, "learning_rate": 2.1492915667013716e-05, "loss": 0.2665, "step": 54374 }, { "epoch": 3.28, "learning_rate": 2.1491406756899492e-05, "loss": 0.3272, "step": 54376 }, { "epoch": 3.28, "learning_rate": 2.148989784678527e-05, "loss": 0.3845, "step": 54378 }, { "epoch": 3.28, "learning_rate": 2.1488388936671045e-05, "loss": 0.4369, "step": 54380 }, { "epoch": 3.28, "learning_rate": 2.148688002655682e-05, "loss": 0.5728, "step": 54382 }, { "epoch": 3.28, "learning_rate": 2.1485371116442595e-05, "loss": 0.4539, "step": 54384 }, { "epoch": 3.28, "learning_rate": 2.148386220632837e-05, "loss": 0.3486, "step": 54386 }, { "epoch": 3.28, "learning_rate": 2.1482353296214145e-05, "loss": 0.3565, "step": 54388 }, { "epoch": 3.28, "learning_rate": 2.148084438609992e-05, "loss": 0.2878, "step": 54390 }, { "epoch": 3.28, "learning_rate": 2.1479335475985694e-05, "loss": 0.2652, "step": 54392 }, { "epoch": 3.28, "learning_rate": 2.1477826565871474e-05, "loss": 0.3481, "step": 54394 }, { "epoch": 3.28, "learning_rate": 2.1476317655757247e-05, "loss": 0.5051, "step": 54396 }, { "epoch": 3.28, "learning_rate": 2.1474808745643024e-05, "loss": 0.495, "step": 54398 }, { "epoch": 3.28, "learning_rate": 2.1473299835528797e-05, "loss": 0.3452, "step": 54400 }, { "epoch": 3.28, "learning_rate": 2.1471790925414573e-05, "loss": 0.3425, "step": 54402 }, { "epoch": 3.28, "learning_rate": 2.147028201530035e-05, "loss": 0.2743, "step": 54404 }, { "epoch": 3.28, "learning_rate": 2.1468773105186123e-05, "loss": 0.414, "step": 54406 }, { "epoch": 3.28, "learning_rate": 2.14672641950719e-05, "loss": 0.3105, "step": 54408 }, { "epoch": 3.28, "learning_rate": 2.1465755284957676e-05, "loss": 0.2556, "step": 54410 }, { "epoch": 3.28, "learning_rate": 2.1464246374843453e-05, "loss": 0.3782, "step": 54412 }, { "epoch": 3.28, "learning_rate": 2.1462737464729226e-05, "loss": 0.3575, "step": 54414 }, { "epoch": 3.28, "learning_rate": 2.1461228554615002e-05, "loss": 0.3538, "step": 54416 }, { "epoch": 3.28, "learning_rate": 2.145971964450078e-05, "loss": 0.5708, "step": 54418 }, { "epoch": 3.28, "learning_rate": 2.1458210734386552e-05, "loss": 0.454, "step": 54420 }, { "epoch": 3.28, "learning_rate": 2.145670182427233e-05, "loss": 0.3411, "step": 54422 }, { "epoch": 3.28, "learning_rate": 2.1455192914158105e-05, "loss": 0.3954, "step": 54424 }, { "epoch": 3.29, "learning_rate": 2.145368400404388e-05, "loss": 0.4306, "step": 54426 }, { "epoch": 3.29, "learning_rate": 2.1452175093929655e-05, "loss": 0.538, "step": 54428 }, { "epoch": 3.29, "learning_rate": 2.145066618381543e-05, "loss": 0.3996, "step": 54430 }, { "epoch": 3.29, "learning_rate": 2.1449157273701208e-05, "loss": 0.4677, "step": 54432 }, { "epoch": 3.29, "learning_rate": 2.1447648363586984e-05, "loss": 0.3449, "step": 54434 }, { "epoch": 3.29, "learning_rate": 2.1446139453472757e-05, "loss": 0.393, "step": 54436 }, { "epoch": 3.29, "learning_rate": 2.144463054335853e-05, "loss": 0.2769, "step": 54438 }, { "epoch": 3.29, "learning_rate": 2.144312163324431e-05, "loss": 0.3765, "step": 54440 }, { "epoch": 3.29, "learning_rate": 2.1441612723130083e-05, "loss": 0.4869, "step": 54442 }, { "epoch": 3.29, "learning_rate": 2.144010381301586e-05, "loss": 0.2351, "step": 54444 }, { "epoch": 3.29, "learning_rate": 2.1438594902901633e-05, "loss": 0.4381, "step": 54446 }, { "epoch": 3.29, "learning_rate": 2.1437085992787413e-05, "loss": 0.4956, "step": 54448 }, { "epoch": 3.29, "learning_rate": 2.1435577082673186e-05, "loss": 0.2812, "step": 54450 }, { "epoch": 3.29, "learning_rate": 2.143406817255896e-05, "loss": 0.2772, "step": 54452 }, { "epoch": 3.29, "learning_rate": 2.143255926244474e-05, "loss": 0.2556, "step": 54454 }, { "epoch": 3.29, "learning_rate": 2.1431050352330512e-05, "loss": 0.224, "step": 54456 }, { "epoch": 3.29, "learning_rate": 2.142954144221629e-05, "loss": 0.6056, "step": 54458 }, { "epoch": 3.29, "learning_rate": 2.1428032532102062e-05, "loss": 0.3051, "step": 54460 }, { "epoch": 3.29, "learning_rate": 2.1426523621987842e-05, "loss": 0.2718, "step": 54462 }, { "epoch": 3.29, "learning_rate": 2.1425014711873615e-05, "loss": 0.5345, "step": 54464 }, { "epoch": 3.29, "learning_rate": 2.142350580175939e-05, "loss": 0.3503, "step": 54466 }, { "epoch": 3.29, "learning_rate": 2.1421996891645165e-05, "loss": 0.269, "step": 54468 }, { "epoch": 3.29, "learning_rate": 2.142048798153094e-05, "loss": 0.5622, "step": 54470 }, { "epoch": 3.29, "learning_rate": 2.1418979071416718e-05, "loss": 0.3511, "step": 54472 }, { "epoch": 3.29, "learning_rate": 2.141747016130249e-05, "loss": 0.3032, "step": 54474 }, { "epoch": 3.29, "learning_rate": 2.1415961251188267e-05, "loss": 0.4851, "step": 54476 }, { "epoch": 3.29, "learning_rate": 2.1414452341074044e-05, "loss": 0.3821, "step": 54478 }, { "epoch": 3.29, "learning_rate": 2.141294343095982e-05, "loss": 0.4026, "step": 54480 }, { "epoch": 3.29, "learning_rate": 2.1411434520845593e-05, "loss": 0.3968, "step": 54482 }, { "epoch": 3.29, "learning_rate": 2.140992561073137e-05, "loss": 0.4082, "step": 54484 }, { "epoch": 3.29, "learning_rate": 2.1408416700617146e-05, "loss": 0.5936, "step": 54486 }, { "epoch": 3.29, "learning_rate": 2.140690779050292e-05, "loss": 0.3978, "step": 54488 }, { "epoch": 3.29, "learning_rate": 2.1405398880388696e-05, "loss": 0.3773, "step": 54490 }, { "epoch": 3.29, "learning_rate": 2.1403889970274473e-05, "loss": 0.4067, "step": 54492 }, { "epoch": 3.29, "learning_rate": 2.140238106016025e-05, "loss": 0.5205, "step": 54494 }, { "epoch": 3.29, "learning_rate": 2.1400872150046022e-05, "loss": 0.4035, "step": 54496 }, { "epoch": 3.29, "learning_rate": 2.13993632399318e-05, "loss": 0.4741, "step": 54498 }, { "epoch": 3.29, "learning_rate": 2.1397854329817575e-05, "loss": 0.4136, "step": 54500 }, { "epoch": 3.29, "learning_rate": 2.139634541970335e-05, "loss": 0.3444, "step": 54502 }, { "epoch": 3.29, "learning_rate": 2.1394836509589125e-05, "loss": 0.4441, "step": 54504 }, { "epoch": 3.29, "learning_rate": 2.1393327599474898e-05, "loss": 0.3499, "step": 54506 }, { "epoch": 3.29, "learning_rate": 2.1391818689360678e-05, "loss": 0.3884, "step": 54508 }, { "epoch": 3.29, "learning_rate": 2.139030977924645e-05, "loss": 0.4722, "step": 54510 }, { "epoch": 3.29, "learning_rate": 2.1388800869132228e-05, "loss": 0.4062, "step": 54512 }, { "epoch": 3.29, "learning_rate": 2.1387291959018e-05, "loss": 0.4754, "step": 54514 }, { "epoch": 3.29, "learning_rate": 2.138578304890378e-05, "loss": 0.327, "step": 54516 }, { "epoch": 3.29, "learning_rate": 2.1384274138789554e-05, "loss": 0.5961, "step": 54518 }, { "epoch": 3.29, "learning_rate": 2.1382765228675327e-05, "loss": 0.288, "step": 54520 }, { "epoch": 3.29, "learning_rate": 2.1381256318561103e-05, "loss": 0.4485, "step": 54522 }, { "epoch": 3.29, "learning_rate": 2.137974740844688e-05, "loss": 0.331, "step": 54524 }, { "epoch": 3.29, "learning_rate": 2.1378238498332656e-05, "loss": 0.3542, "step": 54526 }, { "epoch": 3.29, "learning_rate": 2.137672958821843e-05, "loss": 0.3249, "step": 54528 }, { "epoch": 3.29, "learning_rate": 2.1375220678104206e-05, "loss": 0.2578, "step": 54530 }, { "epoch": 3.29, "learning_rate": 2.1373711767989982e-05, "loss": 0.5678, "step": 54532 }, { "epoch": 3.29, "learning_rate": 2.1372202857875756e-05, "loss": 0.5133, "step": 54534 }, { "epoch": 3.29, "learning_rate": 2.1370693947761532e-05, "loss": 0.4088, "step": 54536 }, { "epoch": 3.29, "learning_rate": 2.136918503764731e-05, "loss": 0.4377, "step": 54538 }, { "epoch": 3.29, "learning_rate": 2.1367676127533085e-05, "loss": 0.3149, "step": 54540 }, { "epoch": 3.29, "learning_rate": 2.1366167217418858e-05, "loss": 0.4759, "step": 54542 }, { "epoch": 3.29, "learning_rate": 2.1364658307304635e-05, "loss": 0.4409, "step": 54544 }, { "epoch": 3.29, "learning_rate": 2.136314939719041e-05, "loss": 0.6657, "step": 54546 }, { "epoch": 3.29, "learning_rate": 2.1361640487076188e-05, "loss": 0.4462, "step": 54548 }, { "epoch": 3.29, "learning_rate": 2.136013157696196e-05, "loss": 0.3753, "step": 54550 }, { "epoch": 3.29, "learning_rate": 2.1358622666847734e-05, "loss": 0.3675, "step": 54552 }, { "epoch": 3.29, "learning_rate": 2.1357113756733514e-05, "loss": 0.451, "step": 54554 }, { "epoch": 3.29, "learning_rate": 2.1355604846619287e-05, "loss": 0.3379, "step": 54556 }, { "epoch": 3.29, "learning_rate": 2.1354095936505064e-05, "loss": 0.4226, "step": 54558 }, { "epoch": 3.29, "learning_rate": 2.1352587026390837e-05, "loss": 0.5934, "step": 54560 }, { "epoch": 3.29, "learning_rate": 2.1351078116276617e-05, "loss": 0.3627, "step": 54562 }, { "epoch": 3.29, "learning_rate": 2.134956920616239e-05, "loss": 0.379, "step": 54564 }, { "epoch": 3.29, "learning_rate": 2.1348060296048166e-05, "loss": 0.3365, "step": 54566 }, { "epoch": 3.29, "learning_rate": 2.134655138593394e-05, "loss": 0.3874, "step": 54568 }, { "epoch": 3.29, "learning_rate": 2.1345042475819716e-05, "loss": 0.3596, "step": 54570 }, { "epoch": 3.29, "learning_rate": 2.1343533565705492e-05, "loss": 0.3698, "step": 54572 }, { "epoch": 3.29, "learning_rate": 2.1342024655591266e-05, "loss": 0.4783, "step": 54574 }, { "epoch": 3.29, "learning_rate": 2.1340515745477045e-05, "loss": 0.3939, "step": 54576 }, { "epoch": 3.29, "learning_rate": 2.133900683536282e-05, "loss": 0.4151, "step": 54578 }, { "epoch": 3.29, "learning_rate": 2.1337497925248595e-05, "loss": 0.5713, "step": 54580 }, { "epoch": 3.29, "learning_rate": 2.1335989015134368e-05, "loss": 0.381, "step": 54582 }, { "epoch": 3.29, "learning_rate": 2.1334480105020145e-05, "loss": 0.3591, "step": 54584 }, { "epoch": 3.29, "learning_rate": 2.133297119490592e-05, "loss": 0.4222, "step": 54586 }, { "epoch": 3.29, "learning_rate": 2.1331462284791694e-05, "loss": 0.3654, "step": 54588 }, { "epoch": 3.29, "learning_rate": 2.132995337467747e-05, "loss": 0.5102, "step": 54590 }, { "epoch": 3.3, "learning_rate": 2.1328444464563247e-05, "loss": 0.2404, "step": 54592 }, { "epoch": 3.3, "learning_rate": 2.1326935554449024e-05, "loss": 0.5261, "step": 54594 }, { "epoch": 3.3, "learning_rate": 2.1325426644334797e-05, "loss": 0.5142, "step": 54596 }, { "epoch": 3.3, "learning_rate": 2.1323917734220574e-05, "loss": 0.3952, "step": 54598 }, { "epoch": 3.3, "learning_rate": 2.132240882410635e-05, "loss": 0.3352, "step": 54600 }, { "epoch": 3.3, "learning_rate": 2.1320899913992123e-05, "loss": 0.3478, "step": 54602 }, { "epoch": 3.3, "learning_rate": 2.13193910038779e-05, "loss": 0.3344, "step": 54604 }, { "epoch": 3.3, "learning_rate": 2.1317882093763676e-05, "loss": 0.3822, "step": 54606 }, { "epoch": 3.3, "learning_rate": 2.1316373183649453e-05, "loss": 0.2487, "step": 54608 }, { "epoch": 3.3, "learning_rate": 2.1314864273535226e-05, "loss": 0.3722, "step": 54610 }, { "epoch": 3.3, "learning_rate": 2.1313355363421002e-05, "loss": 0.3019, "step": 54612 }, { "epoch": 3.3, "learning_rate": 2.131184645330678e-05, "loss": 0.3062, "step": 54614 }, { "epoch": 3.3, "learning_rate": 2.1310337543192552e-05, "loss": 0.3486, "step": 54616 }, { "epoch": 3.3, "learning_rate": 2.130882863307833e-05, "loss": 0.3526, "step": 54618 }, { "epoch": 3.3, "learning_rate": 2.13073197229641e-05, "loss": 0.5396, "step": 54620 }, { "epoch": 3.3, "learning_rate": 2.130581081284988e-05, "loss": 0.4404, "step": 54622 }, { "epoch": 3.3, "learning_rate": 2.1304301902735655e-05, "loss": 0.4709, "step": 54624 }, { "epoch": 3.3, "learning_rate": 2.130279299262143e-05, "loss": 0.3362, "step": 54626 }, { "epoch": 3.3, "learning_rate": 2.1301284082507204e-05, "loss": 0.2785, "step": 54628 }, { "epoch": 3.3, "learning_rate": 2.1299775172392984e-05, "loss": 0.4096, "step": 54630 }, { "epoch": 3.3, "learning_rate": 2.1298266262278757e-05, "loss": 0.3475, "step": 54632 }, { "epoch": 3.3, "learning_rate": 2.129675735216453e-05, "loss": 0.5453, "step": 54634 }, { "epoch": 3.3, "learning_rate": 2.1295248442050307e-05, "loss": 0.5681, "step": 54636 }, { "epoch": 3.3, "learning_rate": 2.1293739531936084e-05, "loss": 0.3582, "step": 54638 }, { "epoch": 3.3, "learning_rate": 2.129223062182186e-05, "loss": 0.4927, "step": 54640 }, { "epoch": 3.3, "learning_rate": 2.1290721711707633e-05, "loss": 0.4692, "step": 54642 }, { "epoch": 3.3, "learning_rate": 2.128921280159341e-05, "loss": 0.4733, "step": 54644 }, { "epoch": 3.3, "learning_rate": 2.1287703891479186e-05, "loss": 0.4394, "step": 54646 }, { "epoch": 3.3, "learning_rate": 2.1286194981364963e-05, "loss": 0.3817, "step": 54648 }, { "epoch": 3.3, "learning_rate": 2.1284686071250736e-05, "loss": 0.2673, "step": 54650 }, { "epoch": 3.3, "learning_rate": 2.1283177161136512e-05, "loss": 0.4178, "step": 54652 }, { "epoch": 3.3, "learning_rate": 2.128166825102229e-05, "loss": 0.2767, "step": 54654 }, { "epoch": 3.3, "learning_rate": 2.1280159340908062e-05, "loss": 0.297, "step": 54656 }, { "epoch": 3.3, "learning_rate": 2.127865043079384e-05, "loss": 0.3153, "step": 54658 }, { "epoch": 3.3, "learning_rate": 2.1277141520679615e-05, "loss": 0.4703, "step": 54660 }, { "epoch": 3.3, "learning_rate": 2.127563261056539e-05, "loss": 0.2849, "step": 54662 }, { "epoch": 3.3, "learning_rate": 2.1274123700451165e-05, "loss": 0.465, "step": 54664 }, { "epoch": 3.3, "learning_rate": 2.1272614790336938e-05, "loss": 0.5486, "step": 54666 }, { "epoch": 3.3, "learning_rate": 2.1271105880222718e-05, "loss": 0.5149, "step": 54668 }, { "epoch": 3.3, "learning_rate": 2.126959697010849e-05, "loss": 0.3551, "step": 54670 }, { "epoch": 3.3, "learning_rate": 2.1268088059994267e-05, "loss": 0.2776, "step": 54672 }, { "epoch": 3.3, "learning_rate": 2.126657914988004e-05, "loss": 0.4444, "step": 54674 }, { "epoch": 3.3, "learning_rate": 2.126507023976582e-05, "loss": 0.3427, "step": 54676 }, { "epoch": 3.3, "learning_rate": 2.1263561329651593e-05, "loss": 0.4858, "step": 54678 }, { "epoch": 3.3, "learning_rate": 2.126205241953737e-05, "loss": 0.3445, "step": 54680 }, { "epoch": 3.3, "learning_rate": 2.1260543509423143e-05, "loss": 0.5121, "step": 54682 }, { "epoch": 3.3, "learning_rate": 2.125903459930892e-05, "loss": 0.4132, "step": 54684 }, { "epoch": 3.3, "learning_rate": 2.1257525689194696e-05, "loss": 0.3814, "step": 54686 }, { "epoch": 3.3, "learning_rate": 2.125601677908047e-05, "loss": 0.2732, "step": 54688 }, { "epoch": 3.3, "learning_rate": 2.125450786896625e-05, "loss": 0.6645, "step": 54690 }, { "epoch": 3.3, "learning_rate": 2.1252998958852022e-05, "loss": 0.3397, "step": 54692 }, { "epoch": 3.3, "learning_rate": 2.12514900487378e-05, "loss": 0.4068, "step": 54694 }, { "epoch": 3.3, "learning_rate": 2.1249981138623572e-05, "loss": 0.3337, "step": 54696 }, { "epoch": 3.3, "learning_rate": 2.124847222850935e-05, "loss": 0.2445, "step": 54698 }, { "epoch": 3.3, "learning_rate": 2.1246963318395125e-05, "loss": 0.3736, "step": 54700 }, { "epoch": 3.3, "learning_rate": 2.1245454408280898e-05, "loss": 0.4014, "step": 54702 }, { "epoch": 3.3, "learning_rate": 2.1243945498166675e-05, "loss": 0.3833, "step": 54704 }, { "epoch": 3.3, "learning_rate": 2.124243658805245e-05, "loss": 0.345, "step": 54706 }, { "epoch": 3.3, "learning_rate": 2.1240927677938228e-05, "loss": 0.4418, "step": 54708 }, { "epoch": 3.3, "learning_rate": 2.1239418767824e-05, "loss": 0.3075, "step": 54710 }, { "epoch": 3.3, "learning_rate": 2.1237909857709777e-05, "loss": 0.2703, "step": 54712 }, { "epoch": 3.3, "learning_rate": 2.1236400947595554e-05, "loss": 0.4125, "step": 54714 }, { "epoch": 3.3, "learning_rate": 2.1234892037481327e-05, "loss": 0.4671, "step": 54716 }, { "epoch": 3.3, "learning_rate": 2.1233383127367103e-05, "loss": 0.586, "step": 54718 }, { "epoch": 3.3, "learning_rate": 2.123187421725288e-05, "loss": 0.3335, "step": 54720 }, { "epoch": 3.3, "learning_rate": 2.1230365307138656e-05, "loss": 0.4389, "step": 54722 }, { "epoch": 3.3, "learning_rate": 2.122885639702443e-05, "loss": 0.4043, "step": 54724 }, { "epoch": 3.3, "learning_rate": 2.1227347486910206e-05, "loss": 0.2595, "step": 54726 }, { "epoch": 3.3, "learning_rate": 2.1225838576795983e-05, "loss": 0.4312, "step": 54728 }, { "epoch": 3.3, "learning_rate": 2.122432966668176e-05, "loss": 0.3543, "step": 54730 }, { "epoch": 3.3, "learning_rate": 2.1222820756567532e-05, "loss": 0.3517, "step": 54732 }, { "epoch": 3.3, "learning_rate": 2.1221311846453305e-05, "loss": 0.3692, "step": 54734 }, { "epoch": 3.3, "learning_rate": 2.1219802936339085e-05, "loss": 0.377, "step": 54736 }, { "epoch": 3.3, "learning_rate": 2.121829402622486e-05, "loss": 0.4039, "step": 54738 }, { "epoch": 3.3, "learning_rate": 2.1216785116110635e-05, "loss": 0.3478, "step": 54740 }, { "epoch": 3.3, "learning_rate": 2.1215276205996408e-05, "loss": 0.4598, "step": 54742 }, { "epoch": 3.3, "learning_rate": 2.1213767295882188e-05, "loss": 0.2228, "step": 54744 }, { "epoch": 3.3, "learning_rate": 2.121225838576796e-05, "loss": 0.3196, "step": 54746 }, { "epoch": 3.3, "learning_rate": 2.1210749475653734e-05, "loss": 0.3282, "step": 54748 }, { "epoch": 3.3, "learning_rate": 2.120924056553951e-05, "loss": 0.2702, "step": 54750 }, { "epoch": 3.3, "learning_rate": 2.1207731655425287e-05, "loss": 0.3566, "step": 54752 }, { "epoch": 3.3, "learning_rate": 2.1206222745311064e-05, "loss": 0.4393, "step": 54754 }, { "epoch": 3.3, "learning_rate": 2.1204713835196837e-05, "loss": 0.3801, "step": 54756 }, { "epoch": 3.31, "learning_rate": 2.1203204925082613e-05, "loss": 0.3377, "step": 54758 }, { "epoch": 3.31, "learning_rate": 2.120169601496839e-05, "loss": 0.3375, "step": 54760 }, { "epoch": 3.31, "learning_rate": 2.1200187104854166e-05, "loss": 0.5319, "step": 54762 }, { "epoch": 3.31, "learning_rate": 2.119867819473994e-05, "loss": 0.3564, "step": 54764 }, { "epoch": 3.31, "learning_rate": 2.1197169284625716e-05, "loss": 0.3179, "step": 54766 }, { "epoch": 3.31, "learning_rate": 2.1195660374511493e-05, "loss": 0.378, "step": 54768 }, { "epoch": 3.31, "learning_rate": 2.1194151464397266e-05, "loss": 0.344, "step": 54770 }, { "epoch": 3.31, "learning_rate": 2.1192642554283042e-05, "loss": 0.3365, "step": 54772 }, { "epoch": 3.31, "learning_rate": 2.119113364416882e-05, "loss": 0.347, "step": 54774 }, { "epoch": 3.31, "learning_rate": 2.1189624734054595e-05, "loss": 0.4423, "step": 54776 }, { "epoch": 3.31, "learning_rate": 2.118811582394037e-05, "loss": 0.3384, "step": 54778 }, { "epoch": 3.31, "learning_rate": 2.118660691382614e-05, "loss": 0.421, "step": 54780 }, { "epoch": 3.31, "learning_rate": 2.118509800371192e-05, "loss": 0.3893, "step": 54782 }, { "epoch": 3.31, "learning_rate": 2.1183589093597694e-05, "loss": 0.4136, "step": 54784 }, { "epoch": 3.31, "learning_rate": 2.118208018348347e-05, "loss": 0.4639, "step": 54786 }, { "epoch": 3.31, "learning_rate": 2.1180571273369244e-05, "loss": 0.4004, "step": 54788 }, { "epoch": 3.31, "learning_rate": 2.1179062363255024e-05, "loss": 0.4595, "step": 54790 }, { "epoch": 3.31, "learning_rate": 2.1177553453140797e-05, "loss": 0.5376, "step": 54792 }, { "epoch": 3.31, "learning_rate": 2.1176044543026574e-05, "loss": 0.3646, "step": 54794 }, { "epoch": 3.31, "learning_rate": 2.1174535632912347e-05, "loss": 0.356, "step": 54796 }, { "epoch": 3.31, "learning_rate": 2.1173026722798123e-05, "loss": 0.2363, "step": 54798 }, { "epoch": 3.31, "learning_rate": 2.11715178126839e-05, "loss": 0.2854, "step": 54800 }, { "epoch": 3.31, "learning_rate": 2.1170008902569673e-05, "loss": 0.2777, "step": 54802 }, { "epoch": 3.31, "learning_rate": 2.116849999245545e-05, "loss": 0.29, "step": 54804 }, { "epoch": 3.31, "learning_rate": 2.1166991082341226e-05, "loss": 0.3955, "step": 54806 }, { "epoch": 3.31, "learning_rate": 2.1165482172227002e-05, "loss": 0.3208, "step": 54808 }, { "epoch": 3.31, "learning_rate": 2.1163973262112776e-05, "loss": 0.3725, "step": 54810 }, { "epoch": 3.31, "learning_rate": 2.1162464351998555e-05, "loss": 0.3826, "step": 54812 }, { "epoch": 3.31, "learning_rate": 2.116095544188433e-05, "loss": 0.4673, "step": 54814 }, { "epoch": 3.31, "learning_rate": 2.1159446531770102e-05, "loss": 0.3914, "step": 54816 }, { "epoch": 3.31, "learning_rate": 2.1157937621655878e-05, "loss": 0.4287, "step": 54818 }, { "epoch": 3.31, "learning_rate": 2.1156428711541655e-05, "loss": 0.3302, "step": 54820 }, { "epoch": 3.31, "learning_rate": 2.115491980142743e-05, "loss": 0.3434, "step": 54822 }, { "epoch": 3.31, "learning_rate": 2.1153410891313204e-05, "loss": 0.324, "step": 54824 }, { "epoch": 3.31, "learning_rate": 2.115190198119898e-05, "loss": 0.2928, "step": 54826 }, { "epoch": 3.31, "learning_rate": 2.1150393071084757e-05, "loss": 0.4262, "step": 54828 }, { "epoch": 3.31, "learning_rate": 2.114888416097053e-05, "loss": 0.3784, "step": 54830 }, { "epoch": 3.31, "learning_rate": 2.1147375250856307e-05, "loss": 0.384, "step": 54832 }, { "epoch": 3.31, "learning_rate": 2.1145866340742084e-05, "loss": 0.2466, "step": 54834 }, { "epoch": 3.31, "learning_rate": 2.114435743062786e-05, "loss": 0.2855, "step": 54836 }, { "epoch": 3.31, "learning_rate": 2.1142848520513633e-05, "loss": 0.3265, "step": 54838 }, { "epoch": 3.31, "learning_rate": 2.114133961039941e-05, "loss": 0.3503, "step": 54840 }, { "epoch": 3.31, "learning_rate": 2.1139830700285186e-05, "loss": 0.3655, "step": 54842 }, { "epoch": 3.31, "learning_rate": 2.1138321790170963e-05, "loss": 0.4855, "step": 54844 }, { "epoch": 3.31, "learning_rate": 2.1136812880056736e-05, "loss": 0.4726, "step": 54846 }, { "epoch": 3.31, "learning_rate": 2.113530396994251e-05, "loss": 0.4547, "step": 54848 }, { "epoch": 3.31, "learning_rate": 2.113379505982829e-05, "loss": 0.3545, "step": 54850 }, { "epoch": 3.31, "learning_rate": 2.1132286149714062e-05, "loss": 0.3746, "step": 54852 }, { "epoch": 3.31, "learning_rate": 2.113077723959984e-05, "loss": 0.3376, "step": 54854 }, { "epoch": 3.31, "learning_rate": 2.112926832948561e-05, "loss": 0.3579, "step": 54856 }, { "epoch": 3.31, "learning_rate": 2.112775941937139e-05, "loss": 0.2524, "step": 54858 }, { "epoch": 3.31, "learning_rate": 2.1126250509257165e-05, "loss": 0.483, "step": 54860 }, { "epoch": 3.31, "learning_rate": 2.1124741599142938e-05, "loss": 0.3361, "step": 54862 }, { "epoch": 3.31, "learning_rate": 2.1123232689028714e-05, "loss": 0.4291, "step": 54864 }, { "epoch": 3.31, "learning_rate": 2.112172377891449e-05, "loss": 0.3754, "step": 54866 }, { "epoch": 3.31, "learning_rate": 2.1120214868800267e-05, "loss": 0.4258, "step": 54868 }, { "epoch": 3.31, "learning_rate": 2.111870595868604e-05, "loss": 0.3973, "step": 54870 }, { "epoch": 3.31, "learning_rate": 2.1117197048571817e-05, "loss": 0.5062, "step": 54872 }, { "epoch": 3.31, "learning_rate": 2.1115688138457594e-05, "loss": 0.3166, "step": 54874 }, { "epoch": 3.31, "learning_rate": 2.111417922834337e-05, "loss": 0.6395, "step": 54876 }, { "epoch": 3.31, "learning_rate": 2.1112670318229143e-05, "loss": 0.3988, "step": 54878 }, { "epoch": 3.31, "learning_rate": 2.111116140811492e-05, "loss": 0.3296, "step": 54880 }, { "epoch": 3.31, "learning_rate": 2.1109652498000696e-05, "loss": 0.4856, "step": 54882 }, { "epoch": 3.31, "learning_rate": 2.110814358788647e-05, "loss": 0.4686, "step": 54884 }, { "epoch": 3.31, "learning_rate": 2.1106634677772246e-05, "loss": 0.3146, "step": 54886 }, { "epoch": 3.31, "learning_rate": 2.1105125767658022e-05, "loss": 0.41, "step": 54888 }, { "epoch": 3.31, "learning_rate": 2.11036168575438e-05, "loss": 0.3497, "step": 54890 }, { "epoch": 3.31, "learning_rate": 2.1102107947429572e-05, "loss": 0.3909, "step": 54892 }, { "epoch": 3.31, "learning_rate": 2.110059903731535e-05, "loss": 0.5534, "step": 54894 }, { "epoch": 3.31, "learning_rate": 2.1099090127201125e-05, "loss": 0.5111, "step": 54896 }, { "epoch": 3.31, "learning_rate": 2.1097581217086898e-05, "loss": 0.3611, "step": 54898 }, { "epoch": 3.31, "learning_rate": 2.1096072306972675e-05, "loss": 0.3723, "step": 54900 }, { "epoch": 3.31, "learning_rate": 2.1094563396858448e-05, "loss": 0.4068, "step": 54902 }, { "epoch": 3.31, "learning_rate": 2.1093054486744228e-05, "loss": 0.4886, "step": 54904 }, { "epoch": 3.31, "learning_rate": 2.109154557663e-05, "loss": 0.382, "step": 54906 }, { "epoch": 3.31, "learning_rate": 2.1090036666515777e-05, "loss": 0.431, "step": 54908 }, { "epoch": 3.31, "learning_rate": 2.108852775640155e-05, "loss": 0.3044, "step": 54910 }, { "epoch": 3.31, "learning_rate": 2.1087018846287327e-05, "loss": 0.3915, "step": 54912 }, { "epoch": 3.31, "learning_rate": 2.1085509936173103e-05, "loss": 0.3418, "step": 54914 }, { "epoch": 3.31, "learning_rate": 2.1084001026058877e-05, "loss": 0.2957, "step": 54916 }, { "epoch": 3.31, "learning_rate": 2.1082492115944653e-05, "loss": 0.5965, "step": 54918 }, { "epoch": 3.31, "learning_rate": 2.108098320583043e-05, "loss": 0.564, "step": 54920 }, { "epoch": 3.31, "learning_rate": 2.1079474295716206e-05, "loss": 0.5096, "step": 54922 }, { "epoch": 3.32, "learning_rate": 2.107796538560198e-05, "loss": 0.3071, "step": 54924 }, { "epoch": 3.32, "learning_rate": 2.107645647548776e-05, "loss": 0.5099, "step": 54926 }, { "epoch": 3.32, "learning_rate": 2.1074947565373532e-05, "loss": 0.272, "step": 54928 }, { "epoch": 3.32, "learning_rate": 2.1073438655259305e-05, "loss": 0.3864, "step": 54930 }, { "epoch": 3.32, "learning_rate": 2.1071929745145082e-05, "loss": 0.3208, "step": 54932 }, { "epoch": 3.32, "learning_rate": 2.107042083503086e-05, "loss": 0.3899, "step": 54934 }, { "epoch": 3.32, "learning_rate": 2.1068911924916635e-05, "loss": 0.3697, "step": 54936 }, { "epoch": 3.32, "learning_rate": 2.1067403014802408e-05, "loss": 0.3207, "step": 54938 }, { "epoch": 3.32, "learning_rate": 2.1065894104688185e-05, "loss": 0.2045, "step": 54940 }, { "epoch": 3.32, "learning_rate": 2.106438519457396e-05, "loss": 0.3419, "step": 54942 }, { "epoch": 3.32, "learning_rate": 2.1062876284459734e-05, "loss": 0.3951, "step": 54944 }, { "epoch": 3.32, "learning_rate": 2.106136737434551e-05, "loss": 0.539, "step": 54946 }, { "epoch": 3.32, "learning_rate": 2.1059858464231287e-05, "loss": 0.4555, "step": 54948 }, { "epoch": 3.32, "learning_rate": 2.1058349554117064e-05, "loss": 0.3635, "step": 54950 }, { "epoch": 3.32, "learning_rate": 2.1056840644002837e-05, "loss": 0.4607, "step": 54952 }, { "epoch": 3.32, "learning_rate": 2.1055331733888613e-05, "loss": 0.5888, "step": 54954 }, { "epoch": 3.32, "learning_rate": 2.105382282377439e-05, "loss": 0.5367, "step": 54956 }, { "epoch": 3.32, "learning_rate": 2.1052313913660166e-05, "loss": 0.3483, "step": 54958 }, { "epoch": 3.32, "learning_rate": 2.105080500354594e-05, "loss": 0.2776, "step": 54960 }, { "epoch": 3.32, "learning_rate": 2.1049296093431713e-05, "loss": 0.4559, "step": 54962 }, { "epoch": 3.32, "learning_rate": 2.1047787183317493e-05, "loss": 0.2696, "step": 54964 }, { "epoch": 3.32, "learning_rate": 2.1046278273203266e-05, "loss": 0.5375, "step": 54966 }, { "epoch": 3.32, "learning_rate": 2.1044769363089042e-05, "loss": 0.5453, "step": 54968 }, { "epoch": 3.32, "learning_rate": 2.1043260452974815e-05, "loss": 0.4445, "step": 54970 }, { "epoch": 3.32, "learning_rate": 2.1041751542860595e-05, "loss": 0.4633, "step": 54972 }, { "epoch": 3.32, "learning_rate": 2.104024263274637e-05, "loss": 0.3356, "step": 54974 }, { "epoch": 3.32, "learning_rate": 2.1038733722632145e-05, "loss": 0.3274, "step": 54976 }, { "epoch": 3.32, "learning_rate": 2.1037224812517918e-05, "loss": 0.3552, "step": 54978 }, { "epoch": 3.32, "learning_rate": 2.1035715902403695e-05, "loss": 0.3967, "step": 54980 }, { "epoch": 3.32, "learning_rate": 2.103420699228947e-05, "loss": 0.3091, "step": 54982 }, { "epoch": 3.32, "learning_rate": 2.1032698082175244e-05, "loss": 0.3548, "step": 54984 }, { "epoch": 3.32, "learning_rate": 2.103118917206102e-05, "loss": 0.3253, "step": 54986 }, { "epoch": 3.32, "learning_rate": 2.1029680261946797e-05, "loss": 0.2616, "step": 54988 }, { "epoch": 3.32, "learning_rate": 2.1028171351832574e-05, "loss": 0.3999, "step": 54990 }, { "epoch": 3.32, "learning_rate": 2.1026662441718347e-05, "loss": 0.4083, "step": 54992 }, { "epoch": 3.32, "learning_rate": 2.1025153531604123e-05, "loss": 0.4301, "step": 54994 }, { "epoch": 3.32, "learning_rate": 2.10236446214899e-05, "loss": 0.3781, "step": 54996 }, { "epoch": 3.32, "learning_rate": 2.1022135711375673e-05, "loss": 0.3953, "step": 54998 }, { "epoch": 3.32, "learning_rate": 2.102062680126145e-05, "loss": 0.3247, "step": 55000 }, { "epoch": 3.32, "learning_rate": 2.1019117891147226e-05, "loss": 0.4108, "step": 55002 }, { "epoch": 3.32, "learning_rate": 2.1017608981033003e-05, "loss": 0.6945, "step": 55004 }, { "epoch": 3.32, "learning_rate": 2.1016100070918776e-05, "loss": 0.4292, "step": 55006 }, { "epoch": 3.32, "learning_rate": 2.1014591160804552e-05, "loss": 0.4587, "step": 55008 }, { "epoch": 3.32, "learning_rate": 2.101308225069033e-05, "loss": 0.4364, "step": 55010 }, { "epoch": 3.32, "learning_rate": 2.1011573340576102e-05, "loss": 0.2403, "step": 55012 }, { "epoch": 3.32, "learning_rate": 2.101006443046188e-05, "loss": 0.3577, "step": 55014 }, { "epoch": 3.32, "learning_rate": 2.100855552034765e-05, "loss": 0.6094, "step": 55016 }, { "epoch": 3.32, "learning_rate": 2.100704661023343e-05, "loss": 0.3482, "step": 55018 }, { "epoch": 3.32, "learning_rate": 2.1005537700119204e-05, "loss": 0.3128, "step": 55020 }, { "epoch": 3.32, "learning_rate": 2.100402879000498e-05, "loss": 0.362, "step": 55022 }, { "epoch": 3.32, "learning_rate": 2.1002519879890754e-05, "loss": 0.4341, "step": 55024 }, { "epoch": 3.32, "learning_rate": 2.100101096977653e-05, "loss": 0.3798, "step": 55026 }, { "epoch": 3.32, "learning_rate": 2.0999502059662307e-05, "loss": 0.3955, "step": 55028 }, { "epoch": 3.32, "learning_rate": 2.099799314954808e-05, "loss": 0.3406, "step": 55030 }, { "epoch": 3.32, "learning_rate": 2.0996484239433857e-05, "loss": 0.4158, "step": 55032 }, { "epoch": 3.32, "learning_rate": 2.0994975329319633e-05, "loss": 0.3568, "step": 55034 }, { "epoch": 3.32, "learning_rate": 2.099346641920541e-05, "loss": 0.397, "step": 55036 }, { "epoch": 3.32, "learning_rate": 2.0991957509091183e-05, "loss": 0.2881, "step": 55038 }, { "epoch": 3.32, "learning_rate": 2.099044859897696e-05, "loss": 0.8046, "step": 55040 }, { "epoch": 3.32, "learning_rate": 2.0988939688862736e-05, "loss": 0.3859, "step": 55042 }, { "epoch": 3.32, "learning_rate": 2.098743077874851e-05, "loss": 0.2008, "step": 55044 }, { "epoch": 3.32, "learning_rate": 2.0985921868634286e-05, "loss": 0.3178, "step": 55046 }, { "epoch": 3.32, "learning_rate": 2.0984412958520062e-05, "loss": 0.486, "step": 55048 }, { "epoch": 3.32, "learning_rate": 2.098290404840584e-05, "loss": 0.403, "step": 55050 }, { "epoch": 3.32, "learning_rate": 2.0981395138291612e-05, "loss": 0.268, "step": 55052 }, { "epoch": 3.32, "learning_rate": 2.0979886228177388e-05, "loss": 0.4213, "step": 55054 }, { "epoch": 3.32, "learning_rate": 2.0978377318063165e-05, "loss": 0.2712, "step": 55056 }, { "epoch": 3.32, "learning_rate": 2.097762286300605e-05, "loss": 0.3325, "step": 55058 }, { "epoch": 3.32, "learning_rate": 2.0976113952891828e-05, "loss": 0.4217, "step": 55060 }, { "epoch": 3.32, "learning_rate": 2.0974605042777604e-05, "loss": 0.4529, "step": 55062 }, { "epoch": 3.32, "learning_rate": 2.0973096132663378e-05, "loss": 0.2585, "step": 55064 }, { "epoch": 3.32, "learning_rate": 2.0971587222549154e-05, "loss": 0.3385, "step": 55066 }, { "epoch": 3.32, "learning_rate": 2.097007831243493e-05, "loss": 0.3362, "step": 55068 }, { "epoch": 3.32, "learning_rate": 2.0968569402320704e-05, "loss": 0.2763, "step": 55070 }, { "epoch": 3.32, "learning_rate": 2.096706049220648e-05, "loss": 0.3583, "step": 55072 }, { "epoch": 3.32, "learning_rate": 2.0965551582092253e-05, "loss": 0.381, "step": 55074 }, { "epoch": 3.32, "learning_rate": 2.0964042671978033e-05, "loss": 0.4596, "step": 55076 }, { "epoch": 3.32, "learning_rate": 2.0962533761863806e-05, "loss": 0.4383, "step": 55078 }, { "epoch": 3.32, "learning_rate": 2.0961024851749583e-05, "loss": 0.4629, "step": 55080 }, { "epoch": 3.32, "learning_rate": 2.0959515941635356e-05, "loss": 0.4468, "step": 55082 }, { "epoch": 3.32, "learning_rate": 2.0958007031521136e-05, "loss": 0.4822, "step": 55084 }, { "epoch": 3.32, "learning_rate": 2.095649812140691e-05, "loss": 0.2291, "step": 55086 }, { "epoch": 3.32, "learning_rate": 2.0954989211292682e-05, "loss": 0.3393, "step": 55088 }, { "epoch": 3.33, "learning_rate": 2.095348030117846e-05, "loss": 0.285, "step": 55090 }, { "epoch": 3.33, "learning_rate": 2.0951971391064235e-05, "loss": 0.4437, "step": 55092 }, { "epoch": 3.33, "learning_rate": 2.095046248095001e-05, "loss": 0.3654, "step": 55094 }, { "epoch": 3.33, "learning_rate": 2.0948953570835785e-05, "loss": 0.6, "step": 55096 }, { "epoch": 3.33, "learning_rate": 2.094744466072156e-05, "loss": 0.335, "step": 55098 }, { "epoch": 3.33, "learning_rate": 2.0945935750607338e-05, "loss": 0.426, "step": 55100 }, { "epoch": 3.33, "learning_rate": 2.094442684049311e-05, "loss": 0.4436, "step": 55102 }, { "epoch": 3.33, "learning_rate": 2.0942917930378887e-05, "loss": 0.4544, "step": 55104 }, { "epoch": 3.33, "learning_rate": 2.0941409020264664e-05, "loss": 0.3788, "step": 55106 }, { "epoch": 3.33, "learning_rate": 2.093990011015044e-05, "loss": 0.3526, "step": 55108 }, { "epoch": 3.33, "learning_rate": 2.0938391200036214e-05, "loss": 0.4279, "step": 55110 }, { "epoch": 3.33, "learning_rate": 2.093688228992199e-05, "loss": 0.3798, "step": 55112 }, { "epoch": 3.33, "learning_rate": 2.0935373379807767e-05, "loss": 0.3178, "step": 55114 }, { "epoch": 3.33, "learning_rate": 2.0933864469693543e-05, "loss": 0.4734, "step": 55116 }, { "epoch": 3.33, "learning_rate": 2.0932355559579316e-05, "loss": 0.2886, "step": 55118 }, { "epoch": 3.33, "learning_rate": 2.0930846649465093e-05, "loss": 0.2713, "step": 55120 }, { "epoch": 3.33, "learning_rate": 2.092933773935087e-05, "loss": 0.3091, "step": 55122 }, { "epoch": 3.33, "learning_rate": 2.0927828829236642e-05, "loss": 0.3707, "step": 55124 }, { "epoch": 3.33, "learning_rate": 2.092631991912242e-05, "loss": 0.2549, "step": 55126 }, { "epoch": 3.33, "learning_rate": 2.0924811009008195e-05, "loss": 0.3538, "step": 55128 }, { "epoch": 3.33, "learning_rate": 2.0923302098893972e-05, "loss": 0.5468, "step": 55130 }, { "epoch": 3.33, "learning_rate": 2.0921793188779745e-05, "loss": 0.5002, "step": 55132 }, { "epoch": 3.33, "learning_rate": 2.0920284278665518e-05, "loss": 0.4433, "step": 55134 }, { "epoch": 3.33, "learning_rate": 2.0918775368551298e-05, "loss": 0.4479, "step": 55136 }, { "epoch": 3.33, "learning_rate": 2.091726645843707e-05, "loss": 0.2819, "step": 55138 }, { "epoch": 3.33, "learning_rate": 2.0915757548322848e-05, "loss": 0.2737, "step": 55140 }, { "epoch": 3.33, "learning_rate": 2.091424863820862e-05, "loss": 0.3282, "step": 55142 }, { "epoch": 3.33, "learning_rate": 2.09127397280944e-05, "loss": 0.3598, "step": 55144 }, { "epoch": 3.33, "learning_rate": 2.0911230817980174e-05, "loss": 0.4131, "step": 55146 }, { "epoch": 3.33, "learning_rate": 2.090972190786595e-05, "loss": 0.317, "step": 55148 }, { "epoch": 3.33, "learning_rate": 2.0908212997751724e-05, "loss": 0.3475, "step": 55150 }, { "epoch": 3.33, "learning_rate": 2.09067040876375e-05, "loss": 0.3421, "step": 55152 }, { "epoch": 3.33, "learning_rate": 2.0905195177523277e-05, "loss": 0.462, "step": 55154 }, { "epoch": 3.33, "learning_rate": 2.090368626740905e-05, "loss": 0.3988, "step": 55156 }, { "epoch": 3.33, "learning_rate": 2.0902177357294826e-05, "loss": 0.3036, "step": 55158 }, { "epoch": 3.33, "learning_rate": 2.0900668447180603e-05, "loss": 0.3668, "step": 55160 }, { "epoch": 3.33, "learning_rate": 2.089915953706638e-05, "loss": 0.2803, "step": 55162 }, { "epoch": 3.33, "learning_rate": 2.0897650626952152e-05, "loss": 0.3984, "step": 55164 }, { "epoch": 3.33, "learning_rate": 2.089614171683793e-05, "loss": 0.3328, "step": 55166 }, { "epoch": 3.33, "learning_rate": 2.0894632806723705e-05, "loss": 0.4526, "step": 55168 }, { "epoch": 3.33, "learning_rate": 2.089312389660948e-05, "loss": 0.3515, "step": 55170 }, { "epoch": 3.33, "learning_rate": 2.0891614986495255e-05, "loss": 0.395, "step": 55172 }, { "epoch": 3.33, "learning_rate": 2.089010607638103e-05, "loss": 0.3566, "step": 55174 }, { "epoch": 3.33, "learning_rate": 2.0888597166266808e-05, "loss": 0.3053, "step": 55176 }, { "epoch": 3.33, "learning_rate": 2.088708825615258e-05, "loss": 0.5384, "step": 55178 }, { "epoch": 3.33, "learning_rate": 2.0885579346038358e-05, "loss": 0.2898, "step": 55180 }, { "epoch": 3.33, "learning_rate": 2.0884070435924134e-05, "loss": 0.3379, "step": 55182 }, { "epoch": 3.33, "learning_rate": 2.0882561525809907e-05, "loss": 0.4654, "step": 55184 }, { "epoch": 3.33, "learning_rate": 2.0881052615695684e-05, "loss": 0.3804, "step": 55186 }, { "epoch": 3.33, "learning_rate": 2.0879543705581457e-05, "loss": 0.4542, "step": 55188 }, { "epoch": 3.33, "learning_rate": 2.0878034795467237e-05, "loss": 0.2479, "step": 55190 }, { "epoch": 3.33, "learning_rate": 2.087652588535301e-05, "loss": 0.4933, "step": 55192 }, { "epoch": 3.33, "learning_rate": 2.0875016975238787e-05, "loss": 0.3032, "step": 55194 }, { "epoch": 3.33, "learning_rate": 2.087350806512456e-05, "loss": 0.3781, "step": 55196 }, { "epoch": 3.33, "learning_rate": 2.087199915501034e-05, "loss": 0.3917, "step": 55198 }, { "epoch": 3.33, "learning_rate": 2.0870490244896113e-05, "loss": 0.2885, "step": 55200 }, { "epoch": 3.33, "learning_rate": 2.0868981334781886e-05, "loss": 0.2863, "step": 55202 }, { "epoch": 3.33, "learning_rate": 2.0867472424667662e-05, "loss": 0.415, "step": 55204 }, { "epoch": 3.33, "learning_rate": 2.086596351455344e-05, "loss": 0.4026, "step": 55206 }, { "epoch": 3.33, "learning_rate": 2.0864454604439215e-05, "loss": 0.3799, "step": 55208 }, { "epoch": 3.33, "learning_rate": 2.086294569432499e-05, "loss": 0.3667, "step": 55210 }, { "epoch": 3.33, "learning_rate": 2.0861436784210765e-05, "loss": 0.3799, "step": 55212 }, { "epoch": 3.33, "learning_rate": 2.085992787409654e-05, "loss": 0.2275, "step": 55214 }, { "epoch": 3.33, "learning_rate": 2.0858418963982315e-05, "loss": 0.4493, "step": 55216 }, { "epoch": 3.33, "learning_rate": 2.085691005386809e-05, "loss": 0.4639, "step": 55218 }, { "epoch": 3.33, "learning_rate": 2.0855401143753868e-05, "loss": 0.425, "step": 55220 }, { "epoch": 3.33, "learning_rate": 2.0853892233639644e-05, "loss": 0.3776, "step": 55222 }, { "epoch": 3.33, "learning_rate": 2.0852383323525417e-05, "loss": 0.3617, "step": 55224 }, { "epoch": 3.33, "learning_rate": 2.0850874413411194e-05, "loss": 0.5347, "step": 55226 }, { "epoch": 3.33, "learning_rate": 2.084936550329697e-05, "loss": 0.4987, "step": 55228 }, { "epoch": 3.33, "learning_rate": 2.0847856593182747e-05, "loss": 0.271, "step": 55230 }, { "epoch": 3.33, "learning_rate": 2.084634768306852e-05, "loss": 0.3514, "step": 55232 }, { "epoch": 3.33, "learning_rate": 2.0844838772954296e-05, "loss": 0.4972, "step": 55234 }, { "epoch": 3.33, "learning_rate": 2.0843329862840073e-05, "loss": 0.2247, "step": 55236 }, { "epoch": 3.33, "learning_rate": 2.0841820952725846e-05, "loss": 0.2698, "step": 55238 }, { "epoch": 3.33, "learning_rate": 2.0840312042611623e-05, "loss": 0.3204, "step": 55240 }, { "epoch": 3.33, "learning_rate": 2.08388031324974e-05, "loss": 0.3526, "step": 55242 }, { "epoch": 3.33, "learning_rate": 2.0837294222383176e-05, "loss": 0.3446, "step": 55244 }, { "epoch": 3.33, "learning_rate": 2.083578531226895e-05, "loss": 0.2975, "step": 55246 }, { "epoch": 3.33, "learning_rate": 2.0834276402154725e-05, "loss": 0.3662, "step": 55248 }, { "epoch": 3.33, "learning_rate": 2.0832767492040502e-05, "loss": 0.3814, "step": 55250 }, { "epoch": 3.33, "learning_rate": 2.0831258581926275e-05, "loss": 0.4072, "step": 55252 }, { "epoch": 3.33, "learning_rate": 2.082974967181205e-05, "loss": 0.3557, "step": 55254 }, { "epoch": 3.34, "learning_rate": 2.0828240761697825e-05, "loss": 0.4812, "step": 55256 }, { "epoch": 3.34, "learning_rate": 2.0826731851583604e-05, "loss": 0.4107, "step": 55258 }, { "epoch": 3.34, "learning_rate": 2.0825222941469378e-05, "loss": 0.4075, "step": 55260 }, { "epoch": 3.34, "learning_rate": 2.0823714031355154e-05, "loss": 0.3506, "step": 55262 }, { "epoch": 3.34, "learning_rate": 2.0822205121240927e-05, "loss": 0.3473, "step": 55264 }, { "epoch": 3.34, "learning_rate": 2.0820696211126704e-05, "loss": 0.2581, "step": 55266 }, { "epoch": 3.34, "learning_rate": 2.081918730101248e-05, "loss": 0.4384, "step": 55268 }, { "epoch": 3.34, "learning_rate": 2.0817678390898253e-05, "loss": 0.4833, "step": 55270 }, { "epoch": 3.34, "learning_rate": 2.081616948078403e-05, "loss": 0.4031, "step": 55272 }, { "epoch": 3.34, "learning_rate": 2.0814660570669806e-05, "loss": 0.3907, "step": 55274 }, { "epoch": 3.34, "learning_rate": 2.0813151660555583e-05, "loss": 0.3545, "step": 55276 }, { "epoch": 3.34, "learning_rate": 2.0811642750441356e-05, "loss": 0.3881, "step": 55278 }, { "epoch": 3.34, "learning_rate": 2.0810133840327133e-05, "loss": 0.346, "step": 55280 }, { "epoch": 3.34, "learning_rate": 2.080862493021291e-05, "loss": 0.2998, "step": 55282 }, { "epoch": 3.34, "learning_rate": 2.0807116020098682e-05, "loss": 0.4354, "step": 55284 }, { "epoch": 3.34, "learning_rate": 2.080560710998446e-05, "loss": 0.2677, "step": 55286 }, { "epoch": 3.34, "learning_rate": 2.0804098199870235e-05, "loss": 0.4154, "step": 55288 }, { "epoch": 3.34, "learning_rate": 2.0802589289756012e-05, "loss": 0.4426, "step": 55290 }, { "epoch": 3.34, "learning_rate": 2.0801080379641785e-05, "loss": 0.3825, "step": 55292 }, { "epoch": 3.34, "learning_rate": 2.079957146952756e-05, "loss": 0.3435, "step": 55294 }, { "epoch": 3.34, "learning_rate": 2.0798062559413338e-05, "loss": 0.1794, "step": 55296 }, { "epoch": 3.34, "learning_rate": 2.079655364929911e-05, "loss": 0.3527, "step": 55298 }, { "epoch": 3.34, "learning_rate": 2.0795044739184888e-05, "loss": 0.4398, "step": 55300 }, { "epoch": 3.34, "learning_rate": 2.079353582907066e-05, "loss": 0.7315, "step": 55302 }, { "epoch": 3.34, "learning_rate": 2.079202691895644e-05, "loss": 0.3643, "step": 55304 }, { "epoch": 3.34, "learning_rate": 2.0790518008842214e-05, "loss": 0.5341, "step": 55306 }, { "epoch": 3.34, "learning_rate": 2.078900909872799e-05, "loss": 0.3752, "step": 55308 }, { "epoch": 3.34, "learning_rate": 2.0787500188613763e-05, "loss": 0.4319, "step": 55310 }, { "epoch": 3.34, "learning_rate": 2.0785991278499543e-05, "loss": 0.3347, "step": 55312 }, { "epoch": 3.34, "learning_rate": 2.0784482368385316e-05, "loss": 0.3699, "step": 55314 }, { "epoch": 3.34, "learning_rate": 2.078297345827109e-05, "loss": 0.2304, "step": 55316 }, { "epoch": 3.34, "learning_rate": 2.0781464548156866e-05, "loss": 0.3103, "step": 55318 }, { "epoch": 3.34, "learning_rate": 2.0779955638042643e-05, "loss": 0.3986, "step": 55320 }, { "epoch": 3.34, "learning_rate": 2.077844672792842e-05, "loss": 0.4296, "step": 55322 }, { "epoch": 3.34, "learning_rate": 2.0776937817814192e-05, "loss": 0.2782, "step": 55324 }, { "epoch": 3.34, "learning_rate": 2.077542890769997e-05, "loss": 0.6148, "step": 55326 }, { "epoch": 3.34, "learning_rate": 2.0773919997585745e-05, "loss": 0.4155, "step": 55328 }, { "epoch": 3.34, "learning_rate": 2.077241108747152e-05, "loss": 0.2988, "step": 55330 }, { "epoch": 3.34, "learning_rate": 2.0770902177357295e-05, "loss": 0.3523, "step": 55332 }, { "epoch": 3.34, "learning_rate": 2.076939326724307e-05, "loss": 0.4324, "step": 55334 }, { "epoch": 3.34, "learning_rate": 2.0767884357128848e-05, "loss": 0.5011, "step": 55336 }, { "epoch": 3.34, "learning_rate": 2.076637544701462e-05, "loss": 0.3107, "step": 55338 }, { "epoch": 3.34, "learning_rate": 2.0764866536900397e-05, "loss": 0.2918, "step": 55340 }, { "epoch": 3.34, "learning_rate": 2.0763357626786174e-05, "loss": 0.3878, "step": 55342 }, { "epoch": 3.34, "learning_rate": 2.076184871667195e-05, "loss": 0.4049, "step": 55344 }, { "epoch": 3.34, "learning_rate": 2.0760339806557724e-05, "loss": 0.3726, "step": 55346 }, { "epoch": 3.34, "learning_rate": 2.07588308964435e-05, "loss": 0.4218, "step": 55348 }, { "epoch": 3.34, "learning_rate": 2.0757321986329277e-05, "loss": 0.2768, "step": 55350 }, { "epoch": 3.34, "learning_rate": 2.075581307621505e-05, "loss": 0.4538, "step": 55352 }, { "epoch": 3.34, "learning_rate": 2.0754304166100826e-05, "loss": 0.3954, "step": 55354 }, { "epoch": 3.34, "learning_rate": 2.0752795255986603e-05, "loss": 0.2678, "step": 55356 }, { "epoch": 3.34, "learning_rate": 2.075128634587238e-05, "loss": 0.2735, "step": 55358 }, { "epoch": 3.34, "learning_rate": 2.0749777435758152e-05, "loss": 0.5123, "step": 55360 }, { "epoch": 3.34, "learning_rate": 2.074826852564393e-05, "loss": 0.5973, "step": 55362 }, { "epoch": 3.34, "learning_rate": 2.0746759615529705e-05, "loss": 0.4363, "step": 55364 }, { "epoch": 3.34, "learning_rate": 2.074525070541548e-05, "loss": 0.3933, "step": 55366 }, { "epoch": 3.34, "learning_rate": 2.0743741795301255e-05, "loss": 0.3358, "step": 55368 }, { "epoch": 3.34, "learning_rate": 2.0742232885187028e-05, "loss": 0.4819, "step": 55370 }, { "epoch": 3.34, "learning_rate": 2.0740723975072808e-05, "loss": 0.432, "step": 55372 }, { "epoch": 3.34, "learning_rate": 2.073921506495858e-05, "loss": 0.4138, "step": 55374 }, { "epoch": 3.34, "learning_rate": 2.0737706154844358e-05, "loss": 0.4202, "step": 55376 }, { "epoch": 3.34, "learning_rate": 2.073619724473013e-05, "loss": 0.2619, "step": 55378 }, { "epoch": 3.34, "learning_rate": 2.0734688334615907e-05, "loss": 0.3791, "step": 55380 }, { "epoch": 3.34, "learning_rate": 2.0733179424501684e-05, "loss": 0.3155, "step": 55382 }, { "epoch": 3.34, "learning_rate": 2.0731670514387457e-05, "loss": 0.4153, "step": 55384 }, { "epoch": 3.34, "learning_rate": 2.0730161604273234e-05, "loss": 0.4926, "step": 55386 }, { "epoch": 3.34, "learning_rate": 2.072865269415901e-05, "loss": 0.4048, "step": 55388 }, { "epoch": 3.34, "learning_rate": 2.0727143784044787e-05, "loss": 0.5959, "step": 55390 }, { "epoch": 3.34, "learning_rate": 2.072563487393056e-05, "loss": 0.4129, "step": 55392 }, { "epoch": 3.34, "learning_rate": 2.0724125963816336e-05, "loss": 0.3286, "step": 55394 }, { "epoch": 3.34, "learning_rate": 2.0722617053702113e-05, "loss": 0.5113, "step": 55396 }, { "epoch": 3.34, "learning_rate": 2.0721108143587886e-05, "loss": 0.3061, "step": 55398 }, { "epoch": 3.34, "learning_rate": 2.0719599233473662e-05, "loss": 0.4903, "step": 55400 }, { "epoch": 3.34, "learning_rate": 2.071809032335944e-05, "loss": 0.3561, "step": 55402 }, { "epoch": 3.34, "learning_rate": 2.0716581413245215e-05, "loss": 0.3085, "step": 55404 }, { "epoch": 3.34, "learning_rate": 2.071507250313099e-05, "loss": 0.2386, "step": 55406 }, { "epoch": 3.34, "learning_rate": 2.0713563593016765e-05, "loss": 0.3394, "step": 55408 }, { "epoch": 3.34, "learning_rate": 2.071205468290254e-05, "loss": 0.2705, "step": 55410 }, { "epoch": 3.34, "learning_rate": 2.0710545772788318e-05, "loss": 0.3445, "step": 55412 }, { "epoch": 3.34, "learning_rate": 2.070903686267409e-05, "loss": 0.2647, "step": 55414 }, { "epoch": 3.34, "learning_rate": 2.0707527952559864e-05, "loss": 0.3682, "step": 55416 }, { "epoch": 3.34, "learning_rate": 2.0706019042445644e-05, "loss": 0.5097, "step": 55418 }, { "epoch": 3.35, "learning_rate": 2.0704510132331417e-05, "loss": 0.4269, "step": 55420 }, { "epoch": 3.35, "learning_rate": 2.0703001222217194e-05, "loss": 0.3789, "step": 55422 }, { "epoch": 3.35, "learning_rate": 2.0701492312102967e-05, "loss": 0.3394, "step": 55424 }, { "epoch": 3.35, "learning_rate": 2.0699983401988747e-05, "loss": 0.3296, "step": 55426 }, { "epoch": 3.35, "learning_rate": 2.069847449187452e-05, "loss": 0.2592, "step": 55428 }, { "epoch": 3.35, "learning_rate": 2.0696965581760293e-05, "loss": 0.3288, "step": 55430 }, { "epoch": 3.35, "learning_rate": 2.069545667164607e-05, "loss": 0.4452, "step": 55432 }, { "epoch": 3.35, "learning_rate": 2.0693947761531846e-05, "loss": 0.2808, "step": 55434 }, { "epoch": 3.35, "learning_rate": 2.0692438851417623e-05, "loss": 0.3285, "step": 55436 }, { "epoch": 3.35, "learning_rate": 2.0690929941303396e-05, "loss": 0.301, "step": 55438 }, { "epoch": 3.35, "learning_rate": 2.0689421031189172e-05, "loss": 0.5597, "step": 55440 }, { "epoch": 3.35, "learning_rate": 2.068791212107495e-05, "loss": 0.4006, "step": 55442 }, { "epoch": 3.35, "learning_rate": 2.0687157666017835e-05, "loss": 0.5069, "step": 55444 }, { "epoch": 3.35, "learning_rate": 2.0685648755903612e-05, "loss": 0.4245, "step": 55446 }, { "epoch": 3.35, "learning_rate": 2.068413984578939e-05, "loss": 0.3109, "step": 55448 }, { "epoch": 3.35, "learning_rate": 2.068263093567516e-05, "loss": 0.508, "step": 55450 }, { "epoch": 3.35, "learning_rate": 2.0681122025560938e-05, "loss": 0.3382, "step": 55452 }, { "epoch": 3.35, "learning_rate": 2.0679613115446715e-05, "loss": 0.3375, "step": 55454 }, { "epoch": 3.35, "learning_rate": 2.0678104205332488e-05, "loss": 0.326, "step": 55456 }, { "epoch": 3.35, "learning_rate": 2.0676595295218264e-05, "loss": 0.369, "step": 55458 }, { "epoch": 3.35, "learning_rate": 2.067508638510404e-05, "loss": 0.424, "step": 55460 }, { "epoch": 3.35, "learning_rate": 2.0673577474989817e-05, "loss": 0.3985, "step": 55462 }, { "epoch": 3.35, "learning_rate": 2.067206856487559e-05, "loss": 0.3701, "step": 55464 }, { "epoch": 3.35, "learning_rate": 2.0670559654761367e-05, "loss": 0.3475, "step": 55466 }, { "epoch": 3.35, "learning_rate": 2.0669050744647143e-05, "loss": 0.4729, "step": 55468 }, { "epoch": 3.35, "learning_rate": 2.066754183453292e-05, "loss": 0.37, "step": 55470 }, { "epoch": 3.35, "learning_rate": 2.0666032924418693e-05, "loss": 0.332, "step": 55472 }, { "epoch": 3.35, "learning_rate": 2.0664524014304466e-05, "loss": 0.4007, "step": 55474 }, { "epoch": 3.35, "learning_rate": 2.0663015104190246e-05, "loss": 0.32, "step": 55476 }, { "epoch": 3.35, "learning_rate": 2.066150619407602e-05, "loss": 0.3245, "step": 55478 }, { "epoch": 3.35, "learning_rate": 2.0659997283961796e-05, "loss": 0.393, "step": 55480 }, { "epoch": 3.35, "learning_rate": 2.065848837384757e-05, "loss": 0.3608, "step": 55482 }, { "epoch": 3.35, "learning_rate": 2.065697946373335e-05, "loss": 0.3587, "step": 55484 }, { "epoch": 3.35, "learning_rate": 2.0655470553619122e-05, "loss": 0.4305, "step": 55486 }, { "epoch": 3.35, "learning_rate": 2.0653961643504895e-05, "loss": 0.4438, "step": 55488 }, { "epoch": 3.35, "learning_rate": 2.065245273339067e-05, "loss": 0.2595, "step": 55490 }, { "epoch": 3.35, "learning_rate": 2.0650943823276448e-05, "loss": 0.3738, "step": 55492 }, { "epoch": 3.35, "learning_rate": 2.0649434913162225e-05, "loss": 0.2642, "step": 55494 }, { "epoch": 3.35, "learning_rate": 2.0647926003047998e-05, "loss": 0.3992, "step": 55496 }, { "epoch": 3.35, "learning_rate": 2.0646417092933774e-05, "loss": 0.3451, "step": 55498 }, { "epoch": 3.35, "learning_rate": 2.064490818281955e-05, "loss": 0.5498, "step": 55500 }, { "epoch": 3.35, "learning_rate": 2.0643399272705327e-05, "loss": 0.3333, "step": 55502 }, { "epoch": 3.35, "learning_rate": 2.06418903625911e-05, "loss": 0.4372, "step": 55504 }, { "epoch": 3.35, "learning_rate": 2.0640381452476877e-05, "loss": 0.5525, "step": 55506 }, { "epoch": 3.35, "learning_rate": 2.0638872542362653e-05, "loss": 0.4822, "step": 55508 }, { "epoch": 3.35, "learning_rate": 2.0637363632248426e-05, "loss": 0.4301, "step": 55510 }, { "epoch": 3.35, "learning_rate": 2.0635854722134203e-05, "loss": 0.4596, "step": 55512 }, { "epoch": 3.35, "learning_rate": 2.063434581201998e-05, "loss": 0.4073, "step": 55514 }, { "epoch": 3.35, "learning_rate": 2.0632836901905756e-05, "loss": 0.4304, "step": 55516 }, { "epoch": 3.35, "learning_rate": 2.063132799179153e-05, "loss": 0.3116, "step": 55518 }, { "epoch": 3.35, "learning_rate": 2.0629819081677306e-05, "loss": 0.4119, "step": 55520 }, { "epoch": 3.35, "learning_rate": 2.0628310171563082e-05, "loss": 0.3094, "step": 55522 }, { "epoch": 3.35, "learning_rate": 2.0626801261448855e-05, "loss": 0.3495, "step": 55524 }, { "epoch": 3.35, "learning_rate": 2.0625292351334632e-05, "loss": 0.3317, "step": 55526 }, { "epoch": 3.35, "learning_rate": 2.062378344122041e-05, "loss": 0.4901, "step": 55528 }, { "epoch": 3.35, "learning_rate": 2.0622274531106185e-05, "loss": 0.5774, "step": 55530 }, { "epoch": 3.35, "learning_rate": 2.0620765620991958e-05, "loss": 0.3884, "step": 55532 }, { "epoch": 3.35, "learning_rate": 2.0619256710877734e-05, "loss": 0.3854, "step": 55534 }, { "epoch": 3.35, "learning_rate": 2.061774780076351e-05, "loss": 0.3289, "step": 55536 }, { "epoch": 3.35, "learning_rate": 2.0616238890649284e-05, "loss": 0.4465, "step": 55538 }, { "epoch": 3.35, "learning_rate": 2.061472998053506e-05, "loss": 0.4955, "step": 55540 }, { "epoch": 3.35, "learning_rate": 2.0613221070420834e-05, "loss": 0.393, "step": 55542 }, { "epoch": 3.35, "learning_rate": 2.0611712160306614e-05, "loss": 0.3257, "step": 55544 }, { "epoch": 3.35, "learning_rate": 2.0610203250192387e-05, "loss": 0.4271, "step": 55546 }, { "epoch": 3.35, "learning_rate": 2.0608694340078163e-05, "loss": 0.2787, "step": 55548 }, { "epoch": 3.35, "learning_rate": 2.0607185429963936e-05, "loss": 0.3919, "step": 55550 }, { "epoch": 3.35, "learning_rate": 2.0605676519849716e-05, "loss": 0.3753, "step": 55552 }, { "epoch": 3.35, "learning_rate": 2.060416760973549e-05, "loss": 0.5556, "step": 55554 }, { "epoch": 3.35, "learning_rate": 2.0602658699621263e-05, "loss": 0.3491, "step": 55556 }, { "epoch": 3.35, "learning_rate": 2.060114978950704e-05, "loss": 0.2725, "step": 55558 }, { "epoch": 3.35, "learning_rate": 2.0599640879392816e-05, "loss": 0.3346, "step": 55560 }, { "epoch": 3.35, "learning_rate": 2.0598131969278592e-05, "loss": 0.3286, "step": 55562 }, { "epoch": 3.35, "learning_rate": 2.0596623059164365e-05, "loss": 0.2599, "step": 55564 }, { "epoch": 3.35, "learning_rate": 2.0595114149050142e-05, "loss": 0.5106, "step": 55566 }, { "epoch": 3.35, "learning_rate": 2.0593605238935918e-05, "loss": 0.2792, "step": 55568 }, { "epoch": 3.35, "learning_rate": 2.059209632882169e-05, "loss": 0.5635, "step": 55570 }, { "epoch": 3.35, "learning_rate": 2.0590587418707468e-05, "loss": 0.2401, "step": 55572 }, { "epoch": 3.35, "learning_rate": 2.0589078508593244e-05, "loss": 0.3789, "step": 55574 }, { "epoch": 3.35, "learning_rate": 2.058756959847902e-05, "loss": 0.2398, "step": 55576 }, { "epoch": 3.35, "learning_rate": 2.0586060688364794e-05, "loss": 0.5105, "step": 55578 }, { "epoch": 3.35, "learning_rate": 2.058455177825057e-05, "loss": 0.4573, "step": 55580 }, { "epoch": 3.35, "learning_rate": 2.0583042868136347e-05, "loss": 0.389, "step": 55582 }, { "epoch": 3.35, "learning_rate": 2.0581533958022124e-05, "loss": 0.4708, "step": 55584 }, { "epoch": 3.36, "learning_rate": 2.0580025047907897e-05, "loss": 0.453, "step": 55586 }, { "epoch": 3.36, "learning_rate": 2.057851613779367e-05, "loss": 0.3098, "step": 55588 }, { "epoch": 3.36, "learning_rate": 2.057700722767945e-05, "loss": 0.2837, "step": 55590 }, { "epoch": 3.36, "learning_rate": 2.0575498317565223e-05, "loss": 0.3129, "step": 55592 }, { "epoch": 3.36, "learning_rate": 2.0573989407451e-05, "loss": 0.3605, "step": 55594 }, { "epoch": 3.36, "learning_rate": 2.0572480497336773e-05, "loss": 0.3765, "step": 55596 }, { "epoch": 3.36, "learning_rate": 2.0570971587222552e-05, "loss": 0.49, "step": 55598 }, { "epoch": 3.36, "learning_rate": 2.0569462677108326e-05, "loss": 0.4739, "step": 55600 }, { "epoch": 3.36, "learning_rate": 2.0567953766994102e-05, "loss": 0.39, "step": 55602 }, { "epoch": 3.36, "learning_rate": 2.0566444856879875e-05, "loss": 0.2982, "step": 55604 }, { "epoch": 3.36, "learning_rate": 2.0564935946765652e-05, "loss": 0.3034, "step": 55606 }, { "epoch": 3.36, "learning_rate": 2.0563427036651428e-05, "loss": 0.2789, "step": 55608 }, { "epoch": 3.36, "learning_rate": 2.05619181265372e-05, "loss": 0.336, "step": 55610 }, { "epoch": 3.36, "learning_rate": 2.0560409216422978e-05, "loss": 0.3229, "step": 55612 }, { "epoch": 3.36, "learning_rate": 2.0558900306308754e-05, "loss": 0.459, "step": 55614 }, { "epoch": 3.36, "learning_rate": 2.055739139619453e-05, "loss": 0.3777, "step": 55616 }, { "epoch": 3.36, "learning_rate": 2.0555882486080304e-05, "loss": 0.5024, "step": 55618 }, { "epoch": 3.36, "learning_rate": 2.055437357596608e-05, "loss": 0.3757, "step": 55620 }, { "epoch": 3.36, "learning_rate": 2.0552864665851857e-05, "loss": 0.4087, "step": 55622 }, { "epoch": 3.36, "learning_rate": 2.055135575573763e-05, "loss": 0.2025, "step": 55624 }, { "epoch": 3.36, "learning_rate": 2.0549846845623407e-05, "loss": 0.3136, "step": 55626 }, { "epoch": 3.36, "learning_rate": 2.0548337935509183e-05, "loss": 0.4446, "step": 55628 }, { "epoch": 3.36, "learning_rate": 2.054682902539496e-05, "loss": 0.3271, "step": 55630 }, { "epoch": 3.36, "learning_rate": 2.0545320115280733e-05, "loss": 0.3485, "step": 55632 }, { "epoch": 3.36, "learning_rate": 2.054381120516651e-05, "loss": 0.4006, "step": 55634 }, { "epoch": 3.36, "learning_rate": 2.0542302295052286e-05, "loss": 0.4356, "step": 55636 }, { "epoch": 3.36, "learning_rate": 2.054079338493806e-05, "loss": 0.3838, "step": 55638 }, { "epoch": 3.36, "learning_rate": 2.0539284474823835e-05, "loss": 0.3115, "step": 55640 }, { "epoch": 3.36, "learning_rate": 2.0537775564709612e-05, "loss": 0.3595, "step": 55642 }, { "epoch": 3.36, "learning_rate": 2.053626665459539e-05, "loss": 0.4246, "step": 55644 }, { "epoch": 3.36, "learning_rate": 2.053475774448116e-05, "loss": 0.3765, "step": 55646 }, { "epoch": 3.36, "learning_rate": 2.0533248834366938e-05, "loss": 0.4419, "step": 55648 }, { "epoch": 3.36, "learning_rate": 2.0531739924252715e-05, "loss": 0.2764, "step": 55650 }, { "epoch": 3.36, "learning_rate": 2.0530231014138488e-05, "loss": 0.3975, "step": 55652 }, { "epoch": 3.36, "learning_rate": 2.0528722104024264e-05, "loss": 0.352, "step": 55654 }, { "epoch": 3.36, "learning_rate": 2.0527213193910037e-05, "loss": 0.327, "step": 55656 }, { "epoch": 3.36, "learning_rate": 2.0525704283795817e-05, "loss": 0.4711, "step": 55658 }, { "epoch": 3.36, "learning_rate": 2.052419537368159e-05, "loss": 0.4585, "step": 55660 }, { "epoch": 3.36, "learning_rate": 2.0522686463567367e-05, "loss": 0.3073, "step": 55662 }, { "epoch": 3.36, "learning_rate": 2.052117755345314e-05, "loss": 0.3728, "step": 55664 }, { "epoch": 3.36, "learning_rate": 2.051966864333892e-05, "loss": 0.4196, "step": 55666 }, { "epoch": 3.36, "learning_rate": 2.0518159733224693e-05, "loss": 0.3397, "step": 55668 }, { "epoch": 3.36, "learning_rate": 2.0516650823110466e-05, "loss": 0.343, "step": 55670 }, { "epoch": 3.36, "learning_rate": 2.0515141912996243e-05, "loss": 0.4169, "step": 55672 }, { "epoch": 3.36, "learning_rate": 2.051363300288202e-05, "loss": 0.3708, "step": 55674 }, { "epoch": 3.36, "learning_rate": 2.0512124092767796e-05, "loss": 0.4474, "step": 55676 }, { "epoch": 3.36, "learning_rate": 2.051061518265357e-05, "loss": 0.2737, "step": 55678 }, { "epoch": 3.36, "learning_rate": 2.0509106272539345e-05, "loss": 0.2622, "step": 55680 }, { "epoch": 3.36, "learning_rate": 2.0507597362425122e-05, "loss": 0.5385, "step": 55682 }, { "epoch": 3.36, "learning_rate": 2.05060884523109e-05, "loss": 0.4988, "step": 55684 }, { "epoch": 3.36, "learning_rate": 2.050457954219667e-05, "loss": 0.3783, "step": 55686 }, { "epoch": 3.36, "learning_rate": 2.0503070632082448e-05, "loss": 0.2396, "step": 55688 }, { "epoch": 3.36, "learning_rate": 2.0501561721968225e-05, "loss": 0.2377, "step": 55690 }, { "epoch": 3.36, "learning_rate": 2.0500052811853998e-05, "loss": 0.2883, "step": 55692 }, { "epoch": 3.36, "learning_rate": 2.0498543901739774e-05, "loss": 0.5808, "step": 55694 }, { "epoch": 3.36, "learning_rate": 2.049703499162555e-05, "loss": 0.3972, "step": 55696 }, { "epoch": 3.36, "learning_rate": 2.0495526081511327e-05, "loss": 0.3616, "step": 55698 }, { "epoch": 3.36, "learning_rate": 2.04940171713971e-05, "loss": 0.3834, "step": 55700 }, { "epoch": 3.36, "learning_rate": 2.0492508261282874e-05, "loss": 0.3333, "step": 55702 }, { "epoch": 3.36, "learning_rate": 2.0490999351168653e-05, "loss": 0.3738, "step": 55704 }, { "epoch": 3.36, "learning_rate": 2.0489490441054427e-05, "loss": 0.2874, "step": 55706 }, { "epoch": 3.36, "learning_rate": 2.0487981530940203e-05, "loss": 0.3895, "step": 55708 }, { "epoch": 3.36, "learning_rate": 2.0486472620825976e-05, "loss": 0.2514, "step": 55710 }, { "epoch": 3.36, "learning_rate": 2.0484963710711756e-05, "loss": 0.44, "step": 55712 }, { "epoch": 3.36, "learning_rate": 2.048345480059753e-05, "loss": 0.4311, "step": 55714 }, { "epoch": 3.36, "learning_rate": 2.0481945890483306e-05, "loss": 0.422, "step": 55716 }, { "epoch": 3.36, "learning_rate": 2.048043698036908e-05, "loss": 0.4083, "step": 55718 }, { "epoch": 3.36, "learning_rate": 2.0478928070254855e-05, "loss": 0.3444, "step": 55720 }, { "epoch": 3.36, "learning_rate": 2.0477419160140632e-05, "loss": 0.4205, "step": 55722 }, { "epoch": 3.36, "learning_rate": 2.0475910250026405e-05, "loss": 0.4886, "step": 55724 }, { "epoch": 3.36, "learning_rate": 2.047440133991218e-05, "loss": 0.2667, "step": 55726 }, { "epoch": 3.36, "learning_rate": 2.0472892429797958e-05, "loss": 0.5901, "step": 55728 }, { "epoch": 3.36, "learning_rate": 2.0471383519683735e-05, "loss": 0.3368, "step": 55730 }, { "epoch": 3.36, "learning_rate": 2.0469874609569508e-05, "loss": 0.2868, "step": 55732 }, { "epoch": 3.36, "learning_rate": 2.0468365699455284e-05, "loss": 0.2804, "step": 55734 }, { "epoch": 3.36, "learning_rate": 2.046685678934106e-05, "loss": 0.459, "step": 55736 }, { "epoch": 3.36, "learning_rate": 2.0465347879226834e-05, "loss": 0.3808, "step": 55738 }, { "epoch": 3.36, "learning_rate": 2.046383896911261e-05, "loss": 0.428, "step": 55740 }, { "epoch": 3.36, "learning_rate": 2.0462330058998387e-05, "loss": 0.3881, "step": 55742 }, { "epoch": 3.36, "learning_rate": 2.0460821148884163e-05, "loss": 0.4234, "step": 55744 }, { "epoch": 3.36, "learning_rate": 2.0459312238769936e-05, "loss": 0.4071, "step": 55746 }, { "epoch": 3.36, "learning_rate": 2.0457803328655713e-05, "loss": 0.3537, "step": 55748 }, { "epoch": 3.36, "learning_rate": 2.045629441854149e-05, "loss": 0.2743, "step": 55750 }, { "epoch": 3.37, "learning_rate": 2.0454785508427263e-05, "loss": 0.2685, "step": 55752 }, { "epoch": 3.37, "learning_rate": 2.045327659831304e-05, "loss": 0.3254, "step": 55754 }, { "epoch": 3.37, "learning_rate": 2.0451767688198816e-05, "loss": 0.318, "step": 55756 }, { "epoch": 3.37, "learning_rate": 2.0450258778084592e-05, "loss": 0.5041, "step": 55758 }, { "epoch": 3.37, "learning_rate": 2.0448749867970365e-05, "loss": 0.2664, "step": 55760 }, { "epoch": 3.37, "learning_rate": 2.0447240957856142e-05, "loss": 0.4784, "step": 55762 }, { "epoch": 3.37, "learning_rate": 2.044573204774192e-05, "loss": 0.311, "step": 55764 }, { "epoch": 3.37, "learning_rate": 2.0444223137627695e-05, "loss": 0.2885, "step": 55766 }, { "epoch": 3.37, "learning_rate": 2.0442714227513468e-05, "loss": 0.3742, "step": 55768 }, { "epoch": 3.37, "learning_rate": 2.044120531739924e-05, "loss": 0.4286, "step": 55770 }, { "epoch": 3.37, "learning_rate": 2.043969640728502e-05, "loss": 0.3495, "step": 55772 }, { "epoch": 3.37, "learning_rate": 2.0438187497170794e-05, "loss": 0.4022, "step": 55774 }, { "epoch": 3.37, "learning_rate": 2.043667858705657e-05, "loss": 0.3611, "step": 55776 }, { "epoch": 3.37, "learning_rate": 2.0435169676942344e-05, "loss": 0.3649, "step": 55778 }, { "epoch": 3.37, "learning_rate": 2.0433660766828124e-05, "loss": 0.3756, "step": 55780 }, { "epoch": 3.37, "learning_rate": 2.0432151856713897e-05, "loss": 0.3966, "step": 55782 }, { "epoch": 3.37, "learning_rate": 2.043064294659967e-05, "loss": 0.3275, "step": 55784 }, { "epoch": 3.37, "learning_rate": 2.0429134036485446e-05, "loss": 0.3877, "step": 55786 }, { "epoch": 3.37, "learning_rate": 2.0427625126371223e-05, "loss": 0.2685, "step": 55788 }, { "epoch": 3.37, "learning_rate": 2.0426116216257e-05, "loss": 0.3721, "step": 55790 }, { "epoch": 3.37, "learning_rate": 2.0424607306142773e-05, "loss": 0.2531, "step": 55792 }, { "epoch": 3.37, "learning_rate": 2.042309839602855e-05, "loss": 0.5533, "step": 55794 }, { "epoch": 3.37, "learning_rate": 2.0421589485914326e-05, "loss": 0.5262, "step": 55796 }, { "epoch": 3.37, "learning_rate": 2.0420080575800102e-05, "loss": 0.3276, "step": 55798 }, { "epoch": 3.37, "learning_rate": 2.0418571665685875e-05, "loss": 0.3468, "step": 55800 }, { "epoch": 3.37, "learning_rate": 2.0417062755571652e-05, "loss": 0.3989, "step": 55802 }, { "epoch": 3.37, "learning_rate": 2.0415553845457428e-05, "loss": 0.3332, "step": 55804 }, { "epoch": 3.37, "learning_rate": 2.04140449353432e-05, "loss": 0.3001, "step": 55806 }, { "epoch": 3.37, "learning_rate": 2.0412536025228978e-05, "loss": 0.4092, "step": 55808 }, { "epoch": 3.37, "learning_rate": 2.0411027115114754e-05, "loss": 0.3723, "step": 55810 }, { "epoch": 3.37, "learning_rate": 2.040951820500053e-05, "loss": 0.3656, "step": 55812 }, { "epoch": 3.37, "learning_rate": 2.0408009294886304e-05, "loss": 0.425, "step": 55814 }, { "epoch": 3.37, "learning_rate": 2.040650038477208e-05, "loss": 0.3753, "step": 55816 }, { "epoch": 3.37, "learning_rate": 2.0404991474657857e-05, "loss": 0.3963, "step": 55818 }, { "epoch": 3.37, "learning_rate": 2.040348256454363e-05, "loss": 0.4271, "step": 55820 }, { "epoch": 3.37, "learning_rate": 2.0401973654429407e-05, "loss": 0.3351, "step": 55822 }, { "epoch": 3.37, "learning_rate": 2.040046474431518e-05, "loss": 0.4853, "step": 55824 }, { "epoch": 3.37, "learning_rate": 2.039895583420096e-05, "loss": 0.6085, "step": 55826 }, { "epoch": 3.37, "learning_rate": 2.0397446924086733e-05, "loss": 0.4268, "step": 55828 }, { "epoch": 3.37, "learning_rate": 2.039593801397251e-05, "loss": 0.2441, "step": 55830 }, { "epoch": 3.37, "learning_rate": 2.0394429103858283e-05, "loss": 0.3505, "step": 55832 }, { "epoch": 3.37, "learning_rate": 2.039292019374406e-05, "loss": 0.2432, "step": 55834 }, { "epoch": 3.37, "learning_rate": 2.0391411283629836e-05, "loss": 0.3736, "step": 55836 }, { "epoch": 3.37, "learning_rate": 2.038990237351561e-05, "loss": 0.4594, "step": 55838 }, { "epoch": 3.37, "learning_rate": 2.0388393463401385e-05, "loss": 0.3502, "step": 55840 }, { "epoch": 3.37, "learning_rate": 2.0386884553287162e-05, "loss": 0.3142, "step": 55842 }, { "epoch": 3.37, "learning_rate": 2.0385375643172938e-05, "loss": 0.2897, "step": 55844 }, { "epoch": 3.37, "learning_rate": 2.038386673305871e-05, "loss": 0.3539, "step": 55846 }, { "epoch": 3.37, "learning_rate": 2.0382357822944488e-05, "loss": 0.3593, "step": 55848 }, { "epoch": 3.37, "learning_rate": 2.0380848912830264e-05, "loss": 0.3266, "step": 55850 }, { "epoch": 3.37, "learning_rate": 2.0379340002716038e-05, "loss": 0.4809, "step": 55852 }, { "epoch": 3.37, "learning_rate": 2.0377831092601814e-05, "loss": 0.3525, "step": 55854 }, { "epoch": 3.37, "learning_rate": 2.037632218248759e-05, "loss": 0.3972, "step": 55856 }, { "epoch": 3.37, "learning_rate": 2.0374813272373367e-05, "loss": 0.3453, "step": 55858 }, { "epoch": 3.37, "learning_rate": 2.037330436225914e-05, "loss": 0.4375, "step": 55860 }, { "epoch": 3.37, "learning_rate": 2.0371795452144917e-05, "loss": 0.3206, "step": 55862 }, { "epoch": 3.37, "learning_rate": 2.0370286542030693e-05, "loss": 0.3871, "step": 55864 }, { "epoch": 3.37, "learning_rate": 2.0368777631916466e-05, "loss": 0.4139, "step": 55866 }, { "epoch": 3.37, "learning_rate": 2.0367268721802243e-05, "loss": 0.4028, "step": 55868 }, { "epoch": 3.37, "learning_rate": 2.036575981168802e-05, "loss": 0.2738, "step": 55870 }, { "epoch": 3.37, "learning_rate": 2.0364250901573796e-05, "loss": 0.2427, "step": 55872 }, { "epoch": 3.37, "learning_rate": 2.036274199145957e-05, "loss": 0.5595, "step": 55874 }, { "epoch": 3.37, "learning_rate": 2.0361233081345346e-05, "loss": 0.4542, "step": 55876 }, { "epoch": 3.37, "learning_rate": 2.0359724171231122e-05, "loss": 0.3181, "step": 55878 }, { "epoch": 3.37, "learning_rate": 2.03582152611169e-05, "loss": 0.3627, "step": 55880 }, { "epoch": 3.37, "learning_rate": 2.035670635100267e-05, "loss": 0.3664, "step": 55882 }, { "epoch": 3.37, "learning_rate": 2.0355197440888445e-05, "loss": 0.3193, "step": 55884 }, { "epoch": 3.37, "learning_rate": 2.0353688530774225e-05, "loss": 0.3221, "step": 55886 }, { "epoch": 3.37, "learning_rate": 2.0352179620659998e-05, "loss": 0.2587, "step": 55888 }, { "epoch": 3.37, "learning_rate": 2.0350670710545774e-05, "loss": 0.3891, "step": 55890 }, { "epoch": 3.37, "learning_rate": 2.0349161800431547e-05, "loss": 0.4311, "step": 55892 }, { "epoch": 3.37, "learning_rate": 2.0347652890317327e-05, "loss": 0.4056, "step": 55894 }, { "epoch": 3.37, "learning_rate": 2.03461439802031e-05, "loss": 0.4175, "step": 55896 }, { "epoch": 3.37, "learning_rate": 2.0344635070088877e-05, "loss": 0.4161, "step": 55898 }, { "epoch": 3.37, "learning_rate": 2.034312615997465e-05, "loss": 0.2352, "step": 55900 }, { "epoch": 3.37, "learning_rate": 2.0341617249860427e-05, "loss": 0.3362, "step": 55902 }, { "epoch": 3.37, "learning_rate": 2.0340108339746203e-05, "loss": 0.3587, "step": 55904 }, { "epoch": 3.37, "learning_rate": 2.0338599429631976e-05, "loss": 0.4429, "step": 55906 }, { "epoch": 3.37, "learning_rate": 2.0337090519517753e-05, "loss": 0.3385, "step": 55908 }, { "epoch": 3.37, "learning_rate": 2.033558160940353e-05, "loss": 0.4191, "step": 55910 }, { "epoch": 3.37, "learning_rate": 2.0334072699289306e-05, "loss": 0.2189, "step": 55912 }, { "epoch": 3.37, "learning_rate": 2.033256378917508e-05, "loss": 0.4443, "step": 55914 }, { "epoch": 3.37, "learning_rate": 2.0331054879060855e-05, "loss": 0.5021, "step": 55916 }, { "epoch": 3.38, "learning_rate": 2.0329545968946632e-05, "loss": 0.39, "step": 55918 }, { "epoch": 3.38, "learning_rate": 2.0328037058832405e-05, "loss": 0.3742, "step": 55920 }, { "epoch": 3.38, "learning_rate": 2.032652814871818e-05, "loss": 0.4292, "step": 55922 }, { "epoch": 3.38, "learning_rate": 2.0325019238603958e-05, "loss": 0.302, "step": 55924 }, { "epoch": 3.38, "learning_rate": 2.0323510328489735e-05, "loss": 0.4594, "step": 55926 }, { "epoch": 3.38, "learning_rate": 2.0322001418375508e-05, "loss": 0.2899, "step": 55928 }, { "epoch": 3.38, "learning_rate": 2.0320492508261284e-05, "loss": 0.5541, "step": 55930 }, { "epoch": 3.38, "learning_rate": 2.031898359814706e-05, "loss": 0.2782, "step": 55932 }, { "epoch": 3.38, "learning_rate": 2.0317474688032834e-05, "loss": 0.4369, "step": 55934 }, { "epoch": 3.38, "learning_rate": 2.031596577791861e-05, "loss": 0.2824, "step": 55936 }, { "epoch": 3.38, "learning_rate": 2.0314456867804384e-05, "loss": 0.2279, "step": 55938 }, { "epoch": 3.38, "learning_rate": 2.0312947957690163e-05, "loss": 0.5232, "step": 55940 }, { "epoch": 3.38, "learning_rate": 2.0311439047575937e-05, "loss": 0.3181, "step": 55942 }, { "epoch": 3.38, "learning_rate": 2.0309930137461713e-05, "loss": 0.3807, "step": 55944 }, { "epoch": 3.38, "learning_rate": 2.0308421227347486e-05, "loss": 0.4203, "step": 55946 }, { "epoch": 3.38, "learning_rate": 2.0306912317233263e-05, "loss": 0.3999, "step": 55948 }, { "epoch": 3.38, "learning_rate": 2.030540340711904e-05, "loss": 0.4018, "step": 55950 }, { "epoch": 3.38, "learning_rate": 2.0303894497004812e-05, "loss": 0.5462, "step": 55952 }, { "epoch": 3.38, "learning_rate": 2.030238558689059e-05, "loss": 0.6041, "step": 55954 }, { "epoch": 3.38, "learning_rate": 2.0300876676776365e-05, "loss": 0.2669, "step": 55956 }, { "epoch": 3.38, "learning_rate": 2.0299367766662142e-05, "loss": 0.1994, "step": 55958 }, { "epoch": 3.38, "learning_rate": 2.0297858856547915e-05, "loss": 0.2472, "step": 55960 }, { "epoch": 3.38, "learning_rate": 2.029634994643369e-05, "loss": 0.292, "step": 55962 }, { "epoch": 3.38, "learning_rate": 2.0294841036319468e-05, "loss": 0.4059, "step": 55964 }, { "epoch": 3.38, "learning_rate": 2.029333212620524e-05, "loss": 0.3138, "step": 55966 }, { "epoch": 3.38, "learning_rate": 2.0291823216091018e-05, "loss": 0.4019, "step": 55968 }, { "epoch": 3.38, "learning_rate": 2.0290314305976794e-05, "loss": 0.491, "step": 55970 }, { "epoch": 3.38, "learning_rate": 2.028880539586257e-05, "loss": 0.3092, "step": 55972 }, { "epoch": 3.38, "learning_rate": 2.0287296485748344e-05, "loss": 0.3813, "step": 55974 }, { "epoch": 3.38, "learning_rate": 2.028578757563412e-05, "loss": 0.279, "step": 55976 }, { "epoch": 3.38, "learning_rate": 2.0284278665519897e-05, "loss": 0.4757, "step": 55978 }, { "epoch": 3.38, "learning_rate": 2.0282769755405673e-05, "loss": 0.4912, "step": 55980 }, { "epoch": 3.38, "learning_rate": 2.0281260845291447e-05, "loss": 0.4592, "step": 55982 }, { "epoch": 3.38, "learning_rate": 2.0279751935177223e-05, "loss": 0.4067, "step": 55984 }, { "epoch": 3.38, "learning_rate": 2.0278243025063e-05, "loss": 0.2956, "step": 55986 }, { "epoch": 3.38, "learning_rate": 2.0276734114948773e-05, "loss": 0.3574, "step": 55988 }, { "epoch": 3.38, "learning_rate": 2.027522520483455e-05, "loss": 0.499, "step": 55990 }, { "epoch": 3.38, "learning_rate": 2.0273716294720326e-05, "loss": 0.2393, "step": 55992 }, { "epoch": 3.38, "learning_rate": 2.0272207384606102e-05, "loss": 0.4392, "step": 55994 }, { "epoch": 3.38, "learning_rate": 2.0270698474491875e-05, "loss": 0.4426, "step": 55996 }, { "epoch": 3.38, "learning_rate": 2.026918956437765e-05, "loss": 0.316, "step": 55998 }, { "epoch": 3.38, "learning_rate": 2.026768065426343e-05, "loss": 0.3437, "step": 56000 }, { "epoch": 3.38, "learning_rate": 2.02661717441492e-05, "loss": 0.3168, "step": 56002 }, { "epoch": 3.38, "learning_rate": 2.0264662834034978e-05, "loss": 0.3809, "step": 56004 }, { "epoch": 3.38, "learning_rate": 2.026315392392075e-05, "loss": 0.2557, "step": 56006 }, { "epoch": 3.38, "learning_rate": 2.026164501380653e-05, "loss": 0.3125, "step": 56008 }, { "epoch": 3.38, "learning_rate": 2.0260136103692304e-05, "loss": 0.3814, "step": 56010 }, { "epoch": 3.38, "learning_rate": 2.025862719357808e-05, "loss": 0.2968, "step": 56012 }, { "epoch": 3.38, "learning_rate": 2.0257118283463854e-05, "loss": 0.3835, "step": 56014 }, { "epoch": 3.38, "learning_rate": 2.025560937334963e-05, "loss": 0.3698, "step": 56016 }, { "epoch": 3.38, "learning_rate": 2.0254100463235407e-05, "loss": 0.4272, "step": 56018 }, { "epoch": 3.38, "learning_rate": 2.025259155312118e-05, "loss": 0.5336, "step": 56020 }, { "epoch": 3.38, "learning_rate": 2.0251082643006956e-05, "loss": 0.4104, "step": 56022 }, { "epoch": 3.38, "learning_rate": 2.0249573732892733e-05, "loss": 0.2443, "step": 56024 }, { "epoch": 3.38, "learning_rate": 2.024806482277851e-05, "loss": 0.2745, "step": 56026 }, { "epoch": 3.38, "learning_rate": 2.0246555912664283e-05, "loss": 0.2797, "step": 56028 }, { "epoch": 3.38, "learning_rate": 2.024504700255006e-05, "loss": 0.3337, "step": 56030 }, { "epoch": 3.38, "learning_rate": 2.0243538092435836e-05, "loss": 0.3133, "step": 56032 }, { "epoch": 3.38, "learning_rate": 2.024202918232161e-05, "loss": 0.3317, "step": 56034 }, { "epoch": 3.38, "learning_rate": 2.0240520272207385e-05, "loss": 0.3385, "step": 56036 }, { "epoch": 3.38, "learning_rate": 2.0239011362093162e-05, "loss": 0.4266, "step": 56038 }, { "epoch": 3.38, "learning_rate": 2.023750245197894e-05, "loss": 0.7472, "step": 56040 }, { "epoch": 3.38, "learning_rate": 2.023599354186471e-05, "loss": 0.4259, "step": 56042 }, { "epoch": 3.38, "learning_rate": 2.0234484631750488e-05, "loss": 0.3316, "step": 56044 }, { "epoch": 3.38, "learning_rate": 2.0232975721636264e-05, "loss": 0.3915, "step": 56046 }, { "epoch": 3.38, "learning_rate": 2.0231466811522038e-05, "loss": 0.3984, "step": 56048 }, { "epoch": 3.38, "learning_rate": 2.0229957901407814e-05, "loss": 0.3364, "step": 56050 }, { "epoch": 3.38, "learning_rate": 2.0228448991293587e-05, "loss": 0.4096, "step": 56052 }, { "epoch": 3.38, "learning_rate": 2.0226940081179367e-05, "loss": 0.3424, "step": 56054 }, { "epoch": 3.38, "learning_rate": 2.022543117106514e-05, "loss": 0.332, "step": 56056 }, { "epoch": 3.38, "learning_rate": 2.0223922260950917e-05, "loss": 0.2962, "step": 56058 }, { "epoch": 3.38, "learning_rate": 2.022241335083669e-05, "loss": 0.2842, "step": 56060 }, { "epoch": 3.38, "learning_rate": 2.022090444072247e-05, "loss": 0.345, "step": 56062 }, { "epoch": 3.38, "learning_rate": 2.0219395530608243e-05, "loss": 0.2126, "step": 56064 }, { "epoch": 3.38, "learning_rate": 2.0217886620494016e-05, "loss": 0.3933, "step": 56066 }, { "epoch": 3.38, "learning_rate": 2.0216377710379793e-05, "loss": 0.5565, "step": 56068 }, { "epoch": 3.38, "learning_rate": 2.021486880026557e-05, "loss": 0.4261, "step": 56070 }, { "epoch": 3.38, "learning_rate": 2.0213359890151346e-05, "loss": 0.4467, "step": 56072 }, { "epoch": 3.38, "learning_rate": 2.021185098003712e-05, "loss": 0.3361, "step": 56074 }, { "epoch": 3.38, "learning_rate": 2.0210342069922895e-05, "loss": 0.3001, "step": 56076 }, { "epoch": 3.38, "learning_rate": 2.0208833159808672e-05, "loss": 0.5281, "step": 56078 }, { "epoch": 3.38, "learning_rate": 2.0207324249694445e-05, "loss": 0.2806, "step": 56080 }, { "epoch": 3.38, "learning_rate": 2.020581533958022e-05, "loss": 0.3898, "step": 56082 }, { "epoch": 3.39, "learning_rate": 2.0204306429465998e-05, "loss": 0.3094, "step": 56084 }, { "epoch": 3.39, "learning_rate": 2.0202797519351774e-05, "loss": 0.2199, "step": 56086 }, { "epoch": 3.39, "learning_rate": 2.0201288609237548e-05, "loss": 0.4248, "step": 56088 }, { "epoch": 3.39, "learning_rate": 2.0199779699123324e-05, "loss": 0.2429, "step": 56090 }, { "epoch": 3.39, "learning_rate": 2.01982707890091e-05, "loss": 0.6032, "step": 56092 }, { "epoch": 3.39, "learning_rate": 2.0196761878894877e-05, "loss": 0.4169, "step": 56094 }, { "epoch": 3.39, "learning_rate": 2.019525296878065e-05, "loss": 0.2799, "step": 56096 }, { "epoch": 3.39, "learning_rate": 2.0193744058666427e-05, "loss": 0.2749, "step": 56098 }, { "epoch": 3.39, "learning_rate": 2.0192235148552203e-05, "loss": 0.2627, "step": 56100 }, { "epoch": 3.39, "learning_rate": 2.0190726238437976e-05, "loss": 0.2986, "step": 56102 }, { "epoch": 3.39, "learning_rate": 2.0189217328323753e-05, "loss": 0.3692, "step": 56104 }, { "epoch": 3.39, "learning_rate": 2.018770841820953e-05, "loss": 0.4249, "step": 56106 }, { "epoch": 3.39, "learning_rate": 2.0186199508095306e-05, "loss": 0.2528, "step": 56108 }, { "epoch": 3.39, "learning_rate": 2.018469059798108e-05, "loss": 0.4575, "step": 56110 }, { "epoch": 3.39, "learning_rate": 2.0183181687866852e-05, "loss": 0.4572, "step": 56112 }, { "epoch": 3.39, "learning_rate": 2.0181672777752632e-05, "loss": 0.4609, "step": 56114 }, { "epoch": 3.39, "learning_rate": 2.0180163867638405e-05, "loss": 0.5372, "step": 56116 }, { "epoch": 3.39, "learning_rate": 2.017865495752418e-05, "loss": 0.4727, "step": 56118 }, { "epoch": 3.39, "learning_rate": 2.0177146047409955e-05, "loss": 0.3236, "step": 56120 }, { "epoch": 3.39, "learning_rate": 2.0175637137295735e-05, "loss": 0.2632, "step": 56122 }, { "epoch": 3.39, "learning_rate": 2.0174128227181508e-05, "loss": 0.2309, "step": 56124 }, { "epoch": 3.39, "learning_rate": 2.0172619317067284e-05, "loss": 0.3253, "step": 56126 }, { "epoch": 3.39, "learning_rate": 2.0171110406953057e-05, "loss": 0.3891, "step": 56128 }, { "epoch": 3.39, "learning_rate": 2.0169601496838834e-05, "loss": 0.3737, "step": 56130 }, { "epoch": 3.39, "learning_rate": 2.016809258672461e-05, "loss": 0.431, "step": 56132 }, { "epoch": 3.39, "learning_rate": 2.0166583676610384e-05, "loss": 0.3002, "step": 56134 }, { "epoch": 3.39, "learning_rate": 2.016507476649616e-05, "loss": 0.3524, "step": 56136 }, { "epoch": 3.39, "learning_rate": 2.0163565856381937e-05, "loss": 0.4788, "step": 56138 }, { "epoch": 3.39, "learning_rate": 2.0162056946267713e-05, "loss": 0.4449, "step": 56140 }, { "epoch": 3.39, "learning_rate": 2.0160548036153486e-05, "loss": 0.4369, "step": 56142 }, { "epoch": 3.39, "learning_rate": 2.0159039126039263e-05, "loss": 0.3702, "step": 56144 }, { "epoch": 3.39, "learning_rate": 2.015753021592504e-05, "loss": 0.2852, "step": 56146 }, { "epoch": 3.39, "learning_rate": 2.0156021305810812e-05, "loss": 0.4527, "step": 56148 }, { "epoch": 3.39, "learning_rate": 2.015451239569659e-05, "loss": 0.2715, "step": 56150 }, { "epoch": 3.39, "learning_rate": 2.0153003485582365e-05, "loss": 0.2923, "step": 56152 }, { "epoch": 3.39, "learning_rate": 2.0151494575468142e-05, "loss": 0.3947, "step": 56154 }, { "epoch": 3.39, "learning_rate": 2.0149985665353915e-05, "loss": 0.3795, "step": 56156 }, { "epoch": 3.39, "learning_rate": 2.014847675523969e-05, "loss": 0.3772, "step": 56158 }, { "epoch": 3.39, "learning_rate": 2.0146967845125468e-05, "loss": 0.2484, "step": 56160 }, { "epoch": 3.39, "learning_rate": 2.014545893501124e-05, "loss": 0.574, "step": 56162 }, { "epoch": 3.39, "learning_rate": 2.0143950024897018e-05, "loss": 0.3616, "step": 56164 }, { "epoch": 3.39, "learning_rate": 2.014244111478279e-05, "loss": 0.2498, "step": 56166 }, { "epoch": 3.39, "learning_rate": 2.014093220466857e-05, "loss": 0.3712, "step": 56168 }, { "epoch": 3.39, "learning_rate": 2.0139423294554344e-05, "loss": 0.3934, "step": 56170 }, { "epoch": 3.39, "learning_rate": 2.013791438444012e-05, "loss": 0.3075, "step": 56172 }, { "epoch": 3.39, "learning_rate": 2.0136405474325894e-05, "loss": 0.4498, "step": 56174 }, { "epoch": 3.39, "learning_rate": 2.0134896564211673e-05, "loss": 0.4223, "step": 56176 }, { "epoch": 3.39, "learning_rate": 2.0133387654097447e-05, "loss": 0.3122, "step": 56178 }, { "epoch": 3.39, "learning_rate": 2.013187874398322e-05, "loss": 0.434, "step": 56180 }, { "epoch": 3.39, "learning_rate": 2.0130369833868996e-05, "loss": 0.4522, "step": 56182 }, { "epoch": 3.39, "learning_rate": 2.0128860923754773e-05, "loss": 0.4042, "step": 56184 }, { "epoch": 3.39, "learning_rate": 2.012735201364055e-05, "loss": 0.3332, "step": 56186 }, { "epoch": 3.39, "learning_rate": 2.0125843103526322e-05, "loss": 0.2981, "step": 56188 }, { "epoch": 3.39, "learning_rate": 2.01243341934121e-05, "loss": 0.2565, "step": 56190 }, { "epoch": 3.39, "learning_rate": 2.0122825283297875e-05, "loss": 0.5082, "step": 56192 }, { "epoch": 3.39, "learning_rate": 2.012131637318365e-05, "loss": 0.2778, "step": 56194 }, { "epoch": 3.39, "learning_rate": 2.0119807463069425e-05, "loss": 0.1837, "step": 56196 }, { "epoch": 3.39, "learning_rate": 2.01182985529552e-05, "loss": 0.3346, "step": 56198 }, { "epoch": 3.39, "learning_rate": 2.0116789642840978e-05, "loss": 0.3619, "step": 56200 }, { "epoch": 3.39, "learning_rate": 2.011528073272675e-05, "loss": 0.5405, "step": 56202 }, { "epoch": 3.39, "learning_rate": 2.0113771822612528e-05, "loss": 0.369, "step": 56204 }, { "epoch": 3.39, "learning_rate": 2.0112262912498304e-05, "loss": 0.421, "step": 56206 }, { "epoch": 3.39, "learning_rate": 2.011075400238408e-05, "loss": 0.6016, "step": 56208 }, { "epoch": 3.39, "learning_rate": 2.0109245092269854e-05, "loss": 0.3435, "step": 56210 }, { "epoch": 3.39, "learning_rate": 2.010773618215563e-05, "loss": 0.3728, "step": 56212 }, { "epoch": 3.39, "learning_rate": 2.0106227272041407e-05, "loss": 0.3254, "step": 56214 }, { "epoch": 3.39, "learning_rate": 2.010471836192718e-05, "loss": 0.4821, "step": 56216 }, { "epoch": 3.39, "learning_rate": 2.0103209451812957e-05, "loss": 0.6112, "step": 56218 }, { "epoch": 3.39, "learning_rate": 2.0101700541698733e-05, "loss": 0.3748, "step": 56220 }, { "epoch": 3.39, "learning_rate": 2.010019163158451e-05, "loss": 0.4355, "step": 56222 }, { "epoch": 3.39, "learning_rate": 2.0098682721470283e-05, "loss": 0.3838, "step": 56224 }, { "epoch": 3.39, "learning_rate": 2.009717381135606e-05, "loss": 0.2146, "step": 56226 }, { "epoch": 3.39, "learning_rate": 2.0095664901241836e-05, "loss": 0.4553, "step": 56228 }, { "epoch": 3.39, "learning_rate": 2.009415599112761e-05, "loss": 0.3498, "step": 56230 }, { "epoch": 3.39, "learning_rate": 2.0092647081013385e-05, "loss": 0.3479, "step": 56232 }, { "epoch": 3.39, "learning_rate": 2.009113817089916e-05, "loss": 0.4468, "step": 56234 }, { "epoch": 3.39, "learning_rate": 2.008962926078494e-05, "loss": 0.3189, "step": 56236 }, { "epoch": 3.39, "learning_rate": 2.008812035067071e-05, "loss": 0.3086, "step": 56238 }, { "epoch": 3.39, "learning_rate": 2.0086611440556488e-05, "loss": 0.4442, "step": 56240 }, { "epoch": 3.39, "learning_rate": 2.008510253044226e-05, "loss": 0.2901, "step": 56242 }, { "epoch": 3.39, "learning_rate": 2.0083593620328038e-05, "loss": 0.4299, "step": 56244 }, { "epoch": 3.39, "learning_rate": 2.0082084710213814e-05, "loss": 0.4962, "step": 56246 }, { "epoch": 3.39, "learning_rate": 2.0080575800099587e-05, "loss": 0.4385, "step": 56248 }, { "epoch": 3.4, "learning_rate": 2.0079066889985364e-05, "loss": 0.3581, "step": 56250 }, { "epoch": 3.4, "learning_rate": 2.007755797987114e-05, "loss": 0.5639, "step": 56252 }, { "epoch": 3.4, "learning_rate": 2.0076049069756917e-05, "loss": 0.4705, "step": 56254 }, { "epoch": 3.4, "learning_rate": 2.007454015964269e-05, "loss": 0.2792, "step": 56256 }, { "epoch": 3.4, "learning_rate": 2.0073031249528466e-05, "loss": 0.3579, "step": 56258 }, { "epoch": 3.4, "learning_rate": 2.0071522339414243e-05, "loss": 0.3615, "step": 56260 }, { "epoch": 3.4, "learning_rate": 2.0070013429300016e-05, "loss": 0.3492, "step": 56262 }, { "epoch": 3.4, "learning_rate": 2.0068504519185793e-05, "loss": 0.3134, "step": 56264 }, { "epoch": 3.4, "learning_rate": 2.006699560907157e-05, "loss": 0.2464, "step": 56266 }, { "epoch": 3.4, "learning_rate": 2.0065486698957346e-05, "loss": 0.3267, "step": 56268 }, { "epoch": 3.4, "learning_rate": 2.006397778884312e-05, "loss": 0.3962, "step": 56270 }, { "epoch": 3.4, "learning_rate": 2.0062468878728895e-05, "loss": 0.3489, "step": 56272 }, { "epoch": 3.4, "learning_rate": 2.0060959968614672e-05, "loss": 0.4609, "step": 56274 }, { "epoch": 3.4, "learning_rate": 2.0059451058500445e-05, "loss": 0.4684, "step": 56276 }, { "epoch": 3.4, "learning_rate": 2.005794214838622e-05, "loss": 0.5014, "step": 56278 }, { "epoch": 3.4, "learning_rate": 2.0056433238271995e-05, "loss": 0.3259, "step": 56280 }, { "epoch": 3.4, "learning_rate": 2.0054924328157774e-05, "loss": 0.2635, "step": 56282 }, { "epoch": 3.4, "learning_rate": 2.0053415418043548e-05, "loss": 0.504, "step": 56284 }, { "epoch": 3.4, "learning_rate": 2.0051906507929324e-05, "loss": 0.3514, "step": 56286 }, { "epoch": 3.4, "learning_rate": 2.0050397597815097e-05, "loss": 0.3269, "step": 56288 }, { "epoch": 3.4, "learning_rate": 2.0048888687700877e-05, "loss": 0.5044, "step": 56290 }, { "epoch": 3.4, "learning_rate": 2.004737977758665e-05, "loss": 0.3983, "step": 56292 }, { "epoch": 3.4, "learning_rate": 2.0045870867472423e-05, "loss": 0.1617, "step": 56294 }, { "epoch": 3.4, "learning_rate": 2.00443619573582e-05, "loss": 0.3814, "step": 56296 }, { "epoch": 3.4, "learning_rate": 2.0042853047243976e-05, "loss": 0.4651, "step": 56298 }, { "epoch": 3.4, "learning_rate": 2.0041344137129753e-05, "loss": 0.3815, "step": 56300 }, { "epoch": 3.4, "learning_rate": 2.0039835227015526e-05, "loss": 0.2389, "step": 56302 }, { "epoch": 3.4, "learning_rate": 2.0038326316901303e-05, "loss": 0.3975, "step": 56304 }, { "epoch": 3.4, "learning_rate": 2.003681740678708e-05, "loss": 0.2501, "step": 56306 }, { "epoch": 3.4, "learning_rate": 2.0035308496672856e-05, "loss": 0.3094, "step": 56308 }, { "epoch": 3.4, "learning_rate": 2.003379958655863e-05, "loss": 0.394, "step": 56310 }, { "epoch": 3.4, "learning_rate": 2.0032290676444405e-05, "loss": 0.4054, "step": 56312 }, { "epoch": 3.4, "learning_rate": 2.0030781766330182e-05, "loss": 0.3823, "step": 56314 }, { "epoch": 3.4, "learning_rate": 2.0029272856215955e-05, "loss": 0.343, "step": 56316 }, { "epoch": 3.4, "learning_rate": 2.002776394610173e-05, "loss": 0.3635, "step": 56318 }, { "epoch": 3.4, "learning_rate": 2.0026255035987508e-05, "loss": 0.3405, "step": 56320 }, { "epoch": 3.4, "learning_rate": 2.0024746125873284e-05, "loss": 0.4555, "step": 56322 }, { "epoch": 3.4, "learning_rate": 2.0023237215759058e-05, "loss": 0.3919, "step": 56324 }, { "epoch": 3.4, "learning_rate": 2.002172830564483e-05, "loss": 0.4136, "step": 56326 }, { "epoch": 3.4, "learning_rate": 2.002021939553061e-05, "loss": 0.2803, "step": 56328 }, { "epoch": 3.4, "learning_rate": 2.0018710485416384e-05, "loss": 0.2027, "step": 56330 }, { "epoch": 3.4, "learning_rate": 2.001720157530216e-05, "loss": 0.2791, "step": 56332 }, { "epoch": 3.4, "learning_rate": 2.0015692665187937e-05, "loss": 0.389, "step": 56334 }, { "epoch": 3.4, "learning_rate": 2.0014183755073713e-05, "loss": 0.3533, "step": 56336 }, { "epoch": 3.4, "learning_rate": 2.0012674844959486e-05, "loss": 0.4071, "step": 56338 }, { "epoch": 3.4, "learning_rate": 2.0011165934845263e-05, "loss": 0.3733, "step": 56340 }, { "epoch": 3.4, "learning_rate": 2.000965702473104e-05, "loss": 0.5244, "step": 56342 }, { "epoch": 3.4, "learning_rate": 2.0008148114616813e-05, "loss": 0.5192, "step": 56344 }, { "epoch": 3.4, "learning_rate": 2.000663920450259e-05, "loss": 0.4233, "step": 56346 }, { "epoch": 3.4, "learning_rate": 2.0005130294388362e-05, "loss": 0.3039, "step": 56348 }, { "epoch": 3.4, "learning_rate": 2.0003621384274142e-05, "loss": 0.2616, "step": 56350 }, { "epoch": 3.4, "learning_rate": 2.0002112474159915e-05, "loss": 0.3387, "step": 56352 }, { "epoch": 3.4, "learning_rate": 2.0000603564045692e-05, "loss": 0.2859, "step": 56354 }, { "epoch": 3.4, "learning_rate": 1.9999094653931465e-05, "loss": 0.3942, "step": 56356 }, { "epoch": 3.4, "learning_rate": 1.9997585743817245e-05, "loss": 0.2872, "step": 56358 }, { "epoch": 3.4, "learning_rate": 1.9996076833703018e-05, "loss": 0.3949, "step": 56360 }, { "epoch": 3.4, "learning_rate": 1.999456792358879e-05, "loss": 0.2823, "step": 56362 }, { "epoch": 3.4, "learning_rate": 1.9993059013474568e-05, "loss": 0.2615, "step": 56364 }, { "epoch": 3.4, "learning_rate": 1.9991550103360344e-05, "loss": 0.3709, "step": 56366 }, { "epoch": 3.4, "learning_rate": 1.999004119324612e-05, "loss": 0.3041, "step": 56368 }, { "epoch": 3.4, "learning_rate": 1.9988532283131894e-05, "loss": 0.3245, "step": 56370 }, { "epoch": 3.4, "learning_rate": 1.998702337301767e-05, "loss": 0.3348, "step": 56372 }, { "epoch": 3.4, "learning_rate": 1.9985514462903447e-05, "loss": 0.4206, "step": 56374 }, { "epoch": 3.4, "learning_rate": 1.998400555278922e-05, "loss": 0.5395, "step": 56376 }, { "epoch": 3.4, "learning_rate": 1.9982496642674996e-05, "loss": 0.3476, "step": 56378 }, { "epoch": 3.4, "learning_rate": 1.9980987732560773e-05, "loss": 0.3476, "step": 56380 }, { "epoch": 3.4, "learning_rate": 1.997947882244655e-05, "loss": 0.3262, "step": 56382 }, { "epoch": 3.4, "learning_rate": 1.9977969912332322e-05, "loss": 0.3228, "step": 56384 }, { "epoch": 3.4, "learning_rate": 1.99764610022181e-05, "loss": 0.2853, "step": 56386 }, { "epoch": 3.4, "learning_rate": 1.9974952092103875e-05, "loss": 0.4818, "step": 56388 }, { "epoch": 3.4, "learning_rate": 1.9973443181989652e-05, "loss": 0.4615, "step": 56390 }, { "epoch": 3.4, "learning_rate": 1.9971934271875425e-05, "loss": 0.4433, "step": 56392 }, { "epoch": 3.4, "learning_rate": 1.9970425361761198e-05, "loss": 0.5147, "step": 56394 }, { "epoch": 3.4, "learning_rate": 1.9968916451646978e-05, "loss": 0.3175, "step": 56396 }, { "epoch": 3.4, "learning_rate": 1.996740754153275e-05, "loss": 0.3479, "step": 56398 }, { "epoch": 3.4, "learning_rate": 1.9965898631418528e-05, "loss": 0.3432, "step": 56400 }, { "epoch": 3.4, "learning_rate": 1.99643897213043e-05, "loss": 0.4552, "step": 56402 }, { "epoch": 3.4, "learning_rate": 1.996288081119008e-05, "loss": 0.416, "step": 56404 }, { "epoch": 3.4, "learning_rate": 1.9961371901075854e-05, "loss": 0.4001, "step": 56406 }, { "epoch": 3.4, "learning_rate": 1.9959862990961627e-05, "loss": 0.3711, "step": 56408 }, { "epoch": 3.4, "learning_rate": 1.9958354080847404e-05, "loss": 0.4012, "step": 56410 }, { "epoch": 3.4, "learning_rate": 1.995684517073318e-05, "loss": 0.4518, "step": 56412 }, { "epoch": 3.4, "learning_rate": 1.9955336260618957e-05, "loss": 0.4043, "step": 56414 }, { "epoch": 3.41, "learning_rate": 1.995382735050473e-05, "loss": 0.3004, "step": 56416 }, { "epoch": 3.41, "learning_rate": 1.9952318440390506e-05, "loss": 0.4792, "step": 56418 }, { "epoch": 3.41, "learning_rate": 1.9950809530276283e-05, "loss": 0.2626, "step": 56420 }, { "epoch": 3.41, "learning_rate": 1.994930062016206e-05, "loss": 0.3086, "step": 56422 }, { "epoch": 3.41, "learning_rate": 1.9947791710047832e-05, "loss": 0.3766, "step": 56424 }, { "epoch": 3.41, "learning_rate": 1.994628279993361e-05, "loss": 0.2506, "step": 56426 }, { "epoch": 3.41, "learning_rate": 1.9944773889819385e-05, "loss": 0.5483, "step": 56428 }, { "epoch": 3.41, "learning_rate": 1.994326497970516e-05, "loss": 0.2201, "step": 56430 }, { "epoch": 3.41, "learning_rate": 1.9941756069590935e-05, "loss": 0.305, "step": 56432 }, { "epoch": 3.41, "learning_rate": 1.994024715947671e-05, "loss": 0.5229, "step": 56434 }, { "epoch": 3.41, "learning_rate": 1.9938738249362488e-05, "loss": 0.3211, "step": 56436 }, { "epoch": 3.41, "learning_rate": 1.993722933924826e-05, "loss": 0.3165, "step": 56438 }, { "epoch": 3.41, "learning_rate": 1.9935720429134038e-05, "loss": 0.3297, "step": 56440 }, { "epoch": 3.41, "learning_rate": 1.9934211519019814e-05, "loss": 0.4282, "step": 56442 }, { "epoch": 3.41, "learning_rate": 1.9932702608905587e-05, "loss": 0.3222, "step": 56444 }, { "epoch": 3.41, "learning_rate": 1.9931193698791364e-05, "loss": 0.4158, "step": 56446 }, { "epoch": 3.41, "learning_rate": 1.992968478867714e-05, "loss": 0.3663, "step": 56448 }, { "epoch": 3.41, "learning_rate": 1.9928175878562917e-05, "loss": 0.513, "step": 56450 }, { "epoch": 3.41, "learning_rate": 1.992666696844869e-05, "loss": 0.3778, "step": 56452 }, { "epoch": 3.41, "learning_rate": 1.9925158058334467e-05, "loss": 0.3816, "step": 56454 }, { "epoch": 3.41, "learning_rate": 1.9923649148220243e-05, "loss": 0.2832, "step": 56456 }, { "epoch": 3.41, "learning_rate": 1.9922140238106016e-05, "loss": 0.4347, "step": 56458 }, { "epoch": 3.41, "learning_rate": 1.9920631327991793e-05, "loss": 0.361, "step": 56460 }, { "epoch": 3.41, "learning_rate": 1.9919122417877566e-05, "loss": 0.374, "step": 56462 }, { "epoch": 3.41, "learning_rate": 1.9917613507763346e-05, "loss": 0.3773, "step": 56464 }, { "epoch": 3.41, "learning_rate": 1.991610459764912e-05, "loss": 0.3904, "step": 56466 }, { "epoch": 3.41, "learning_rate": 1.9914595687534895e-05, "loss": 0.3993, "step": 56468 }, { "epoch": 3.41, "learning_rate": 1.991308677742067e-05, "loss": 0.3042, "step": 56470 }, { "epoch": 3.41, "learning_rate": 1.991157786730645e-05, "loss": 0.2758, "step": 56472 }, { "epoch": 3.41, "learning_rate": 1.991006895719222e-05, "loss": 0.6199, "step": 56474 }, { "epoch": 3.41, "learning_rate": 1.9908560047077995e-05, "loss": 0.2925, "step": 56476 }, { "epoch": 3.41, "learning_rate": 1.990705113696377e-05, "loss": 0.316, "step": 56478 }, { "epoch": 3.41, "learning_rate": 1.9905542226849548e-05, "loss": 0.3398, "step": 56480 }, { "epoch": 3.41, "learning_rate": 1.9904033316735324e-05, "loss": 0.3842, "step": 56482 }, { "epoch": 3.41, "learning_rate": 1.9902524406621097e-05, "loss": 0.3676, "step": 56484 }, { "epoch": 3.41, "learning_rate": 1.9901015496506874e-05, "loss": 0.3218, "step": 56486 }, { "epoch": 3.41, "learning_rate": 1.989950658639265e-05, "loss": 0.5747, "step": 56488 }, { "epoch": 3.41, "learning_rate": 1.9897997676278423e-05, "loss": 0.3933, "step": 56490 }, { "epoch": 3.41, "learning_rate": 1.98964887661642e-05, "loss": 0.5578, "step": 56492 }, { "epoch": 3.41, "learning_rate": 1.9894979856049977e-05, "loss": 0.2512, "step": 56494 }, { "epoch": 3.41, "learning_rate": 1.9893470945935753e-05, "loss": 0.3209, "step": 56496 }, { "epoch": 3.41, "learning_rate": 1.9891962035821526e-05, "loss": 0.3283, "step": 56498 }, { "epoch": 3.41, "learning_rate": 1.9890453125707303e-05, "loss": 0.3143, "step": 56500 }, { "epoch": 3.41, "learning_rate": 1.988894421559308e-05, "loss": 0.2566, "step": 56502 }, { "epoch": 3.41, "learning_rate": 1.9887435305478856e-05, "loss": 0.2723, "step": 56504 }, { "epoch": 3.41, "learning_rate": 1.988592639536463e-05, "loss": 0.4517, "step": 56506 }, { "epoch": 3.41, "learning_rate": 1.9884417485250402e-05, "loss": 0.3596, "step": 56508 }, { "epoch": 3.41, "learning_rate": 1.9882908575136182e-05, "loss": 0.5184, "step": 56510 }, { "epoch": 3.41, "learning_rate": 1.9881399665021955e-05, "loss": 0.3856, "step": 56512 }, { "epoch": 3.41, "learning_rate": 1.987989075490773e-05, "loss": 0.2935, "step": 56514 }, { "epoch": 3.41, "learning_rate": 1.9878381844793505e-05, "loss": 0.4122, "step": 56516 }, { "epoch": 3.41, "learning_rate": 1.9876872934679285e-05, "loss": 0.6695, "step": 56518 }, { "epoch": 3.41, "learning_rate": 1.9875364024565058e-05, "loss": 0.3477, "step": 56520 }, { "epoch": 3.41, "learning_rate": 1.9873855114450834e-05, "loss": 0.4689, "step": 56522 }, { "epoch": 3.41, "learning_rate": 1.9872346204336607e-05, "loss": 0.4292, "step": 56524 }, { "epoch": 3.41, "learning_rate": 1.9870837294222384e-05, "loss": 0.3816, "step": 56526 }, { "epoch": 3.41, "learning_rate": 1.986932838410816e-05, "loss": 0.453, "step": 56528 }, { "epoch": 3.41, "learning_rate": 1.9867819473993933e-05, "loss": 0.3897, "step": 56530 }, { "epoch": 3.41, "learning_rate": 1.986631056387971e-05, "loss": 0.3514, "step": 56532 }, { "epoch": 3.41, "learning_rate": 1.9864801653765486e-05, "loss": 0.3847, "step": 56534 }, { "epoch": 3.41, "learning_rate": 1.9863292743651263e-05, "loss": 0.2972, "step": 56536 }, { "epoch": 3.41, "learning_rate": 1.9861783833537036e-05, "loss": 0.3189, "step": 56538 }, { "epoch": 3.41, "learning_rate": 1.9860274923422813e-05, "loss": 0.2922, "step": 56540 }, { "epoch": 3.41, "learning_rate": 1.985876601330859e-05, "loss": 0.3579, "step": 56542 }, { "epoch": 3.41, "learning_rate": 1.9857257103194362e-05, "loss": 0.3113, "step": 56544 }, { "epoch": 3.41, "learning_rate": 1.985574819308014e-05, "loss": 0.4214, "step": 56546 }, { "epoch": 3.41, "learning_rate": 1.9854239282965915e-05, "loss": 0.2073, "step": 56548 }, { "epoch": 3.41, "learning_rate": 1.9852730372851692e-05, "loss": 0.4348, "step": 56550 }, { "epoch": 3.41, "learning_rate": 1.9851221462737465e-05, "loss": 0.424, "step": 56552 }, { "epoch": 3.41, "learning_rate": 1.984971255262324e-05, "loss": 0.3031, "step": 56554 }, { "epoch": 3.41, "learning_rate": 1.9848203642509018e-05, "loss": 0.609, "step": 56556 }, { "epoch": 3.41, "learning_rate": 1.984669473239479e-05, "loss": 0.3666, "step": 56558 }, { "epoch": 3.41, "learning_rate": 1.9845185822280568e-05, "loss": 0.4489, "step": 56560 }, { "epoch": 3.41, "learning_rate": 1.984367691216634e-05, "loss": 0.4706, "step": 56562 }, { "epoch": 3.41, "learning_rate": 1.984216800205212e-05, "loss": 0.5841, "step": 56564 }, { "epoch": 3.41, "learning_rate": 1.9840659091937894e-05, "loss": 0.3914, "step": 56566 }, { "epoch": 3.41, "learning_rate": 1.983915018182367e-05, "loss": 0.4135, "step": 56568 }, { "epoch": 3.41, "learning_rate": 1.9837641271709447e-05, "loss": 0.291, "step": 56570 }, { "epoch": 3.41, "learning_rate": 1.983613236159522e-05, "loss": 0.2767, "step": 56572 }, { "epoch": 3.41, "learning_rate": 1.9834623451480996e-05, "loss": 0.3516, "step": 56574 }, { "epoch": 3.41, "learning_rate": 1.983311454136677e-05, "loss": 0.4555, "step": 56576 }, { "epoch": 3.41, "learning_rate": 1.983160563125255e-05, "loss": 0.4805, "step": 56578 }, { "epoch": 3.42, "learning_rate": 1.9830096721138323e-05, "loss": 0.4965, "step": 56580 }, { "epoch": 3.42, "learning_rate": 1.98285878110241e-05, "loss": 0.4527, "step": 56582 }, { "epoch": 3.42, "learning_rate": 1.9827078900909872e-05, "loss": 0.4059, "step": 56584 }, { "epoch": 3.42, "learning_rate": 1.9825569990795652e-05, "loss": 0.3127, "step": 56586 }, { "epoch": 3.42, "learning_rate": 1.9824061080681425e-05, "loss": 0.5038, "step": 56588 }, { "epoch": 3.42, "learning_rate": 1.98225521705672e-05, "loss": 0.4361, "step": 56590 }, { "epoch": 3.42, "learning_rate": 1.9821043260452975e-05, "loss": 0.6056, "step": 56592 }, { "epoch": 3.42, "learning_rate": 1.981953435033875e-05, "loss": 0.4673, "step": 56594 }, { "epoch": 3.42, "learning_rate": 1.9818025440224528e-05, "loss": 0.4274, "step": 56596 }, { "epoch": 3.42, "learning_rate": 1.98165165301103e-05, "loss": 0.3644, "step": 56598 }, { "epoch": 3.42, "learning_rate": 1.9815007619996078e-05, "loss": 0.2476, "step": 56600 }, { "epoch": 3.42, "learning_rate": 1.9813498709881854e-05, "loss": 0.4213, "step": 56602 }, { "epoch": 3.42, "learning_rate": 1.981198979976763e-05, "loss": 0.4808, "step": 56604 }, { "epoch": 3.42, "learning_rate": 1.9810480889653404e-05, "loss": 0.2969, "step": 56606 }, { "epoch": 3.42, "learning_rate": 1.980897197953918e-05, "loss": 0.4808, "step": 56608 }, { "epoch": 3.42, "learning_rate": 1.9807463069424957e-05, "loss": 0.3707, "step": 56610 }, { "epoch": 3.42, "learning_rate": 1.980595415931073e-05, "loss": 0.4232, "step": 56612 }, { "epoch": 3.42, "learning_rate": 1.9804445249196506e-05, "loss": 0.3219, "step": 56614 }, { "epoch": 3.42, "learning_rate": 1.9802936339082283e-05, "loss": 0.3831, "step": 56616 }, { "epoch": 3.42, "learning_rate": 1.980142742896806e-05, "loss": 0.3094, "step": 56618 }, { "epoch": 3.42, "learning_rate": 1.9799918518853832e-05, "loss": 0.4403, "step": 56620 }, { "epoch": 3.42, "learning_rate": 1.9798409608739606e-05, "loss": 0.4416, "step": 56622 }, { "epoch": 3.42, "learning_rate": 1.9796900698625386e-05, "loss": 0.4426, "step": 56624 }, { "epoch": 3.42, "learning_rate": 1.979539178851116e-05, "loss": 0.3073, "step": 56626 }, { "epoch": 3.42, "learning_rate": 1.9793882878396935e-05, "loss": 0.3614, "step": 56628 }, { "epoch": 3.42, "learning_rate": 1.9792373968282708e-05, "loss": 0.3706, "step": 56630 }, { "epoch": 3.42, "learning_rate": 1.9790865058168488e-05, "loss": 0.261, "step": 56632 }, { "epoch": 3.42, "learning_rate": 1.978935614805426e-05, "loss": 0.5066, "step": 56634 }, { "epoch": 3.42, "learning_rate": 1.9787847237940038e-05, "loss": 0.3815, "step": 56636 }, { "epoch": 3.42, "learning_rate": 1.978633832782581e-05, "loss": 0.304, "step": 56638 }, { "epoch": 3.42, "learning_rate": 1.9784829417711587e-05, "loss": 0.3587, "step": 56640 }, { "epoch": 3.42, "learning_rate": 1.9783320507597364e-05, "loss": 0.3993, "step": 56642 }, { "epoch": 3.42, "learning_rate": 1.9781811597483137e-05, "loss": 0.3792, "step": 56644 }, { "epoch": 3.42, "learning_rate": 1.9780302687368914e-05, "loss": 0.4025, "step": 56646 }, { "epoch": 3.42, "learning_rate": 1.977879377725469e-05, "loss": 0.5072, "step": 56648 }, { "epoch": 3.42, "learning_rate": 1.9777284867140467e-05, "loss": 0.2762, "step": 56650 }, { "epoch": 3.42, "learning_rate": 1.977577595702624e-05, "loss": 0.2869, "step": 56652 }, { "epoch": 3.42, "learning_rate": 1.9774267046912016e-05, "loss": 0.4296, "step": 56654 }, { "epoch": 3.42, "learning_rate": 1.9772758136797793e-05, "loss": 0.3308, "step": 56656 }, { "epoch": 3.42, "learning_rate": 1.9771249226683566e-05, "loss": 0.4177, "step": 56658 }, { "epoch": 3.42, "learning_rate": 1.9769740316569342e-05, "loss": 0.4793, "step": 56660 }, { "epoch": 3.42, "learning_rate": 1.976823140645512e-05, "loss": 0.4096, "step": 56662 }, { "epoch": 3.42, "learning_rate": 1.9766722496340895e-05, "loss": 0.2778, "step": 56664 }, { "epoch": 3.42, "learning_rate": 1.976521358622667e-05, "loss": 0.2892, "step": 56666 }, { "epoch": 3.42, "learning_rate": 1.9763704676112445e-05, "loss": 0.6888, "step": 56668 }, { "epoch": 3.42, "learning_rate": 1.976219576599822e-05, "loss": 0.5485, "step": 56670 }, { "epoch": 3.42, "learning_rate": 1.9760686855883995e-05, "loss": 0.3497, "step": 56672 }, { "epoch": 3.42, "learning_rate": 1.975917794576977e-05, "loss": 0.4614, "step": 56674 }, { "epoch": 3.42, "learning_rate": 1.9757669035655544e-05, "loss": 0.5593, "step": 56676 }, { "epoch": 3.42, "learning_rate": 1.9756160125541324e-05, "loss": 0.4343, "step": 56678 }, { "epoch": 3.42, "learning_rate": 1.9754651215427097e-05, "loss": 0.3629, "step": 56680 }, { "epoch": 3.42, "learning_rate": 1.9753142305312874e-05, "loss": 0.3404, "step": 56682 }, { "epoch": 3.42, "learning_rate": 1.975163339519865e-05, "loss": 0.3927, "step": 56684 }, { "epoch": 3.42, "learning_rate": 1.9750124485084427e-05, "loss": 0.7036, "step": 56686 }, { "epoch": 3.42, "learning_rate": 1.97486155749702e-05, "loss": 0.28, "step": 56688 }, { "epoch": 3.42, "learning_rate": 1.9747106664855973e-05, "loss": 0.2717, "step": 56690 }, { "epoch": 3.42, "learning_rate": 1.9745597754741753e-05, "loss": 0.5175, "step": 56692 }, { "epoch": 3.42, "learning_rate": 1.9744088844627526e-05, "loss": 0.3633, "step": 56694 }, { "epoch": 3.42, "learning_rate": 1.9742579934513303e-05, "loss": 0.4095, "step": 56696 }, { "epoch": 3.42, "learning_rate": 1.9741071024399076e-05, "loss": 0.428, "step": 56698 }, { "epoch": 3.42, "learning_rate": 1.9739562114284856e-05, "loss": 0.5528, "step": 56700 }, { "epoch": 3.42, "learning_rate": 1.973805320417063e-05, "loss": 0.3553, "step": 56702 }, { "epoch": 3.42, "learning_rate": 1.9736544294056402e-05, "loss": 0.5042, "step": 56704 }, { "epoch": 3.42, "learning_rate": 1.973503538394218e-05, "loss": 0.3094, "step": 56706 }, { "epoch": 3.42, "learning_rate": 1.9733526473827955e-05, "loss": 0.316, "step": 56708 }, { "epoch": 3.42, "learning_rate": 1.973201756371373e-05, "loss": 0.3469, "step": 56710 }, { "epoch": 3.42, "learning_rate": 1.9730508653599505e-05, "loss": 0.3024, "step": 56712 }, { "epoch": 3.42, "learning_rate": 1.972899974348528e-05, "loss": 0.5624, "step": 56714 }, { "epoch": 3.42, "learning_rate": 1.9727490833371058e-05, "loss": 0.438, "step": 56716 }, { "epoch": 3.42, "learning_rate": 1.9725981923256834e-05, "loss": 0.3336, "step": 56718 }, { "epoch": 3.42, "learning_rate": 1.9724473013142607e-05, "loss": 0.2415, "step": 56720 }, { "epoch": 3.42, "learning_rate": 1.9722964103028384e-05, "loss": 0.3717, "step": 56722 }, { "epoch": 3.42, "learning_rate": 1.972145519291416e-05, "loss": 0.2247, "step": 56724 }, { "epoch": 3.42, "learning_rate": 1.9719946282799934e-05, "loss": 0.4219, "step": 56726 }, { "epoch": 3.42, "learning_rate": 1.971843737268571e-05, "loss": 0.3757, "step": 56728 }, { "epoch": 3.42, "learning_rate": 1.9716928462571487e-05, "loss": 0.271, "step": 56730 }, { "epoch": 3.42, "learning_rate": 1.9715419552457263e-05, "loss": 0.3053, "step": 56732 }, { "epoch": 3.42, "learning_rate": 1.9713910642343036e-05, "loss": 0.3991, "step": 56734 }, { "epoch": 3.42, "learning_rate": 1.971240173222881e-05, "loss": 0.2747, "step": 56736 }, { "epoch": 3.42, "learning_rate": 1.971089282211459e-05, "loss": 0.3371, "step": 56738 }, { "epoch": 3.42, "learning_rate": 1.9709383912000362e-05, "loss": 0.3723, "step": 56740 }, { "epoch": 3.42, "learning_rate": 1.970787500188614e-05, "loss": 0.5415, "step": 56742 }, { "epoch": 3.42, "learning_rate": 1.9706366091771912e-05, "loss": 0.3354, "step": 56744 }, { "epoch": 3.43, "learning_rate": 1.9704857181657692e-05, "loss": 0.4967, "step": 56746 }, { "epoch": 3.43, "learning_rate": 1.9703348271543465e-05, "loss": 0.4065, "step": 56748 }, { "epoch": 3.43, "learning_rate": 1.970183936142924e-05, "loss": 0.487, "step": 56750 }, { "epoch": 3.43, "learning_rate": 1.9700330451315015e-05, "loss": 0.3727, "step": 56752 }, { "epoch": 3.43, "learning_rate": 1.969882154120079e-05, "loss": 0.45, "step": 56754 }, { "epoch": 3.43, "learning_rate": 1.9697312631086568e-05, "loss": 0.3018, "step": 56756 }, { "epoch": 3.43, "learning_rate": 1.969580372097234e-05, "loss": 0.3876, "step": 56758 }, { "epoch": 3.43, "learning_rate": 1.9694294810858117e-05, "loss": 0.3607, "step": 56760 }, { "epoch": 3.43, "learning_rate": 1.9692785900743894e-05, "loss": 0.3228, "step": 56762 }, { "epoch": 3.43, "learning_rate": 1.969127699062967e-05, "loss": 0.3064, "step": 56764 }, { "epoch": 3.43, "learning_rate": 1.9689768080515443e-05, "loss": 0.438, "step": 56766 }, { "epoch": 3.43, "learning_rate": 1.968825917040122e-05, "loss": 0.2556, "step": 56768 }, { "epoch": 3.43, "learning_rate": 1.9686750260286996e-05, "loss": 0.4245, "step": 56770 }, { "epoch": 3.43, "learning_rate": 1.968524135017277e-05, "loss": 0.2607, "step": 56772 }, { "epoch": 3.43, "learning_rate": 1.9683732440058546e-05, "loss": 0.265, "step": 56774 }, { "epoch": 3.43, "learning_rate": 1.9682223529944323e-05, "loss": 0.6072, "step": 56776 }, { "epoch": 3.43, "learning_rate": 1.96807146198301e-05, "loss": 0.5165, "step": 56778 }, { "epoch": 3.43, "learning_rate": 1.9679205709715872e-05, "loss": 0.4944, "step": 56780 }, { "epoch": 3.43, "learning_rate": 1.967769679960165e-05, "loss": 0.2933, "step": 56782 }, { "epoch": 3.43, "learning_rate": 1.9676187889487425e-05, "loss": 0.3878, "step": 56784 }, { "epoch": 3.43, "learning_rate": 1.96746789793732e-05, "loss": 0.2497, "step": 56786 }, { "epoch": 3.43, "learning_rate": 1.9673170069258975e-05, "loss": 0.4175, "step": 56788 }, { "epoch": 3.43, "learning_rate": 1.9671661159144748e-05, "loss": 0.4146, "step": 56790 }, { "epoch": 3.43, "learning_rate": 1.9670152249030528e-05, "loss": 0.3195, "step": 56792 }, { "epoch": 3.43, "learning_rate": 1.96686433389163e-05, "loss": 0.3035, "step": 56794 }, { "epoch": 3.43, "learning_rate": 1.9667134428802078e-05, "loss": 0.3417, "step": 56796 }, { "epoch": 3.43, "learning_rate": 1.966562551868785e-05, "loss": 0.4467, "step": 56798 }, { "epoch": 3.43, "learning_rate": 1.966411660857363e-05, "loss": 0.3452, "step": 56800 }, { "epoch": 3.43, "learning_rate": 1.9662607698459404e-05, "loss": 0.4604, "step": 56802 }, { "epoch": 3.43, "learning_rate": 1.9661098788345177e-05, "loss": 0.3601, "step": 56804 }, { "epoch": 3.43, "learning_rate": 1.9659589878230957e-05, "loss": 0.442, "step": 56806 }, { "epoch": 3.43, "learning_rate": 1.965808096811673e-05, "loss": 0.3872, "step": 56808 }, { "epoch": 3.43, "learning_rate": 1.9656572058002506e-05, "loss": 0.4025, "step": 56810 }, { "epoch": 3.43, "learning_rate": 1.965506314788828e-05, "loss": 0.279, "step": 56812 }, { "epoch": 3.43, "learning_rate": 1.965355423777406e-05, "loss": 0.4726, "step": 56814 }, { "epoch": 3.43, "learning_rate": 1.9652045327659833e-05, "loss": 0.3941, "step": 56816 }, { "epoch": 3.43, "learning_rate": 1.9650536417545606e-05, "loss": 0.3369, "step": 56818 }, { "epoch": 3.43, "learning_rate": 1.9649027507431382e-05, "loss": 0.32, "step": 56820 }, { "epoch": 3.43, "learning_rate": 1.964751859731716e-05, "loss": 0.3156, "step": 56822 }, { "epoch": 3.43, "learning_rate": 1.9646009687202935e-05, "loss": 0.3688, "step": 56824 }, { "epoch": 3.43, "learning_rate": 1.964450077708871e-05, "loss": 0.462, "step": 56826 }, { "epoch": 3.43, "learning_rate": 1.9642991866974485e-05, "loss": 0.4063, "step": 56828 }, { "epoch": 3.43, "learning_rate": 1.964148295686026e-05, "loss": 0.3172, "step": 56830 }, { "epoch": 3.43, "learning_rate": 1.9639974046746038e-05, "loss": 0.3642, "step": 56832 }, { "epoch": 3.43, "learning_rate": 1.963846513663181e-05, "loss": 0.2388, "step": 56834 }, { "epoch": 3.43, "learning_rate": 1.9636956226517588e-05, "loss": 0.3961, "step": 56836 }, { "epoch": 3.43, "learning_rate": 1.9635447316403364e-05, "loss": 0.2414, "step": 56838 }, { "epoch": 3.43, "learning_rate": 1.9633938406289137e-05, "loss": 0.2522, "step": 56840 }, { "epoch": 3.43, "learning_rate": 1.9632429496174914e-05, "loss": 0.3272, "step": 56842 }, { "epoch": 3.43, "learning_rate": 1.963092058606069e-05, "loss": 0.3058, "step": 56844 }, { "epoch": 3.43, "learning_rate": 1.9629411675946467e-05, "loss": 0.3814, "step": 56846 }, { "epoch": 3.43, "learning_rate": 1.962790276583224e-05, "loss": 0.4027, "step": 56848 }, { "epoch": 3.43, "learning_rate": 1.9626393855718016e-05, "loss": 0.3964, "step": 56850 }, { "epoch": 3.43, "learning_rate": 1.9624884945603793e-05, "loss": 0.4795, "step": 56852 }, { "epoch": 3.43, "learning_rate": 1.9623376035489566e-05, "loss": 0.376, "step": 56854 }, { "epoch": 3.43, "learning_rate": 1.9621867125375343e-05, "loss": 0.432, "step": 56856 }, { "epoch": 3.43, "learning_rate": 1.9620358215261116e-05, "loss": 0.3886, "step": 56858 }, { "epoch": 3.43, "learning_rate": 1.9618849305146896e-05, "loss": 0.3631, "step": 56860 }, { "epoch": 3.43, "learning_rate": 1.961734039503267e-05, "loss": 0.4162, "step": 56862 }, { "epoch": 3.43, "learning_rate": 1.9615831484918445e-05, "loss": 0.4056, "step": 56864 }, { "epoch": 3.43, "learning_rate": 1.961432257480422e-05, "loss": 0.5046, "step": 56866 }, { "epoch": 3.43, "learning_rate": 1.9612813664689995e-05, "loss": 0.2796, "step": 56868 }, { "epoch": 3.43, "learning_rate": 1.961130475457577e-05, "loss": 0.3865, "step": 56870 }, { "epoch": 3.43, "learning_rate": 1.9609795844461544e-05, "loss": 0.5053, "step": 56872 }, { "epoch": 3.43, "learning_rate": 1.960828693434732e-05, "loss": 0.39, "step": 56874 }, { "epoch": 3.43, "learning_rate": 1.9606778024233097e-05, "loss": 0.4615, "step": 56876 }, { "epoch": 3.43, "learning_rate": 1.9605269114118874e-05, "loss": 0.4063, "step": 56878 }, { "epoch": 3.43, "learning_rate": 1.9603760204004647e-05, "loss": 0.3891, "step": 56880 }, { "epoch": 3.43, "learning_rate": 1.9602251293890424e-05, "loss": 0.2404, "step": 56882 }, { "epoch": 3.43, "learning_rate": 1.96007423837762e-05, "loss": 0.3688, "step": 56884 }, { "epoch": 3.43, "learning_rate": 1.9599233473661973e-05, "loss": 0.3938, "step": 56886 }, { "epoch": 3.43, "learning_rate": 1.959772456354775e-05, "loss": 0.2503, "step": 56888 }, { "epoch": 3.43, "learning_rate": 1.9596215653433526e-05, "loss": 0.4432, "step": 56890 }, { "epoch": 3.43, "learning_rate": 1.9594706743319303e-05, "loss": 0.4344, "step": 56892 }, { "epoch": 3.43, "learning_rate": 1.9593197833205076e-05, "loss": 0.341, "step": 56894 }, { "epoch": 3.43, "learning_rate": 1.9591688923090852e-05, "loss": 0.3498, "step": 56896 }, { "epoch": 3.43, "learning_rate": 1.959018001297663e-05, "loss": 0.3352, "step": 56898 }, { "epoch": 3.43, "learning_rate": 1.9588671102862405e-05, "loss": 0.3428, "step": 56900 }, { "epoch": 3.43, "learning_rate": 1.958716219274818e-05, "loss": 0.3125, "step": 56902 }, { "epoch": 3.43, "learning_rate": 1.9585653282633952e-05, "loss": 0.2414, "step": 56904 }, { "epoch": 3.43, "learning_rate": 1.958414437251973e-05, "loss": 0.2503, "step": 56906 }, { "epoch": 3.43, "learning_rate": 1.9582635462405505e-05, "loss": 0.3136, "step": 56908 }, { "epoch": 3.43, "learning_rate": 1.958112655229128e-05, "loss": 0.3608, "step": 56910 }, { "epoch": 3.44, "learning_rate": 1.9579617642177054e-05, "loss": 0.2686, "step": 56912 }, { "epoch": 3.44, "learning_rate": 1.9578108732062834e-05, "loss": 0.309, "step": 56914 }, { "epoch": 3.44, "learning_rate": 1.9576599821948607e-05, "loss": 0.2906, "step": 56916 }, { "epoch": 3.44, "learning_rate": 1.957509091183438e-05, "loss": 0.3389, "step": 56918 }, { "epoch": 3.44, "learning_rate": 1.957358200172016e-05, "loss": 0.2169, "step": 56920 }, { "epoch": 3.44, "learning_rate": 1.9572073091605934e-05, "loss": 0.4912, "step": 56922 }, { "epoch": 3.44, "learning_rate": 1.957056418149171e-05, "loss": 0.391, "step": 56924 }, { "epoch": 3.44, "learning_rate": 1.9569055271377483e-05, "loss": 0.3066, "step": 56926 }, { "epoch": 3.44, "learning_rate": 1.9567546361263263e-05, "loss": 0.235, "step": 56928 }, { "epoch": 3.44, "learning_rate": 1.9566037451149036e-05, "loss": 0.3698, "step": 56930 }, { "epoch": 3.44, "learning_rate": 1.9564528541034813e-05, "loss": 0.4247, "step": 56932 }, { "epoch": 3.44, "learning_rate": 1.9563019630920586e-05, "loss": 0.5204, "step": 56934 }, { "epoch": 3.44, "learning_rate": 1.9561510720806362e-05, "loss": 0.4704, "step": 56936 }, { "epoch": 3.44, "learning_rate": 1.956000181069214e-05, "loss": 0.469, "step": 56938 }, { "epoch": 3.44, "learning_rate": 1.9558492900577912e-05, "loss": 0.3183, "step": 56940 }, { "epoch": 3.44, "learning_rate": 1.955698399046369e-05, "loss": 0.3263, "step": 56942 }, { "epoch": 3.44, "learning_rate": 1.9555475080349465e-05, "loss": 0.4017, "step": 56944 }, { "epoch": 3.44, "learning_rate": 1.955396617023524e-05, "loss": 0.4347, "step": 56946 }, { "epoch": 3.44, "learning_rate": 1.9552457260121015e-05, "loss": 0.325, "step": 56948 }, { "epoch": 3.44, "learning_rate": 1.955094835000679e-05, "loss": 0.5913, "step": 56950 }, { "epoch": 3.44, "learning_rate": 1.9549439439892568e-05, "loss": 0.4805, "step": 56952 }, { "epoch": 3.44, "learning_rate": 1.954793052977834e-05, "loss": 0.3518, "step": 56954 }, { "epoch": 3.44, "learning_rate": 1.9546421619664117e-05, "loss": 0.3795, "step": 56956 }, { "epoch": 3.44, "learning_rate": 1.9544912709549894e-05, "loss": 0.4424, "step": 56958 }, { "epoch": 3.44, "learning_rate": 1.954340379943567e-05, "loss": 0.3962, "step": 56960 }, { "epoch": 3.44, "learning_rate": 1.9541894889321444e-05, "loss": 0.3457, "step": 56962 }, { "epoch": 3.44, "learning_rate": 1.954038597920722e-05, "loss": 0.3282, "step": 56964 }, { "epoch": 3.44, "learning_rate": 1.9538877069092997e-05, "loss": 0.3724, "step": 56966 }, { "epoch": 3.44, "learning_rate": 1.953736815897877e-05, "loss": 0.2375, "step": 56968 }, { "epoch": 3.44, "learning_rate": 1.9535859248864546e-05, "loss": 0.3316, "step": 56970 }, { "epoch": 3.44, "learning_rate": 1.953435033875032e-05, "loss": 0.3502, "step": 56972 }, { "epoch": 3.44, "learning_rate": 1.95328414286361e-05, "loss": 0.2703, "step": 56974 }, { "epoch": 3.44, "learning_rate": 1.9531332518521872e-05, "loss": 0.3402, "step": 56976 }, { "epoch": 3.44, "learning_rate": 1.952982360840765e-05, "loss": 0.4727, "step": 56978 }, { "epoch": 3.44, "learning_rate": 1.9528314698293422e-05, "loss": 0.2252, "step": 56980 }, { "epoch": 3.44, "learning_rate": 1.9526805788179202e-05, "loss": 0.4214, "step": 56982 }, { "epoch": 3.44, "learning_rate": 1.9525296878064975e-05, "loss": 0.3844, "step": 56984 }, { "epoch": 3.44, "learning_rate": 1.9523787967950748e-05, "loss": 0.4764, "step": 56986 }, { "epoch": 3.44, "learning_rate": 1.9522279057836525e-05, "loss": 0.2715, "step": 56988 }, { "epoch": 3.44, "learning_rate": 1.95207701477223e-05, "loss": 0.5396, "step": 56990 }, { "epoch": 3.44, "learning_rate": 1.9519261237608078e-05, "loss": 0.3109, "step": 56992 }, { "epoch": 3.44, "learning_rate": 1.951775232749385e-05, "loss": 0.4658, "step": 56994 }, { "epoch": 3.44, "learning_rate": 1.9516243417379627e-05, "loss": 0.3757, "step": 56996 }, { "epoch": 3.44, "learning_rate": 1.9514734507265404e-05, "loss": 0.3981, "step": 56998 }, { "epoch": 3.44, "learning_rate": 1.9513225597151177e-05, "loss": 0.5417, "step": 57000 }, { "epoch": 3.44, "learning_rate": 1.9511716687036953e-05, "loss": 0.4766, "step": 57002 }, { "epoch": 3.44, "learning_rate": 1.951020777692273e-05, "loss": 0.2503, "step": 57004 }, { "epoch": 3.44, "learning_rate": 1.9508698866808506e-05, "loss": 0.3657, "step": 57006 }, { "epoch": 3.44, "learning_rate": 1.950718995669428e-05, "loss": 0.3897, "step": 57008 }, { "epoch": 3.44, "learning_rate": 1.9505681046580056e-05, "loss": 0.4632, "step": 57010 }, { "epoch": 3.44, "learning_rate": 1.9504172136465833e-05, "loss": 0.3658, "step": 57012 }, { "epoch": 3.44, "learning_rate": 1.950266322635161e-05, "loss": 0.486, "step": 57014 }, { "epoch": 3.44, "learning_rate": 1.9501154316237382e-05, "loss": 0.3684, "step": 57016 }, { "epoch": 3.44, "learning_rate": 1.9499645406123155e-05, "loss": 0.3014, "step": 57018 }, { "epoch": 3.44, "learning_rate": 1.9498136496008935e-05, "loss": 0.4427, "step": 57020 }, { "epoch": 3.44, "learning_rate": 1.949662758589471e-05, "loss": 0.3618, "step": 57022 }, { "epoch": 3.44, "learning_rate": 1.9495118675780485e-05, "loss": 0.375, "step": 57024 }, { "epoch": 3.44, "learning_rate": 1.9493609765666258e-05, "loss": 0.3255, "step": 57026 }, { "epoch": 3.44, "learning_rate": 1.9492100855552038e-05, "loss": 0.3536, "step": 57028 }, { "epoch": 3.44, "learning_rate": 1.949059194543781e-05, "loss": 0.5308, "step": 57030 }, { "epoch": 3.44, "learning_rate": 1.9489083035323584e-05, "loss": 0.2474, "step": 57032 }, { "epoch": 3.44, "learning_rate": 1.948757412520936e-05, "loss": 0.3267, "step": 57034 }, { "epoch": 3.44, "learning_rate": 1.9486065215095137e-05, "loss": 0.3974, "step": 57036 }, { "epoch": 3.44, "learning_rate": 1.9484556304980914e-05, "loss": 0.4998, "step": 57038 }, { "epoch": 3.44, "learning_rate": 1.9483047394866687e-05, "loss": 0.3865, "step": 57040 }, { "epoch": 3.44, "learning_rate": 1.9481538484752467e-05, "loss": 0.3353, "step": 57042 }, { "epoch": 3.44, "learning_rate": 1.948002957463824e-05, "loss": 0.3126, "step": 57044 }, { "epoch": 3.44, "learning_rate": 1.9478520664524016e-05, "loss": 0.3312, "step": 57046 }, { "epoch": 3.44, "learning_rate": 1.947701175440979e-05, "loss": 0.3086, "step": 57048 }, { "epoch": 3.44, "learning_rate": 1.9475502844295566e-05, "loss": 0.4053, "step": 57050 }, { "epoch": 3.44, "learning_rate": 1.9473993934181343e-05, "loss": 0.351, "step": 57052 }, { "epoch": 3.44, "learning_rate": 1.9472485024067116e-05, "loss": 0.3987, "step": 57054 }, { "epoch": 3.44, "learning_rate": 1.9470976113952892e-05, "loss": 0.489, "step": 57056 }, { "epoch": 3.44, "learning_rate": 1.946946720383867e-05, "loss": 0.4629, "step": 57058 }, { "epoch": 3.44, "learning_rate": 1.9467958293724445e-05, "loss": 0.3423, "step": 57060 }, { "epoch": 3.44, "learning_rate": 1.946644938361022e-05, "loss": 0.3168, "step": 57062 }, { "epoch": 3.44, "learning_rate": 1.9464940473495995e-05, "loss": 0.3134, "step": 57064 }, { "epoch": 3.44, "learning_rate": 1.946343156338177e-05, "loss": 0.422, "step": 57066 }, { "epoch": 3.44, "learning_rate": 1.9461922653267545e-05, "loss": 0.4627, "step": 57068 }, { "epoch": 3.44, "learning_rate": 1.946041374315332e-05, "loss": 0.335, "step": 57070 }, { "epoch": 3.44, "learning_rate": 1.9458904833039098e-05, "loss": 0.46, "step": 57072 }, { "epoch": 3.44, "learning_rate": 1.9457395922924874e-05, "loss": 0.3837, "step": 57074 }, { "epoch": 3.44, "learning_rate": 1.9455887012810647e-05, "loss": 0.3297, "step": 57076 }, { "epoch": 3.45, "learning_rate": 1.9454378102696424e-05, "loss": 0.3229, "step": 57078 }, { "epoch": 3.45, "learning_rate": 1.94528691925822e-05, "loss": 0.3532, "step": 57080 }, { "epoch": 3.45, "learning_rate": 1.9451360282467973e-05, "loss": 0.3933, "step": 57082 }, { "epoch": 3.45, "learning_rate": 1.944985137235375e-05, "loss": 0.3833, "step": 57084 }, { "epoch": 3.45, "learning_rate": 1.9448342462239523e-05, "loss": 0.3223, "step": 57086 }, { "epoch": 3.45, "learning_rate": 1.9446833552125303e-05, "loss": 0.4304, "step": 57088 }, { "epoch": 3.45, "learning_rate": 1.9445324642011076e-05, "loss": 0.3027, "step": 57090 }, { "epoch": 3.45, "learning_rate": 1.9443815731896853e-05, "loss": 0.4161, "step": 57092 }, { "epoch": 3.45, "learning_rate": 1.9442306821782626e-05, "loss": 0.5061, "step": 57094 }, { "epoch": 3.45, "learning_rate": 1.9440797911668406e-05, "loss": 0.5022, "step": 57096 }, { "epoch": 3.45, "learning_rate": 1.943928900155418e-05, "loss": 0.3898, "step": 57098 }, { "epoch": 3.45, "learning_rate": 1.9437780091439952e-05, "loss": 0.3435, "step": 57100 }, { "epoch": 3.45, "learning_rate": 1.943627118132573e-05, "loss": 0.2747, "step": 57102 }, { "epoch": 3.45, "learning_rate": 1.9434762271211505e-05, "loss": 0.3569, "step": 57104 }, { "epoch": 3.45, "learning_rate": 1.943325336109728e-05, "loss": 0.2804, "step": 57106 }, { "epoch": 3.45, "learning_rate": 1.9431744450983054e-05, "loss": 0.3249, "step": 57108 }, { "epoch": 3.45, "learning_rate": 1.943023554086883e-05, "loss": 0.3315, "step": 57110 }, { "epoch": 3.45, "learning_rate": 1.9428726630754608e-05, "loss": 0.2775, "step": 57112 }, { "epoch": 3.45, "learning_rate": 1.942721772064038e-05, "loss": 0.4741, "step": 57114 }, { "epoch": 3.45, "learning_rate": 1.9425708810526157e-05, "loss": 0.3153, "step": 57116 }, { "epoch": 3.45, "learning_rate": 1.9424199900411934e-05, "loss": 0.3214, "step": 57118 }, { "epoch": 3.45, "learning_rate": 1.942269099029771e-05, "loss": 0.2334, "step": 57120 }, { "epoch": 3.45, "learning_rate": 1.9421182080183483e-05, "loss": 0.3658, "step": 57122 }, { "epoch": 3.45, "learning_rate": 1.941967317006926e-05, "loss": 0.2371, "step": 57124 }, { "epoch": 3.45, "learning_rate": 1.9418164259955036e-05, "loss": 0.2845, "step": 57126 }, { "epoch": 3.45, "learning_rate": 1.9416655349840813e-05, "loss": 0.4774, "step": 57128 }, { "epoch": 3.45, "learning_rate": 1.9415146439726586e-05, "loss": 0.4574, "step": 57130 }, { "epoch": 3.45, "learning_rate": 1.941363752961236e-05, "loss": 0.49, "step": 57132 }, { "epoch": 3.45, "learning_rate": 1.941212861949814e-05, "loss": 0.3665, "step": 57134 }, { "epoch": 3.45, "learning_rate": 1.9410619709383912e-05, "loss": 0.3523, "step": 57136 }, { "epoch": 3.45, "learning_rate": 1.940911079926969e-05, "loss": 0.3358, "step": 57138 }, { "epoch": 3.45, "learning_rate": 1.9407601889155462e-05, "loss": 0.3241, "step": 57140 }, { "epoch": 3.45, "learning_rate": 1.940609297904124e-05, "loss": 0.3911, "step": 57142 }, { "epoch": 3.45, "learning_rate": 1.9404584068927015e-05, "loss": 0.3417, "step": 57144 }, { "epoch": 3.45, "learning_rate": 1.940307515881279e-05, "loss": 0.4241, "step": 57146 }, { "epoch": 3.45, "learning_rate": 1.9401566248698564e-05, "loss": 0.391, "step": 57148 }, { "epoch": 3.45, "learning_rate": 1.940005733858434e-05, "loss": 0.3416, "step": 57150 }, { "epoch": 3.45, "learning_rate": 1.9398548428470117e-05, "loss": 0.2931, "step": 57152 }, { "epoch": 3.45, "learning_rate": 1.939703951835589e-05, "loss": 0.3319, "step": 57154 }, { "epoch": 3.45, "learning_rate": 1.9395530608241667e-05, "loss": 0.3735, "step": 57156 }, { "epoch": 3.45, "learning_rate": 1.9394021698127444e-05, "loss": 0.3416, "step": 57158 }, { "epoch": 3.45, "learning_rate": 1.939251278801322e-05, "loss": 0.3148, "step": 57160 }, { "epoch": 3.45, "learning_rate": 1.9391003877898993e-05, "loss": 0.42, "step": 57162 }, { "epoch": 3.45, "learning_rate": 1.938949496778477e-05, "loss": 0.375, "step": 57164 }, { "epoch": 3.45, "learning_rate": 1.9387986057670546e-05, "loss": 0.3258, "step": 57166 }, { "epoch": 3.45, "learning_rate": 1.938647714755632e-05, "loss": 0.3063, "step": 57168 }, { "epoch": 3.45, "learning_rate": 1.9384968237442096e-05, "loss": 0.5994, "step": 57170 }, { "epoch": 3.45, "learning_rate": 1.9383459327327872e-05, "loss": 0.6877, "step": 57172 }, { "epoch": 3.45, "learning_rate": 1.938195041721365e-05, "loss": 0.4348, "step": 57174 }, { "epoch": 3.45, "learning_rate": 1.9380441507099422e-05, "loss": 0.3275, "step": 57176 }, { "epoch": 3.45, "learning_rate": 1.93789325969852e-05, "loss": 0.3457, "step": 57178 }, { "epoch": 3.45, "learning_rate": 1.9377423686870975e-05, "loss": 0.1787, "step": 57180 }, { "epoch": 3.45, "learning_rate": 1.9375914776756748e-05, "loss": 0.4988, "step": 57182 }, { "epoch": 3.45, "learning_rate": 1.9374405866642525e-05, "loss": 0.5363, "step": 57184 }, { "epoch": 3.45, "learning_rate": 1.93728969565283e-05, "loss": 0.3821, "step": 57186 }, { "epoch": 3.45, "learning_rate": 1.9371388046414078e-05, "loss": 0.4215, "step": 57188 }, { "epoch": 3.45, "learning_rate": 1.936987913629985e-05, "loss": 0.2908, "step": 57190 }, { "epoch": 3.45, "learning_rate": 1.9368370226185627e-05, "loss": 0.3904, "step": 57192 }, { "epoch": 3.45, "learning_rate": 1.9366861316071404e-05, "loss": 0.4618, "step": 57194 }, { "epoch": 3.45, "learning_rate": 1.9365352405957177e-05, "loss": 0.4612, "step": 57196 }, { "epoch": 3.45, "learning_rate": 1.9363843495842954e-05, "loss": 0.2717, "step": 57198 }, { "epoch": 3.45, "learning_rate": 1.9362334585728727e-05, "loss": 0.3595, "step": 57200 }, { "epoch": 3.45, "learning_rate": 1.9360825675614507e-05, "loss": 0.3464, "step": 57202 }, { "epoch": 3.45, "learning_rate": 1.935931676550028e-05, "loss": 0.3228, "step": 57204 }, { "epoch": 3.45, "learning_rate": 1.9357807855386056e-05, "loss": 0.3171, "step": 57206 }, { "epoch": 3.45, "learning_rate": 1.935629894527183e-05, "loss": 0.3092, "step": 57208 }, { "epoch": 3.45, "learning_rate": 1.935479003515761e-05, "loss": 0.3574, "step": 57210 }, { "epoch": 3.45, "learning_rate": 1.9353281125043382e-05, "loss": 0.4451, "step": 57212 }, { "epoch": 3.45, "learning_rate": 1.9351772214929155e-05, "loss": 0.2325, "step": 57214 }, { "epoch": 3.45, "learning_rate": 1.9350263304814932e-05, "loss": 0.3224, "step": 57216 }, { "epoch": 3.45, "learning_rate": 1.934875439470071e-05, "loss": 0.4106, "step": 57218 }, { "epoch": 3.45, "learning_rate": 1.9347245484586485e-05, "loss": 0.3001, "step": 57220 }, { "epoch": 3.45, "learning_rate": 1.9345736574472258e-05, "loss": 0.5796, "step": 57222 }, { "epoch": 3.45, "learning_rate": 1.9344227664358035e-05, "loss": 0.4839, "step": 57224 }, { "epoch": 3.45, "learning_rate": 1.934271875424381e-05, "loss": 0.3491, "step": 57226 }, { "epoch": 3.45, "learning_rate": 1.9341209844129588e-05, "loss": 0.3628, "step": 57228 }, { "epoch": 3.45, "learning_rate": 1.933970093401536e-05, "loss": 0.3711, "step": 57230 }, { "epoch": 3.45, "learning_rate": 1.9338192023901137e-05, "loss": 0.5337, "step": 57232 }, { "epoch": 3.45, "learning_rate": 1.9336683113786914e-05, "loss": 0.4015, "step": 57234 }, { "epoch": 3.45, "learning_rate": 1.9335174203672687e-05, "loss": 0.3511, "step": 57236 }, { "epoch": 3.45, "learning_rate": 1.9333665293558463e-05, "loss": 0.3106, "step": 57238 }, { "epoch": 3.45, "learning_rate": 1.933215638344424e-05, "loss": 0.509, "step": 57240 }, { "epoch": 3.45, "learning_rate": 1.9330647473330017e-05, "loss": 0.3971, "step": 57242 }, { "epoch": 3.46, "learning_rate": 1.932913856321579e-05, "loss": 0.3122, "step": 57244 }, { "epoch": 3.46, "learning_rate": 1.9327629653101563e-05, "loss": 0.2709, "step": 57246 }, { "epoch": 3.46, "learning_rate": 1.9326120742987343e-05, "loss": 0.469, "step": 57248 }, { "epoch": 3.46, "learning_rate": 1.9324611832873116e-05, "loss": 0.2883, "step": 57250 }, { "epoch": 3.46, "learning_rate": 1.9323102922758892e-05, "loss": 0.5247, "step": 57252 }, { "epoch": 3.46, "learning_rate": 1.9321594012644665e-05, "loss": 0.3801, "step": 57254 }, { "epoch": 3.46, "learning_rate": 1.9320085102530445e-05, "loss": 0.3568, "step": 57256 }, { "epoch": 3.46, "learning_rate": 1.931857619241622e-05, "loss": 0.4327, "step": 57258 }, { "epoch": 3.46, "learning_rate": 1.9317067282301995e-05, "loss": 0.3815, "step": 57260 }, { "epoch": 3.46, "learning_rate": 1.9315558372187768e-05, "loss": 0.3596, "step": 57262 }, { "epoch": 3.46, "learning_rate": 1.9314049462073545e-05, "loss": 0.394, "step": 57264 }, { "epoch": 3.46, "learning_rate": 1.931254055195932e-05, "loss": 0.4733, "step": 57266 }, { "epoch": 3.46, "learning_rate": 1.9311031641845094e-05, "loss": 0.5112, "step": 57268 }, { "epoch": 3.46, "learning_rate": 1.930952273173087e-05, "loss": 0.349, "step": 57270 }, { "epoch": 3.46, "learning_rate": 1.9308013821616647e-05, "loss": 0.3564, "step": 57272 }, { "epoch": 3.46, "learning_rate": 1.9306504911502424e-05, "loss": 0.5111, "step": 57274 }, { "epoch": 3.46, "learning_rate": 1.9304996001388197e-05, "loss": 0.4081, "step": 57276 }, { "epoch": 3.46, "learning_rate": 1.9303487091273973e-05, "loss": 0.7055, "step": 57278 }, { "epoch": 3.46, "learning_rate": 1.930197818115975e-05, "loss": 0.2452, "step": 57280 }, { "epoch": 3.46, "learning_rate": 1.9300469271045523e-05, "loss": 0.3907, "step": 57282 }, { "epoch": 3.46, "learning_rate": 1.92989603609313e-05, "loss": 0.3867, "step": 57284 }, { "epoch": 3.46, "learning_rate": 1.9297451450817076e-05, "loss": 0.3564, "step": 57286 }, { "epoch": 3.46, "learning_rate": 1.9295942540702853e-05, "loss": 0.2984, "step": 57288 }, { "epoch": 3.46, "learning_rate": 1.9294433630588626e-05, "loss": 0.2419, "step": 57290 }, { "epoch": 3.46, "learning_rate": 1.9292924720474402e-05, "loss": 0.3503, "step": 57292 }, { "epoch": 3.46, "learning_rate": 1.929141581036018e-05, "loss": 0.4685, "step": 57294 }, { "epoch": 3.46, "learning_rate": 1.9289906900245952e-05, "loss": 0.4713, "step": 57296 }, { "epoch": 3.46, "learning_rate": 1.928839799013173e-05, "loss": 0.4302, "step": 57298 }, { "epoch": 3.46, "learning_rate": 1.9286889080017505e-05, "loss": 0.3312, "step": 57300 }, { "epoch": 3.46, "learning_rate": 1.928538016990328e-05, "loss": 0.5414, "step": 57302 }, { "epoch": 3.46, "learning_rate": 1.9283871259789055e-05, "loss": 0.3384, "step": 57304 }, { "epoch": 3.46, "learning_rate": 1.928236234967483e-05, "loss": 0.3748, "step": 57306 }, { "epoch": 3.46, "learning_rate": 1.9280853439560608e-05, "loss": 0.4048, "step": 57308 }, { "epoch": 3.46, "learning_rate": 1.9279344529446384e-05, "loss": 0.3032, "step": 57310 }, { "epoch": 3.46, "learning_rate": 1.9277835619332157e-05, "loss": 0.6066, "step": 57312 }, { "epoch": 3.46, "learning_rate": 1.927632670921793e-05, "loss": 0.3165, "step": 57314 }, { "epoch": 3.46, "learning_rate": 1.927481779910371e-05, "loss": 0.4065, "step": 57316 }, { "epoch": 3.46, "learning_rate": 1.9273308888989483e-05, "loss": 0.4294, "step": 57318 }, { "epoch": 3.46, "learning_rate": 1.927179997887526e-05, "loss": 0.4507, "step": 57320 }, { "epoch": 3.46, "learning_rate": 1.9270291068761033e-05, "loss": 0.3861, "step": 57322 }, { "epoch": 3.46, "learning_rate": 1.9268782158646813e-05, "loss": 0.2946, "step": 57324 }, { "epoch": 3.46, "learning_rate": 1.9267273248532586e-05, "loss": 0.2561, "step": 57326 }, { "epoch": 3.46, "learning_rate": 1.926576433841836e-05, "loss": 0.3267, "step": 57328 }, { "epoch": 3.46, "learning_rate": 1.9264255428304136e-05, "loss": 0.2333, "step": 57330 }, { "epoch": 3.46, "learning_rate": 1.9262746518189912e-05, "loss": 0.4648, "step": 57332 }, { "epoch": 3.46, "learning_rate": 1.926123760807569e-05, "loss": 0.3833, "step": 57334 }, { "epoch": 3.46, "learning_rate": 1.9259728697961462e-05, "loss": 0.518, "step": 57336 }, { "epoch": 3.46, "learning_rate": 1.925821978784724e-05, "loss": 0.2504, "step": 57338 }, { "epoch": 3.46, "learning_rate": 1.9256710877733015e-05, "loss": 0.4608, "step": 57340 }, { "epoch": 3.46, "learning_rate": 1.925520196761879e-05, "loss": 0.3868, "step": 57342 }, { "epoch": 3.46, "learning_rate": 1.9253693057504565e-05, "loss": 0.2916, "step": 57344 }, { "epoch": 3.46, "learning_rate": 1.925218414739034e-05, "loss": 0.2787, "step": 57346 }, { "epoch": 3.46, "learning_rate": 1.9250675237276118e-05, "loss": 0.3554, "step": 57348 }, { "epoch": 3.46, "learning_rate": 1.924916632716189e-05, "loss": 0.3088, "step": 57350 }, { "epoch": 3.46, "learning_rate": 1.9247657417047667e-05, "loss": 0.2534, "step": 57352 }, { "epoch": 3.46, "learning_rate": 1.9246148506933444e-05, "loss": 0.2361, "step": 57354 }, { "epoch": 3.46, "learning_rate": 1.924463959681922e-05, "loss": 0.2997, "step": 57356 }, { "epoch": 3.46, "learning_rate": 1.9243130686704993e-05, "loss": 0.3541, "step": 57358 }, { "epoch": 3.46, "learning_rate": 1.9241621776590766e-05, "loss": 0.36, "step": 57360 }, { "epoch": 3.46, "learning_rate": 1.9240112866476546e-05, "loss": 0.3245, "step": 57362 }, { "epoch": 3.46, "learning_rate": 1.923860395636232e-05, "loss": 0.3386, "step": 57364 }, { "epoch": 3.46, "learning_rate": 1.9237095046248096e-05, "loss": 0.332, "step": 57366 }, { "epoch": 3.46, "learning_rate": 1.923558613613387e-05, "loss": 0.3544, "step": 57368 }, { "epoch": 3.46, "learning_rate": 1.923407722601965e-05, "loss": 0.2651, "step": 57370 }, { "epoch": 3.46, "learning_rate": 1.9232568315905422e-05, "loss": 0.2255, "step": 57372 }, { "epoch": 3.46, "learning_rate": 1.92310594057912e-05, "loss": 0.4006, "step": 57374 }, { "epoch": 3.46, "learning_rate": 1.9229550495676972e-05, "loss": 0.3358, "step": 57376 }, { "epoch": 3.46, "learning_rate": 1.9228041585562748e-05, "loss": 0.3946, "step": 57378 }, { "epoch": 3.46, "learning_rate": 1.9226532675448525e-05, "loss": 0.3476, "step": 57380 }, { "epoch": 3.46, "learning_rate": 1.9225023765334298e-05, "loss": 0.2952, "step": 57382 }, { "epoch": 3.46, "learning_rate": 1.9223514855220074e-05, "loss": 0.282, "step": 57384 }, { "epoch": 3.46, "learning_rate": 1.922200594510585e-05, "loss": 0.3148, "step": 57386 }, { "epoch": 3.46, "learning_rate": 1.9220497034991627e-05, "loss": 0.4708, "step": 57388 }, { "epoch": 3.46, "learning_rate": 1.92189881248774e-05, "loss": 0.2911, "step": 57390 }, { "epoch": 3.46, "learning_rate": 1.9217479214763177e-05, "loss": 0.4666, "step": 57392 }, { "epoch": 3.46, "learning_rate": 1.9215970304648954e-05, "loss": 0.376, "step": 57394 }, { "epoch": 3.46, "learning_rate": 1.9214461394534727e-05, "loss": 0.3913, "step": 57396 }, { "epoch": 3.46, "learning_rate": 1.9212952484420503e-05, "loss": 0.3065, "step": 57398 }, { "epoch": 3.46, "learning_rate": 1.921144357430628e-05, "loss": 0.3918, "step": 57400 }, { "epoch": 3.46, "learning_rate": 1.9209934664192056e-05, "loss": 0.3699, "step": 57402 }, { "epoch": 3.46, "learning_rate": 1.920842575407783e-05, "loss": 0.3951, "step": 57404 }, { "epoch": 3.46, "learning_rate": 1.9206916843963606e-05, "loss": 0.3557, "step": 57406 }, { "epoch": 3.46, "learning_rate": 1.9205407933849382e-05, "loss": 0.1951, "step": 57408 }, { "epoch": 3.47, "learning_rate": 1.9203899023735156e-05, "loss": 0.3423, "step": 57410 }, { "epoch": 3.47, "learning_rate": 1.9202390113620932e-05, "loss": 0.3864, "step": 57412 }, { "epoch": 3.47, "learning_rate": 1.920088120350671e-05, "loss": 0.5913, "step": 57414 }, { "epoch": 3.47, "learning_rate": 1.9199372293392485e-05, "loss": 0.2822, "step": 57416 }, { "epoch": 3.47, "learning_rate": 1.9197863383278258e-05, "loss": 0.4325, "step": 57418 }, { "epoch": 3.47, "learning_rate": 1.9196354473164035e-05, "loss": 0.419, "step": 57420 }, { "epoch": 3.47, "learning_rate": 1.919484556304981e-05, "loss": 0.4937, "step": 57422 }, { "epoch": 3.47, "learning_rate": 1.9193336652935588e-05, "loss": 0.4058, "step": 57424 }, { "epoch": 3.47, "learning_rate": 1.919182774282136e-05, "loss": 0.355, "step": 57426 }, { "epoch": 3.47, "learning_rate": 1.9190318832707134e-05, "loss": 0.3468, "step": 57428 }, { "epoch": 3.47, "learning_rate": 1.9188809922592914e-05, "loss": 0.5448, "step": 57430 }, { "epoch": 3.47, "learning_rate": 1.9187301012478687e-05, "loss": 0.2749, "step": 57432 }, { "epoch": 3.47, "learning_rate": 1.9185792102364464e-05, "loss": 0.4677, "step": 57434 }, { "epoch": 3.47, "learning_rate": 1.9184283192250237e-05, "loss": 0.4368, "step": 57436 }, { "epoch": 3.47, "learning_rate": 1.9182774282136017e-05, "loss": 0.3681, "step": 57438 }, { "epoch": 3.47, "learning_rate": 1.918126537202179e-05, "loss": 0.3273, "step": 57440 }, { "epoch": 3.47, "learning_rate": 1.9179756461907563e-05, "loss": 0.4496, "step": 57442 }, { "epoch": 3.47, "learning_rate": 1.917824755179334e-05, "loss": 0.2407, "step": 57444 }, { "epoch": 3.47, "learning_rate": 1.9176738641679116e-05, "loss": 0.3964, "step": 57446 }, { "epoch": 3.47, "learning_rate": 1.9175229731564892e-05, "loss": 0.4009, "step": 57448 }, { "epoch": 3.47, "learning_rate": 1.9173720821450666e-05, "loss": 0.3852, "step": 57450 }, { "epoch": 3.47, "learning_rate": 1.9172211911336442e-05, "loss": 0.3463, "step": 57452 }, { "epoch": 3.47, "learning_rate": 1.917070300122222e-05, "loss": 0.4189, "step": 57454 }, { "epoch": 3.47, "learning_rate": 1.9169194091107995e-05, "loss": 0.3445, "step": 57456 }, { "epoch": 3.47, "learning_rate": 1.9167685180993768e-05, "loss": 0.5047, "step": 57458 }, { "epoch": 3.47, "learning_rate": 1.9166176270879545e-05, "loss": 0.3357, "step": 57460 }, { "epoch": 3.47, "learning_rate": 1.916466736076532e-05, "loss": 0.255, "step": 57462 }, { "epoch": 3.47, "learning_rate": 1.9163158450651094e-05, "loss": 0.2953, "step": 57464 }, { "epoch": 3.47, "learning_rate": 1.916164954053687e-05, "loss": 0.2647, "step": 57466 }, { "epoch": 3.47, "learning_rate": 1.9160140630422647e-05, "loss": 0.3523, "step": 57468 }, { "epoch": 3.47, "learning_rate": 1.9158631720308424e-05, "loss": 0.3062, "step": 57470 }, { "epoch": 3.47, "learning_rate": 1.9157122810194197e-05, "loss": 0.2223, "step": 57472 }, { "epoch": 3.47, "learning_rate": 1.9155613900079974e-05, "loss": 0.2346, "step": 57474 }, { "epoch": 3.47, "learning_rate": 1.915410498996575e-05, "loss": 0.3202, "step": 57476 }, { "epoch": 3.47, "learning_rate": 1.9152596079851523e-05, "loss": 0.3131, "step": 57478 }, { "epoch": 3.47, "learning_rate": 1.91510871697373e-05, "loss": 0.45, "step": 57480 }, { "epoch": 3.47, "learning_rate": 1.9149578259623073e-05, "loss": 0.3498, "step": 57482 }, { "epoch": 3.47, "learning_rate": 1.9148069349508853e-05, "loss": 0.4319, "step": 57484 }, { "epoch": 3.47, "learning_rate": 1.9146560439394626e-05, "loss": 0.3543, "step": 57486 }, { "epoch": 3.47, "learning_rate": 1.9145051529280402e-05, "loss": 0.3577, "step": 57488 }, { "epoch": 3.47, "learning_rate": 1.9143542619166175e-05, "loss": 0.3973, "step": 57490 }, { "epoch": 3.47, "learning_rate": 1.9142033709051952e-05, "loss": 0.3517, "step": 57492 }, { "epoch": 3.47, "learning_rate": 1.914052479893773e-05, "loss": 0.2843, "step": 57494 }, { "epoch": 3.47, "learning_rate": 1.91390158888235e-05, "loss": 0.3468, "step": 57496 }, { "epoch": 3.47, "learning_rate": 1.9137506978709278e-05, "loss": 0.5283, "step": 57498 }, { "epoch": 3.47, "learning_rate": 1.9135998068595055e-05, "loss": 0.4015, "step": 57500 }, { "epoch": 3.47, "learning_rate": 1.913448915848083e-05, "loss": 0.4233, "step": 57502 }, { "epoch": 3.47, "learning_rate": 1.9132980248366604e-05, "loss": 0.3912, "step": 57504 }, { "epoch": 3.47, "learning_rate": 1.913147133825238e-05, "loss": 0.3038, "step": 57506 }, { "epoch": 3.47, "learning_rate": 1.9129962428138157e-05, "loss": 0.2823, "step": 57508 }, { "epoch": 3.47, "learning_rate": 1.912845351802393e-05, "loss": 0.4462, "step": 57510 }, { "epoch": 3.47, "learning_rate": 1.9126944607909707e-05, "loss": 0.3844, "step": 57512 }, { "epoch": 3.47, "learning_rate": 1.9125435697795483e-05, "loss": 0.3958, "step": 57514 }, { "epoch": 3.47, "learning_rate": 1.912392678768126e-05, "loss": 0.3516, "step": 57516 }, { "epoch": 3.47, "learning_rate": 1.9122417877567033e-05, "loss": 0.2683, "step": 57518 }, { "epoch": 3.47, "learning_rate": 1.912090896745281e-05, "loss": 0.4202, "step": 57520 }, { "epoch": 3.47, "learning_rate": 1.9119400057338586e-05, "loss": 0.3483, "step": 57522 }, { "epoch": 3.47, "learning_rate": 1.9117891147224363e-05, "loss": 0.5584, "step": 57524 }, { "epoch": 3.47, "learning_rate": 1.9116382237110136e-05, "loss": 0.2849, "step": 57526 }, { "epoch": 3.47, "learning_rate": 1.9114873326995912e-05, "loss": 0.3978, "step": 57528 }, { "epoch": 3.47, "learning_rate": 1.911336441688169e-05, "loss": 0.4001, "step": 57530 }, { "epoch": 3.47, "learning_rate": 1.9111855506767462e-05, "loss": 0.3231, "step": 57532 }, { "epoch": 3.47, "learning_rate": 1.911034659665324e-05, "loss": 0.2772, "step": 57534 }, { "epoch": 3.47, "learning_rate": 1.9108837686539015e-05, "loss": 0.3057, "step": 57536 }, { "epoch": 3.47, "learning_rate": 1.910732877642479e-05, "loss": 0.3739, "step": 57538 }, { "epoch": 3.47, "learning_rate": 1.9105819866310565e-05, "loss": 0.4373, "step": 57540 }, { "epoch": 3.47, "learning_rate": 1.9104310956196338e-05, "loss": 0.3101, "step": 57542 }, { "epoch": 3.47, "learning_rate": 1.9102802046082118e-05, "loss": 0.3155, "step": 57544 }, { "epoch": 3.47, "learning_rate": 1.910129313596789e-05, "loss": 0.4866, "step": 57546 }, { "epoch": 3.47, "learning_rate": 1.9099784225853667e-05, "loss": 0.3021, "step": 57548 }, { "epoch": 3.47, "learning_rate": 1.909827531573944e-05, "loss": 0.4157, "step": 57550 }, { "epoch": 3.47, "learning_rate": 1.909676640562522e-05, "loss": 0.268, "step": 57552 }, { "epoch": 3.47, "learning_rate": 1.9095257495510993e-05, "loss": 0.388, "step": 57554 }, { "epoch": 3.47, "learning_rate": 1.909374858539677e-05, "loss": 0.3844, "step": 57556 }, { "epoch": 3.47, "learning_rate": 1.9092239675282543e-05, "loss": 0.3538, "step": 57558 }, { "epoch": 3.47, "learning_rate": 1.909073076516832e-05, "loss": 0.3492, "step": 57560 }, { "epoch": 3.47, "learning_rate": 1.9089221855054096e-05, "loss": 0.5868, "step": 57562 }, { "epoch": 3.47, "learning_rate": 1.908771294493987e-05, "loss": 0.261, "step": 57564 }, { "epoch": 3.47, "learning_rate": 1.9086204034825646e-05, "loss": 0.3047, "step": 57566 }, { "epoch": 3.47, "learning_rate": 1.9084695124711422e-05, "loss": 0.3889, "step": 57568 }, { "epoch": 3.47, "learning_rate": 1.90831862145972e-05, "loss": 0.356, "step": 57570 }, { "epoch": 3.47, "learning_rate": 1.9081677304482972e-05, "loss": 0.2905, "step": 57572 }, { "epoch": 3.48, "learning_rate": 1.908016839436875e-05, "loss": 0.2939, "step": 57574 }, { "epoch": 3.48, "learning_rate": 1.9078659484254525e-05, "loss": 0.5239, "step": 57576 }, { "epoch": 3.48, "learning_rate": 1.9077150574140298e-05, "loss": 0.5181, "step": 57578 }, { "epoch": 3.48, "learning_rate": 1.9075641664026075e-05, "loss": 0.3265, "step": 57580 }, { "epoch": 3.48, "learning_rate": 1.907413275391185e-05, "loss": 0.5499, "step": 57582 }, { "epoch": 3.48, "learning_rate": 1.9072623843797628e-05, "loss": 0.2764, "step": 57584 }, { "epoch": 3.48, "learning_rate": 1.90711149336834e-05, "loss": 0.4292, "step": 57586 }, { "epoch": 3.48, "learning_rate": 1.9069606023569177e-05, "loss": 0.5221, "step": 57588 }, { "epoch": 3.48, "learning_rate": 1.9068097113454954e-05, "loss": 0.2473, "step": 57590 }, { "epoch": 3.48, "learning_rate": 1.9066588203340727e-05, "loss": 0.2727, "step": 57592 }, { "epoch": 3.48, "learning_rate": 1.9065079293226503e-05, "loss": 0.469, "step": 57594 }, { "epoch": 3.48, "learning_rate": 1.9063570383112276e-05, "loss": 0.4365, "step": 57596 }, { "epoch": 3.48, "learning_rate": 1.9062061472998056e-05, "loss": 0.2411, "step": 57598 }, { "epoch": 3.48, "learning_rate": 1.906055256288383e-05, "loss": 0.3498, "step": 57600 }, { "epoch": 3.48, "learning_rate": 1.9059043652769606e-05, "loss": 0.4526, "step": 57602 }, { "epoch": 3.48, "learning_rate": 1.905753474265538e-05, "loss": 0.4843, "step": 57604 }, { "epoch": 3.48, "learning_rate": 1.905602583254116e-05, "loss": 0.573, "step": 57606 }, { "epoch": 3.48, "learning_rate": 1.9054516922426932e-05, "loss": 0.4037, "step": 57608 }, { "epoch": 3.48, "learning_rate": 1.9053008012312705e-05, "loss": 0.4207, "step": 57610 }, { "epoch": 3.48, "learning_rate": 1.9051499102198482e-05, "loss": 0.4026, "step": 57612 }, { "epoch": 3.48, "learning_rate": 1.904999019208426e-05, "loss": 0.3102, "step": 57614 }, { "epoch": 3.48, "learning_rate": 1.9048481281970035e-05, "loss": 0.4232, "step": 57616 }, { "epoch": 3.48, "learning_rate": 1.9046972371855808e-05, "loss": 0.4535, "step": 57618 }, { "epoch": 3.48, "learning_rate": 1.9045463461741584e-05, "loss": 0.4787, "step": 57620 }, { "epoch": 3.48, "learning_rate": 1.904395455162736e-05, "loss": 0.3348, "step": 57622 }, { "epoch": 3.48, "learning_rate": 1.9042445641513134e-05, "loss": 0.5728, "step": 57624 }, { "epoch": 3.48, "learning_rate": 1.904093673139891e-05, "loss": 0.5194, "step": 57626 }, { "epoch": 3.48, "learning_rate": 1.9039427821284687e-05, "loss": 0.3543, "step": 57628 }, { "epoch": 3.48, "learning_rate": 1.9037918911170464e-05, "loss": 0.4275, "step": 57630 }, { "epoch": 3.48, "learning_rate": 1.9036410001056237e-05, "loss": 0.339, "step": 57632 }, { "epoch": 3.48, "learning_rate": 1.9034901090942013e-05, "loss": 0.3207, "step": 57634 }, { "epoch": 3.48, "learning_rate": 1.903339218082779e-05, "loss": 0.4195, "step": 57636 }, { "epoch": 3.48, "learning_rate": 1.9031883270713566e-05, "loss": 0.3082, "step": 57638 }, { "epoch": 3.48, "learning_rate": 1.903037436059934e-05, "loss": 0.2845, "step": 57640 }, { "epoch": 3.48, "learning_rate": 1.9028865450485116e-05, "loss": 0.3191, "step": 57642 }, { "epoch": 3.48, "learning_rate": 1.9027356540370892e-05, "loss": 0.3906, "step": 57644 }, { "epoch": 3.48, "learning_rate": 1.9025847630256666e-05, "loss": 0.2984, "step": 57646 }, { "epoch": 3.48, "learning_rate": 1.9024338720142442e-05, "loss": 0.3104, "step": 57648 }, { "epoch": 3.48, "learning_rate": 1.902282981002822e-05, "loss": 0.3937, "step": 57650 }, { "epoch": 3.48, "learning_rate": 1.9021320899913995e-05, "loss": 0.4784, "step": 57652 }, { "epoch": 3.48, "learning_rate": 1.9019811989799768e-05, "loss": 0.3058, "step": 57654 }, { "epoch": 3.48, "learning_rate": 1.901830307968554e-05, "loss": 0.2634, "step": 57656 }, { "epoch": 3.48, "learning_rate": 1.901679416957132e-05, "loss": 0.2547, "step": 57658 }, { "epoch": 3.48, "learning_rate": 1.9015285259457094e-05, "loss": 0.3645, "step": 57660 }, { "epoch": 3.48, "learning_rate": 1.901377634934287e-05, "loss": 0.365, "step": 57662 }, { "epoch": 3.48, "learning_rate": 1.9012267439228644e-05, "loss": 0.3568, "step": 57664 }, { "epoch": 3.48, "learning_rate": 1.9010758529114424e-05, "loss": 0.2629, "step": 57666 }, { "epoch": 3.48, "learning_rate": 1.9009249619000197e-05, "loss": 0.4403, "step": 57668 }, { "epoch": 3.48, "learning_rate": 1.9007740708885974e-05, "loss": 0.4533, "step": 57670 }, { "epoch": 3.48, "learning_rate": 1.9006231798771747e-05, "loss": 0.3618, "step": 57672 }, { "epoch": 3.48, "learning_rate": 1.9004722888657523e-05, "loss": 0.2692, "step": 57674 }, { "epoch": 3.48, "learning_rate": 1.90032139785433e-05, "loss": 0.36, "step": 57676 }, { "epoch": 3.48, "learning_rate": 1.9001705068429073e-05, "loss": 0.2887, "step": 57678 }, { "epoch": 3.48, "learning_rate": 1.900019615831485e-05, "loss": 0.353, "step": 57680 }, { "epoch": 3.48, "learning_rate": 1.8998687248200626e-05, "loss": 0.3692, "step": 57682 }, { "epoch": 3.48, "learning_rate": 1.8997178338086402e-05, "loss": 0.3694, "step": 57684 }, { "epoch": 3.48, "learning_rate": 1.8995669427972176e-05, "loss": 0.3188, "step": 57686 }, { "epoch": 3.48, "learning_rate": 1.8994160517857952e-05, "loss": 0.4501, "step": 57688 }, { "epoch": 3.48, "learning_rate": 1.899265160774373e-05, "loss": 0.4242, "step": 57690 }, { "epoch": 3.48, "learning_rate": 1.89911426976295e-05, "loss": 0.4722, "step": 57692 }, { "epoch": 3.48, "learning_rate": 1.8989633787515278e-05, "loss": 0.3865, "step": 57694 }, { "epoch": 3.48, "learning_rate": 1.8988124877401055e-05, "loss": 0.2978, "step": 57696 }, { "epoch": 3.48, "learning_rate": 1.898661596728683e-05, "loss": 0.3584, "step": 57698 }, { "epoch": 3.48, "learning_rate": 1.8985107057172604e-05, "loss": 0.2678, "step": 57700 }, { "epoch": 3.48, "learning_rate": 1.898359814705838e-05, "loss": 0.3327, "step": 57702 }, { "epoch": 3.48, "learning_rate": 1.8982089236944157e-05, "loss": 0.2518, "step": 57704 }, { "epoch": 3.48, "learning_rate": 1.898058032682993e-05, "loss": 0.4297, "step": 57706 }, { "epoch": 3.48, "learning_rate": 1.8979071416715707e-05, "loss": 0.3093, "step": 57708 }, { "epoch": 3.48, "learning_rate": 1.897756250660148e-05, "loss": 0.4199, "step": 57710 }, { "epoch": 3.48, "learning_rate": 1.897605359648726e-05, "loss": 0.3457, "step": 57712 }, { "epoch": 3.48, "learning_rate": 1.8974544686373033e-05, "loss": 0.3371, "step": 57714 }, { "epoch": 3.48, "learning_rate": 1.897303577625881e-05, "loss": 0.3458, "step": 57716 }, { "epoch": 3.48, "learning_rate": 1.8971526866144583e-05, "loss": 0.4516, "step": 57718 }, { "epoch": 3.48, "learning_rate": 1.8970017956030363e-05, "loss": 0.2523, "step": 57720 }, { "epoch": 3.48, "learning_rate": 1.8968509045916136e-05, "loss": 0.4661, "step": 57722 }, { "epoch": 3.48, "learning_rate": 1.896700013580191e-05, "loss": 0.2528, "step": 57724 }, { "epoch": 3.48, "learning_rate": 1.8965491225687685e-05, "loss": 0.2595, "step": 57726 }, { "epoch": 3.48, "learning_rate": 1.8963982315573462e-05, "loss": 0.2944, "step": 57728 }, { "epoch": 3.48, "learning_rate": 1.896247340545924e-05, "loss": 0.3448, "step": 57730 }, { "epoch": 3.48, "learning_rate": 1.896096449534501e-05, "loss": 0.3186, "step": 57732 }, { "epoch": 3.48, "learning_rate": 1.8959455585230788e-05, "loss": 0.3985, "step": 57734 }, { "epoch": 3.48, "learning_rate": 1.8957946675116565e-05, "loss": 0.4101, "step": 57736 }, { "epoch": 3.48, "learning_rate": 1.8956437765002338e-05, "loss": 0.4639, "step": 57738 }, { "epoch": 3.49, "learning_rate": 1.8954928854888114e-05, "loss": 0.5185, "step": 57740 }, { "epoch": 3.49, "learning_rate": 1.895341994477389e-05, "loss": 0.3745, "step": 57742 }, { "epoch": 3.49, "learning_rate": 1.8951911034659667e-05, "loss": 0.3529, "step": 57744 }, { "epoch": 3.49, "learning_rate": 1.895040212454544e-05, "loss": 0.3036, "step": 57746 }, { "epoch": 3.49, "learning_rate": 1.8948893214431217e-05, "loss": 0.353, "step": 57748 }, { "epoch": 3.49, "learning_rate": 1.8947384304316993e-05, "loss": 0.4708, "step": 57750 }, { "epoch": 3.49, "learning_rate": 1.894587539420277e-05, "loss": 0.3706, "step": 57752 }, { "epoch": 3.49, "learning_rate": 1.8944366484088543e-05, "loss": 0.5326, "step": 57754 }, { "epoch": 3.49, "learning_rate": 1.894285757397432e-05, "loss": 0.3379, "step": 57756 }, { "epoch": 3.49, "learning_rate": 1.8941348663860096e-05, "loss": 0.3085, "step": 57758 }, { "epoch": 3.49, "learning_rate": 1.893983975374587e-05, "loss": 0.3486, "step": 57760 }, { "epoch": 3.49, "learning_rate": 1.8938330843631646e-05, "loss": 0.4788, "step": 57762 }, { "epoch": 3.49, "learning_rate": 1.8936821933517422e-05, "loss": 0.1864, "step": 57764 }, { "epoch": 3.49, "learning_rate": 1.89353130234032e-05, "loss": 0.3936, "step": 57766 }, { "epoch": 3.49, "learning_rate": 1.8933804113288972e-05, "loss": 0.3255, "step": 57768 }, { "epoch": 3.49, "learning_rate": 1.893229520317475e-05, "loss": 0.3178, "step": 57770 }, { "epoch": 3.49, "learning_rate": 1.8930786293060525e-05, "loss": 0.2753, "step": 57772 }, { "epoch": 3.49, "learning_rate": 1.8929277382946298e-05, "loss": 0.4047, "step": 57774 }, { "epoch": 3.49, "learning_rate": 1.8927768472832075e-05, "loss": 0.4357, "step": 57776 }, { "epoch": 3.49, "learning_rate": 1.8926259562717848e-05, "loss": 0.515, "step": 57778 }, { "epoch": 3.49, "learning_rate": 1.8924750652603628e-05, "loss": 0.3825, "step": 57780 }, { "epoch": 3.49, "learning_rate": 1.89232417424894e-05, "loss": 0.3612, "step": 57782 }, { "epoch": 3.49, "learning_rate": 1.8921732832375177e-05, "loss": 0.5098, "step": 57784 }, { "epoch": 3.49, "learning_rate": 1.892022392226095e-05, "loss": 0.4044, "step": 57786 }, { "epoch": 3.49, "learning_rate": 1.8918715012146727e-05, "loss": 0.4806, "step": 57788 }, { "epoch": 3.49, "learning_rate": 1.8917206102032503e-05, "loss": 0.4617, "step": 57790 }, { "epoch": 3.49, "learning_rate": 1.8915697191918277e-05, "loss": 0.3749, "step": 57792 }, { "epoch": 3.49, "learning_rate": 1.8914188281804053e-05, "loss": 0.3647, "step": 57794 }, { "epoch": 3.49, "learning_rate": 1.891267937168983e-05, "loss": 0.5163, "step": 57796 }, { "epoch": 3.49, "learning_rate": 1.8911170461575606e-05, "loss": 0.413, "step": 57798 }, { "epoch": 3.49, "learning_rate": 1.890966155146138e-05, "loss": 0.2343, "step": 57800 }, { "epoch": 3.49, "learning_rate": 1.8908152641347156e-05, "loss": 0.5605, "step": 57802 }, { "epoch": 3.49, "learning_rate": 1.8906643731232932e-05, "loss": 0.3853, "step": 57804 }, { "epoch": 3.49, "learning_rate": 1.8905134821118705e-05, "loss": 0.3465, "step": 57806 }, { "epoch": 3.49, "learning_rate": 1.8903625911004482e-05, "loss": 0.5191, "step": 57808 }, { "epoch": 3.49, "learning_rate": 1.890211700089026e-05, "loss": 0.3282, "step": 57810 }, { "epoch": 3.49, "learning_rate": 1.8900608090776035e-05, "loss": 0.312, "step": 57812 }, { "epoch": 3.49, "learning_rate": 1.8899099180661808e-05, "loss": 0.3892, "step": 57814 }, { "epoch": 3.49, "learning_rate": 1.8897590270547585e-05, "loss": 0.4435, "step": 57816 }, { "epoch": 3.49, "learning_rate": 1.889608136043336e-05, "loss": 0.3127, "step": 57818 }, { "epoch": 3.49, "learning_rate": 1.8894572450319134e-05, "loss": 0.2742, "step": 57820 }, { "epoch": 3.49, "learning_rate": 1.889306354020491e-05, "loss": 0.5113, "step": 57822 }, { "epoch": 3.49, "learning_rate": 1.8891554630090684e-05, "loss": 0.2945, "step": 57824 }, { "epoch": 3.49, "learning_rate": 1.8890045719976464e-05, "loss": 0.3411, "step": 57826 }, { "epoch": 3.49, "learning_rate": 1.8888536809862237e-05, "loss": 0.4931, "step": 57828 }, { "epoch": 3.49, "learning_rate": 1.8887027899748013e-05, "loss": 0.451, "step": 57830 }, { "epoch": 3.49, "learning_rate": 1.8885518989633786e-05, "loss": 0.2098, "step": 57832 }, { "epoch": 3.49, "learning_rate": 1.8884010079519566e-05, "loss": 0.3816, "step": 57834 }, { "epoch": 3.49, "learning_rate": 1.888250116940534e-05, "loss": 0.3233, "step": 57836 }, { "epoch": 3.49, "learning_rate": 1.8880992259291113e-05, "loss": 0.4192, "step": 57838 }, { "epoch": 3.49, "learning_rate": 1.887948334917689e-05, "loss": 0.3544, "step": 57840 }, { "epoch": 3.49, "learning_rate": 1.8877974439062666e-05, "loss": 0.4959, "step": 57842 }, { "epoch": 3.49, "learning_rate": 1.8876465528948442e-05, "loss": 0.3843, "step": 57844 }, { "epoch": 3.49, "learning_rate": 1.8874956618834215e-05, "loss": 0.2836, "step": 57846 }, { "epoch": 3.49, "learning_rate": 1.8873447708719992e-05, "loss": 0.4154, "step": 57848 }, { "epoch": 3.49, "learning_rate": 1.887193879860577e-05, "loss": 0.3303, "step": 57850 }, { "epoch": 3.49, "learning_rate": 1.8870429888491545e-05, "loss": 0.3528, "step": 57852 }, { "epoch": 3.49, "learning_rate": 1.8868920978377318e-05, "loss": 0.4823, "step": 57854 }, { "epoch": 3.49, "learning_rate": 1.8867412068263094e-05, "loss": 0.3109, "step": 57856 }, { "epoch": 3.49, "learning_rate": 1.886590315814887e-05, "loss": 0.3547, "step": 57858 }, { "epoch": 3.49, "learning_rate": 1.8864394248034644e-05, "loss": 0.3964, "step": 57860 }, { "epoch": 3.49, "learning_rate": 1.886288533792042e-05, "loss": 0.4148, "step": 57862 }, { "epoch": 3.49, "learning_rate": 1.8861376427806197e-05, "loss": 0.3439, "step": 57864 }, { "epoch": 3.49, "learning_rate": 1.8859867517691974e-05, "loss": 0.2831, "step": 57866 }, { "epoch": 3.49, "learning_rate": 1.8858358607577747e-05, "loss": 0.4457, "step": 57868 }, { "epoch": 3.49, "learning_rate": 1.8856849697463523e-05, "loss": 0.4051, "step": 57870 }, { "epoch": 3.49, "learning_rate": 1.88553407873493e-05, "loss": 0.3793, "step": 57872 }, { "epoch": 3.49, "learning_rate": 1.8853831877235073e-05, "loss": 0.3653, "step": 57874 }, { "epoch": 3.49, "learning_rate": 1.885232296712085e-05, "loss": 0.3281, "step": 57876 }, { "epoch": 3.49, "learning_rate": 1.8850814057006626e-05, "loss": 0.4345, "step": 57878 }, { "epoch": 3.49, "learning_rate": 1.8849305146892402e-05, "loss": 0.3003, "step": 57880 }, { "epoch": 3.49, "learning_rate": 1.8847796236778176e-05, "loss": 0.4663, "step": 57882 }, { "epoch": 3.49, "learning_rate": 1.8846287326663952e-05, "loss": 0.4224, "step": 57884 }, { "epoch": 3.49, "learning_rate": 1.884477841654973e-05, "loss": 0.3931, "step": 57886 }, { "epoch": 3.49, "learning_rate": 1.8843269506435502e-05, "loss": 0.3697, "step": 57888 }, { "epoch": 3.49, "learning_rate": 1.8841760596321278e-05, "loss": 0.3678, "step": 57890 }, { "epoch": 3.49, "learning_rate": 1.884025168620705e-05, "loss": 0.4025, "step": 57892 }, { "epoch": 3.49, "learning_rate": 1.883874277609283e-05, "loss": 0.4708, "step": 57894 }, { "epoch": 3.49, "learning_rate": 1.8837233865978604e-05, "loss": 0.498, "step": 57896 }, { "epoch": 3.49, "learning_rate": 1.883572495586438e-05, "loss": 0.2768, "step": 57898 }, { "epoch": 3.49, "learning_rate": 1.8834216045750154e-05, "loss": 0.4504, "step": 57900 }, { "epoch": 3.49, "learning_rate": 1.883270713563593e-05, "loss": 0.487, "step": 57902 }, { "epoch": 3.49, "learning_rate": 1.8831198225521707e-05, "loss": 0.3205, "step": 57904 }, { "epoch": 3.5, "learning_rate": 1.882968931540748e-05, "loss": 0.4859, "step": 57906 }, { "epoch": 3.5, "learning_rate": 1.8828180405293257e-05, "loss": 0.3603, "step": 57908 }, { "epoch": 3.5, "learning_rate": 1.8826671495179033e-05, "loss": 0.2967, "step": 57910 }, { "epoch": 3.5, "learning_rate": 1.882516258506481e-05, "loss": 0.3883, "step": 57912 }, { "epoch": 3.5, "learning_rate": 1.8823653674950583e-05, "loss": 0.6015, "step": 57914 }, { "epoch": 3.5, "learning_rate": 1.882214476483636e-05, "loss": 0.3074, "step": 57916 }, { "epoch": 3.5, "learning_rate": 1.8820635854722136e-05, "loss": 0.3634, "step": 57918 }, { "epoch": 3.5, "learning_rate": 1.881912694460791e-05, "loss": 0.7093, "step": 57920 }, { "epoch": 3.5, "learning_rate": 1.8817618034493686e-05, "loss": 0.4757, "step": 57922 }, { "epoch": 3.5, "learning_rate": 1.8816109124379462e-05, "loss": 0.2486, "step": 57924 }, { "epoch": 3.5, "learning_rate": 1.881460021426524e-05, "loss": 0.4346, "step": 57926 }, { "epoch": 3.5, "learning_rate": 1.8813091304151012e-05, "loss": 0.5143, "step": 57928 }, { "epoch": 3.5, "learning_rate": 1.8811582394036788e-05, "loss": 0.322, "step": 57930 }, { "epoch": 3.5, "learning_rate": 1.8810073483922565e-05, "loss": 0.4135, "step": 57932 }, { "epoch": 3.5, "learning_rate": 1.880856457380834e-05, "loss": 0.3566, "step": 57934 }, { "epoch": 3.5, "learning_rate": 1.8807055663694114e-05, "loss": 0.3694, "step": 57936 }, { "epoch": 3.5, "learning_rate": 1.8805546753579888e-05, "loss": 0.2542, "step": 57938 }, { "epoch": 3.5, "learning_rate": 1.8804037843465667e-05, "loss": 0.5616, "step": 57940 }, { "epoch": 3.5, "learning_rate": 1.880252893335144e-05, "loss": 0.2733, "step": 57942 }, { "epoch": 3.5, "learning_rate": 1.8801020023237217e-05, "loss": 0.2634, "step": 57944 }, { "epoch": 3.5, "learning_rate": 1.879951111312299e-05, "loss": 0.3721, "step": 57946 }, { "epoch": 3.5, "learning_rate": 1.879800220300877e-05, "loss": 0.3837, "step": 57948 }, { "epoch": 3.5, "learning_rate": 1.8796493292894543e-05, "loss": 0.409, "step": 57950 }, { "epoch": 3.5, "learning_rate": 1.8794984382780316e-05, "loss": 0.3617, "step": 57952 }, { "epoch": 3.5, "learning_rate": 1.8793475472666093e-05, "loss": 0.4354, "step": 57954 }, { "epoch": 3.5, "learning_rate": 1.879196656255187e-05, "loss": 0.5028, "step": 57956 }, { "epoch": 3.5, "learning_rate": 1.8790457652437646e-05, "loss": 0.3587, "step": 57958 }, { "epoch": 3.5, "learning_rate": 1.878894874232342e-05, "loss": 0.3529, "step": 57960 }, { "epoch": 3.5, "learning_rate": 1.8787439832209196e-05, "loss": 0.2665, "step": 57962 }, { "epoch": 3.5, "learning_rate": 1.8785930922094972e-05, "loss": 0.3702, "step": 57964 }, { "epoch": 3.5, "learning_rate": 1.878442201198075e-05, "loss": 0.4838, "step": 57966 }, { "epoch": 3.5, "learning_rate": 1.878291310186652e-05, "loss": 0.5705, "step": 57968 }, { "epoch": 3.5, "learning_rate": 1.8781404191752298e-05, "loss": 0.3927, "step": 57970 }, { "epoch": 3.5, "learning_rate": 1.8779895281638075e-05, "loss": 0.4571, "step": 57972 }, { "epoch": 3.5, "learning_rate": 1.8778386371523848e-05, "loss": 0.3322, "step": 57974 }, { "epoch": 3.5, "learning_rate": 1.8776877461409624e-05, "loss": 0.3661, "step": 57976 }, { "epoch": 3.5, "learning_rate": 1.87753685512954e-05, "loss": 0.349, "step": 57978 }, { "epoch": 3.5, "learning_rate": 1.8773859641181177e-05, "loss": 0.4088, "step": 57980 }, { "epoch": 3.5, "learning_rate": 1.877235073106695e-05, "loss": 0.3051, "step": 57982 }, { "epoch": 3.5, "learning_rate": 1.8770841820952727e-05, "loss": 0.3237, "step": 57984 }, { "epoch": 3.5, "learning_rate": 1.8769332910838504e-05, "loss": 0.3796, "step": 57986 }, { "epoch": 3.5, "learning_rate": 1.8767824000724277e-05, "loss": 0.3636, "step": 57988 }, { "epoch": 3.5, "learning_rate": 1.8766315090610053e-05, "loss": 0.3, "step": 57990 }, { "epoch": 3.5, "learning_rate": 1.876480618049583e-05, "loss": 0.462, "step": 57992 }, { "epoch": 3.5, "learning_rate": 1.8763297270381606e-05, "loss": 0.4467, "step": 57994 }, { "epoch": 3.5, "learning_rate": 1.876178836026738e-05, "loss": 0.4505, "step": 57996 }, { "epoch": 3.5, "learning_rate": 1.8760279450153156e-05, "loss": 0.2579, "step": 57998 }, { "epoch": 3.5, "learning_rate": 1.8758770540038932e-05, "loss": 0.2511, "step": 58000 }, { "epoch": 3.5, "learning_rate": 1.8757261629924705e-05, "loss": 0.2367, "step": 58002 }, { "epoch": 3.5, "learning_rate": 1.8755752719810482e-05, "loss": 0.4741, "step": 58004 }, { "epoch": 3.5, "learning_rate": 1.8754243809696255e-05, "loss": 0.355, "step": 58006 }, { "epoch": 3.5, "learning_rate": 1.8752734899582035e-05, "loss": 0.2457, "step": 58008 }, { "epoch": 3.5, "learning_rate": 1.8751225989467808e-05, "loss": 0.3569, "step": 58010 }, { "epoch": 3.5, "learning_rate": 1.8749717079353585e-05, "loss": 0.3761, "step": 58012 }, { "epoch": 3.5, "learning_rate": 1.8748208169239358e-05, "loss": 0.3304, "step": 58014 }, { "epoch": 3.5, "learning_rate": 1.8746699259125138e-05, "loss": 0.3387, "step": 58016 }, { "epoch": 3.5, "learning_rate": 1.874519034901091e-05, "loss": 0.4722, "step": 58018 }, { "epoch": 3.5, "learning_rate": 1.8743681438896684e-05, "loss": 0.3036, "step": 58020 }, { "epoch": 3.5, "learning_rate": 1.874217252878246e-05, "loss": 0.4608, "step": 58022 }, { "epoch": 3.5, "learning_rate": 1.8740663618668237e-05, "loss": 0.4002, "step": 58024 }, { "epoch": 3.5, "learning_rate": 1.8739154708554013e-05, "loss": 0.352, "step": 58026 }, { "epoch": 3.5, "learning_rate": 1.8737645798439787e-05, "loss": 0.4325, "step": 58028 }, { "epoch": 3.5, "learning_rate": 1.8736136888325563e-05, "loss": 0.3716, "step": 58030 }, { "epoch": 3.5, "learning_rate": 1.873462797821134e-05, "loss": 0.3598, "step": 58032 }, { "epoch": 3.5, "learning_rate": 1.8733119068097113e-05, "loss": 0.3622, "step": 58034 }, { "epoch": 3.5, "learning_rate": 1.873161015798289e-05, "loss": 0.329, "step": 58036 }, { "epoch": 3.5, "learning_rate": 1.8730101247868666e-05, "loss": 0.5314, "step": 58038 }, { "epoch": 3.5, "learning_rate": 1.8728592337754442e-05, "loss": 0.4217, "step": 58040 }, { "epoch": 3.5, "learning_rate": 1.8727083427640215e-05, "loss": 0.2902, "step": 58042 }, { "epoch": 3.5, "learning_rate": 1.8725574517525992e-05, "loss": 0.5471, "step": 58044 }, { "epoch": 3.5, "learning_rate": 1.872406560741177e-05, "loss": 0.3937, "step": 58046 }, { "epoch": 3.5, "learning_rate": 1.8722556697297545e-05, "loss": 0.4714, "step": 58048 }, { "epoch": 3.5, "learning_rate": 1.8721047787183318e-05, "loss": 0.4623, "step": 58050 }, { "epoch": 3.5, "learning_rate": 1.871953887706909e-05, "loss": 0.3722, "step": 58052 }, { "epoch": 3.5, "learning_rate": 1.871802996695487e-05, "loss": 0.3209, "step": 58054 }, { "epoch": 3.5, "learning_rate": 1.8716521056840644e-05, "loss": 0.2467, "step": 58056 }, { "epoch": 3.5, "learning_rate": 1.871501214672642e-05, "loss": 0.527, "step": 58058 }, { "epoch": 3.5, "learning_rate": 1.8713503236612194e-05, "loss": 0.4135, "step": 58060 }, { "epoch": 3.5, "learning_rate": 1.8711994326497974e-05, "loss": 0.3281, "step": 58062 }, { "epoch": 3.5, "learning_rate": 1.8710485416383747e-05, "loss": 0.3551, "step": 58064 }, { "epoch": 3.5, "learning_rate": 1.8708976506269523e-05, "loss": 0.5351, "step": 58066 }, { "epoch": 3.5, "learning_rate": 1.8707467596155297e-05, "loss": 0.2939, "step": 58068 }, { "epoch": 3.5, "learning_rate": 1.8705958686041073e-05, "loss": 0.3971, "step": 58070 }, { "epoch": 3.51, "learning_rate": 1.870444977592685e-05, "loss": 0.4056, "step": 58072 }, { "epoch": 3.51, "learning_rate": 1.8702940865812623e-05, "loss": 0.338, "step": 58074 }, { "epoch": 3.51, "learning_rate": 1.87014319556984e-05, "loss": 0.3043, "step": 58076 }, { "epoch": 3.51, "learning_rate": 1.8699923045584176e-05, "loss": 0.3724, "step": 58078 }, { "epoch": 3.51, "learning_rate": 1.8698414135469952e-05, "loss": 0.2821, "step": 58080 }, { "epoch": 3.51, "learning_rate": 1.8696905225355725e-05, "loss": 0.3992, "step": 58082 }, { "epoch": 3.51, "learning_rate": 1.8695396315241502e-05, "loss": 0.2892, "step": 58084 }, { "epoch": 3.51, "learning_rate": 1.869388740512728e-05, "loss": 0.3651, "step": 58086 }, { "epoch": 3.51, "learning_rate": 1.869237849501305e-05, "loss": 0.45, "step": 58088 }, { "epoch": 3.51, "learning_rate": 1.8690869584898828e-05, "loss": 0.432, "step": 58090 }, { "epoch": 3.51, "learning_rate": 1.8689360674784605e-05, "loss": 0.3181, "step": 58092 }, { "epoch": 3.51, "learning_rate": 1.868785176467038e-05, "loss": 0.2607, "step": 58094 }, { "epoch": 3.51, "learning_rate": 1.8686342854556154e-05, "loss": 0.3529, "step": 58096 }, { "epoch": 3.51, "learning_rate": 1.868483394444193e-05, "loss": 0.3168, "step": 58098 }, { "epoch": 3.51, "learning_rate": 1.8683325034327707e-05, "loss": 0.2781, "step": 58100 }, { "epoch": 3.51, "learning_rate": 1.868181612421348e-05, "loss": 0.3179, "step": 58102 }, { "epoch": 3.51, "learning_rate": 1.8680307214099257e-05, "loss": 0.3968, "step": 58104 }, { "epoch": 3.51, "learning_rate": 1.8678798303985033e-05, "loss": 0.3613, "step": 58106 }, { "epoch": 3.51, "learning_rate": 1.867728939387081e-05, "loss": 0.4469, "step": 58108 }, { "epoch": 3.51, "learning_rate": 1.8675780483756583e-05, "loss": 0.2468, "step": 58110 }, { "epoch": 3.51, "learning_rate": 1.867427157364236e-05, "loss": 0.3706, "step": 58112 }, { "epoch": 3.51, "learning_rate": 1.8672762663528136e-05, "loss": 0.3576, "step": 58114 }, { "epoch": 3.51, "learning_rate": 1.867125375341391e-05, "loss": 0.2346, "step": 58116 }, { "epoch": 3.51, "learning_rate": 1.8669744843299686e-05, "loss": 0.2965, "step": 58118 }, { "epoch": 3.51, "learning_rate": 1.866823593318546e-05, "loss": 0.3645, "step": 58120 }, { "epoch": 3.51, "learning_rate": 1.866672702307124e-05, "loss": 0.3401, "step": 58122 }, { "epoch": 3.51, "learning_rate": 1.8665218112957012e-05, "loss": 0.4213, "step": 58124 }, { "epoch": 3.51, "learning_rate": 1.866370920284279e-05, "loss": 0.2992, "step": 58126 }, { "epoch": 3.51, "learning_rate": 1.866220029272856e-05, "loss": 0.4787, "step": 58128 }, { "epoch": 3.51, "learning_rate": 1.866069138261434e-05, "loss": 0.4341, "step": 58130 }, { "epoch": 3.51, "learning_rate": 1.8659182472500114e-05, "loss": 0.3417, "step": 58132 }, { "epoch": 3.51, "learning_rate": 1.8657673562385888e-05, "loss": 0.3861, "step": 58134 }, { "epoch": 3.51, "learning_rate": 1.8656164652271664e-05, "loss": 0.3055, "step": 58136 }, { "epoch": 3.51, "learning_rate": 1.865465574215744e-05, "loss": 0.3269, "step": 58138 }, { "epoch": 3.51, "learning_rate": 1.8653146832043217e-05, "loss": 0.3172, "step": 58140 }, { "epoch": 3.51, "learning_rate": 1.865163792192899e-05, "loss": 0.2687, "step": 58142 }, { "epoch": 3.51, "learning_rate": 1.8650129011814767e-05, "loss": 0.2797, "step": 58144 }, { "epoch": 3.51, "learning_rate": 1.8648620101700543e-05, "loss": 0.3136, "step": 58146 }, { "epoch": 3.51, "learning_rate": 1.864711119158632e-05, "loss": 0.4013, "step": 58148 }, { "epoch": 3.51, "learning_rate": 1.8645602281472093e-05, "loss": 0.3995, "step": 58150 }, { "epoch": 3.51, "learning_rate": 1.864409337135787e-05, "loss": 0.3107, "step": 58152 }, { "epoch": 3.51, "learning_rate": 1.8642584461243646e-05, "loss": 0.2685, "step": 58154 }, { "epoch": 3.51, "learning_rate": 1.864107555112942e-05, "loss": 0.2119, "step": 58156 }, { "epoch": 3.51, "learning_rate": 1.8639566641015196e-05, "loss": 0.3737, "step": 58158 }, { "epoch": 3.51, "learning_rate": 1.8638057730900972e-05, "loss": 0.2174, "step": 58160 }, { "epoch": 3.51, "learning_rate": 1.863654882078675e-05, "loss": 0.2813, "step": 58162 }, { "epoch": 3.51, "learning_rate": 1.8635039910672522e-05, "loss": 0.3264, "step": 58164 }, { "epoch": 3.51, "learning_rate": 1.8633531000558295e-05, "loss": 0.4748, "step": 58166 }, { "epoch": 3.51, "learning_rate": 1.8632022090444075e-05, "loss": 0.264, "step": 58168 }, { "epoch": 3.51, "learning_rate": 1.8630513180329848e-05, "loss": 0.3316, "step": 58170 }, { "epoch": 3.51, "learning_rate": 1.8629004270215624e-05, "loss": 0.5367, "step": 58172 }, { "epoch": 3.51, "learning_rate": 1.8627495360101398e-05, "loss": 0.4466, "step": 58174 }, { "epoch": 3.51, "learning_rate": 1.8625986449987177e-05, "loss": 0.4484, "step": 58176 }, { "epoch": 3.51, "learning_rate": 1.862447753987295e-05, "loss": 0.3661, "step": 58178 }, { "epoch": 3.51, "learning_rate": 1.8622968629758727e-05, "loss": 0.3186, "step": 58180 }, { "epoch": 3.51, "learning_rate": 1.86214597196445e-05, "loss": 0.3544, "step": 58182 }, { "epoch": 3.51, "learning_rate": 1.8619950809530277e-05, "loss": 0.268, "step": 58184 }, { "epoch": 3.51, "learning_rate": 1.8618441899416053e-05, "loss": 0.2647, "step": 58186 }, { "epoch": 3.51, "learning_rate": 1.8616932989301826e-05, "loss": 0.3369, "step": 58188 }, { "epoch": 3.51, "learning_rate": 1.8615424079187603e-05, "loss": 0.3288, "step": 58190 }, { "epoch": 3.51, "learning_rate": 1.861391516907338e-05, "loss": 0.248, "step": 58192 }, { "epoch": 3.51, "learning_rate": 1.8612406258959156e-05, "loss": 0.3526, "step": 58194 }, { "epoch": 3.51, "learning_rate": 1.861089734884493e-05, "loss": 0.2982, "step": 58196 }, { "epoch": 3.51, "learning_rate": 1.8609388438730706e-05, "loss": 0.49, "step": 58198 }, { "epoch": 3.51, "learning_rate": 1.8607879528616482e-05, "loss": 0.3182, "step": 58200 }, { "epoch": 3.51, "learning_rate": 1.8606370618502255e-05, "loss": 0.4846, "step": 58202 }, { "epoch": 3.51, "learning_rate": 1.860486170838803e-05, "loss": 0.3227, "step": 58204 }, { "epoch": 3.51, "learning_rate": 1.8603352798273808e-05, "loss": 0.3467, "step": 58206 }, { "epoch": 3.51, "learning_rate": 1.8601843888159585e-05, "loss": 0.4307, "step": 58208 }, { "epoch": 3.51, "learning_rate": 1.8600334978045358e-05, "loss": 0.3473, "step": 58210 }, { "epoch": 3.51, "learning_rate": 1.8598826067931134e-05, "loss": 0.3774, "step": 58212 }, { "epoch": 3.51, "learning_rate": 1.859731715781691e-05, "loss": 0.497, "step": 58214 }, { "epoch": 3.51, "learning_rate": 1.8595808247702684e-05, "loss": 0.4188, "step": 58216 }, { "epoch": 3.51, "learning_rate": 1.859429933758846e-05, "loss": 0.4837, "step": 58218 }, { "epoch": 3.51, "learning_rate": 1.8592790427474237e-05, "loss": 0.4206, "step": 58220 }, { "epoch": 3.51, "learning_rate": 1.8591281517360014e-05, "loss": 0.2311, "step": 58222 }, { "epoch": 3.51, "learning_rate": 1.8589772607245787e-05, "loss": 0.2632, "step": 58224 }, { "epoch": 3.51, "learning_rate": 1.8588263697131563e-05, "loss": 0.2649, "step": 58226 }, { "epoch": 3.51, "learning_rate": 1.858675478701734e-05, "loss": 0.3782, "step": 58228 }, { "epoch": 3.51, "learning_rate": 1.8585245876903116e-05, "loss": 0.2621, "step": 58230 }, { "epoch": 3.51, "learning_rate": 1.858373696678889e-05, "loss": 0.3098, "step": 58232 }, { "epoch": 3.51, "learning_rate": 1.8582228056674662e-05, "loss": 0.3305, "step": 58234 }, { "epoch": 3.51, "learning_rate": 1.8580719146560442e-05, "loss": 0.4201, "step": 58236 }, { "epoch": 3.52, "learning_rate": 1.8579210236446215e-05, "loss": 0.3879, "step": 58238 }, { "epoch": 3.52, "learning_rate": 1.8577701326331992e-05, "loss": 0.3344, "step": 58240 }, { "epoch": 3.52, "learning_rate": 1.8576192416217765e-05, "loss": 0.3054, "step": 58242 }, { "epoch": 3.52, "learning_rate": 1.8574683506103545e-05, "loss": 0.3676, "step": 58244 }, { "epoch": 3.52, "learning_rate": 1.8573174595989318e-05, "loss": 0.6659, "step": 58246 }, { "epoch": 3.52, "learning_rate": 1.857166568587509e-05, "loss": 0.305, "step": 58248 }, { "epoch": 3.52, "learning_rate": 1.8570156775760868e-05, "loss": 0.368, "step": 58250 }, { "epoch": 3.52, "learning_rate": 1.8568647865646644e-05, "loss": 0.5304, "step": 58252 }, { "epoch": 3.52, "learning_rate": 1.856713895553242e-05, "loss": 0.4074, "step": 58254 }, { "epoch": 3.52, "learning_rate": 1.8565630045418194e-05, "loss": 0.2518, "step": 58256 }, { "epoch": 3.52, "learning_rate": 1.856412113530397e-05, "loss": 0.3117, "step": 58258 }, { "epoch": 3.52, "learning_rate": 1.8562612225189747e-05, "loss": 0.3202, "step": 58260 }, { "epoch": 3.52, "learning_rate": 1.8561103315075523e-05, "loss": 0.3479, "step": 58262 }, { "epoch": 3.52, "learning_rate": 1.8559594404961297e-05, "loss": 0.5544, "step": 58264 }, { "epoch": 3.52, "learning_rate": 1.8558085494847073e-05, "loss": 0.4073, "step": 58266 }, { "epoch": 3.52, "learning_rate": 1.855657658473285e-05, "loss": 0.4435, "step": 58268 }, { "epoch": 3.52, "learning_rate": 1.8555067674618623e-05, "loss": 0.3332, "step": 58270 }, { "epoch": 3.52, "learning_rate": 1.85535587645044e-05, "loss": 0.4414, "step": 58272 }, { "epoch": 3.52, "learning_rate": 1.8552049854390176e-05, "loss": 0.2546, "step": 58274 }, { "epoch": 3.52, "learning_rate": 1.8550540944275952e-05, "loss": 0.3637, "step": 58276 }, { "epoch": 3.52, "learning_rate": 1.8549032034161725e-05, "loss": 0.4498, "step": 58278 }, { "epoch": 3.52, "learning_rate": 1.85475231240475e-05, "loss": 0.4196, "step": 58280 }, { "epoch": 3.52, "learning_rate": 1.854601421393328e-05, "loss": 0.4856, "step": 58282 }, { "epoch": 3.52, "learning_rate": 1.854450530381905e-05, "loss": 0.2971, "step": 58284 }, { "epoch": 3.52, "learning_rate": 1.8542996393704828e-05, "loss": 0.3767, "step": 58286 }, { "epoch": 3.52, "learning_rate": 1.85414874835906e-05, "loss": 0.3118, "step": 58288 }, { "epoch": 3.52, "learning_rate": 1.853997857347638e-05, "loss": 0.4539, "step": 58290 }, { "epoch": 3.52, "learning_rate": 1.8538469663362154e-05, "loss": 0.3385, "step": 58292 }, { "epoch": 3.52, "learning_rate": 1.853696075324793e-05, "loss": 0.3012, "step": 58294 }, { "epoch": 3.52, "learning_rate": 1.8535451843133704e-05, "loss": 0.3324, "step": 58296 }, { "epoch": 3.52, "learning_rate": 1.853394293301948e-05, "loss": 0.4308, "step": 58298 }, { "epoch": 3.52, "learning_rate": 1.8532434022905257e-05, "loss": 0.3241, "step": 58300 }, { "epoch": 3.52, "learning_rate": 1.853092511279103e-05, "loss": 0.5147, "step": 58302 }, { "epoch": 3.52, "learning_rate": 1.8529416202676807e-05, "loss": 0.3953, "step": 58304 }, { "epoch": 3.52, "learning_rate": 1.8527907292562583e-05, "loss": 0.3153, "step": 58306 }, { "epoch": 3.52, "learning_rate": 1.852639838244836e-05, "loss": 0.2434, "step": 58308 }, { "epoch": 3.52, "learning_rate": 1.8524889472334133e-05, "loss": 0.5025, "step": 58310 }, { "epoch": 3.52, "learning_rate": 1.852338056221991e-05, "loss": 0.3306, "step": 58312 }, { "epoch": 3.52, "learning_rate": 1.8521871652105686e-05, "loss": 0.354, "step": 58314 }, { "epoch": 3.52, "learning_rate": 1.852036274199146e-05, "loss": 0.4305, "step": 58316 }, { "epoch": 3.52, "learning_rate": 1.8518853831877235e-05, "loss": 0.2617, "step": 58318 }, { "epoch": 3.52, "learning_rate": 1.8517344921763012e-05, "loss": 0.3866, "step": 58320 }, { "epoch": 3.52, "learning_rate": 1.851583601164879e-05, "loss": 0.4271, "step": 58322 }, { "epoch": 3.52, "learning_rate": 1.851432710153456e-05, "loss": 0.4097, "step": 58324 }, { "epoch": 3.52, "learning_rate": 1.8512818191420338e-05, "loss": 0.4107, "step": 58326 }, { "epoch": 3.52, "learning_rate": 1.8511309281306115e-05, "loss": 0.4251, "step": 58328 }, { "epoch": 3.52, "learning_rate": 1.8509800371191888e-05, "loss": 0.3659, "step": 58330 }, { "epoch": 3.52, "learning_rate": 1.8508291461077664e-05, "loss": 0.3597, "step": 58332 }, { "epoch": 3.52, "learning_rate": 1.850678255096344e-05, "loss": 0.2788, "step": 58334 }, { "epoch": 3.52, "learning_rate": 1.8505273640849217e-05, "loss": 0.3523, "step": 58336 }, { "epoch": 3.52, "learning_rate": 1.850376473073499e-05, "loss": 0.3339, "step": 58338 }, { "epoch": 3.52, "learning_rate": 1.8502255820620767e-05, "loss": 0.292, "step": 58340 }, { "epoch": 3.52, "learning_rate": 1.8500746910506543e-05, "loss": 0.4624, "step": 58342 }, { "epoch": 3.52, "learning_rate": 1.849923800039232e-05, "loss": 0.3347, "step": 58344 }, { "epoch": 3.52, "learning_rate": 1.8497729090278093e-05, "loss": 0.3662, "step": 58346 }, { "epoch": 3.52, "learning_rate": 1.8496220180163866e-05, "loss": 0.186, "step": 58348 }, { "epoch": 3.52, "learning_rate": 1.8494711270049646e-05, "loss": 0.3641, "step": 58350 }, { "epoch": 3.52, "learning_rate": 1.849320235993542e-05, "loss": 0.3538, "step": 58352 }, { "epoch": 3.52, "learning_rate": 1.8491693449821196e-05, "loss": 0.4497, "step": 58354 }, { "epoch": 3.52, "learning_rate": 1.849018453970697e-05, "loss": 0.3979, "step": 58356 }, { "epoch": 3.52, "learning_rate": 1.848867562959275e-05, "loss": 0.385, "step": 58358 }, { "epoch": 3.52, "learning_rate": 1.8487166719478522e-05, "loss": 0.3968, "step": 58360 }, { "epoch": 3.52, "learning_rate": 1.8485657809364295e-05, "loss": 0.3778, "step": 58362 }, { "epoch": 3.52, "learning_rate": 1.848414889925007e-05, "loss": 0.3748, "step": 58364 }, { "epoch": 3.52, "learning_rate": 1.8482639989135848e-05, "loss": 0.393, "step": 58366 }, { "epoch": 3.52, "learning_rate": 1.8481131079021624e-05, "loss": 0.3744, "step": 58368 }, { "epoch": 3.52, "learning_rate": 1.8479622168907398e-05, "loss": 0.3897, "step": 58370 }, { "epoch": 3.52, "learning_rate": 1.8478113258793174e-05, "loss": 0.4248, "step": 58372 }, { "epoch": 3.52, "learning_rate": 1.847660434867895e-05, "loss": 0.3943, "step": 58374 }, { "epoch": 3.52, "learning_rate": 1.8475095438564727e-05, "loss": 0.3126, "step": 58376 }, { "epoch": 3.52, "learning_rate": 1.84735865284505e-05, "loss": 0.4902, "step": 58378 }, { "epoch": 3.52, "learning_rate": 1.8472077618336277e-05, "loss": 0.3801, "step": 58380 }, { "epoch": 3.52, "learning_rate": 1.8470568708222053e-05, "loss": 0.3363, "step": 58382 }, { "epoch": 3.52, "learning_rate": 1.8469059798107826e-05, "loss": 0.1986, "step": 58384 }, { "epoch": 3.52, "learning_rate": 1.8467550887993603e-05, "loss": 0.3445, "step": 58386 }, { "epoch": 3.52, "learning_rate": 1.846604197787938e-05, "loss": 0.3885, "step": 58388 }, { "epoch": 3.52, "learning_rate": 1.8464533067765156e-05, "loss": 0.3224, "step": 58390 }, { "epoch": 3.52, "learning_rate": 1.846302415765093e-05, "loss": 0.3976, "step": 58392 }, { "epoch": 3.52, "learning_rate": 1.8461515247536706e-05, "loss": 0.3463, "step": 58394 }, { "epoch": 3.52, "learning_rate": 1.8460006337422482e-05, "loss": 0.4296, "step": 58396 }, { "epoch": 3.52, "learning_rate": 1.8458497427308255e-05, "loss": 0.2159, "step": 58398 }, { "epoch": 3.52, "learning_rate": 1.8456988517194032e-05, "loss": 0.4059, "step": 58400 }, { "epoch": 3.52, "learning_rate": 1.8455479607079805e-05, "loss": 0.3739, "step": 58402 }, { "epoch": 3.53, "learning_rate": 1.8453970696965585e-05, "loss": 0.3745, "step": 58404 }, { "epoch": 3.53, "learning_rate": 1.8452461786851358e-05, "loss": 0.2296, "step": 58406 }, { "epoch": 3.53, "learning_rate": 1.8450952876737134e-05, "loss": 0.3928, "step": 58408 }, { "epoch": 3.53, "learning_rate": 1.8449443966622908e-05, "loss": 0.4183, "step": 58410 }, { "epoch": 3.53, "learning_rate": 1.8447935056508684e-05, "loss": 0.4276, "step": 58412 }, { "epoch": 3.53, "learning_rate": 1.844642614639446e-05, "loss": 0.4411, "step": 58414 }, { "epoch": 3.53, "learning_rate": 1.8444917236280234e-05, "loss": 0.2268, "step": 58416 }, { "epoch": 3.53, "learning_rate": 1.844340832616601e-05, "loss": 0.4464, "step": 58418 }, { "epoch": 3.53, "learning_rate": 1.8441899416051787e-05, "loss": 0.3482, "step": 58420 }, { "epoch": 3.53, "learning_rate": 1.8440390505937563e-05, "loss": 0.4655, "step": 58422 }, { "epoch": 3.53, "learning_rate": 1.8438881595823336e-05, "loss": 0.2928, "step": 58424 }, { "epoch": 3.53, "learning_rate": 1.8437372685709113e-05, "loss": 0.3197, "step": 58426 }, { "epoch": 3.53, "learning_rate": 1.843586377559489e-05, "loss": 0.3084, "step": 58428 }, { "epoch": 3.53, "learning_rate": 1.8434354865480663e-05, "loss": 0.3525, "step": 58430 }, { "epoch": 3.53, "learning_rate": 1.843284595536644e-05, "loss": 0.4618, "step": 58432 }, { "epoch": 3.53, "learning_rate": 1.8431337045252216e-05, "loss": 0.5029, "step": 58434 }, { "epoch": 3.53, "learning_rate": 1.8429828135137992e-05, "loss": 0.4171, "step": 58436 }, { "epoch": 3.53, "learning_rate": 1.8428319225023765e-05, "loss": 0.3288, "step": 58438 }, { "epoch": 3.53, "learning_rate": 1.8426810314909542e-05, "loss": 0.7028, "step": 58440 }, { "epoch": 3.53, "learning_rate": 1.8425301404795318e-05, "loss": 0.3583, "step": 58442 }, { "epoch": 3.53, "learning_rate": 1.842379249468109e-05, "loss": 0.3293, "step": 58444 }, { "epoch": 3.53, "learning_rate": 1.8422283584566868e-05, "loss": 0.2987, "step": 58446 }, { "epoch": 3.53, "learning_rate": 1.8420774674452644e-05, "loss": 0.5387, "step": 58448 }, { "epoch": 3.53, "learning_rate": 1.841926576433842e-05, "loss": 0.5061, "step": 58450 }, { "epoch": 3.53, "learning_rate": 1.8417756854224194e-05, "loss": 0.3576, "step": 58452 }, { "epoch": 3.53, "learning_rate": 1.841624794410997e-05, "loss": 0.4837, "step": 58454 }, { "epoch": 3.53, "learning_rate": 1.8414739033995747e-05, "loss": 0.2942, "step": 58456 }, { "epoch": 3.53, "learning_rate": 1.8413230123881524e-05, "loss": 0.3505, "step": 58458 }, { "epoch": 3.53, "learning_rate": 1.8411721213767297e-05, "loss": 0.3771, "step": 58460 }, { "epoch": 3.53, "learning_rate": 1.841021230365307e-05, "loss": 0.2844, "step": 58462 }, { "epoch": 3.53, "learning_rate": 1.840870339353885e-05, "loss": 0.3171, "step": 58464 }, { "epoch": 3.53, "learning_rate": 1.8407194483424623e-05, "loss": 0.2407, "step": 58466 }, { "epoch": 3.53, "learning_rate": 1.84056855733104e-05, "loss": 0.3699, "step": 58468 }, { "epoch": 3.53, "learning_rate": 1.8404176663196172e-05, "loss": 0.2776, "step": 58470 }, { "epoch": 3.53, "learning_rate": 1.8402667753081952e-05, "loss": 0.3518, "step": 58472 }, { "epoch": 3.53, "learning_rate": 1.8401158842967725e-05, "loss": 0.3317, "step": 58474 }, { "epoch": 3.53, "learning_rate": 1.8399649932853502e-05, "loss": 0.3146, "step": 58476 }, { "epoch": 3.53, "learning_rate": 1.8398141022739275e-05, "loss": 0.3386, "step": 58478 }, { "epoch": 3.53, "learning_rate": 1.839663211262505e-05, "loss": 0.552, "step": 58480 }, { "epoch": 3.53, "learning_rate": 1.8395123202510828e-05, "loss": 0.3157, "step": 58482 }, { "epoch": 3.53, "learning_rate": 1.83936142923966e-05, "loss": 0.2923, "step": 58484 }, { "epoch": 3.53, "learning_rate": 1.8392105382282378e-05, "loss": 0.4979, "step": 58486 }, { "epoch": 3.53, "learning_rate": 1.8390596472168154e-05, "loss": 0.3953, "step": 58488 }, { "epoch": 3.53, "learning_rate": 1.838908756205393e-05, "loss": 0.4719, "step": 58490 }, { "epoch": 3.53, "learning_rate": 1.8387578651939704e-05, "loss": 0.4239, "step": 58492 }, { "epoch": 3.53, "learning_rate": 1.838606974182548e-05, "loss": 0.2911, "step": 58494 }, { "epoch": 3.53, "learning_rate": 1.8384560831711257e-05, "loss": 0.4033, "step": 58496 }, { "epoch": 3.53, "learning_rate": 1.838305192159703e-05, "loss": 0.2567, "step": 58498 }, { "epoch": 3.53, "learning_rate": 1.8381543011482807e-05, "loss": 0.3699, "step": 58500 }, { "epoch": 3.53, "learning_rate": 1.8380034101368583e-05, "loss": 0.3751, "step": 58502 }, { "epoch": 3.53, "learning_rate": 1.837852519125436e-05, "loss": 0.2673, "step": 58504 }, { "epoch": 3.53, "learning_rate": 1.8377016281140133e-05, "loss": 0.5454, "step": 58506 }, { "epoch": 3.53, "learning_rate": 1.837550737102591e-05, "loss": 0.3175, "step": 58508 }, { "epoch": 3.53, "learning_rate": 1.8373998460911686e-05, "loss": 0.3731, "step": 58510 }, { "epoch": 3.53, "learning_rate": 1.837248955079746e-05, "loss": 0.3039, "step": 58512 }, { "epoch": 3.53, "learning_rate": 1.8370980640683235e-05, "loss": 0.3227, "step": 58514 }, { "epoch": 3.53, "learning_rate": 1.836947173056901e-05, "loss": 0.2251, "step": 58516 }, { "epoch": 3.53, "learning_rate": 1.836796282045479e-05, "loss": 0.3069, "step": 58518 }, { "epoch": 3.53, "learning_rate": 1.836645391034056e-05, "loss": 0.2067, "step": 58520 }, { "epoch": 3.53, "learning_rate": 1.8364945000226338e-05, "loss": 0.4114, "step": 58522 }, { "epoch": 3.53, "learning_rate": 1.836343609011211e-05, "loss": 0.4363, "step": 58524 }, { "epoch": 3.53, "learning_rate": 1.8361927179997888e-05, "loss": 0.3878, "step": 58526 }, { "epoch": 3.53, "learning_rate": 1.8360418269883664e-05, "loss": 0.3539, "step": 58528 }, { "epoch": 3.53, "learning_rate": 1.8358909359769437e-05, "loss": 0.3624, "step": 58530 }, { "epoch": 3.53, "learning_rate": 1.8357400449655214e-05, "loss": 0.388, "step": 58532 }, { "epoch": 3.53, "learning_rate": 1.835589153954099e-05, "loss": 0.3946, "step": 58534 }, { "epoch": 3.53, "learning_rate": 1.8354382629426767e-05, "loss": 0.3473, "step": 58536 }, { "epoch": 3.53, "learning_rate": 1.835287371931254e-05, "loss": 0.3677, "step": 58538 }, { "epoch": 3.53, "learning_rate": 1.8351364809198317e-05, "loss": 0.3562, "step": 58540 }, { "epoch": 3.53, "learning_rate": 1.8349855899084093e-05, "loss": 0.3143, "step": 58542 }, { "epoch": 3.53, "learning_rate": 1.8348346988969866e-05, "loss": 0.3584, "step": 58544 }, { "epoch": 3.53, "learning_rate": 1.8346838078855643e-05, "loss": 0.4015, "step": 58546 }, { "epoch": 3.53, "learning_rate": 1.834532916874142e-05, "loss": 0.2894, "step": 58548 }, { "epoch": 3.53, "learning_rate": 1.8343820258627196e-05, "loss": 0.3475, "step": 58550 }, { "epoch": 3.53, "learning_rate": 1.834231134851297e-05, "loss": 0.4006, "step": 58552 }, { "epoch": 3.53, "learning_rate": 1.8340802438398745e-05, "loss": 0.2995, "step": 58554 }, { "epoch": 3.53, "learning_rate": 1.8339293528284522e-05, "loss": 0.3468, "step": 58556 }, { "epoch": 3.53, "learning_rate": 1.83377846181703e-05, "loss": 0.3615, "step": 58558 }, { "epoch": 3.53, "learning_rate": 1.833627570805607e-05, "loss": 0.4623, "step": 58560 }, { "epoch": 3.53, "learning_rate": 1.8334766797941848e-05, "loss": 0.3508, "step": 58562 }, { "epoch": 3.53, "learning_rate": 1.8333257887827625e-05, "loss": 0.3181, "step": 58564 }, { "epoch": 3.53, "learning_rate": 1.8331748977713398e-05, "loss": 0.3823, "step": 58566 }, { "epoch": 3.54, "learning_rate": 1.8330240067599174e-05, "loss": 0.3986, "step": 58568 }, { "epoch": 3.54, "learning_rate": 1.832873115748495e-05, "loss": 0.2308, "step": 58570 }, { "epoch": 3.54, "learning_rate": 1.8327222247370727e-05, "loss": 0.461, "step": 58572 }, { "epoch": 3.54, "learning_rate": 1.83257133372565e-05, "loss": 0.4242, "step": 58574 }, { "epoch": 3.54, "learning_rate": 1.8324204427142273e-05, "loss": 0.2702, "step": 58576 }, { "epoch": 3.54, "learning_rate": 1.8322695517028053e-05, "loss": 0.3788, "step": 58578 }, { "epoch": 3.54, "learning_rate": 1.8321186606913827e-05, "loss": 0.3156, "step": 58580 }, { "epoch": 3.54, "learning_rate": 1.8319677696799603e-05, "loss": 0.1872, "step": 58582 }, { "epoch": 3.54, "learning_rate": 1.8318168786685376e-05, "loss": 0.3745, "step": 58584 }, { "epoch": 3.54, "learning_rate": 1.8316659876571156e-05, "loss": 0.1731, "step": 58586 }, { "epoch": 3.54, "learning_rate": 1.831515096645693e-05, "loss": 0.303, "step": 58588 }, { "epoch": 3.54, "learning_rate": 1.8313642056342706e-05, "loss": 0.3511, "step": 58590 }, { "epoch": 3.54, "learning_rate": 1.831213314622848e-05, "loss": 0.3883, "step": 58592 }, { "epoch": 3.54, "learning_rate": 1.8310624236114255e-05, "loss": 0.4138, "step": 58594 }, { "epoch": 3.54, "learning_rate": 1.8309115326000032e-05, "loss": 0.4719, "step": 58596 }, { "epoch": 3.54, "learning_rate": 1.8307606415885805e-05, "loss": 0.3133, "step": 58598 }, { "epoch": 3.54, "learning_rate": 1.830609750577158e-05, "loss": 0.3721, "step": 58600 }, { "epoch": 3.54, "learning_rate": 1.8304588595657358e-05, "loss": 0.3846, "step": 58602 }, { "epoch": 3.54, "learning_rate": 1.8303079685543135e-05, "loss": 0.284, "step": 58604 }, { "epoch": 3.54, "learning_rate": 1.8301570775428908e-05, "loss": 0.5158, "step": 58606 }, { "epoch": 3.54, "learning_rate": 1.8300061865314684e-05, "loss": 0.192, "step": 58608 }, { "epoch": 3.54, "learning_rate": 1.829855295520046e-05, "loss": 0.3072, "step": 58610 }, { "epoch": 3.54, "learning_rate": 1.8297044045086234e-05, "loss": 0.4705, "step": 58612 }, { "epoch": 3.54, "learning_rate": 1.829553513497201e-05, "loss": 0.4258, "step": 58614 }, { "epoch": 3.54, "learning_rate": 1.8294026224857787e-05, "loss": 0.4136, "step": 58616 }, { "epoch": 3.54, "learning_rate": 1.8292517314743563e-05, "loss": 0.4033, "step": 58618 }, { "epoch": 3.54, "learning_rate": 1.8291008404629336e-05, "loss": 0.3484, "step": 58620 }, { "epoch": 3.54, "learning_rate": 1.8289499494515113e-05, "loss": 0.2958, "step": 58622 }, { "epoch": 3.54, "learning_rate": 1.828799058440089e-05, "loss": 0.5946, "step": 58624 }, { "epoch": 3.54, "learning_rate": 1.8286481674286663e-05, "loss": 0.322, "step": 58626 }, { "epoch": 3.54, "learning_rate": 1.828497276417244e-05, "loss": 0.5341, "step": 58628 }, { "epoch": 3.54, "learning_rate": 1.8283463854058212e-05, "loss": 0.5929, "step": 58630 }, { "epoch": 3.54, "learning_rate": 1.8281954943943992e-05, "loss": 0.3674, "step": 58632 }, { "epoch": 3.54, "learning_rate": 1.8280446033829765e-05, "loss": 0.4639, "step": 58634 }, { "epoch": 3.54, "learning_rate": 1.8278937123715542e-05, "loss": 0.4824, "step": 58636 }, { "epoch": 3.54, "learning_rate": 1.8277428213601315e-05, "loss": 0.3272, "step": 58638 }, { "epoch": 3.54, "learning_rate": 1.8275919303487095e-05, "loss": 0.3425, "step": 58640 }, { "epoch": 3.54, "learning_rate": 1.8274410393372868e-05, "loss": 0.3143, "step": 58642 }, { "epoch": 3.54, "learning_rate": 1.827290148325864e-05, "loss": 0.5349, "step": 58644 }, { "epoch": 3.54, "learning_rate": 1.8271392573144418e-05, "loss": 0.5897, "step": 58646 }, { "epoch": 3.54, "learning_rate": 1.8269883663030194e-05, "loss": 0.3364, "step": 58648 }, { "epoch": 3.54, "learning_rate": 1.826837475291597e-05, "loss": 0.3023, "step": 58650 }, { "epoch": 3.54, "learning_rate": 1.8266865842801744e-05, "loss": 0.2638, "step": 58652 }, { "epoch": 3.54, "learning_rate": 1.826535693268752e-05, "loss": 0.357, "step": 58654 }, { "epoch": 3.54, "learning_rate": 1.8263848022573297e-05, "loss": 0.4102, "step": 58656 }, { "epoch": 3.54, "learning_rate": 1.826233911245907e-05, "loss": 0.3592, "step": 58658 }, { "epoch": 3.54, "learning_rate": 1.8260830202344846e-05, "loss": 0.193, "step": 58660 }, { "epoch": 3.54, "learning_rate": 1.8259321292230623e-05, "loss": 0.4006, "step": 58662 }, { "epoch": 3.54, "learning_rate": 1.82578123821164e-05, "loss": 0.3746, "step": 58664 }, { "epoch": 3.54, "learning_rate": 1.8256303472002173e-05, "loss": 0.2742, "step": 58666 }, { "epoch": 3.54, "learning_rate": 1.825479456188795e-05, "loss": 0.44, "step": 58668 }, { "epoch": 3.54, "learning_rate": 1.8253285651773726e-05, "loss": 0.4665, "step": 58670 }, { "epoch": 3.54, "learning_rate": 1.8251776741659502e-05, "loss": 0.4672, "step": 58672 }, { "epoch": 3.54, "learning_rate": 1.8250267831545275e-05, "loss": 0.3677, "step": 58674 }, { "epoch": 3.54, "learning_rate": 1.8248758921431052e-05, "loss": 0.3632, "step": 58676 }, { "epoch": 3.54, "learning_rate": 1.8247250011316828e-05, "loss": 0.3368, "step": 58678 }, { "epoch": 3.54, "learning_rate": 1.82457411012026e-05, "loss": 0.3479, "step": 58680 }, { "epoch": 3.54, "learning_rate": 1.8244232191088378e-05, "loss": 0.3917, "step": 58682 }, { "epoch": 3.54, "learning_rate": 1.8242723280974154e-05, "loss": 0.4073, "step": 58684 }, { "epoch": 3.54, "learning_rate": 1.824121437085993e-05, "loss": 0.2697, "step": 58686 }, { "epoch": 3.54, "learning_rate": 1.8239705460745704e-05, "loss": 0.428, "step": 58688 }, { "epoch": 3.54, "learning_rate": 1.823819655063148e-05, "loss": 0.2402, "step": 58690 }, { "epoch": 3.54, "learning_rate": 1.8236687640517257e-05, "loss": 0.4942, "step": 58692 }, { "epoch": 3.54, "learning_rate": 1.823517873040303e-05, "loss": 0.3712, "step": 58694 }, { "epoch": 3.54, "learning_rate": 1.8233669820288807e-05, "loss": 0.3628, "step": 58696 }, { "epoch": 3.54, "learning_rate": 1.823216091017458e-05, "loss": 0.4553, "step": 58698 }, { "epoch": 3.54, "learning_rate": 1.823065200006036e-05, "loss": 0.4012, "step": 58700 }, { "epoch": 3.54, "learning_rate": 1.8229143089946133e-05, "loss": 0.3548, "step": 58702 }, { "epoch": 3.54, "learning_rate": 1.822763417983191e-05, "loss": 0.3096, "step": 58704 }, { "epoch": 3.54, "learning_rate": 1.8226125269717682e-05, "loss": 0.4463, "step": 58706 }, { "epoch": 3.54, "learning_rate": 1.822461635960346e-05, "loss": 0.24, "step": 58708 }, { "epoch": 3.54, "learning_rate": 1.8223107449489236e-05, "loss": 0.3576, "step": 58710 }, { "epoch": 3.54, "learning_rate": 1.822159853937501e-05, "loss": 0.3381, "step": 58712 }, { "epoch": 3.54, "learning_rate": 1.8220089629260785e-05, "loss": 0.3237, "step": 58714 }, { "epoch": 3.54, "learning_rate": 1.821858071914656e-05, "loss": 0.4611, "step": 58716 }, { "epoch": 3.54, "learning_rate": 1.8217071809032338e-05, "loss": 0.5852, "step": 58718 }, { "epoch": 3.54, "learning_rate": 1.821556289891811e-05, "loss": 0.3213, "step": 58720 }, { "epoch": 3.54, "learning_rate": 1.8214053988803888e-05, "loss": 0.3243, "step": 58722 }, { "epoch": 3.54, "learning_rate": 1.8212545078689664e-05, "loss": 0.3557, "step": 58724 }, { "epoch": 3.54, "learning_rate": 1.8211036168575437e-05, "loss": 0.589, "step": 58726 }, { "epoch": 3.54, "learning_rate": 1.8209527258461214e-05, "loss": 0.5206, "step": 58728 }, { "epoch": 3.54, "learning_rate": 1.820801834834699e-05, "loss": 0.3883, "step": 58730 }, { "epoch": 3.54, "learning_rate": 1.8206509438232767e-05, "loss": 0.2289, "step": 58732 }, { "epoch": 3.55, "learning_rate": 1.820500052811854e-05, "loss": 0.4407, "step": 58734 }, { "epoch": 3.55, "learning_rate": 1.8203491618004317e-05, "loss": 0.3348, "step": 58736 }, { "epoch": 3.55, "learning_rate": 1.8201982707890093e-05, "loss": 0.3766, "step": 58738 }, { "epoch": 3.55, "learning_rate": 1.8200473797775866e-05, "loss": 0.3181, "step": 58740 }, { "epoch": 3.55, "learning_rate": 1.8198964887661643e-05, "loss": 0.4297, "step": 58742 }, { "epoch": 3.55, "learning_rate": 1.8197455977547416e-05, "loss": 0.3099, "step": 58744 }, { "epoch": 3.55, "learning_rate": 1.8195947067433196e-05, "loss": 0.4937, "step": 58746 }, { "epoch": 3.55, "learning_rate": 1.819443815731897e-05, "loss": 0.3975, "step": 58748 }, { "epoch": 3.55, "learning_rate": 1.8192929247204745e-05, "loss": 0.2235, "step": 58750 }, { "epoch": 3.55, "learning_rate": 1.819142033709052e-05, "loss": 0.5485, "step": 58752 }, { "epoch": 3.55, "learning_rate": 1.81899114269763e-05, "loss": 0.4642, "step": 58754 }, { "epoch": 3.55, "learning_rate": 1.818840251686207e-05, "loss": 0.4133, "step": 58756 }, { "epoch": 3.55, "learning_rate": 1.8186893606747845e-05, "loss": 0.255, "step": 58758 }, { "epoch": 3.55, "learning_rate": 1.818538469663362e-05, "loss": 0.3799, "step": 58760 }, { "epoch": 3.55, "learning_rate": 1.8183875786519398e-05, "loss": 0.3146, "step": 58762 }, { "epoch": 3.55, "learning_rate": 1.8182366876405174e-05, "loss": 0.3785, "step": 58764 }, { "epoch": 3.55, "learning_rate": 1.8180857966290947e-05, "loss": 0.3721, "step": 58766 }, { "epoch": 3.55, "learning_rate": 1.8179349056176724e-05, "loss": 0.3699, "step": 58768 }, { "epoch": 3.55, "learning_rate": 1.81778401460625e-05, "loss": 0.3895, "step": 58770 }, { "epoch": 3.55, "learning_rate": 1.8176331235948277e-05, "loss": 0.4324, "step": 58772 }, { "epoch": 3.55, "learning_rate": 1.817482232583405e-05, "loss": 0.5675, "step": 58774 }, { "epoch": 3.55, "learning_rate": 1.8173313415719827e-05, "loss": 0.2587, "step": 58776 }, { "epoch": 3.55, "learning_rate": 1.8171804505605603e-05, "loss": 0.2711, "step": 58778 }, { "epoch": 3.55, "learning_rate": 1.8170295595491376e-05, "loss": 0.441, "step": 58780 }, { "epoch": 3.55, "learning_rate": 1.8168786685377153e-05, "loss": 0.305, "step": 58782 }, { "epoch": 3.55, "learning_rate": 1.816727777526293e-05, "loss": 0.5328, "step": 58784 }, { "epoch": 3.55, "learning_rate": 1.8165768865148706e-05, "loss": 0.3793, "step": 58786 }, { "epoch": 3.55, "learning_rate": 1.816425995503448e-05, "loss": 0.4252, "step": 58788 }, { "epoch": 3.55, "learning_rate": 1.8162751044920252e-05, "loss": 0.3585, "step": 58790 }, { "epoch": 3.55, "learning_rate": 1.8161242134806032e-05, "loss": 0.3791, "step": 58792 }, { "epoch": 3.55, "learning_rate": 1.8159733224691805e-05, "loss": 0.27, "step": 58794 }, { "epoch": 3.55, "learning_rate": 1.815822431457758e-05, "loss": 0.3692, "step": 58796 }, { "epoch": 3.55, "learning_rate": 1.8156715404463358e-05, "loss": 0.3778, "step": 58798 }, { "epoch": 3.55, "learning_rate": 1.8155206494349135e-05, "loss": 0.3421, "step": 58800 }, { "epoch": 3.55, "learning_rate": 1.8153697584234908e-05, "loss": 0.577, "step": 58802 }, { "epoch": 3.55, "learning_rate": 1.8152188674120684e-05, "loss": 0.3304, "step": 58804 }, { "epoch": 3.55, "learning_rate": 1.815067976400646e-05, "loss": 0.4781, "step": 58806 }, { "epoch": 3.55, "learning_rate": 1.8149170853892234e-05, "loss": 0.4098, "step": 58808 }, { "epoch": 3.55, "learning_rate": 1.814766194377801e-05, "loss": 0.3171, "step": 58810 }, { "epoch": 3.55, "learning_rate": 1.8146153033663784e-05, "loss": 0.2714, "step": 58812 }, { "epoch": 3.55, "learning_rate": 1.8144644123549563e-05, "loss": 0.4039, "step": 58814 }, { "epoch": 3.55, "learning_rate": 1.8143135213435337e-05, "loss": 0.4254, "step": 58816 }, { "epoch": 3.55, "learning_rate": 1.8141626303321113e-05, "loss": 0.4437, "step": 58818 }, { "epoch": 3.55, "learning_rate": 1.8140117393206886e-05, "loss": 0.3897, "step": 58820 }, { "epoch": 3.55, "learning_rate": 1.8138608483092663e-05, "loss": 0.2946, "step": 58822 }, { "epoch": 3.55, "learning_rate": 1.813709957297844e-05, "loss": 0.4458, "step": 58824 }, { "epoch": 3.55, "learning_rate": 1.8135590662864212e-05, "loss": 0.3466, "step": 58826 }, { "epoch": 3.55, "learning_rate": 1.813408175274999e-05, "loss": 0.33, "step": 58828 }, { "epoch": 3.55, "learning_rate": 1.8132572842635765e-05, "loss": 0.2973, "step": 58830 }, { "epoch": 3.55, "learning_rate": 1.8131063932521542e-05, "loss": 0.2062, "step": 58832 }, { "epoch": 3.55, "learning_rate": 1.8129555022407315e-05, "loss": 0.3287, "step": 58834 }, { "epoch": 3.55, "learning_rate": 1.812804611229309e-05, "loss": 0.4392, "step": 58836 }, { "epoch": 3.55, "learning_rate": 1.8126537202178868e-05, "loss": 0.3605, "step": 58838 }, { "epoch": 3.55, "learning_rate": 1.812502829206464e-05, "loss": 0.2664, "step": 58840 }, { "epoch": 3.55, "learning_rate": 1.8123519381950418e-05, "loss": 0.4409, "step": 58842 }, { "epoch": 3.55, "learning_rate": 1.8122010471836194e-05, "loss": 0.2287, "step": 58844 }, { "epoch": 3.55, "learning_rate": 1.812050156172197e-05, "loss": 0.1867, "step": 58846 }, { "epoch": 3.55, "learning_rate": 1.8118992651607744e-05, "loss": 0.336, "step": 58848 }, { "epoch": 3.55, "learning_rate": 1.811748374149352e-05, "loss": 0.4607, "step": 58850 }, { "epoch": 3.55, "learning_rate": 1.8115974831379297e-05, "loss": 0.3003, "step": 58852 }, { "epoch": 3.55, "learning_rate": 1.8114465921265073e-05, "loss": 0.3374, "step": 58854 }, { "epoch": 3.55, "learning_rate": 1.8112957011150846e-05, "loss": 0.361, "step": 58856 }, { "epoch": 3.55, "learning_rate": 1.811144810103662e-05, "loss": 0.3822, "step": 58858 }, { "epoch": 3.55, "learning_rate": 1.81099391909224e-05, "loss": 0.3722, "step": 58860 }, { "epoch": 3.55, "learning_rate": 1.8108430280808173e-05, "loss": 0.3691, "step": 58862 }, { "epoch": 3.55, "learning_rate": 1.810692137069395e-05, "loss": 0.3496, "step": 58864 }, { "epoch": 3.55, "learning_rate": 1.8105412460579722e-05, "loss": 0.3264, "step": 58866 }, { "epoch": 3.55, "learning_rate": 1.8103903550465502e-05, "loss": 0.3302, "step": 58868 }, { "epoch": 3.55, "learning_rate": 1.8102394640351275e-05, "loss": 0.4417, "step": 58870 }, { "epoch": 3.55, "learning_rate": 1.810088573023705e-05, "loss": 0.3684, "step": 58872 }, { "epoch": 3.55, "learning_rate": 1.8099376820122825e-05, "loss": 0.2408, "step": 58874 }, { "epoch": 3.55, "learning_rate": 1.80978679100086e-05, "loss": 0.3374, "step": 58876 }, { "epoch": 3.55, "learning_rate": 1.8096358999894378e-05, "loss": 0.3922, "step": 58878 }, { "epoch": 3.55, "learning_rate": 1.809485008978015e-05, "loss": 0.371, "step": 58880 }, { "epoch": 3.55, "learning_rate": 1.8093341179665928e-05, "loss": 0.2197, "step": 58882 }, { "epoch": 3.55, "learning_rate": 1.8091832269551704e-05, "loss": 0.3221, "step": 58884 }, { "epoch": 3.55, "learning_rate": 1.809032335943748e-05, "loss": 0.5557, "step": 58886 }, { "epoch": 3.55, "learning_rate": 1.8088814449323254e-05, "loss": 0.2216, "step": 58888 }, { "epoch": 3.55, "learning_rate": 1.808730553920903e-05, "loss": 0.4118, "step": 58890 }, { "epoch": 3.55, "learning_rate": 1.8085796629094807e-05, "loss": 0.4254, "step": 58892 }, { "epoch": 3.55, "learning_rate": 1.808428771898058e-05, "loss": 0.3304, "step": 58894 }, { "epoch": 3.55, "learning_rate": 1.8082778808866356e-05, "loss": 0.3328, "step": 58896 }, { "epoch": 3.55, "learning_rate": 1.8081269898752133e-05, "loss": 0.4515, "step": 58898 }, { "epoch": 3.56, "learning_rate": 1.807976098863791e-05, "loss": 0.3708, "step": 58900 }, { "epoch": 3.56, "learning_rate": 1.8078252078523683e-05, "loss": 0.3118, "step": 58902 }, { "epoch": 3.56, "learning_rate": 1.8076743168409456e-05, "loss": 0.3855, "step": 58904 }, { "epoch": 3.56, "learning_rate": 1.8075234258295236e-05, "loss": 0.3923, "step": 58906 }, { "epoch": 3.56, "learning_rate": 1.807372534818101e-05, "loss": 0.4242, "step": 58908 }, { "epoch": 3.56, "learning_rate": 1.8072216438066785e-05, "loss": 0.4179, "step": 58910 }, { "epoch": 3.56, "learning_rate": 1.8070707527952562e-05, "loss": 0.2847, "step": 58912 }, { "epoch": 3.56, "learning_rate": 1.8069198617838338e-05, "loss": 0.3218, "step": 58914 }, { "epoch": 3.56, "learning_rate": 1.806768970772411e-05, "loss": 0.4202, "step": 58916 }, { "epoch": 3.56, "learning_rate": 1.8066180797609888e-05, "loss": 0.4185, "step": 58918 }, { "epoch": 3.56, "learning_rate": 1.8064671887495664e-05, "loss": 0.3514, "step": 58920 }, { "epoch": 3.56, "learning_rate": 1.8063162977381438e-05, "loss": 0.2927, "step": 58922 }, { "epoch": 3.56, "learning_rate": 1.8061654067267214e-05, "loss": 0.3181, "step": 58924 }, { "epoch": 3.56, "learning_rate": 1.8060145157152987e-05, "loss": 0.3165, "step": 58926 }, { "epoch": 3.56, "learning_rate": 1.8058636247038767e-05, "loss": 0.5157, "step": 58928 }, { "epoch": 3.56, "learning_rate": 1.805712733692454e-05, "loss": 0.3076, "step": 58930 }, { "epoch": 3.56, "learning_rate": 1.8055618426810317e-05, "loss": 0.5479, "step": 58932 }, { "epoch": 3.56, "learning_rate": 1.805410951669609e-05, "loss": 0.2449, "step": 58934 }, { "epoch": 3.56, "learning_rate": 1.805260060658187e-05, "loss": 0.3168, "step": 58936 }, { "epoch": 3.56, "learning_rate": 1.8051091696467643e-05, "loss": 0.2259, "step": 58938 }, { "epoch": 3.56, "learning_rate": 1.8049582786353416e-05, "loss": 0.3245, "step": 58940 }, { "epoch": 3.56, "learning_rate": 1.8048073876239193e-05, "loss": 0.3298, "step": 58942 }, { "epoch": 3.56, "learning_rate": 1.804656496612497e-05, "loss": 0.3549, "step": 58944 }, { "epoch": 3.56, "learning_rate": 1.8045056056010746e-05, "loss": 0.315, "step": 58946 }, { "epoch": 3.56, "learning_rate": 1.804354714589652e-05, "loss": 0.1754, "step": 58948 }, { "epoch": 3.56, "learning_rate": 1.8042038235782295e-05, "loss": 0.2297, "step": 58950 }, { "epoch": 3.56, "learning_rate": 1.804052932566807e-05, "loss": 0.4041, "step": 58952 }, { "epoch": 3.56, "learning_rate": 1.8039020415553845e-05, "loss": 0.2893, "step": 58954 }, { "epoch": 3.56, "learning_rate": 1.803751150543962e-05, "loss": 0.3506, "step": 58956 }, { "epoch": 3.56, "learning_rate": 1.8036002595325398e-05, "loss": 0.4933, "step": 58958 }, { "epoch": 3.56, "learning_rate": 1.8034493685211174e-05, "loss": 0.3275, "step": 58960 }, { "epoch": 3.56, "learning_rate": 1.8032984775096947e-05, "loss": 0.3172, "step": 58962 }, { "epoch": 3.56, "learning_rate": 1.8031475864982724e-05, "loss": 0.3216, "step": 58964 }, { "epoch": 3.56, "learning_rate": 1.80299669548685e-05, "loss": 0.4122, "step": 58966 }, { "epoch": 3.56, "learning_rate": 1.8028458044754277e-05, "loss": 0.3562, "step": 58968 }, { "epoch": 3.56, "learning_rate": 1.802694913464005e-05, "loss": 0.3559, "step": 58970 }, { "epoch": 3.56, "learning_rate": 1.8025440224525823e-05, "loss": 0.374, "step": 58972 }, { "epoch": 3.56, "learning_rate": 1.8023931314411603e-05, "loss": 0.2605, "step": 58974 }, { "epoch": 3.56, "learning_rate": 1.8022422404297376e-05, "loss": 0.3451, "step": 58976 }, { "epoch": 3.56, "learning_rate": 1.8020913494183153e-05, "loss": 0.4273, "step": 58978 }, { "epoch": 3.56, "learning_rate": 1.8019404584068926e-05, "loss": 0.3473, "step": 58980 }, { "epoch": 3.56, "learning_rate": 1.8017895673954706e-05, "loss": 0.2791, "step": 58982 }, { "epoch": 3.56, "learning_rate": 1.801638676384048e-05, "loss": 0.3958, "step": 58984 }, { "epoch": 3.56, "learning_rate": 1.8014877853726252e-05, "loss": 0.4073, "step": 58986 }, { "epoch": 3.56, "learning_rate": 1.801336894361203e-05, "loss": 0.3418, "step": 58988 }, { "epoch": 3.56, "learning_rate": 1.8011860033497805e-05, "loss": 0.3801, "step": 58990 }, { "epoch": 3.56, "learning_rate": 1.801035112338358e-05, "loss": 0.539, "step": 58992 }, { "epoch": 3.56, "learning_rate": 1.8008842213269355e-05, "loss": 0.3831, "step": 58994 }, { "epoch": 3.56, "learning_rate": 1.800733330315513e-05, "loss": 0.3919, "step": 58996 }, { "epoch": 3.56, "learning_rate": 1.8005824393040908e-05, "loss": 0.359, "step": 58998 }, { "epoch": 3.56, "learning_rate": 1.8004315482926684e-05, "loss": 0.3101, "step": 59000 }, { "epoch": 3.56, "learning_rate": 1.8002806572812457e-05, "loss": 0.4295, "step": 59002 }, { "epoch": 3.56, "learning_rate": 1.8001297662698234e-05, "loss": 0.3465, "step": 59004 }, { "epoch": 3.56, "learning_rate": 1.799978875258401e-05, "loss": 0.2093, "step": 59006 }, { "epoch": 3.56, "learning_rate": 1.7998279842469784e-05, "loss": 0.2657, "step": 59008 }, { "epoch": 3.56, "learning_rate": 1.799677093235556e-05, "loss": 0.4754, "step": 59010 }, { "epoch": 3.56, "learning_rate": 1.7995262022241337e-05, "loss": 0.2888, "step": 59012 }, { "epoch": 3.56, "learning_rate": 1.7993753112127113e-05, "loss": 0.2824, "step": 59014 }, { "epoch": 3.56, "learning_rate": 1.7992244202012886e-05, "loss": 0.424, "step": 59016 }, { "epoch": 3.56, "learning_rate": 1.7990735291898663e-05, "loss": 0.417, "step": 59018 }, { "epoch": 3.56, "learning_rate": 1.798922638178444e-05, "loss": 0.2967, "step": 59020 }, { "epoch": 3.56, "learning_rate": 1.7987717471670212e-05, "loss": 0.4044, "step": 59022 }, { "epoch": 3.56, "learning_rate": 1.798620856155599e-05, "loss": 0.2774, "step": 59024 }, { "epoch": 3.56, "learning_rate": 1.7984699651441762e-05, "loss": 0.3194, "step": 59026 }, { "epoch": 3.56, "learning_rate": 1.7983190741327542e-05, "loss": 0.2789, "step": 59028 }, { "epoch": 3.56, "learning_rate": 1.7981681831213315e-05, "loss": 0.2493, "step": 59030 }, { "epoch": 3.56, "learning_rate": 1.798017292109909e-05, "loss": 0.4097, "step": 59032 }, { "epoch": 3.56, "learning_rate": 1.7978664010984868e-05, "loss": 0.372, "step": 59034 }, { "epoch": 3.56, "learning_rate": 1.797715510087064e-05, "loss": 0.4442, "step": 59036 }, { "epoch": 3.56, "learning_rate": 1.7975646190756418e-05, "loss": 0.3113, "step": 59038 }, { "epoch": 3.56, "learning_rate": 1.797413728064219e-05, "loss": 0.4264, "step": 59040 }, { "epoch": 3.56, "learning_rate": 1.797262837052797e-05, "loss": 0.4812, "step": 59042 }, { "epoch": 3.56, "learning_rate": 1.7971119460413744e-05, "loss": 0.3416, "step": 59044 }, { "epoch": 3.56, "learning_rate": 1.796961055029952e-05, "loss": 0.4505, "step": 59046 }, { "epoch": 3.56, "learning_rate": 1.7968101640185294e-05, "loss": 0.3749, "step": 59048 }, { "epoch": 3.56, "learning_rate": 1.7966592730071073e-05, "loss": 0.3988, "step": 59050 }, { "epoch": 3.56, "learning_rate": 1.7965083819956847e-05, "loss": 0.3618, "step": 59052 }, { "epoch": 3.56, "learning_rate": 1.796357490984262e-05, "loss": 0.377, "step": 59054 }, { "epoch": 3.56, "learning_rate": 1.7962065999728396e-05, "loss": 0.2321, "step": 59056 }, { "epoch": 3.56, "learning_rate": 1.7960557089614173e-05, "loss": 0.5294, "step": 59058 }, { "epoch": 3.56, "learning_rate": 1.795904817949995e-05, "loss": 0.2983, "step": 59060 }, { "epoch": 3.56, "learning_rate": 1.7957539269385722e-05, "loss": 0.4789, "step": 59062 }, { "epoch": 3.56, "learning_rate": 1.79560303592715e-05, "loss": 0.3423, "step": 59064 }, { "epoch": 3.57, "learning_rate": 1.7954521449157275e-05, "loss": 0.3822, "step": 59066 }, { "epoch": 3.57, "learning_rate": 1.795301253904305e-05, "loss": 0.3461, "step": 59068 }, { "epoch": 3.57, "learning_rate": 1.7951503628928825e-05, "loss": 0.3157, "step": 59070 }, { "epoch": 3.57, "learning_rate": 1.79499947188146e-05, "loss": 0.3566, "step": 59072 }, { "epoch": 3.57, "learning_rate": 1.7948485808700378e-05, "loss": 0.3136, "step": 59074 }, { "epoch": 3.57, "learning_rate": 1.794697689858615e-05, "loss": 0.3355, "step": 59076 }, { "epoch": 3.57, "learning_rate": 1.7945467988471928e-05, "loss": 0.3622, "step": 59078 }, { "epoch": 3.57, "learning_rate": 1.7943959078357704e-05, "loss": 0.3317, "step": 59080 }, { "epoch": 3.57, "learning_rate": 1.794245016824348e-05, "loss": 0.3482, "step": 59082 }, { "epoch": 3.57, "learning_rate": 1.7940941258129254e-05, "loss": 0.3828, "step": 59084 }, { "epoch": 3.57, "learning_rate": 1.7939432348015027e-05, "loss": 0.5356, "step": 59086 }, { "epoch": 3.57, "learning_rate": 1.7937923437900807e-05, "loss": 0.333, "step": 59088 }, { "epoch": 3.57, "learning_rate": 1.793641452778658e-05, "loss": 0.3957, "step": 59090 }, { "epoch": 3.57, "learning_rate": 1.7934905617672356e-05, "loss": 0.2714, "step": 59092 }, { "epoch": 3.57, "learning_rate": 1.793339670755813e-05, "loss": 0.4762, "step": 59094 }, { "epoch": 3.57, "learning_rate": 1.793188779744391e-05, "loss": 0.364, "step": 59096 }, { "epoch": 3.57, "learning_rate": 1.7930378887329683e-05, "loss": 0.3006, "step": 59098 }, { "epoch": 3.57, "learning_rate": 1.792886997721546e-05, "loss": 0.4636, "step": 59100 }, { "epoch": 3.57, "learning_rate": 1.7927361067101232e-05, "loss": 0.4022, "step": 59102 }, { "epoch": 3.57, "learning_rate": 1.792585215698701e-05, "loss": 0.4935, "step": 59104 }, { "epoch": 3.57, "learning_rate": 1.7924343246872785e-05, "loss": 0.3078, "step": 59106 }, { "epoch": 3.57, "learning_rate": 1.792283433675856e-05, "loss": 0.3914, "step": 59108 }, { "epoch": 3.57, "learning_rate": 1.7921325426644335e-05, "loss": 0.3777, "step": 59110 }, { "epoch": 3.57, "learning_rate": 1.791981651653011e-05, "loss": 0.3656, "step": 59112 }, { "epoch": 3.57, "learning_rate": 1.7918307606415888e-05, "loss": 0.4617, "step": 59114 }, { "epoch": 3.57, "learning_rate": 1.791679869630166e-05, "loss": 0.2901, "step": 59116 }, { "epoch": 3.57, "learning_rate": 1.7915289786187438e-05, "loss": 0.2928, "step": 59118 }, { "epoch": 3.57, "learning_rate": 1.7913780876073214e-05, "loss": 0.3032, "step": 59120 }, { "epoch": 3.57, "learning_rate": 1.7912271965958987e-05, "loss": 0.2922, "step": 59122 }, { "epoch": 3.57, "learning_rate": 1.7910763055844764e-05, "loss": 0.4252, "step": 59124 }, { "epoch": 3.57, "learning_rate": 1.790925414573054e-05, "loss": 0.3389, "step": 59126 }, { "epoch": 3.57, "learning_rate": 1.7907745235616317e-05, "loss": 0.4676, "step": 59128 }, { "epoch": 3.57, "learning_rate": 1.790623632550209e-05, "loss": 0.3813, "step": 59130 }, { "epoch": 3.57, "learning_rate": 1.7904727415387866e-05, "loss": 0.3742, "step": 59132 }, { "epoch": 3.57, "learning_rate": 1.7903218505273643e-05, "loss": 0.3237, "step": 59134 }, { "epoch": 3.57, "learning_rate": 1.7901709595159416e-05, "loss": 0.342, "step": 59136 }, { "epoch": 3.57, "learning_rate": 1.7900200685045193e-05, "loss": 0.2686, "step": 59138 }, { "epoch": 3.57, "learning_rate": 1.7898691774930966e-05, "loss": 0.277, "step": 59140 }, { "epoch": 3.57, "learning_rate": 1.7897182864816746e-05, "loss": 0.2677, "step": 59142 }, { "epoch": 3.57, "learning_rate": 1.789567395470252e-05, "loss": 0.4079, "step": 59144 }, { "epoch": 3.57, "learning_rate": 1.7894165044588295e-05, "loss": 0.1546, "step": 59146 }, { "epoch": 3.57, "learning_rate": 1.789265613447407e-05, "loss": 0.2136, "step": 59148 }, { "epoch": 3.57, "learning_rate": 1.7891147224359845e-05, "loss": 0.3786, "step": 59150 }, { "epoch": 3.57, "learning_rate": 1.788963831424562e-05, "loss": 0.3528, "step": 59152 }, { "epoch": 3.57, "learning_rate": 1.7888129404131395e-05, "loss": 0.4105, "step": 59154 }, { "epoch": 3.57, "learning_rate": 1.7886620494017174e-05, "loss": 0.3275, "step": 59156 }, { "epoch": 3.57, "learning_rate": 1.7885111583902948e-05, "loss": 0.3538, "step": 59158 }, { "epoch": 3.57, "learning_rate": 1.7883602673788724e-05, "loss": 0.3851, "step": 59160 }, { "epoch": 3.57, "learning_rate": 1.7882093763674497e-05, "loss": 0.314, "step": 59162 }, { "epoch": 3.57, "learning_rate": 1.7880584853560277e-05, "loss": 0.3243, "step": 59164 }, { "epoch": 3.57, "learning_rate": 1.787907594344605e-05, "loss": 0.3581, "step": 59166 }, { "epoch": 3.57, "learning_rate": 1.7877567033331823e-05, "loss": 0.5187, "step": 59168 }, { "epoch": 3.57, "learning_rate": 1.78760581232176e-05, "loss": 0.3216, "step": 59170 }, { "epoch": 3.57, "learning_rate": 1.7874549213103376e-05, "loss": 0.5101, "step": 59172 }, { "epoch": 3.57, "learning_rate": 1.7873040302989153e-05, "loss": 0.417, "step": 59174 }, { "epoch": 3.57, "learning_rate": 1.7871531392874926e-05, "loss": 0.4094, "step": 59176 }, { "epoch": 3.57, "learning_rate": 1.7870022482760703e-05, "loss": 0.3229, "step": 59178 }, { "epoch": 3.57, "learning_rate": 1.786851357264648e-05, "loss": 0.4077, "step": 59180 }, { "epoch": 3.57, "learning_rate": 1.7867004662532256e-05, "loss": 0.2561, "step": 59182 }, { "epoch": 3.57, "learning_rate": 1.786549575241803e-05, "loss": 0.6769, "step": 59184 }, { "epoch": 3.57, "learning_rate": 1.7863986842303805e-05, "loss": 0.36, "step": 59186 }, { "epoch": 3.57, "learning_rate": 1.7862477932189582e-05, "loss": 0.3578, "step": 59188 }, { "epoch": 3.57, "learning_rate": 1.7860969022075355e-05, "loss": 0.3545, "step": 59190 }, { "epoch": 3.57, "learning_rate": 1.785946011196113e-05, "loss": 0.2838, "step": 59192 }, { "epoch": 3.57, "learning_rate": 1.7857951201846908e-05, "loss": 0.3139, "step": 59194 }, { "epoch": 3.57, "learning_rate": 1.7856442291732684e-05, "loss": 0.3677, "step": 59196 }, { "epoch": 3.57, "learning_rate": 1.7854933381618458e-05, "loss": 0.5266, "step": 59198 }, { "epoch": 3.57, "learning_rate": 1.785342447150423e-05, "loss": 0.3205, "step": 59200 }, { "epoch": 3.57, "learning_rate": 1.785191556139001e-05, "loss": 0.2956, "step": 59202 }, { "epoch": 3.57, "learning_rate": 1.7850406651275784e-05, "loss": 0.262, "step": 59204 }, { "epoch": 3.57, "learning_rate": 1.784889774116156e-05, "loss": 0.2591, "step": 59206 }, { "epoch": 3.57, "learning_rate": 1.7847388831047333e-05, "loss": 0.4182, "step": 59208 }, { "epoch": 3.57, "learning_rate": 1.7845879920933113e-05, "loss": 0.2843, "step": 59210 }, { "epoch": 3.57, "learning_rate": 1.7844371010818886e-05, "loss": 0.3802, "step": 59212 }, { "epoch": 3.57, "learning_rate": 1.7842862100704663e-05, "loss": 0.5076, "step": 59214 }, { "epoch": 3.57, "learning_rate": 1.7841353190590436e-05, "loss": 0.2365, "step": 59216 }, { "epoch": 3.57, "learning_rate": 1.7839844280476212e-05, "loss": 0.4525, "step": 59218 }, { "epoch": 3.57, "learning_rate": 1.783833537036199e-05, "loss": 0.2714, "step": 59220 }, { "epoch": 3.57, "learning_rate": 1.7836826460247762e-05, "loss": 0.2676, "step": 59222 }, { "epoch": 3.57, "learning_rate": 1.783531755013354e-05, "loss": 0.3994, "step": 59224 }, { "epoch": 3.57, "learning_rate": 1.7833808640019315e-05, "loss": 0.2784, "step": 59226 }, { "epoch": 3.57, "learning_rate": 1.783229972990509e-05, "loss": 0.3756, "step": 59228 }, { "epoch": 3.57, "learning_rate": 1.7830790819790865e-05, "loss": 0.4555, "step": 59230 }, { "epoch": 3.58, "learning_rate": 1.782928190967664e-05, "loss": 0.5323, "step": 59232 }, { "epoch": 3.58, "learning_rate": 1.7827772999562418e-05, "loss": 0.332, "step": 59234 }, { "epoch": 3.58, "learning_rate": 1.782626408944819e-05, "loss": 0.3245, "step": 59236 }, { "epoch": 3.58, "learning_rate": 1.7824755179333967e-05, "loss": 0.2922, "step": 59238 }, { "epoch": 3.58, "learning_rate": 1.7823246269219744e-05, "loss": 0.411, "step": 59240 }, { "epoch": 3.58, "learning_rate": 1.782173735910552e-05, "loss": 0.3557, "step": 59242 }, { "epoch": 3.58, "learning_rate": 1.7820228448991294e-05, "loss": 0.3342, "step": 59244 }, { "epoch": 3.58, "learning_rate": 1.781871953887707e-05, "loss": 0.4174, "step": 59246 }, { "epoch": 3.58, "learning_rate": 1.7817210628762847e-05, "loss": 0.295, "step": 59248 }, { "epoch": 3.58, "learning_rate": 1.781570171864862e-05, "loss": 0.321, "step": 59250 }, { "epoch": 3.58, "learning_rate": 1.7814192808534396e-05, "loss": 0.3063, "step": 59252 }, { "epoch": 3.58, "learning_rate": 1.781268389842017e-05, "loss": 0.2812, "step": 59254 }, { "epoch": 3.58, "learning_rate": 1.781117498830595e-05, "loss": 0.4344, "step": 59256 }, { "epoch": 3.58, "learning_rate": 1.7809666078191722e-05, "loss": 0.4261, "step": 59258 }, { "epoch": 3.58, "learning_rate": 1.78081571680775e-05, "loss": 0.3742, "step": 59260 }, { "epoch": 3.58, "learning_rate": 1.7806648257963272e-05, "loss": 0.3104, "step": 59262 }, { "epoch": 3.58, "learning_rate": 1.7805139347849052e-05, "loss": 0.4938, "step": 59264 }, { "epoch": 3.58, "learning_rate": 1.7803630437734825e-05, "loss": 0.2799, "step": 59266 }, { "epoch": 3.58, "learning_rate": 1.7802121527620598e-05, "loss": 0.2441, "step": 59268 }, { "epoch": 3.58, "learning_rate": 1.7800612617506378e-05, "loss": 0.2781, "step": 59270 }, { "epoch": 3.58, "learning_rate": 1.779910370739215e-05, "loss": 0.3129, "step": 59272 }, { "epoch": 3.58, "learning_rate": 1.7797594797277928e-05, "loss": 0.3685, "step": 59274 }, { "epoch": 3.58, "learning_rate": 1.77960858871637e-05, "loss": 0.4201, "step": 59276 }, { "epoch": 3.58, "learning_rate": 1.779457697704948e-05, "loss": 0.3943, "step": 59278 }, { "epoch": 3.58, "learning_rate": 1.7793068066935254e-05, "loss": 0.307, "step": 59280 }, { "epoch": 3.58, "learning_rate": 1.7791559156821027e-05, "loss": 0.2523, "step": 59282 }, { "epoch": 3.58, "learning_rate": 1.7790050246706804e-05, "loss": 0.4715, "step": 59284 }, { "epoch": 3.58, "learning_rate": 1.778854133659258e-05, "loss": 0.408, "step": 59286 }, { "epoch": 3.58, "learning_rate": 1.7787032426478357e-05, "loss": 0.3264, "step": 59288 }, { "epoch": 3.58, "learning_rate": 1.778552351636413e-05, "loss": 0.3668, "step": 59290 }, { "epoch": 3.58, "learning_rate": 1.7784014606249906e-05, "loss": 0.4149, "step": 59292 }, { "epoch": 3.58, "learning_rate": 1.7782505696135683e-05, "loss": 0.3191, "step": 59294 }, { "epoch": 3.58, "learning_rate": 1.778099678602146e-05, "loss": 0.4196, "step": 59296 }, { "epoch": 3.58, "learning_rate": 1.7779487875907232e-05, "loss": 0.5037, "step": 59298 }, { "epoch": 3.58, "learning_rate": 1.777797896579301e-05, "loss": 0.3443, "step": 59300 }, { "epoch": 3.58, "learning_rate": 1.7776470055678785e-05, "loss": 0.3236, "step": 59302 }, { "epoch": 3.58, "learning_rate": 1.777496114556456e-05, "loss": 0.2341, "step": 59304 }, { "epoch": 3.58, "learning_rate": 1.7773452235450335e-05, "loss": 0.3869, "step": 59306 }, { "epoch": 3.58, "learning_rate": 1.777194332533611e-05, "loss": 0.3642, "step": 59308 }, { "epoch": 3.58, "learning_rate": 1.7770434415221888e-05, "loss": 0.3237, "step": 59310 }, { "epoch": 3.58, "learning_rate": 1.776892550510766e-05, "loss": 0.3427, "step": 59312 }, { "epoch": 3.58, "learning_rate": 1.7767416594993438e-05, "loss": 0.4523, "step": 59314 }, { "epoch": 3.58, "learning_rate": 1.7765907684879214e-05, "loss": 0.3342, "step": 59316 }, { "epoch": 3.58, "learning_rate": 1.7764398774764987e-05, "loss": 0.3429, "step": 59318 }, { "epoch": 3.58, "learning_rate": 1.7762889864650764e-05, "loss": 0.2702, "step": 59320 }, { "epoch": 3.58, "learning_rate": 1.7761380954536537e-05, "loss": 0.4102, "step": 59322 }, { "epoch": 3.58, "learning_rate": 1.7759872044422317e-05, "loss": 0.4857, "step": 59324 }, { "epoch": 3.58, "learning_rate": 1.775836313430809e-05, "loss": 0.3393, "step": 59326 }, { "epoch": 3.58, "learning_rate": 1.7756854224193867e-05, "loss": 0.3988, "step": 59328 }, { "epoch": 3.58, "learning_rate": 1.775534531407964e-05, "loss": 0.3691, "step": 59330 }, { "epoch": 3.58, "learning_rate": 1.7753836403965416e-05, "loss": 0.2865, "step": 59332 }, { "epoch": 3.58, "learning_rate": 1.7752327493851193e-05, "loss": 0.4661, "step": 59334 }, { "epoch": 3.58, "learning_rate": 1.7750818583736966e-05, "loss": 0.372, "step": 59336 }, { "epoch": 3.58, "learning_rate": 1.7749309673622742e-05, "loss": 0.3326, "step": 59338 }, { "epoch": 3.58, "learning_rate": 1.774780076350852e-05, "loss": 0.3938, "step": 59340 }, { "epoch": 3.58, "learning_rate": 1.7746291853394295e-05, "loss": 0.2794, "step": 59342 }, { "epoch": 3.58, "learning_rate": 1.774478294328007e-05, "loss": 0.3727, "step": 59344 }, { "epoch": 3.58, "learning_rate": 1.7743274033165845e-05, "loss": 0.385, "step": 59346 }, { "epoch": 3.58, "learning_rate": 1.774176512305162e-05, "loss": 0.3939, "step": 59348 }, { "epoch": 3.58, "learning_rate": 1.7740256212937395e-05, "loss": 0.4469, "step": 59350 }, { "epoch": 3.58, "learning_rate": 1.773874730282317e-05, "loss": 0.3755, "step": 59352 }, { "epoch": 3.58, "learning_rate": 1.7737238392708948e-05, "loss": 0.5142, "step": 59354 }, { "epoch": 3.58, "learning_rate": 1.7735729482594724e-05, "loss": 0.3571, "step": 59356 }, { "epoch": 3.58, "learning_rate": 1.7734220572480497e-05, "loss": 0.376, "step": 59358 }, { "epoch": 3.58, "learning_rate": 1.7732711662366274e-05, "loss": 0.3759, "step": 59360 }, { "epoch": 3.58, "learning_rate": 1.773120275225205e-05, "loss": 0.4139, "step": 59362 }, { "epoch": 3.58, "learning_rate": 1.7729693842137823e-05, "loss": 0.3939, "step": 59364 }, { "epoch": 3.58, "learning_rate": 1.77281849320236e-05, "loss": 0.4308, "step": 59366 }, { "epoch": 3.58, "learning_rate": 1.7726676021909373e-05, "loss": 0.4353, "step": 59368 }, { "epoch": 3.58, "learning_rate": 1.7725167111795153e-05, "loss": 0.3425, "step": 59370 }, { "epoch": 3.58, "learning_rate": 1.7723658201680926e-05, "loss": 0.3308, "step": 59372 }, { "epoch": 3.58, "learning_rate": 1.7722149291566703e-05, "loss": 0.2691, "step": 59374 }, { "epoch": 3.58, "learning_rate": 1.7720640381452476e-05, "loss": 0.3143, "step": 59376 }, { "epoch": 3.58, "learning_rate": 1.7719131471338256e-05, "loss": 0.4285, "step": 59378 }, { "epoch": 3.58, "learning_rate": 1.771762256122403e-05, "loss": 0.4412, "step": 59380 }, { "epoch": 3.58, "learning_rate": 1.7716113651109802e-05, "loss": 0.3188, "step": 59382 }, { "epoch": 3.58, "learning_rate": 1.771460474099558e-05, "loss": 0.4351, "step": 59384 }, { "epoch": 3.58, "learning_rate": 1.7713095830881355e-05, "loss": 0.3207, "step": 59386 }, { "epoch": 3.58, "learning_rate": 1.771158692076713e-05, "loss": 0.3358, "step": 59388 }, { "epoch": 3.58, "learning_rate": 1.7710078010652905e-05, "loss": 0.3903, "step": 59390 }, { "epoch": 3.58, "learning_rate": 1.7708569100538684e-05, "loss": 0.4355, "step": 59392 }, { "epoch": 3.58, "learning_rate": 1.7707060190424458e-05, "loss": 0.2748, "step": 59394 }, { "epoch": 3.58, "learning_rate": 1.7705551280310234e-05, "loss": 0.3187, "step": 59396 }, { "epoch": 3.59, "learning_rate": 1.7704042370196007e-05, "loss": 0.3935, "step": 59398 }, { "epoch": 3.59, "learning_rate": 1.7702533460081784e-05, "loss": 0.3536, "step": 59400 }, { "epoch": 3.59, "learning_rate": 1.770102454996756e-05, "loss": 0.5919, "step": 59402 }, { "epoch": 3.59, "learning_rate": 1.7699515639853333e-05, "loss": 0.4061, "step": 59404 }, { "epoch": 3.59, "learning_rate": 1.769800672973911e-05, "loss": 0.4474, "step": 59406 }, { "epoch": 3.59, "learning_rate": 1.7696497819624886e-05, "loss": 0.4192, "step": 59408 }, { "epoch": 3.59, "learning_rate": 1.7694988909510663e-05, "loss": 0.3259, "step": 59410 }, { "epoch": 3.59, "learning_rate": 1.7693479999396436e-05, "loss": 0.3333, "step": 59412 }, { "epoch": 3.59, "learning_rate": 1.7691971089282213e-05, "loss": 0.3145, "step": 59414 }, { "epoch": 3.59, "learning_rate": 1.769046217916799e-05, "loss": 0.2427, "step": 59416 }, { "epoch": 3.59, "learning_rate": 1.7688953269053762e-05, "loss": 0.3474, "step": 59418 }, { "epoch": 3.59, "learning_rate": 1.768744435893954e-05, "loss": 0.3857, "step": 59420 }, { "epoch": 3.59, "learning_rate": 1.7685935448825315e-05, "loss": 0.3861, "step": 59422 }, { "epoch": 3.59, "learning_rate": 1.7684426538711092e-05, "loss": 0.3704, "step": 59424 }, { "epoch": 3.59, "learning_rate": 1.7682917628596865e-05, "loss": 0.4145, "step": 59426 }, { "epoch": 3.59, "learning_rate": 1.768140871848264e-05, "loss": 0.4557, "step": 59428 }, { "epoch": 3.59, "learning_rate": 1.7679899808368418e-05, "loss": 0.4353, "step": 59430 }, { "epoch": 3.59, "learning_rate": 1.767839089825419e-05, "loss": 0.3774, "step": 59432 }, { "epoch": 3.59, "learning_rate": 1.7676881988139968e-05, "loss": 0.399, "step": 59434 }, { "epoch": 3.59, "learning_rate": 1.767537307802574e-05, "loss": 0.3647, "step": 59436 }, { "epoch": 3.59, "learning_rate": 1.767386416791152e-05, "loss": 0.3031, "step": 59438 }, { "epoch": 3.59, "learning_rate": 1.7672355257797294e-05, "loss": 0.3023, "step": 59440 }, { "epoch": 3.59, "learning_rate": 1.767084634768307e-05, "loss": 0.3804, "step": 59442 }, { "epoch": 3.59, "learning_rate": 1.7669337437568843e-05, "loss": 0.3175, "step": 59444 }, { "epoch": 3.59, "learning_rate": 1.766782852745462e-05, "loss": 0.3368, "step": 59446 }, { "epoch": 3.59, "learning_rate": 1.7666319617340396e-05, "loss": 0.3589, "step": 59448 }, { "epoch": 3.59, "learning_rate": 1.766481070722617e-05, "loss": 0.4426, "step": 59450 }, { "epoch": 3.59, "learning_rate": 1.7663301797111946e-05, "loss": 0.2766, "step": 59452 }, { "epoch": 3.59, "learning_rate": 1.7661792886997723e-05, "loss": 0.2494, "step": 59454 }, { "epoch": 3.59, "learning_rate": 1.76602839768835e-05, "loss": 0.2448, "step": 59456 }, { "epoch": 3.59, "learning_rate": 1.7658775066769272e-05, "loss": 0.4386, "step": 59458 }, { "epoch": 3.59, "learning_rate": 1.765726615665505e-05, "loss": 0.245, "step": 59460 }, { "epoch": 3.59, "learning_rate": 1.7655757246540825e-05, "loss": 0.4421, "step": 59462 }, { "epoch": 3.59, "learning_rate": 1.7654248336426598e-05, "loss": 0.362, "step": 59464 }, { "epoch": 3.59, "learning_rate": 1.7652739426312375e-05, "loss": 0.236, "step": 59466 }, { "epoch": 3.59, "learning_rate": 1.765123051619815e-05, "loss": 0.3172, "step": 59468 }, { "epoch": 3.59, "learning_rate": 1.7649721606083928e-05, "loss": 0.5492, "step": 59470 }, { "epoch": 3.59, "learning_rate": 1.76482126959697e-05, "loss": 0.4155, "step": 59472 }, { "epoch": 3.59, "learning_rate": 1.7646703785855477e-05, "loss": 0.2811, "step": 59474 }, { "epoch": 3.59, "learning_rate": 1.7645194875741254e-05, "loss": 0.4011, "step": 59476 }, { "epoch": 3.59, "learning_rate": 1.764368596562703e-05, "loss": 0.3056, "step": 59478 }, { "epoch": 3.59, "learning_rate": 1.7642177055512804e-05, "loss": 0.3517, "step": 59480 }, { "epoch": 3.59, "learning_rate": 1.7640668145398577e-05, "loss": 0.5481, "step": 59482 }, { "epoch": 3.59, "learning_rate": 1.7639159235284357e-05, "loss": 0.3001, "step": 59484 }, { "epoch": 3.59, "learning_rate": 1.763765032517013e-05, "loss": 0.4205, "step": 59486 }, { "epoch": 3.59, "learning_rate": 1.7636141415055906e-05, "loss": 0.331, "step": 59488 }, { "epoch": 3.59, "learning_rate": 1.763463250494168e-05, "loss": 0.3372, "step": 59490 }, { "epoch": 3.59, "learning_rate": 1.763312359482746e-05, "loss": 0.4436, "step": 59492 }, { "epoch": 3.59, "learning_rate": 1.7631614684713232e-05, "loss": 0.4903, "step": 59494 }, { "epoch": 3.59, "learning_rate": 1.7630105774599006e-05, "loss": 0.3566, "step": 59496 }, { "epoch": 3.59, "learning_rate": 1.7628596864484782e-05, "loss": 0.3337, "step": 59498 }, { "epoch": 3.59, "learning_rate": 1.762708795437056e-05, "loss": 0.3255, "step": 59500 }, { "epoch": 3.59, "learning_rate": 1.7625579044256335e-05, "loss": 0.2412, "step": 59502 }, { "epoch": 3.59, "learning_rate": 1.7624070134142108e-05, "loss": 0.4617, "step": 59504 }, { "epoch": 3.59, "learning_rate": 1.7622561224027888e-05, "loss": 0.3414, "step": 59506 }, { "epoch": 3.59, "learning_rate": 1.762105231391366e-05, "loss": 0.163, "step": 59508 }, { "epoch": 3.59, "learning_rate": 1.7619543403799438e-05, "loss": 0.3317, "step": 59510 }, { "epoch": 3.59, "learning_rate": 1.761803449368521e-05, "loss": 0.5832, "step": 59512 }, { "epoch": 3.59, "learning_rate": 1.7616525583570987e-05, "loss": 0.4286, "step": 59514 }, { "epoch": 3.59, "learning_rate": 1.7615016673456764e-05, "loss": 0.5986, "step": 59516 }, { "epoch": 3.59, "learning_rate": 1.7613507763342537e-05, "loss": 0.3192, "step": 59518 }, { "epoch": 3.59, "learning_rate": 1.7611998853228314e-05, "loss": 0.4728, "step": 59520 }, { "epoch": 3.59, "learning_rate": 1.761048994311409e-05, "loss": 0.3867, "step": 59522 }, { "epoch": 3.59, "learning_rate": 1.7608981032999867e-05, "loss": 0.335, "step": 59524 }, { "epoch": 3.59, "learning_rate": 1.760747212288564e-05, "loss": 0.3959, "step": 59526 }, { "epoch": 3.59, "learning_rate": 1.7605963212771416e-05, "loss": 0.2595, "step": 59528 }, { "epoch": 3.59, "learning_rate": 1.7604454302657193e-05, "loss": 0.3898, "step": 59530 }, { "epoch": 3.59, "learning_rate": 1.7602945392542966e-05, "loss": 0.5388, "step": 59532 }, { "epoch": 3.59, "learning_rate": 1.7601436482428742e-05, "loss": 0.2935, "step": 59534 }, { "epoch": 3.59, "learning_rate": 1.759992757231452e-05, "loss": 0.3188, "step": 59536 }, { "epoch": 3.59, "learning_rate": 1.7598418662200295e-05, "loss": 0.4049, "step": 59538 }, { "epoch": 3.59, "learning_rate": 1.759690975208607e-05, "loss": 0.2874, "step": 59540 }, { "epoch": 3.59, "learning_rate": 1.7595400841971845e-05, "loss": 0.369, "step": 59542 }, { "epoch": 3.59, "learning_rate": 1.759389193185762e-05, "loss": 0.4202, "step": 59544 }, { "epoch": 3.59, "learning_rate": 1.7592383021743395e-05, "loss": 0.3244, "step": 59546 }, { "epoch": 3.59, "learning_rate": 1.759087411162917e-05, "loss": 0.2911, "step": 59548 }, { "epoch": 3.59, "learning_rate": 1.7589365201514944e-05, "loss": 0.4735, "step": 59550 }, { "epoch": 3.59, "learning_rate": 1.7587856291400724e-05, "loss": 0.4102, "step": 59552 }, { "epoch": 3.59, "learning_rate": 1.7586347381286497e-05, "loss": 0.3389, "step": 59554 }, { "epoch": 3.59, "learning_rate": 1.7584838471172274e-05, "loss": 0.4383, "step": 59556 }, { "epoch": 3.59, "learning_rate": 1.7583329561058047e-05, "loss": 0.4134, "step": 59558 }, { "epoch": 3.59, "learning_rate": 1.7581820650943827e-05, "loss": 0.3745, "step": 59560 }, { "epoch": 3.6, "learning_rate": 1.75803117408296e-05, "loss": 0.2279, "step": 59562 }, { "epoch": 3.6, "learning_rate": 1.7578802830715373e-05, "loss": 0.2681, "step": 59564 }, { "epoch": 3.6, "learning_rate": 1.757729392060115e-05, "loss": 0.2992, "step": 59566 }, { "epoch": 3.6, "learning_rate": 1.7575785010486926e-05, "loss": 0.3111, "step": 59568 }, { "epoch": 3.6, "learning_rate": 1.7574276100372703e-05, "loss": 0.4325, "step": 59570 }, { "epoch": 3.6, "learning_rate": 1.7572767190258476e-05, "loss": 0.4099, "step": 59572 }, { "epoch": 3.6, "learning_rate": 1.7571258280144252e-05, "loss": 0.2661, "step": 59574 }, { "epoch": 3.6, "learning_rate": 1.756974937003003e-05, "loss": 0.3983, "step": 59576 }, { "epoch": 3.6, "learning_rate": 1.7568240459915802e-05, "loss": 0.3832, "step": 59578 }, { "epoch": 3.6, "learning_rate": 1.756673154980158e-05, "loss": 0.345, "step": 59580 }, { "epoch": 3.6, "learning_rate": 1.7565222639687355e-05, "loss": 0.541, "step": 59582 }, { "epoch": 3.6, "learning_rate": 1.756371372957313e-05, "loss": 0.4641, "step": 59584 }, { "epoch": 3.6, "learning_rate": 1.7562204819458905e-05, "loss": 0.2948, "step": 59586 }, { "epoch": 3.6, "learning_rate": 1.756069590934468e-05, "loss": 0.3336, "step": 59588 }, { "epoch": 3.6, "learning_rate": 1.7559186999230458e-05, "loss": 0.3581, "step": 59590 }, { "epoch": 3.6, "learning_rate": 1.7557678089116234e-05, "loss": 0.3383, "step": 59592 }, { "epoch": 3.6, "learning_rate": 1.7556169179002007e-05, "loss": 0.402, "step": 59594 }, { "epoch": 3.6, "learning_rate": 1.755466026888778e-05, "loss": 0.3461, "step": 59596 }, { "epoch": 3.6, "learning_rate": 1.755315135877356e-05, "loss": 0.2371, "step": 59598 }, { "epoch": 3.6, "learning_rate": 1.7551642448659333e-05, "loss": 0.3625, "step": 59600 }, { "epoch": 3.6, "learning_rate": 1.755013353854511e-05, "loss": 0.3418, "step": 59602 }, { "epoch": 3.6, "learning_rate": 1.7548624628430883e-05, "loss": 0.5483, "step": 59604 }, { "epoch": 3.6, "learning_rate": 1.7547115718316663e-05, "loss": 0.282, "step": 59606 }, { "epoch": 3.6, "learning_rate": 1.7545606808202436e-05, "loss": 0.3309, "step": 59608 }, { "epoch": 3.6, "learning_rate": 1.754409789808821e-05, "loss": 0.584, "step": 59610 }, { "epoch": 3.6, "learning_rate": 1.7542588987973986e-05, "loss": 0.4839, "step": 59612 }, { "epoch": 3.6, "learning_rate": 1.7541080077859762e-05, "loss": 0.2781, "step": 59614 }, { "epoch": 3.6, "learning_rate": 1.753957116774554e-05, "loss": 0.3657, "step": 59616 }, { "epoch": 3.6, "learning_rate": 1.7538062257631312e-05, "loss": 0.341, "step": 59618 }, { "epoch": 3.6, "learning_rate": 1.753655334751709e-05, "loss": 0.513, "step": 59620 }, { "epoch": 3.6, "learning_rate": 1.7535044437402865e-05, "loss": 0.4316, "step": 59622 }, { "epoch": 3.6, "learning_rate": 1.753353552728864e-05, "loss": 0.2728, "step": 59624 }, { "epoch": 3.6, "learning_rate": 1.7532026617174415e-05, "loss": 0.2286, "step": 59626 }, { "epoch": 3.6, "learning_rate": 1.753051770706019e-05, "loss": 0.3005, "step": 59628 }, { "epoch": 3.6, "learning_rate": 1.7529008796945968e-05, "loss": 0.3021, "step": 59630 }, { "epoch": 3.6, "learning_rate": 1.752749988683174e-05, "loss": 0.2805, "step": 59632 }, { "epoch": 3.6, "learning_rate": 1.7525990976717517e-05, "loss": 0.2152, "step": 59634 }, { "epoch": 3.6, "learning_rate": 1.7524482066603294e-05, "loss": 0.4456, "step": 59636 }, { "epoch": 3.6, "learning_rate": 1.752297315648907e-05, "loss": 0.3734, "step": 59638 }, { "epoch": 3.6, "learning_rate": 1.7521464246374843e-05, "loss": 0.3425, "step": 59640 }, { "epoch": 3.6, "learning_rate": 1.751995533626062e-05, "loss": 0.2657, "step": 59642 }, { "epoch": 3.6, "learning_rate": 1.7518446426146396e-05, "loss": 0.4222, "step": 59644 }, { "epoch": 3.6, "learning_rate": 1.751693751603217e-05, "loss": 0.4093, "step": 59646 }, { "epoch": 3.6, "learning_rate": 1.7515428605917946e-05, "loss": 0.3936, "step": 59648 }, { "epoch": 3.6, "learning_rate": 1.7513919695803723e-05, "loss": 0.2734, "step": 59650 }, { "epoch": 3.6, "learning_rate": 1.75124107856895e-05, "loss": 0.3892, "step": 59652 }, { "epoch": 3.6, "learning_rate": 1.7510901875575272e-05, "loss": 0.4165, "step": 59654 }, { "epoch": 3.6, "learning_rate": 1.750939296546105e-05, "loss": 0.3234, "step": 59656 }, { "epoch": 3.6, "learning_rate": 1.7507884055346825e-05, "loss": 0.3404, "step": 59658 }, { "epoch": 3.6, "learning_rate": 1.75063751452326e-05, "loss": 0.5343, "step": 59660 }, { "epoch": 3.6, "learning_rate": 1.7504866235118375e-05, "loss": 0.4141, "step": 59662 }, { "epoch": 3.6, "learning_rate": 1.7503357325004148e-05, "loss": 0.3313, "step": 59664 }, { "epoch": 3.6, "learning_rate": 1.7501848414889928e-05, "loss": 0.4212, "step": 59666 }, { "epoch": 3.6, "learning_rate": 1.75003395047757e-05, "loss": 0.4514, "step": 59668 }, { "epoch": 3.6, "learning_rate": 1.7498830594661478e-05, "loss": 0.4351, "step": 59670 }, { "epoch": 3.6, "learning_rate": 1.749732168454725e-05, "loss": 0.3387, "step": 59672 }, { "epoch": 3.6, "learning_rate": 1.749581277443303e-05, "loss": 0.4192, "step": 59674 }, { "epoch": 3.6, "learning_rate": 1.7494303864318804e-05, "loss": 0.4985, "step": 59676 }, { "epoch": 3.6, "learning_rate": 1.7492794954204577e-05, "loss": 0.375, "step": 59678 }, { "epoch": 3.6, "learning_rate": 1.7491286044090353e-05, "loss": 0.4557, "step": 59680 }, { "epoch": 3.6, "learning_rate": 1.748977713397613e-05, "loss": 0.3035, "step": 59682 }, { "epoch": 3.6, "learning_rate": 1.7488268223861906e-05, "loss": 0.459, "step": 59684 }, { "epoch": 3.6, "learning_rate": 1.748675931374768e-05, "loss": 0.2822, "step": 59686 }, { "epoch": 3.6, "learning_rate": 1.7485250403633456e-05, "loss": 0.4483, "step": 59688 }, { "epoch": 3.6, "learning_rate": 1.7483741493519233e-05, "loss": 0.3305, "step": 59690 }, { "epoch": 3.6, "learning_rate": 1.7482232583405006e-05, "loss": 0.3127, "step": 59692 }, { "epoch": 3.6, "learning_rate": 1.7480723673290782e-05, "loss": 0.3799, "step": 59694 }, { "epoch": 3.6, "learning_rate": 1.747921476317656e-05, "loss": 0.3662, "step": 59696 }, { "epoch": 3.6, "learning_rate": 1.7477705853062335e-05, "loss": 0.2743, "step": 59698 }, { "epoch": 3.6, "learning_rate": 1.747619694294811e-05, "loss": 0.3111, "step": 59700 }, { "epoch": 3.6, "learning_rate": 1.7474688032833885e-05, "loss": 0.4756, "step": 59702 }, { "epoch": 3.6, "learning_rate": 1.747317912271966e-05, "loss": 0.4041, "step": 59704 }, { "epoch": 3.6, "learning_rate": 1.7471670212605438e-05, "loss": 0.3533, "step": 59706 }, { "epoch": 3.6, "learning_rate": 1.747016130249121e-05, "loss": 0.3089, "step": 59708 }, { "epoch": 3.6, "learning_rate": 1.7468652392376984e-05, "loss": 0.464, "step": 59710 }, { "epoch": 3.6, "learning_rate": 1.7467143482262764e-05, "loss": 0.2912, "step": 59712 }, { "epoch": 3.6, "learning_rate": 1.7465634572148537e-05, "loss": 0.2858, "step": 59714 }, { "epoch": 3.6, "learning_rate": 1.7464125662034314e-05, "loss": 0.3449, "step": 59716 }, { "epoch": 3.6, "learning_rate": 1.7462616751920087e-05, "loss": 0.3907, "step": 59718 }, { "epoch": 3.6, "learning_rate": 1.7461107841805867e-05, "loss": 0.4428, "step": 59720 }, { "epoch": 3.6, "learning_rate": 1.745959893169164e-05, "loss": 0.2567, "step": 59722 }, { "epoch": 3.6, "learning_rate": 1.7458090021577416e-05, "loss": 0.2596, "step": 59724 }, { "epoch": 3.6, "learning_rate": 1.745658111146319e-05, "loss": 0.4003, "step": 59726 }, { "epoch": 3.61, "learning_rate": 1.7455072201348966e-05, "loss": 0.362, "step": 59728 }, { "epoch": 3.61, "learning_rate": 1.7453563291234742e-05, "loss": 0.3454, "step": 59730 }, { "epoch": 3.61, "learning_rate": 1.7452054381120516e-05, "loss": 0.3517, "step": 59732 }, { "epoch": 3.61, "learning_rate": 1.7450545471006292e-05, "loss": 0.4305, "step": 59734 }, { "epoch": 3.61, "learning_rate": 1.744903656089207e-05, "loss": 0.4544, "step": 59736 }, { "epoch": 3.61, "learning_rate": 1.7447527650777845e-05, "loss": 0.2845, "step": 59738 }, { "epoch": 3.61, "learning_rate": 1.7446018740663618e-05, "loss": 0.2419, "step": 59740 }, { "epoch": 3.61, "learning_rate": 1.7444509830549395e-05, "loss": 0.3757, "step": 59742 }, { "epoch": 3.61, "learning_rate": 1.744300092043517e-05, "loss": 0.328, "step": 59744 }, { "epoch": 3.61, "learning_rate": 1.7441492010320944e-05, "loss": 0.4818, "step": 59746 }, { "epoch": 3.61, "learning_rate": 1.743998310020672e-05, "loss": 0.4234, "step": 59748 }, { "epoch": 3.61, "learning_rate": 1.7438474190092497e-05, "loss": 0.2016, "step": 59750 }, { "epoch": 3.61, "learning_rate": 1.7436965279978274e-05, "loss": 0.2314, "step": 59752 }, { "epoch": 3.61, "learning_rate": 1.7435456369864047e-05, "loss": 0.5634, "step": 59754 }, { "epoch": 3.61, "learning_rate": 1.7433947459749824e-05, "loss": 0.3925, "step": 59756 }, { "epoch": 3.61, "learning_rate": 1.74324385496356e-05, "loss": 0.3015, "step": 59758 }, { "epoch": 3.61, "learning_rate": 1.7430929639521373e-05, "loss": 0.4309, "step": 59760 }, { "epoch": 3.61, "learning_rate": 1.742942072940715e-05, "loss": 0.441, "step": 59762 }, { "epoch": 3.61, "learning_rate": 1.7427911819292926e-05, "loss": 0.4128, "step": 59764 }, { "epoch": 3.61, "learning_rate": 1.7426402909178703e-05, "loss": 0.4548, "step": 59766 }, { "epoch": 3.61, "learning_rate": 1.7424893999064476e-05, "loss": 0.2614, "step": 59768 }, { "epoch": 3.61, "learning_rate": 1.7423385088950252e-05, "loss": 0.3573, "step": 59770 }, { "epoch": 3.61, "learning_rate": 1.742187617883603e-05, "loss": 0.4248, "step": 59772 }, { "epoch": 3.61, "learning_rate": 1.7420367268721802e-05, "loss": 0.3911, "step": 59774 }, { "epoch": 3.61, "learning_rate": 1.741885835860758e-05, "loss": 0.5522, "step": 59776 }, { "epoch": 3.61, "learning_rate": 1.741734944849335e-05, "loss": 0.3061, "step": 59778 }, { "epoch": 3.61, "learning_rate": 1.741584053837913e-05, "loss": 0.4051, "step": 59780 }, { "epoch": 3.61, "learning_rate": 1.7414331628264905e-05, "loss": 0.2341, "step": 59782 }, { "epoch": 3.61, "learning_rate": 1.741282271815068e-05, "loss": 0.4266, "step": 59784 }, { "epoch": 3.61, "learning_rate": 1.7411313808036454e-05, "loss": 0.2614, "step": 59786 }, { "epoch": 3.61, "learning_rate": 1.7409804897922234e-05, "loss": 0.3101, "step": 59788 }, { "epoch": 3.61, "learning_rate": 1.7408295987808007e-05, "loss": 0.4193, "step": 59790 }, { "epoch": 3.61, "learning_rate": 1.740678707769378e-05, "loss": 0.2825, "step": 59792 }, { "epoch": 3.61, "learning_rate": 1.7405278167579557e-05, "loss": 0.2478, "step": 59794 }, { "epoch": 3.61, "learning_rate": 1.7403769257465334e-05, "loss": 0.3083, "step": 59796 }, { "epoch": 3.61, "learning_rate": 1.740226034735111e-05, "loss": 0.4735, "step": 59798 }, { "epoch": 3.61, "learning_rate": 1.7400751437236883e-05, "loss": 0.7251, "step": 59800 }, { "epoch": 3.61, "learning_rate": 1.739924252712266e-05, "loss": 0.2724, "step": 59802 }, { "epoch": 3.61, "learning_rate": 1.7397733617008436e-05, "loss": 0.5223, "step": 59804 }, { "epoch": 3.61, "learning_rate": 1.7396224706894213e-05, "loss": 0.3412, "step": 59806 }, { "epoch": 3.61, "learning_rate": 1.7394715796779986e-05, "loss": 0.3556, "step": 59808 }, { "epoch": 3.61, "learning_rate": 1.7393206886665762e-05, "loss": 0.4763, "step": 59810 }, { "epoch": 3.61, "learning_rate": 1.739169797655154e-05, "loss": 0.3766, "step": 59812 }, { "epoch": 3.61, "learning_rate": 1.7390189066437312e-05, "loss": 0.4571, "step": 59814 }, { "epoch": 3.61, "learning_rate": 1.738868015632309e-05, "loss": 0.3924, "step": 59816 }, { "epoch": 3.61, "learning_rate": 1.7387171246208865e-05, "loss": 0.4148, "step": 59818 }, { "epoch": 3.61, "learning_rate": 1.738566233609464e-05, "loss": 0.2801, "step": 59820 }, { "epoch": 3.61, "learning_rate": 1.7384153425980415e-05, "loss": 0.4917, "step": 59822 }, { "epoch": 3.61, "learning_rate": 1.7382644515866188e-05, "loss": 0.3452, "step": 59824 }, { "epoch": 3.61, "learning_rate": 1.7381135605751968e-05, "loss": 0.4191, "step": 59826 }, { "epoch": 3.61, "learning_rate": 1.737962669563774e-05, "loss": 0.4495, "step": 59828 }, { "epoch": 3.61, "learning_rate": 1.7378117785523517e-05, "loss": 0.5093, "step": 59830 }, { "epoch": 3.61, "learning_rate": 1.737660887540929e-05, "loss": 0.2853, "step": 59832 }, { "epoch": 3.61, "learning_rate": 1.737509996529507e-05, "loss": 0.4917, "step": 59834 }, { "epoch": 3.61, "learning_rate": 1.7373591055180843e-05, "loss": 0.3571, "step": 59836 }, { "epoch": 3.61, "learning_rate": 1.737208214506662e-05, "loss": 0.5555, "step": 59838 }, { "epoch": 3.61, "learning_rate": 1.7370573234952393e-05, "loss": 0.3305, "step": 59840 }, { "epoch": 3.61, "learning_rate": 1.736906432483817e-05, "loss": 0.2918, "step": 59842 }, { "epoch": 3.61, "learning_rate": 1.7367555414723946e-05, "loss": 0.4527, "step": 59844 }, { "epoch": 3.61, "learning_rate": 1.736604650460972e-05, "loss": 0.3726, "step": 59846 }, { "epoch": 3.61, "learning_rate": 1.7364537594495496e-05, "loss": 0.35, "step": 59848 }, { "epoch": 3.61, "learning_rate": 1.7363028684381272e-05, "loss": 0.2896, "step": 59850 }, { "epoch": 3.61, "learning_rate": 1.736151977426705e-05, "loss": 0.3037, "step": 59852 }, { "epoch": 3.61, "learning_rate": 1.7360010864152822e-05, "loss": 0.437, "step": 59854 }, { "epoch": 3.61, "learning_rate": 1.73585019540386e-05, "loss": 0.4916, "step": 59856 }, { "epoch": 3.61, "learning_rate": 1.7356993043924375e-05, "loss": 0.2837, "step": 59858 }, { "epoch": 3.61, "learning_rate": 1.7355484133810148e-05, "loss": 0.2995, "step": 59860 }, { "epoch": 3.61, "learning_rate": 1.7353975223695925e-05, "loss": 0.3812, "step": 59862 }, { "epoch": 3.61, "learning_rate": 1.73524663135817e-05, "loss": 0.4598, "step": 59864 }, { "epoch": 3.61, "learning_rate": 1.7350957403467478e-05, "loss": 0.3367, "step": 59866 }, { "epoch": 3.61, "learning_rate": 1.734944849335325e-05, "loss": 0.2494, "step": 59868 }, { "epoch": 3.61, "learning_rate": 1.7347939583239027e-05, "loss": 0.4728, "step": 59870 }, { "epoch": 3.61, "learning_rate": 1.7346430673124804e-05, "loss": 0.2963, "step": 59872 }, { "epoch": 3.61, "learning_rate": 1.7344921763010577e-05, "loss": 0.3828, "step": 59874 }, { "epoch": 3.61, "learning_rate": 1.7343412852896353e-05, "loss": 0.3014, "step": 59876 }, { "epoch": 3.61, "learning_rate": 1.734190394278213e-05, "loss": 0.3401, "step": 59878 }, { "epoch": 3.61, "learning_rate": 1.7340395032667906e-05, "loss": 0.5111, "step": 59880 }, { "epoch": 3.61, "learning_rate": 1.733888612255368e-05, "loss": 0.415, "step": 59882 }, { "epoch": 3.61, "learning_rate": 1.7337377212439456e-05, "loss": 0.4829, "step": 59884 }, { "epoch": 3.61, "learning_rate": 1.7335868302325233e-05, "loss": 0.3878, "step": 59886 }, { "epoch": 3.61, "learning_rate": 1.733435939221101e-05, "loss": 0.3842, "step": 59888 }, { "epoch": 3.61, "learning_rate": 1.7332850482096782e-05, "loss": 0.2879, "step": 59890 }, { "epoch": 3.61, "learning_rate": 1.7331341571982555e-05, "loss": 0.4467, "step": 59892 }, { "epoch": 3.62, "learning_rate": 1.7329832661868335e-05, "loss": 0.2784, "step": 59894 }, { "epoch": 3.62, "learning_rate": 1.732832375175411e-05, "loss": 0.3817, "step": 59896 }, { "epoch": 3.62, "learning_rate": 1.7326814841639885e-05, "loss": 0.2819, "step": 59898 }, { "epoch": 3.62, "learning_rate": 1.7325305931525658e-05, "loss": 0.4811, "step": 59900 }, { "epoch": 3.62, "learning_rate": 1.7323797021411438e-05, "loss": 0.2543, "step": 59902 }, { "epoch": 3.62, "learning_rate": 1.732228811129721e-05, "loss": 0.4169, "step": 59904 }, { "epoch": 3.62, "learning_rate": 1.7320779201182984e-05, "loss": 0.3441, "step": 59906 }, { "epoch": 3.62, "learning_rate": 1.731927029106876e-05, "loss": 0.3324, "step": 59908 }, { "epoch": 3.62, "learning_rate": 1.7317761380954537e-05, "loss": 0.3734, "step": 59910 }, { "epoch": 3.62, "learning_rate": 1.7316252470840314e-05, "loss": 0.4398, "step": 59912 }, { "epoch": 3.62, "learning_rate": 1.7314743560726087e-05, "loss": 0.2934, "step": 59914 }, { "epoch": 3.62, "learning_rate": 1.7313234650611863e-05, "loss": 0.3111, "step": 59916 }, { "epoch": 3.62, "learning_rate": 1.731172574049764e-05, "loss": 0.4596, "step": 59918 }, { "epoch": 3.62, "learning_rate": 1.7310216830383416e-05, "loss": 0.1884, "step": 59920 }, { "epoch": 3.62, "learning_rate": 1.730870792026919e-05, "loss": 0.225, "step": 59922 }, { "epoch": 3.62, "learning_rate": 1.7307199010154966e-05, "loss": 0.4168, "step": 59924 }, { "epoch": 3.62, "learning_rate": 1.7305690100040743e-05, "loss": 0.3086, "step": 59926 }, { "epoch": 3.62, "learning_rate": 1.7304181189926516e-05, "loss": 0.3027, "step": 59928 }, { "epoch": 3.62, "learning_rate": 1.7302672279812292e-05, "loss": 0.3632, "step": 59930 }, { "epoch": 3.62, "learning_rate": 1.730116336969807e-05, "loss": 0.3775, "step": 59932 }, { "epoch": 3.62, "learning_rate": 1.7299654459583845e-05, "loss": 0.1975, "step": 59934 }, { "epoch": 3.62, "learning_rate": 1.729814554946962e-05, "loss": 0.4033, "step": 59936 }, { "epoch": 3.62, "learning_rate": 1.7296636639355395e-05, "loss": 0.3159, "step": 59938 }, { "epoch": 3.62, "learning_rate": 1.729512772924117e-05, "loss": 0.3454, "step": 59940 }, { "epoch": 3.62, "learning_rate": 1.7293618819126944e-05, "loss": 0.3482, "step": 59942 }, { "epoch": 3.62, "learning_rate": 1.729210990901272e-05, "loss": 0.3215, "step": 59944 }, { "epoch": 3.62, "learning_rate": 1.7290600998898494e-05, "loss": 0.3738, "step": 59946 }, { "epoch": 3.62, "learning_rate": 1.7289092088784274e-05, "loss": 0.462, "step": 59948 }, { "epoch": 3.62, "learning_rate": 1.7287583178670047e-05, "loss": 0.4562, "step": 59950 }, { "epoch": 3.62, "learning_rate": 1.7286074268555824e-05, "loss": 0.4287, "step": 59952 }, { "epoch": 3.62, "learning_rate": 1.7284565358441597e-05, "loss": 0.3671, "step": 59954 }, { "epoch": 3.62, "learning_rate": 1.7283056448327373e-05, "loss": 0.4486, "step": 59956 }, { "epoch": 3.62, "learning_rate": 1.728154753821315e-05, "loss": 0.4215, "step": 59958 }, { "epoch": 3.62, "learning_rate": 1.7280038628098923e-05, "loss": 0.2636, "step": 59960 }, { "epoch": 3.62, "learning_rate": 1.72785297179847e-05, "loss": 0.3734, "step": 59962 }, { "epoch": 3.62, "learning_rate": 1.7277020807870476e-05, "loss": 0.2897, "step": 59964 }, { "epoch": 3.62, "learning_rate": 1.7275511897756252e-05, "loss": 0.2101, "step": 59966 }, { "epoch": 3.62, "learning_rate": 1.7274002987642026e-05, "loss": 0.501, "step": 59968 }, { "epoch": 3.62, "learning_rate": 1.7272494077527802e-05, "loss": 0.5376, "step": 59970 }, { "epoch": 3.62, "learning_rate": 1.727098516741358e-05, "loss": 0.4319, "step": 59972 }, { "epoch": 3.62, "learning_rate": 1.7269476257299352e-05, "loss": 0.5609, "step": 59974 }, { "epoch": 3.62, "learning_rate": 1.7267967347185128e-05, "loss": 0.3256, "step": 59976 }, { "epoch": 3.62, "learning_rate": 1.7266458437070905e-05, "loss": 0.5034, "step": 59978 }, { "epoch": 3.62, "learning_rate": 1.726494952695668e-05, "loss": 0.4998, "step": 59980 }, { "epoch": 3.62, "learning_rate": 1.7263440616842454e-05, "loss": 0.3804, "step": 59982 }, { "epoch": 3.62, "learning_rate": 1.726193170672823e-05, "loss": 0.3068, "step": 59984 }, { "epoch": 3.62, "learning_rate": 1.7260422796614007e-05, "loss": 0.4331, "step": 59986 }, { "epoch": 3.62, "learning_rate": 1.725891388649978e-05, "loss": 0.3189, "step": 59988 }, { "epoch": 3.62, "learning_rate": 1.7257404976385557e-05, "loss": 0.3247, "step": 59990 }, { "epoch": 3.62, "learning_rate": 1.7255896066271334e-05, "loss": 0.2424, "step": 59992 }, { "epoch": 3.62, "learning_rate": 1.725438715615711e-05, "loss": 0.4663, "step": 59994 }, { "epoch": 3.62, "learning_rate": 1.7252878246042883e-05, "loss": 0.4784, "step": 59996 }, { "epoch": 3.62, "learning_rate": 1.725136933592866e-05, "loss": 0.3613, "step": 59998 }, { "epoch": 3.62, "learning_rate": 1.7249860425814436e-05, "loss": 0.3658, "step": 60000 }, { "epoch": 3.62, "learning_rate": 1.7248351515700213e-05, "loss": 0.2471, "step": 60002 }, { "epoch": 3.62, "learning_rate": 1.7246842605585986e-05, "loss": 0.2893, "step": 60004 }, { "epoch": 3.62, "learning_rate": 1.724533369547176e-05, "loss": 0.3441, "step": 60006 }, { "epoch": 3.62, "learning_rate": 1.724382478535754e-05, "loss": 0.4417, "step": 60008 }, { "epoch": 3.62, "learning_rate": 1.7242315875243312e-05, "loss": 0.2705, "step": 60010 }, { "epoch": 3.62, "learning_rate": 1.724080696512909e-05, "loss": 0.2728, "step": 60012 }, { "epoch": 3.62, "learning_rate": 1.7239298055014862e-05, "loss": 0.2676, "step": 60014 }, { "epoch": 3.62, "learning_rate": 1.723778914490064e-05, "loss": 0.2422, "step": 60016 }, { "epoch": 3.62, "learning_rate": 1.7236280234786415e-05, "loss": 0.3254, "step": 60018 }, { "epoch": 3.62, "learning_rate": 1.723477132467219e-05, "loss": 0.4079, "step": 60020 }, { "epoch": 3.62, "learning_rate": 1.7233262414557964e-05, "loss": 0.2907, "step": 60022 }, { "epoch": 3.62, "learning_rate": 1.723175350444374e-05, "loss": 0.3526, "step": 60024 }, { "epoch": 3.62, "learning_rate": 1.7230244594329517e-05, "loss": 0.2959, "step": 60026 }, { "epoch": 3.62, "learning_rate": 1.722873568421529e-05, "loss": 0.3059, "step": 60028 }, { "epoch": 3.62, "learning_rate": 1.7227226774101067e-05, "loss": 0.5615, "step": 60030 }, { "epoch": 3.62, "learning_rate": 1.7225717863986844e-05, "loss": 0.2825, "step": 60032 }, { "epoch": 3.62, "learning_rate": 1.722420895387262e-05, "loss": 0.3788, "step": 60034 }, { "epoch": 3.62, "learning_rate": 1.7222700043758393e-05, "loss": 0.4153, "step": 60036 }, { "epoch": 3.62, "learning_rate": 1.722119113364417e-05, "loss": 0.5243, "step": 60038 }, { "epoch": 3.62, "learning_rate": 1.7219682223529946e-05, "loss": 0.3613, "step": 60040 }, { "epoch": 3.62, "learning_rate": 1.721817331341572e-05, "loss": 0.2718, "step": 60042 }, { "epoch": 3.62, "learning_rate": 1.7216664403301496e-05, "loss": 0.3804, "step": 60044 }, { "epoch": 3.62, "learning_rate": 1.7215155493187272e-05, "loss": 0.3485, "step": 60046 }, { "epoch": 3.62, "learning_rate": 1.721364658307305e-05, "loss": 0.2497, "step": 60048 }, { "epoch": 3.62, "learning_rate": 1.7212137672958822e-05, "loss": 0.3098, "step": 60050 }, { "epoch": 3.62, "learning_rate": 1.72106287628446e-05, "loss": 0.3439, "step": 60052 }, { "epoch": 3.62, "learning_rate": 1.7209119852730375e-05, "loss": 0.3296, "step": 60054 }, { "epoch": 3.62, "learning_rate": 1.7207610942616148e-05, "loss": 0.3127, "step": 60056 }, { "epoch": 3.62, "learning_rate": 1.7206102032501925e-05, "loss": 0.3007, "step": 60058 }, { "epoch": 3.63, "learning_rate": 1.7204593122387698e-05, "loss": 0.3484, "step": 60060 }, { "epoch": 3.63, "learning_rate": 1.7203084212273478e-05, "loss": 0.3399, "step": 60062 }, { "epoch": 3.63, "learning_rate": 1.720157530215925e-05, "loss": 0.3314, "step": 60064 }, { "epoch": 3.63, "learning_rate": 1.7200066392045027e-05, "loss": 0.2983, "step": 60066 }, { "epoch": 3.63, "learning_rate": 1.71985574819308e-05, "loss": 0.4062, "step": 60068 }, { "epoch": 3.63, "learning_rate": 1.7197048571816577e-05, "loss": 0.353, "step": 60070 }, { "epoch": 3.63, "learning_rate": 1.7195539661702354e-05, "loss": 0.3262, "step": 60072 }, { "epoch": 3.63, "learning_rate": 1.7194030751588127e-05, "loss": 0.2727, "step": 60074 }, { "epoch": 3.63, "learning_rate": 1.7192521841473903e-05, "loss": 0.4471, "step": 60076 }, { "epoch": 3.63, "learning_rate": 1.719101293135968e-05, "loss": 0.4886, "step": 60078 }, { "epoch": 3.63, "learning_rate": 1.7189504021245456e-05, "loss": 0.2313, "step": 60080 }, { "epoch": 3.63, "learning_rate": 1.718799511113123e-05, "loss": 0.4146, "step": 60082 }, { "epoch": 3.63, "learning_rate": 1.7186486201017006e-05, "loss": 0.4543, "step": 60084 }, { "epoch": 3.63, "learning_rate": 1.7184977290902782e-05, "loss": 0.4579, "step": 60086 }, { "epoch": 3.63, "learning_rate": 1.7183468380788555e-05, "loss": 0.3778, "step": 60088 }, { "epoch": 3.63, "learning_rate": 1.7181959470674332e-05, "loss": 0.4268, "step": 60090 }, { "epoch": 3.63, "learning_rate": 1.718045056056011e-05, "loss": 0.3998, "step": 60092 }, { "epoch": 3.63, "learning_rate": 1.7178941650445885e-05, "loss": 0.265, "step": 60094 }, { "epoch": 3.63, "learning_rate": 1.7177432740331658e-05, "loss": 0.2698, "step": 60096 }, { "epoch": 3.63, "learning_rate": 1.7175923830217435e-05, "loss": 0.4582, "step": 60098 }, { "epoch": 3.63, "learning_rate": 1.717441492010321e-05, "loss": 0.3178, "step": 60100 }, { "epoch": 3.63, "learning_rate": 1.7172906009988988e-05, "loss": 0.3324, "step": 60102 }, { "epoch": 3.63, "learning_rate": 1.717139709987476e-05, "loss": 0.4608, "step": 60104 }, { "epoch": 3.63, "learning_rate": 1.7169888189760537e-05, "loss": 0.4622, "step": 60106 }, { "epoch": 3.63, "learning_rate": 1.7168379279646314e-05, "loss": 0.4476, "step": 60108 }, { "epoch": 3.63, "learning_rate": 1.7166870369532087e-05, "loss": 0.2473, "step": 60110 }, { "epoch": 3.63, "learning_rate": 1.7165361459417863e-05, "loss": 0.3691, "step": 60112 }, { "epoch": 3.63, "learning_rate": 1.716385254930364e-05, "loss": 0.357, "step": 60114 }, { "epoch": 3.63, "learning_rate": 1.7162343639189416e-05, "loss": 0.3198, "step": 60116 }, { "epoch": 3.63, "learning_rate": 1.716083472907519e-05, "loss": 0.252, "step": 60118 }, { "epoch": 3.63, "learning_rate": 1.7159325818960963e-05, "loss": 0.5553, "step": 60120 }, { "epoch": 3.63, "learning_rate": 1.7157816908846743e-05, "loss": 0.3389, "step": 60122 }, { "epoch": 3.63, "learning_rate": 1.7156307998732516e-05, "loss": 0.3252, "step": 60124 }, { "epoch": 3.63, "learning_rate": 1.7154799088618292e-05, "loss": 0.4693, "step": 60126 }, { "epoch": 3.63, "learning_rate": 1.7153290178504065e-05, "loss": 0.3821, "step": 60128 }, { "epoch": 3.63, "learning_rate": 1.7151781268389845e-05, "loss": 0.3542, "step": 60130 }, { "epoch": 3.63, "learning_rate": 1.715027235827562e-05, "loss": 0.3817, "step": 60132 }, { "epoch": 3.63, "learning_rate": 1.7148763448161395e-05, "loss": 0.4142, "step": 60134 }, { "epoch": 3.63, "learning_rate": 1.7147254538047168e-05, "loss": 0.2732, "step": 60136 }, { "epoch": 3.63, "learning_rate": 1.7145745627932945e-05, "loss": 0.4758, "step": 60138 }, { "epoch": 3.63, "learning_rate": 1.714423671781872e-05, "loss": 0.3285, "step": 60140 }, { "epoch": 3.63, "learning_rate": 1.7142727807704494e-05, "loss": 0.2508, "step": 60142 }, { "epoch": 3.63, "learning_rate": 1.714121889759027e-05, "loss": 0.4016, "step": 60144 }, { "epoch": 3.63, "learning_rate": 1.7139709987476047e-05, "loss": 0.4164, "step": 60146 }, { "epoch": 3.63, "learning_rate": 1.7138201077361824e-05, "loss": 0.2452, "step": 60148 }, { "epoch": 3.63, "learning_rate": 1.7136692167247597e-05, "loss": 0.3608, "step": 60150 }, { "epoch": 3.63, "learning_rate": 1.7135183257133373e-05, "loss": 0.4907, "step": 60152 }, { "epoch": 3.63, "learning_rate": 1.713367434701915e-05, "loss": 0.3611, "step": 60154 }, { "epoch": 3.63, "learning_rate": 1.7132165436904923e-05, "loss": 0.3403, "step": 60156 }, { "epoch": 3.63, "learning_rate": 1.71306565267907e-05, "loss": 0.389, "step": 60158 }, { "epoch": 3.63, "learning_rate": 1.7129147616676476e-05, "loss": 0.4162, "step": 60160 }, { "epoch": 3.63, "learning_rate": 1.7127638706562253e-05, "loss": 0.272, "step": 60162 }, { "epoch": 3.63, "learning_rate": 1.7126129796448026e-05, "loss": 0.3528, "step": 60164 }, { "epoch": 3.63, "learning_rate": 1.7124620886333802e-05, "loss": 0.435, "step": 60166 }, { "epoch": 3.63, "learning_rate": 1.712311197621958e-05, "loss": 0.2526, "step": 60168 }, { "epoch": 3.63, "learning_rate": 1.7121603066105352e-05, "loss": 0.2508, "step": 60170 }, { "epoch": 3.63, "learning_rate": 1.712009415599113e-05, "loss": 0.5043, "step": 60172 }, { "epoch": 3.63, "learning_rate": 1.71185852458769e-05, "loss": 0.3175, "step": 60174 }, { "epoch": 3.63, "learning_rate": 1.711707633576268e-05, "loss": 0.4738, "step": 60176 }, { "epoch": 3.63, "learning_rate": 1.7115567425648455e-05, "loss": 0.3873, "step": 60178 }, { "epoch": 3.63, "learning_rate": 1.711405851553423e-05, "loss": 0.5837, "step": 60180 }, { "epoch": 3.63, "learning_rate": 1.7112549605420004e-05, "loss": 0.3652, "step": 60182 }, { "epoch": 3.63, "learning_rate": 1.7111040695305784e-05, "loss": 0.2621, "step": 60184 }, { "epoch": 3.63, "learning_rate": 1.7109531785191557e-05, "loss": 0.3961, "step": 60186 }, { "epoch": 3.63, "learning_rate": 1.710802287507733e-05, "loss": 0.4627, "step": 60188 }, { "epoch": 3.63, "learning_rate": 1.7106513964963107e-05, "loss": 0.3301, "step": 60190 }, { "epoch": 3.63, "learning_rate": 1.7105005054848883e-05, "loss": 0.4723, "step": 60192 }, { "epoch": 3.63, "learning_rate": 1.710349614473466e-05, "loss": 0.4525, "step": 60194 }, { "epoch": 3.63, "learning_rate": 1.7101987234620433e-05, "loss": 0.4433, "step": 60196 }, { "epoch": 3.63, "learning_rate": 1.710047832450621e-05, "loss": 0.2917, "step": 60198 }, { "epoch": 3.63, "learning_rate": 1.7098969414391986e-05, "loss": 0.2209, "step": 60200 }, { "epoch": 3.63, "learning_rate": 1.709746050427776e-05, "loss": 0.2507, "step": 60202 }, { "epoch": 3.63, "learning_rate": 1.7095951594163536e-05, "loss": 0.2583, "step": 60204 }, { "epoch": 3.63, "learning_rate": 1.7094442684049312e-05, "loss": 0.4269, "step": 60206 }, { "epoch": 3.63, "learning_rate": 1.709293377393509e-05, "loss": 0.2314, "step": 60208 }, { "epoch": 3.63, "learning_rate": 1.7091424863820862e-05, "loss": 0.3353, "step": 60210 }, { "epoch": 3.63, "learning_rate": 1.7089915953706638e-05, "loss": 0.3742, "step": 60212 }, { "epoch": 3.63, "learning_rate": 1.7088407043592415e-05, "loss": 0.3679, "step": 60214 }, { "epoch": 3.63, "learning_rate": 1.708689813347819e-05, "loss": 0.312, "step": 60216 }, { "epoch": 3.63, "learning_rate": 1.7085389223363964e-05, "loss": 0.3548, "step": 60218 }, { "epoch": 3.63, "learning_rate": 1.708388031324974e-05, "loss": 0.1945, "step": 60220 }, { "epoch": 3.63, "learning_rate": 1.7082371403135517e-05, "loss": 0.4723, "step": 60222 }, { "epoch": 3.63, "learning_rate": 1.708086249302129e-05, "loss": 0.4128, "step": 60224 }, { "epoch": 3.64, "learning_rate": 1.7079353582907067e-05, "loss": 0.4749, "step": 60226 }, { "epoch": 3.64, "learning_rate": 1.7077844672792844e-05, "loss": 0.5223, "step": 60228 }, { "epoch": 3.64, "learning_rate": 1.707633576267862e-05, "loss": 0.1967, "step": 60230 }, { "epoch": 3.64, "learning_rate": 1.7074826852564393e-05, "loss": 0.335, "step": 60232 }, { "epoch": 3.64, "learning_rate": 1.7073317942450166e-05, "loss": 0.4474, "step": 60234 }, { "epoch": 3.64, "learning_rate": 1.7071809032335946e-05, "loss": 0.3597, "step": 60236 }, { "epoch": 3.64, "learning_rate": 1.707030012222172e-05, "loss": 0.4018, "step": 60238 }, { "epoch": 3.64, "learning_rate": 1.7068791212107496e-05, "loss": 0.3678, "step": 60240 }, { "epoch": 3.64, "learning_rate": 1.706728230199327e-05, "loss": 0.3419, "step": 60242 }, { "epoch": 3.64, "learning_rate": 1.706577339187905e-05, "loss": 0.5193, "step": 60244 }, { "epoch": 3.64, "learning_rate": 1.7064264481764822e-05, "loss": 0.3053, "step": 60246 }, { "epoch": 3.64, "learning_rate": 1.70627555716506e-05, "loss": 0.3456, "step": 60248 }, { "epoch": 3.64, "learning_rate": 1.7061246661536372e-05, "loss": 0.4553, "step": 60250 }, { "epoch": 3.64, "learning_rate": 1.7059737751422148e-05, "loss": 0.4088, "step": 60252 }, { "epoch": 3.64, "learning_rate": 1.7058228841307925e-05, "loss": 0.2437, "step": 60254 }, { "epoch": 3.64, "learning_rate": 1.7056719931193698e-05, "loss": 0.5387, "step": 60256 }, { "epoch": 3.64, "learning_rate": 1.7055211021079474e-05, "loss": 0.4398, "step": 60258 }, { "epoch": 3.64, "learning_rate": 1.705370211096525e-05, "loss": 0.5336, "step": 60260 }, { "epoch": 3.64, "learning_rate": 1.7052193200851027e-05, "loss": 0.316, "step": 60262 }, { "epoch": 3.64, "learning_rate": 1.70506842907368e-05, "loss": 0.3655, "step": 60264 }, { "epoch": 3.64, "learning_rate": 1.7049175380622577e-05, "loss": 0.3914, "step": 60266 }, { "epoch": 3.64, "learning_rate": 1.7047666470508354e-05, "loss": 0.4836, "step": 60268 }, { "epoch": 3.64, "learning_rate": 1.7046157560394127e-05, "loss": 0.3598, "step": 60270 }, { "epoch": 3.64, "learning_rate": 1.7044648650279903e-05, "loss": 0.2797, "step": 60272 }, { "epoch": 3.64, "learning_rate": 1.704313974016568e-05, "loss": 0.3745, "step": 60274 }, { "epoch": 3.64, "learning_rate": 1.7041630830051456e-05, "loss": 0.4249, "step": 60276 }, { "epoch": 3.64, "learning_rate": 1.704012191993723e-05, "loss": 0.3544, "step": 60278 }, { "epoch": 3.64, "learning_rate": 1.7038613009823006e-05, "loss": 0.3786, "step": 60280 }, { "epoch": 3.64, "learning_rate": 1.7037104099708782e-05, "loss": 0.359, "step": 60282 }, { "epoch": 3.64, "learning_rate": 1.7035595189594556e-05, "loss": 0.4514, "step": 60284 }, { "epoch": 3.64, "learning_rate": 1.7034086279480332e-05, "loss": 0.2915, "step": 60286 }, { "epoch": 3.64, "learning_rate": 1.7032577369366105e-05, "loss": 0.5437, "step": 60288 }, { "epoch": 3.64, "learning_rate": 1.7031068459251885e-05, "loss": 0.3818, "step": 60290 }, { "epoch": 3.64, "learning_rate": 1.7029559549137658e-05, "loss": 0.4922, "step": 60292 }, { "epoch": 3.64, "learning_rate": 1.7028050639023435e-05, "loss": 0.4077, "step": 60294 }, { "epoch": 3.64, "learning_rate": 1.7026541728909208e-05, "loss": 0.209, "step": 60296 }, { "epoch": 3.64, "learning_rate": 1.7025032818794988e-05, "loss": 0.3574, "step": 60298 }, { "epoch": 3.64, "learning_rate": 1.702352390868076e-05, "loss": 0.4166, "step": 60300 }, { "epoch": 3.64, "learning_rate": 1.7022014998566534e-05, "loss": 0.3914, "step": 60302 }, { "epoch": 3.64, "learning_rate": 1.702050608845231e-05, "loss": 0.2816, "step": 60304 }, { "epoch": 3.64, "learning_rate": 1.7018997178338087e-05, "loss": 0.4159, "step": 60306 }, { "epoch": 3.64, "learning_rate": 1.7017488268223864e-05, "loss": 0.5995, "step": 60308 }, { "epoch": 3.64, "learning_rate": 1.7015979358109637e-05, "loss": 0.35, "step": 60310 }, { "epoch": 3.64, "learning_rate": 1.7014470447995413e-05, "loss": 0.3375, "step": 60312 }, { "epoch": 3.64, "learning_rate": 1.701296153788119e-05, "loss": 0.3624, "step": 60314 }, { "epoch": 3.64, "learning_rate": 1.7011452627766963e-05, "loss": 0.2666, "step": 60316 }, { "epoch": 3.64, "learning_rate": 1.700994371765274e-05, "loss": 0.4193, "step": 60318 }, { "epoch": 3.64, "learning_rate": 1.7008434807538516e-05, "loss": 0.2928, "step": 60320 }, { "epoch": 3.64, "learning_rate": 1.7006925897424292e-05, "loss": 0.3875, "step": 60322 }, { "epoch": 3.64, "learning_rate": 1.7005416987310065e-05, "loss": 0.5083, "step": 60324 }, { "epoch": 3.64, "learning_rate": 1.7003908077195842e-05, "loss": 0.2872, "step": 60326 }, { "epoch": 3.64, "learning_rate": 1.700239916708162e-05, "loss": 0.2795, "step": 60328 }, { "epoch": 3.64, "learning_rate": 1.7000890256967395e-05, "loss": 0.514, "step": 60330 }, { "epoch": 3.64, "learning_rate": 1.6999381346853168e-05, "loss": 0.3102, "step": 60332 }, { "epoch": 3.64, "learning_rate": 1.6997872436738945e-05, "loss": 0.3877, "step": 60334 }, { "epoch": 3.64, "learning_rate": 1.699636352662472e-05, "loss": 0.3159, "step": 60336 }, { "epoch": 3.64, "learning_rate": 1.6994854616510494e-05, "loss": 0.3458, "step": 60338 }, { "epoch": 3.64, "learning_rate": 1.699334570639627e-05, "loss": 0.2564, "step": 60340 }, { "epoch": 3.64, "learning_rate": 1.6991836796282047e-05, "loss": 0.32, "step": 60342 }, { "epoch": 3.64, "learning_rate": 1.6990327886167824e-05, "loss": 0.3737, "step": 60344 }, { "epoch": 3.64, "learning_rate": 1.6988818976053597e-05, "loss": 0.3691, "step": 60346 }, { "epoch": 3.64, "learning_rate": 1.6987310065939373e-05, "loss": 0.3547, "step": 60348 }, { "epoch": 3.64, "learning_rate": 1.698580115582515e-05, "loss": 0.2668, "step": 60350 }, { "epoch": 3.64, "learning_rate": 1.6984292245710923e-05, "loss": 0.4726, "step": 60352 }, { "epoch": 3.64, "learning_rate": 1.69827833355967e-05, "loss": 0.4406, "step": 60354 }, { "epoch": 3.64, "learning_rate": 1.6981274425482473e-05, "loss": 0.4942, "step": 60356 }, { "epoch": 3.64, "learning_rate": 1.6979765515368253e-05, "loss": 0.2268, "step": 60358 }, { "epoch": 3.64, "learning_rate": 1.6978256605254026e-05, "loss": 0.389, "step": 60360 }, { "epoch": 3.64, "learning_rate": 1.6976747695139802e-05, "loss": 0.3565, "step": 60362 }, { "epoch": 3.64, "learning_rate": 1.6975238785025575e-05, "loss": 0.4905, "step": 60364 }, { "epoch": 3.64, "learning_rate": 1.6973729874911352e-05, "loss": 0.3466, "step": 60366 }, { "epoch": 3.64, "learning_rate": 1.697222096479713e-05, "loss": 0.4247, "step": 60368 }, { "epoch": 3.64, "learning_rate": 1.69707120546829e-05, "loss": 0.2533, "step": 60370 }, { "epoch": 3.64, "learning_rate": 1.6969203144568678e-05, "loss": 0.399, "step": 60372 }, { "epoch": 3.64, "learning_rate": 1.6967694234454455e-05, "loss": 0.409, "step": 60374 }, { "epoch": 3.64, "learning_rate": 1.696618532434023e-05, "loss": 0.2422, "step": 60376 }, { "epoch": 3.64, "learning_rate": 1.6964676414226004e-05, "loss": 0.4337, "step": 60378 }, { "epoch": 3.64, "learning_rate": 1.696316750411178e-05, "loss": 0.3134, "step": 60380 }, { "epoch": 3.64, "learning_rate": 1.6961658593997557e-05, "loss": 0.4625, "step": 60382 }, { "epoch": 3.64, "learning_rate": 1.696014968388333e-05, "loss": 0.6265, "step": 60384 }, { "epoch": 3.64, "learning_rate": 1.6958640773769107e-05, "loss": 0.2393, "step": 60386 }, { "epoch": 3.64, "learning_rate": 1.6957131863654883e-05, "loss": 0.5173, "step": 60388 }, { "epoch": 3.64, "learning_rate": 1.695562295354066e-05, "loss": 0.2613, "step": 60390 }, { "epoch": 3.65, "learning_rate": 1.6954114043426433e-05, "loss": 0.3753, "step": 60392 }, { "epoch": 3.65, "learning_rate": 1.695260513331221e-05, "loss": 0.2813, "step": 60394 }, { "epoch": 3.65, "learning_rate": 1.6951096223197986e-05, "loss": 0.3615, "step": 60396 }, { "epoch": 3.65, "learning_rate": 1.6949587313083763e-05, "loss": 0.3451, "step": 60398 }, { "epoch": 3.65, "learning_rate": 1.6948078402969536e-05, "loss": 0.3288, "step": 60400 }, { "epoch": 3.65, "learning_rate": 1.694656949285531e-05, "loss": 0.3032, "step": 60402 }, { "epoch": 3.65, "learning_rate": 1.694506058274109e-05, "loss": 0.4135, "step": 60404 }, { "epoch": 3.65, "learning_rate": 1.6943551672626862e-05, "loss": 0.1649, "step": 60406 }, { "epoch": 3.65, "learning_rate": 1.694204276251264e-05, "loss": 0.3794, "step": 60408 }, { "epoch": 3.65, "learning_rate": 1.694053385239841e-05, "loss": 0.4479, "step": 60410 }, { "epoch": 3.65, "learning_rate": 1.693902494228419e-05, "loss": 0.482, "step": 60412 }, { "epoch": 3.65, "learning_rate": 1.6937516032169965e-05, "loss": 0.3677, "step": 60414 }, { "epoch": 3.65, "learning_rate": 1.6936007122055738e-05, "loss": 0.4112, "step": 60416 }, { "epoch": 3.65, "learning_rate": 1.6934498211941514e-05, "loss": 0.3255, "step": 60418 }, { "epoch": 3.65, "learning_rate": 1.693298930182729e-05, "loss": 0.2948, "step": 60420 }, { "epoch": 3.65, "learning_rate": 1.6931480391713067e-05, "loss": 0.2945, "step": 60422 }, { "epoch": 3.65, "learning_rate": 1.692997148159884e-05, "loss": 0.4939, "step": 60424 }, { "epoch": 3.65, "learning_rate": 1.6928462571484617e-05, "loss": 0.5639, "step": 60426 }, { "epoch": 3.65, "learning_rate": 1.6926953661370393e-05, "loss": 0.2988, "step": 60428 }, { "epoch": 3.65, "learning_rate": 1.692544475125617e-05, "loss": 0.3459, "step": 60430 }, { "epoch": 3.65, "learning_rate": 1.6923935841141943e-05, "loss": 0.2528, "step": 60432 }, { "epoch": 3.65, "learning_rate": 1.692242693102772e-05, "loss": 0.2196, "step": 60434 }, { "epoch": 3.65, "learning_rate": 1.6920918020913496e-05, "loss": 0.3205, "step": 60436 }, { "epoch": 3.65, "learning_rate": 1.691940911079927e-05, "loss": 0.3645, "step": 60438 }, { "epoch": 3.65, "learning_rate": 1.6917900200685046e-05, "loss": 0.3452, "step": 60440 }, { "epoch": 3.65, "learning_rate": 1.6916391290570822e-05, "loss": 0.3866, "step": 60442 }, { "epoch": 3.65, "learning_rate": 1.69148823804566e-05, "loss": 0.3211, "step": 60444 }, { "epoch": 3.65, "learning_rate": 1.6913373470342372e-05, "loss": 0.4402, "step": 60446 }, { "epoch": 3.65, "learning_rate": 1.691186456022815e-05, "loss": 0.4196, "step": 60448 }, { "epoch": 3.65, "learning_rate": 1.6910355650113925e-05, "loss": 0.4304, "step": 60450 }, { "epoch": 3.65, "learning_rate": 1.6908846739999698e-05, "loss": 0.2181, "step": 60452 }, { "epoch": 3.65, "learning_rate": 1.6907337829885474e-05, "loss": 0.4366, "step": 60454 }, { "epoch": 3.65, "learning_rate": 1.690582891977125e-05, "loss": 0.3638, "step": 60456 }, { "epoch": 3.65, "learning_rate": 1.6904320009657028e-05, "loss": 0.5042, "step": 60458 }, { "epoch": 3.65, "learning_rate": 1.69028110995428e-05, "loss": 0.3656, "step": 60460 }, { "epoch": 3.65, "learning_rate": 1.6901302189428577e-05, "loss": 0.2938, "step": 60462 }, { "epoch": 3.65, "learning_rate": 1.6899793279314354e-05, "loss": 0.2686, "step": 60464 }, { "epoch": 3.65, "learning_rate": 1.6898284369200127e-05, "loss": 0.319, "step": 60466 }, { "epoch": 3.65, "learning_rate": 1.6896775459085903e-05, "loss": 0.2231, "step": 60468 }, { "epoch": 3.65, "learning_rate": 1.6895266548971676e-05, "loss": 0.4455, "step": 60470 }, { "epoch": 3.65, "learning_rate": 1.6893757638857456e-05, "loss": 0.4969, "step": 60472 }, { "epoch": 3.65, "learning_rate": 1.689224872874323e-05, "loss": 0.463, "step": 60474 }, { "epoch": 3.65, "learning_rate": 1.6890739818629006e-05, "loss": 0.4203, "step": 60476 }, { "epoch": 3.65, "learning_rate": 1.688923090851478e-05, "loss": 0.4357, "step": 60478 }, { "epoch": 3.65, "learning_rate": 1.688772199840056e-05, "loss": 0.3386, "step": 60480 }, { "epoch": 3.65, "learning_rate": 1.6886213088286332e-05, "loss": 0.5921, "step": 60482 }, { "epoch": 3.65, "learning_rate": 1.6884704178172105e-05, "loss": 0.38, "step": 60484 }, { "epoch": 3.65, "learning_rate": 1.6883195268057882e-05, "loss": 0.3339, "step": 60486 }, { "epoch": 3.65, "learning_rate": 1.6881686357943658e-05, "loss": 0.3344, "step": 60488 }, { "epoch": 3.65, "learning_rate": 1.6880177447829435e-05, "loss": 0.2174, "step": 60490 }, { "epoch": 3.65, "learning_rate": 1.6878668537715208e-05, "loss": 0.4367, "step": 60492 }, { "epoch": 3.65, "learning_rate": 1.6877159627600984e-05, "loss": 0.2901, "step": 60494 }, { "epoch": 3.65, "learning_rate": 1.687565071748676e-05, "loss": 0.3338, "step": 60496 }, { "epoch": 3.65, "learning_rate": 1.6874141807372534e-05, "loss": 0.4814, "step": 60498 }, { "epoch": 3.65, "learning_rate": 1.687263289725831e-05, "loss": 0.4708, "step": 60500 }, { "epoch": 3.65, "learning_rate": 1.6871123987144087e-05, "loss": 0.1829, "step": 60502 }, { "epoch": 3.65, "learning_rate": 1.6869615077029864e-05, "loss": 0.4168, "step": 60504 }, { "epoch": 3.65, "learning_rate": 1.6868106166915637e-05, "loss": 0.4245, "step": 60506 }, { "epoch": 3.65, "learning_rate": 1.6866597256801413e-05, "loss": 0.3778, "step": 60508 }, { "epoch": 3.65, "learning_rate": 1.686508834668719e-05, "loss": 0.3761, "step": 60510 }, { "epoch": 3.65, "learning_rate": 1.6863579436572966e-05, "loss": 0.3066, "step": 60512 }, { "epoch": 3.65, "learning_rate": 1.686207052645874e-05, "loss": 0.3886, "step": 60514 }, { "epoch": 3.65, "learning_rate": 1.6860561616344513e-05, "loss": 0.3727, "step": 60516 }, { "epoch": 3.65, "learning_rate": 1.6859052706230292e-05, "loss": 0.2176, "step": 60518 }, { "epoch": 3.65, "learning_rate": 1.6857543796116066e-05, "loss": 0.3883, "step": 60520 }, { "epoch": 3.65, "learning_rate": 1.6856034886001842e-05, "loss": 0.4132, "step": 60522 }, { "epoch": 3.65, "learning_rate": 1.6854525975887615e-05, "loss": 0.3054, "step": 60524 }, { "epoch": 3.65, "learning_rate": 1.6853017065773395e-05, "loss": 0.4616, "step": 60526 }, { "epoch": 3.65, "learning_rate": 1.6851508155659168e-05, "loss": 0.3884, "step": 60528 }, { "epoch": 3.65, "learning_rate": 1.684999924554494e-05, "loss": 0.309, "step": 60530 }, { "epoch": 3.65, "learning_rate": 1.6848490335430718e-05, "loss": 0.3382, "step": 60532 }, { "epoch": 3.65, "learning_rate": 1.6846981425316494e-05, "loss": 0.334, "step": 60534 }, { "epoch": 3.65, "learning_rate": 1.684547251520227e-05, "loss": 0.387, "step": 60536 }, { "epoch": 3.65, "learning_rate": 1.6843963605088044e-05, "loss": 0.2838, "step": 60538 }, { "epoch": 3.65, "learning_rate": 1.684245469497382e-05, "loss": 0.3124, "step": 60540 }, { "epoch": 3.65, "learning_rate": 1.6840945784859597e-05, "loss": 0.275, "step": 60542 }, { "epoch": 3.65, "learning_rate": 1.6839436874745374e-05, "loss": 0.4334, "step": 60544 }, { "epoch": 3.65, "learning_rate": 1.6837927964631147e-05, "loss": 0.4182, "step": 60546 }, { "epoch": 3.65, "learning_rate": 1.6836419054516923e-05, "loss": 0.3367, "step": 60548 }, { "epoch": 3.65, "learning_rate": 1.68349101444027e-05, "loss": 0.5523, "step": 60550 }, { "epoch": 3.65, "learning_rate": 1.6833401234288473e-05, "loss": 0.3597, "step": 60552 }, { "epoch": 3.65, "learning_rate": 1.683189232417425e-05, "loss": 0.2138, "step": 60554 }, { "epoch": 3.65, "learning_rate": 1.6830383414060026e-05, "loss": 0.2173, "step": 60556 }, { "epoch": 3.66, "learning_rate": 1.6828874503945802e-05, "loss": 0.2487, "step": 60558 }, { "epoch": 3.66, "learning_rate": 1.6827365593831575e-05, "loss": 0.3104, "step": 60560 }, { "epoch": 3.66, "learning_rate": 1.6825856683717352e-05, "loss": 0.3241, "step": 60562 }, { "epoch": 3.66, "learning_rate": 1.682434777360313e-05, "loss": 0.4899, "step": 60564 }, { "epoch": 3.66, "learning_rate": 1.68228388634889e-05, "loss": 0.401, "step": 60566 }, { "epoch": 3.66, "learning_rate": 1.6821329953374678e-05, "loss": 0.4531, "step": 60568 }, { "epoch": 3.66, "learning_rate": 1.6819821043260455e-05, "loss": 0.339, "step": 60570 }, { "epoch": 3.66, "learning_rate": 1.681831213314623e-05, "loss": 0.3858, "step": 60572 }, { "epoch": 3.66, "learning_rate": 1.6816803223032004e-05, "loss": 0.2912, "step": 60574 }, { "epoch": 3.66, "learning_rate": 1.681529431291778e-05, "loss": 0.3423, "step": 60576 }, { "epoch": 3.66, "learning_rate": 1.6813785402803557e-05, "loss": 0.3747, "step": 60578 }, { "epoch": 3.66, "learning_rate": 1.681227649268933e-05, "loss": 0.2576, "step": 60580 }, { "epoch": 3.66, "learning_rate": 1.6810767582575107e-05, "loss": 0.2559, "step": 60582 }, { "epoch": 3.66, "learning_rate": 1.680925867246088e-05, "loss": 0.3272, "step": 60584 }, { "epoch": 3.66, "learning_rate": 1.680774976234666e-05, "loss": 0.5065, "step": 60586 }, { "epoch": 3.66, "learning_rate": 1.6806240852232433e-05, "loss": 0.3751, "step": 60588 }, { "epoch": 3.66, "learning_rate": 1.680473194211821e-05, "loss": 0.5251, "step": 60590 }, { "epoch": 3.66, "learning_rate": 1.6803223032003983e-05, "loss": 0.3595, "step": 60592 }, { "epoch": 3.66, "learning_rate": 1.6801714121889763e-05, "loss": 0.4505, "step": 60594 }, { "epoch": 3.66, "learning_rate": 1.6800205211775536e-05, "loss": 0.3263, "step": 60596 }, { "epoch": 3.66, "learning_rate": 1.679869630166131e-05, "loss": 0.4086, "step": 60598 }, { "epoch": 3.66, "learning_rate": 1.6797187391547085e-05, "loss": 0.2706, "step": 60600 }, { "epoch": 3.66, "learning_rate": 1.6795678481432862e-05, "loss": 0.421, "step": 60602 }, { "epoch": 3.66, "learning_rate": 1.679416957131864e-05, "loss": 0.4173, "step": 60604 }, { "epoch": 3.66, "learning_rate": 1.679266066120441e-05, "loss": 0.3416, "step": 60606 }, { "epoch": 3.66, "learning_rate": 1.6791151751090188e-05, "loss": 0.6504, "step": 60608 }, { "epoch": 3.66, "learning_rate": 1.6789642840975965e-05, "loss": 0.3236, "step": 60610 }, { "epoch": 3.66, "learning_rate": 1.6788133930861738e-05, "loss": 0.407, "step": 60612 }, { "epoch": 3.66, "learning_rate": 1.6786625020747514e-05, "loss": 0.3645, "step": 60614 }, { "epoch": 3.66, "learning_rate": 1.678511611063329e-05, "loss": 0.4527, "step": 60616 }, { "epoch": 3.66, "learning_rate": 1.6783607200519067e-05, "loss": 0.3913, "step": 60618 }, { "epoch": 3.66, "learning_rate": 1.678209829040484e-05, "loss": 0.4181, "step": 60620 }, { "epoch": 3.66, "learning_rate": 1.6780589380290617e-05, "loss": 0.2893, "step": 60622 }, { "epoch": 3.66, "learning_rate": 1.6779080470176393e-05, "loss": 0.3242, "step": 60624 }, { "epoch": 3.66, "learning_rate": 1.677757156006217e-05, "loss": 0.2365, "step": 60626 }, { "epoch": 3.66, "learning_rate": 1.6776062649947943e-05, "loss": 0.388, "step": 60628 }, { "epoch": 3.66, "learning_rate": 1.6774553739833716e-05, "loss": 0.2695, "step": 60630 }, { "epoch": 3.66, "learning_rate": 1.6773044829719496e-05, "loss": 0.2598, "step": 60632 }, { "epoch": 3.66, "learning_rate": 1.677153591960527e-05, "loss": 0.3361, "step": 60634 }, { "epoch": 3.66, "learning_rate": 1.6770027009491046e-05, "loss": 0.359, "step": 60636 }, { "epoch": 3.66, "learning_rate": 1.676851809937682e-05, "loss": 0.3857, "step": 60638 }, { "epoch": 3.66, "learning_rate": 1.67670091892626e-05, "loss": 0.3341, "step": 60640 }, { "epoch": 3.66, "learning_rate": 1.6765500279148372e-05, "loss": 0.3853, "step": 60642 }, { "epoch": 3.66, "learning_rate": 1.676399136903415e-05, "loss": 0.435, "step": 60644 }, { "epoch": 3.66, "learning_rate": 1.676248245891992e-05, "loss": 0.299, "step": 60646 }, { "epoch": 3.66, "learning_rate": 1.6760973548805698e-05, "loss": 0.2789, "step": 60648 }, { "epoch": 3.66, "learning_rate": 1.6759464638691475e-05, "loss": 0.7452, "step": 60650 }, { "epoch": 3.66, "learning_rate": 1.6757955728577248e-05, "loss": 0.4173, "step": 60652 }, { "epoch": 3.66, "learning_rate": 1.6756446818463024e-05, "loss": 0.2687, "step": 60654 }, { "epoch": 3.66, "learning_rate": 1.67549379083488e-05, "loss": 0.3631, "step": 60656 }, { "epoch": 3.66, "learning_rate": 1.6753428998234577e-05, "loss": 0.478, "step": 60658 }, { "epoch": 3.66, "learning_rate": 1.675192008812035e-05, "loss": 0.3615, "step": 60660 }, { "epoch": 3.66, "learning_rate": 1.6750411178006127e-05, "loss": 0.2519, "step": 60662 }, { "epoch": 3.66, "learning_rate": 1.6748902267891903e-05, "loss": 0.3475, "step": 60664 }, { "epoch": 3.66, "learning_rate": 1.6747393357777677e-05, "loss": 0.3044, "step": 60666 }, { "epoch": 3.66, "learning_rate": 1.6745884447663453e-05, "loss": 0.4308, "step": 60668 }, { "epoch": 3.66, "learning_rate": 1.674437553754923e-05, "loss": 0.3785, "step": 60670 }, { "epoch": 3.66, "learning_rate": 1.6742866627435006e-05, "loss": 0.2137, "step": 60672 }, { "epoch": 3.66, "learning_rate": 1.674135771732078e-05, "loss": 0.4307, "step": 60674 }, { "epoch": 3.66, "learning_rate": 1.6739848807206556e-05, "loss": 0.3296, "step": 60676 }, { "epoch": 3.66, "learning_rate": 1.6738339897092332e-05, "loss": 0.2775, "step": 60678 }, { "epoch": 3.66, "learning_rate": 1.6736830986978105e-05, "loss": 0.2346, "step": 60680 }, { "epoch": 3.66, "learning_rate": 1.6735322076863882e-05, "loss": 0.3049, "step": 60682 }, { "epoch": 3.66, "learning_rate": 1.673381316674966e-05, "loss": 0.4967, "step": 60684 }, { "epoch": 3.66, "learning_rate": 1.6732304256635435e-05, "loss": 0.4419, "step": 60686 }, { "epoch": 3.66, "learning_rate": 1.6730795346521208e-05, "loss": 0.384, "step": 60688 }, { "epoch": 3.66, "learning_rate": 1.6729286436406985e-05, "loss": 0.3998, "step": 60690 }, { "epoch": 3.66, "learning_rate": 1.672777752629276e-05, "loss": 0.2566, "step": 60692 }, { "epoch": 3.66, "learning_rate": 1.6726268616178534e-05, "loss": 0.4603, "step": 60694 }, { "epoch": 3.66, "learning_rate": 1.672475970606431e-05, "loss": 0.2621, "step": 60696 }, { "epoch": 3.66, "learning_rate": 1.6723250795950084e-05, "loss": 0.3937, "step": 60698 }, { "epoch": 3.66, "learning_rate": 1.6721741885835864e-05, "loss": 0.3309, "step": 60700 }, { "epoch": 3.66, "learning_rate": 1.6720232975721637e-05, "loss": 0.4316, "step": 60702 }, { "epoch": 3.66, "learning_rate": 1.6718724065607413e-05, "loss": 0.4534, "step": 60704 }, { "epoch": 3.66, "learning_rate": 1.6717215155493186e-05, "loss": 0.3637, "step": 60706 }, { "epoch": 3.66, "learning_rate": 1.6715706245378966e-05, "loss": 0.4976, "step": 60708 }, { "epoch": 3.66, "learning_rate": 1.671419733526474e-05, "loss": 0.2883, "step": 60710 }, { "epoch": 3.66, "learning_rate": 1.6712688425150513e-05, "loss": 0.4627, "step": 60712 }, { "epoch": 3.66, "learning_rate": 1.671117951503629e-05, "loss": 0.2461, "step": 60714 }, { "epoch": 3.66, "learning_rate": 1.6709670604922066e-05, "loss": 0.1918, "step": 60716 }, { "epoch": 3.66, "learning_rate": 1.6708161694807842e-05, "loss": 0.2666, "step": 60718 }, { "epoch": 3.66, "learning_rate": 1.6706652784693615e-05, "loss": 0.4606, "step": 60720 }, { "epoch": 3.67, "learning_rate": 1.6705143874579392e-05, "loss": 0.3326, "step": 60722 }, { "epoch": 3.67, "learning_rate": 1.6703634964465168e-05, "loss": 0.4091, "step": 60724 }, { "epoch": 3.67, "learning_rate": 1.6702126054350945e-05, "loss": 0.4319, "step": 60726 }, { "epoch": 3.67, "learning_rate": 1.6700617144236718e-05, "loss": 0.4021, "step": 60728 }, { "epoch": 3.67, "learning_rate": 1.6699108234122494e-05, "loss": 0.295, "step": 60730 }, { "epoch": 3.67, "learning_rate": 1.669759932400827e-05, "loss": 0.3474, "step": 60732 }, { "epoch": 3.67, "learning_rate": 1.6696090413894044e-05, "loss": 0.5321, "step": 60734 }, { "epoch": 3.67, "learning_rate": 1.669458150377982e-05, "loss": 0.632, "step": 60736 }, { "epoch": 3.67, "learning_rate": 1.6693072593665597e-05, "loss": 0.3678, "step": 60738 }, { "epoch": 3.67, "learning_rate": 1.6691563683551374e-05, "loss": 0.4158, "step": 60740 }, { "epoch": 3.67, "learning_rate": 1.6690054773437147e-05, "loss": 0.4301, "step": 60742 }, { "epoch": 3.67, "learning_rate": 1.668854586332292e-05, "loss": 0.3538, "step": 60744 }, { "epoch": 3.67, "learning_rate": 1.66870369532087e-05, "loss": 0.3347, "step": 60746 }, { "epoch": 3.67, "learning_rate": 1.6685528043094473e-05, "loss": 0.2905, "step": 60748 }, { "epoch": 3.67, "learning_rate": 1.668401913298025e-05, "loss": 0.247, "step": 60750 }, { "epoch": 3.67, "learning_rate": 1.6682510222866023e-05, "loss": 0.4937, "step": 60752 }, { "epoch": 3.67, "learning_rate": 1.6681001312751802e-05, "loss": 0.3019, "step": 60754 }, { "epoch": 3.67, "learning_rate": 1.6679492402637576e-05, "loss": 0.463, "step": 60756 }, { "epoch": 3.67, "learning_rate": 1.6677983492523352e-05, "loss": 0.3552, "step": 60758 }, { "epoch": 3.67, "learning_rate": 1.6676474582409125e-05, "loss": 0.2973, "step": 60760 }, { "epoch": 3.67, "learning_rate": 1.6674965672294902e-05, "loss": 0.3129, "step": 60762 }, { "epoch": 3.67, "learning_rate": 1.6673456762180678e-05, "loss": 0.4138, "step": 60764 }, { "epoch": 3.67, "learning_rate": 1.667194785206645e-05, "loss": 0.3117, "step": 60766 }, { "epoch": 3.67, "learning_rate": 1.6670438941952228e-05, "loss": 0.2118, "step": 60768 }, { "epoch": 3.67, "learning_rate": 1.6668930031838004e-05, "loss": 0.3031, "step": 60770 }, { "epoch": 3.67, "learning_rate": 1.666742112172378e-05, "loss": 0.4021, "step": 60772 }, { "epoch": 3.67, "learning_rate": 1.6665912211609554e-05, "loss": 0.3288, "step": 60774 }, { "epoch": 3.67, "learning_rate": 1.666440330149533e-05, "loss": 0.3029, "step": 60776 }, { "epoch": 3.67, "learning_rate": 1.6662894391381107e-05, "loss": 0.3818, "step": 60778 }, { "epoch": 3.67, "learning_rate": 1.666138548126688e-05, "loss": 0.3871, "step": 60780 }, { "epoch": 3.67, "learning_rate": 1.6659876571152657e-05, "loss": 0.5437, "step": 60782 }, { "epoch": 3.67, "learning_rate": 1.6658367661038433e-05, "loss": 0.3752, "step": 60784 }, { "epoch": 3.67, "learning_rate": 1.665685875092421e-05, "loss": 0.4019, "step": 60786 }, { "epoch": 3.67, "learning_rate": 1.6655349840809983e-05, "loss": 0.3347, "step": 60788 }, { "epoch": 3.67, "learning_rate": 1.665384093069576e-05, "loss": 0.2628, "step": 60790 }, { "epoch": 3.67, "learning_rate": 1.6652332020581536e-05, "loss": 0.193, "step": 60792 }, { "epoch": 3.67, "learning_rate": 1.665082311046731e-05, "loss": 0.3277, "step": 60794 }, { "epoch": 3.67, "learning_rate": 1.6649314200353086e-05, "loss": 0.36, "step": 60796 }, { "epoch": 3.67, "learning_rate": 1.6647805290238862e-05, "loss": 0.2686, "step": 60798 }, { "epoch": 3.67, "learning_rate": 1.664629638012464e-05, "loss": 0.3267, "step": 60800 }, { "epoch": 3.67, "learning_rate": 1.664478747001041e-05, "loss": 0.3862, "step": 60802 }, { "epoch": 3.67, "learning_rate": 1.6643278559896188e-05, "loss": 0.3689, "step": 60804 }, { "epoch": 3.67, "learning_rate": 1.6641769649781965e-05, "loss": 0.3929, "step": 60806 }, { "epoch": 3.67, "learning_rate": 1.664026073966774e-05, "loss": 0.3153, "step": 60808 }, { "epoch": 3.67, "learning_rate": 1.6638751829553514e-05, "loss": 0.2803, "step": 60810 }, { "epoch": 3.67, "learning_rate": 1.6637242919439287e-05, "loss": 0.241, "step": 60812 }, { "epoch": 3.67, "learning_rate": 1.6635734009325067e-05, "loss": 0.2243, "step": 60814 }, { "epoch": 3.67, "learning_rate": 1.663422509921084e-05, "loss": 0.3846, "step": 60816 }, { "epoch": 3.67, "learning_rate": 1.6632716189096617e-05, "loss": 0.3496, "step": 60818 }, { "epoch": 3.67, "learning_rate": 1.663120727898239e-05, "loss": 0.3205, "step": 60820 }, { "epoch": 3.67, "learning_rate": 1.662969836886817e-05, "loss": 0.522, "step": 60822 }, { "epoch": 3.67, "learning_rate": 1.6628189458753943e-05, "loss": 0.296, "step": 60824 }, { "epoch": 3.67, "learning_rate": 1.6626680548639716e-05, "loss": 0.2446, "step": 60826 }, { "epoch": 3.67, "learning_rate": 1.6625171638525493e-05, "loss": 0.2718, "step": 60828 }, { "epoch": 3.67, "learning_rate": 1.662366272841127e-05, "loss": 0.2039, "step": 60830 }, { "epoch": 3.67, "learning_rate": 1.6622153818297046e-05, "loss": 0.4604, "step": 60832 }, { "epoch": 3.67, "learning_rate": 1.662064490818282e-05, "loss": 0.3723, "step": 60834 }, { "epoch": 3.67, "learning_rate": 1.6619135998068595e-05, "loss": 0.3082, "step": 60836 }, { "epoch": 3.67, "learning_rate": 1.6617627087954372e-05, "loss": 0.3201, "step": 60838 }, { "epoch": 3.67, "learning_rate": 1.661611817784015e-05, "loss": 0.4047, "step": 60840 }, { "epoch": 3.67, "learning_rate": 1.661460926772592e-05, "loss": 0.4874, "step": 60842 }, { "epoch": 3.67, "learning_rate": 1.6613100357611698e-05, "loss": 0.2454, "step": 60844 }, { "epoch": 3.67, "learning_rate": 1.6611591447497475e-05, "loss": 0.3294, "step": 60846 }, { "epoch": 3.67, "learning_rate": 1.6610082537383248e-05, "loss": 0.3619, "step": 60848 }, { "epoch": 3.67, "learning_rate": 1.6608573627269024e-05, "loss": 0.4701, "step": 60850 }, { "epoch": 3.67, "learning_rate": 1.66070647171548e-05, "loss": 0.4599, "step": 60852 }, { "epoch": 3.67, "learning_rate": 1.6605555807040577e-05, "loss": 0.3945, "step": 60854 }, { "epoch": 3.67, "learning_rate": 1.660404689692635e-05, "loss": 0.5175, "step": 60856 }, { "epoch": 3.67, "learning_rate": 1.6602537986812124e-05, "loss": 0.2808, "step": 60858 }, { "epoch": 3.67, "learning_rate": 1.6601029076697903e-05, "loss": 0.3288, "step": 60860 }, { "epoch": 3.67, "learning_rate": 1.6599520166583677e-05, "loss": 0.2524, "step": 60862 }, { "epoch": 3.67, "learning_rate": 1.6598011256469453e-05, "loss": 0.3017, "step": 60864 }, { "epoch": 3.67, "learning_rate": 1.6596502346355226e-05, "loss": 0.204, "step": 60866 }, { "epoch": 3.67, "learning_rate": 1.6594993436241006e-05, "loss": 0.3155, "step": 60868 }, { "epoch": 3.67, "learning_rate": 1.659348452612678e-05, "loss": 0.2733, "step": 60870 }, { "epoch": 3.67, "learning_rate": 1.6591975616012556e-05, "loss": 0.4127, "step": 60872 }, { "epoch": 3.67, "learning_rate": 1.659046670589833e-05, "loss": 0.3375, "step": 60874 }, { "epoch": 3.67, "learning_rate": 1.6588957795784105e-05, "loss": 0.5836, "step": 60876 }, { "epoch": 3.67, "learning_rate": 1.6587448885669882e-05, "loss": 0.3198, "step": 60878 }, { "epoch": 3.67, "learning_rate": 1.6585939975555655e-05, "loss": 0.4334, "step": 60880 }, { "epoch": 3.67, "learning_rate": 1.658443106544143e-05, "loss": 0.308, "step": 60882 }, { "epoch": 3.67, "learning_rate": 1.6582922155327208e-05, "loss": 0.3427, "step": 60884 }, { "epoch": 3.67, "learning_rate": 1.6581413245212985e-05, "loss": 0.3733, "step": 60886 }, { "epoch": 3.68, "learning_rate": 1.6579904335098758e-05, "loss": 0.3389, "step": 60888 }, { "epoch": 3.68, "learning_rate": 1.6578395424984534e-05, "loss": 0.2839, "step": 60890 }, { "epoch": 3.68, "learning_rate": 1.657688651487031e-05, "loss": 0.4127, "step": 60892 }, { "epoch": 3.68, "learning_rate": 1.6575377604756084e-05, "loss": 0.2809, "step": 60894 }, { "epoch": 3.68, "learning_rate": 1.657386869464186e-05, "loss": 0.3687, "step": 60896 }, { "epoch": 3.68, "learning_rate": 1.6572359784527637e-05, "loss": 0.2479, "step": 60898 }, { "epoch": 3.68, "learning_rate": 1.6570850874413413e-05, "loss": 0.3376, "step": 60900 }, { "epoch": 3.68, "learning_rate": 1.6569341964299187e-05, "loss": 0.5264, "step": 60902 }, { "epoch": 3.68, "learning_rate": 1.6567833054184963e-05, "loss": 0.3256, "step": 60904 }, { "epoch": 3.68, "learning_rate": 1.656632414407074e-05, "loss": 0.3105, "step": 60906 }, { "epoch": 3.68, "learning_rate": 1.6564815233956513e-05, "loss": 0.2714, "step": 60908 }, { "epoch": 3.68, "learning_rate": 1.656330632384229e-05, "loss": 0.2606, "step": 60910 }, { "epoch": 3.68, "learning_rate": 1.6561797413728066e-05, "loss": 0.2715, "step": 60912 }, { "epoch": 3.68, "learning_rate": 1.6560288503613842e-05, "loss": 0.344, "step": 60914 }, { "epoch": 3.68, "learning_rate": 1.6558779593499615e-05, "loss": 0.266, "step": 60916 }, { "epoch": 3.68, "learning_rate": 1.6557270683385392e-05, "loss": 0.3629, "step": 60918 }, { "epoch": 3.68, "learning_rate": 1.655576177327117e-05, "loss": 0.3201, "step": 60920 }, { "epoch": 3.68, "learning_rate": 1.6554252863156945e-05, "loss": 0.3479, "step": 60922 }, { "epoch": 3.68, "learning_rate": 1.6552743953042718e-05, "loss": 0.3823, "step": 60924 }, { "epoch": 3.68, "learning_rate": 1.655123504292849e-05, "loss": 0.3807, "step": 60926 }, { "epoch": 3.68, "learning_rate": 1.654972613281427e-05, "loss": 0.4689, "step": 60928 }, { "epoch": 3.68, "learning_rate": 1.6548217222700044e-05, "loss": 0.2741, "step": 60930 }, { "epoch": 3.68, "learning_rate": 1.654670831258582e-05, "loss": 0.3051, "step": 60932 }, { "epoch": 3.68, "learning_rate": 1.6545199402471594e-05, "loss": 0.3684, "step": 60934 }, { "epoch": 3.68, "learning_rate": 1.6543690492357374e-05, "loss": 0.4434, "step": 60936 }, { "epoch": 3.68, "learning_rate": 1.6542181582243147e-05, "loss": 0.2227, "step": 60938 }, { "epoch": 3.68, "learning_rate": 1.654067267212892e-05, "loss": 0.5103, "step": 60940 }, { "epoch": 3.68, "learning_rate": 1.6539163762014696e-05, "loss": 0.2973, "step": 60942 }, { "epoch": 3.68, "learning_rate": 1.6537654851900473e-05, "loss": 0.4368, "step": 60944 }, { "epoch": 3.68, "learning_rate": 1.653614594178625e-05, "loss": 0.3772, "step": 60946 }, { "epoch": 3.68, "learning_rate": 1.6534637031672023e-05, "loss": 0.4377, "step": 60948 }, { "epoch": 3.68, "learning_rate": 1.65331281215578e-05, "loss": 0.5574, "step": 60950 }, { "epoch": 3.68, "learning_rate": 1.6531619211443576e-05, "loss": 0.3399, "step": 60952 }, { "epoch": 3.68, "learning_rate": 1.6530110301329352e-05, "loss": 0.403, "step": 60954 }, { "epoch": 3.68, "learning_rate": 1.6528601391215125e-05, "loss": 0.6541, "step": 60956 }, { "epoch": 3.68, "learning_rate": 1.6527092481100902e-05, "loss": 0.4349, "step": 60958 }, { "epoch": 3.68, "learning_rate": 1.652558357098668e-05, "loss": 0.4531, "step": 60960 }, { "epoch": 3.68, "learning_rate": 1.652407466087245e-05, "loss": 0.3184, "step": 60962 }, { "epoch": 3.68, "learning_rate": 1.6522565750758228e-05, "loss": 0.2504, "step": 60964 }, { "epoch": 3.68, "learning_rate": 1.6521056840644004e-05, "loss": 0.3392, "step": 60966 }, { "epoch": 3.68, "learning_rate": 1.651954793052978e-05, "loss": 0.2981, "step": 60968 }, { "epoch": 3.68, "learning_rate": 1.6518039020415554e-05, "loss": 0.2491, "step": 60970 }, { "epoch": 3.68, "learning_rate": 1.651653011030133e-05, "loss": 0.4403, "step": 60972 }, { "epoch": 3.68, "learning_rate": 1.6515021200187107e-05, "loss": 0.2943, "step": 60974 }, { "epoch": 3.68, "learning_rate": 1.651351229007288e-05, "loss": 0.5084, "step": 60976 }, { "epoch": 3.68, "learning_rate": 1.6512003379958657e-05, "loss": 0.2788, "step": 60978 }, { "epoch": 3.68, "learning_rate": 1.651049446984443e-05, "loss": 0.3284, "step": 60980 }, { "epoch": 3.68, "learning_rate": 1.650898555973021e-05, "loss": 0.2831, "step": 60982 }, { "epoch": 3.68, "learning_rate": 1.6507476649615983e-05, "loss": 0.3214, "step": 60984 }, { "epoch": 3.68, "learning_rate": 1.650596773950176e-05, "loss": 0.217, "step": 60986 }, { "epoch": 3.68, "learning_rate": 1.6504458829387533e-05, "loss": 0.2363, "step": 60988 }, { "epoch": 3.68, "learning_rate": 1.650294991927331e-05, "loss": 0.3067, "step": 60990 }, { "epoch": 3.68, "learning_rate": 1.6501441009159086e-05, "loss": 0.2435, "step": 60992 }, { "epoch": 3.68, "learning_rate": 1.649993209904486e-05, "loss": 0.3712, "step": 60994 }, { "epoch": 3.68, "learning_rate": 1.6498423188930635e-05, "loss": 0.3153, "step": 60996 }, { "epoch": 3.68, "learning_rate": 1.6496914278816412e-05, "loss": 0.5052, "step": 60998 }, { "epoch": 3.68, "learning_rate": 1.6495405368702188e-05, "loss": 0.5087, "step": 61000 }, { "epoch": 3.68, "learning_rate": 1.649389645858796e-05, "loss": 0.4052, "step": 61002 }, { "epoch": 3.68, "learning_rate": 1.6492387548473738e-05, "loss": 0.4183, "step": 61004 }, { "epoch": 3.68, "learning_rate": 1.6490878638359514e-05, "loss": 0.2445, "step": 61006 }, { "epoch": 3.68, "learning_rate": 1.6489369728245288e-05, "loss": 0.331, "step": 61008 }, { "epoch": 3.68, "learning_rate": 1.6487860818131064e-05, "loss": 0.3451, "step": 61010 }, { "epoch": 3.68, "learning_rate": 1.648635190801684e-05, "loss": 0.2113, "step": 61012 }, { "epoch": 3.68, "learning_rate": 1.6484842997902617e-05, "loss": 0.3063, "step": 61014 }, { "epoch": 3.68, "learning_rate": 1.648333408778839e-05, "loss": 0.2788, "step": 61016 }, { "epoch": 3.68, "learning_rate": 1.6481825177674167e-05, "loss": 0.3608, "step": 61018 }, { "epoch": 3.68, "learning_rate": 1.6480316267559943e-05, "loss": 0.3629, "step": 61020 }, { "epoch": 3.68, "learning_rate": 1.647880735744572e-05, "loss": 0.4128, "step": 61022 }, { "epoch": 3.68, "learning_rate": 1.6477298447331493e-05, "loss": 0.2881, "step": 61024 }, { "epoch": 3.68, "learning_rate": 1.647578953721727e-05, "loss": 0.2966, "step": 61026 }, { "epoch": 3.68, "learning_rate": 1.6474280627103046e-05, "loss": 0.4213, "step": 61028 }, { "epoch": 3.68, "learning_rate": 1.647277171698882e-05, "loss": 0.3287, "step": 61030 }, { "epoch": 3.68, "learning_rate": 1.6471262806874596e-05, "loss": 0.4727, "step": 61032 }, { "epoch": 3.68, "learning_rate": 1.6469753896760372e-05, "loss": 0.4177, "step": 61034 }, { "epoch": 3.68, "learning_rate": 1.646824498664615e-05, "loss": 0.2903, "step": 61036 }, { "epoch": 3.68, "learning_rate": 1.646673607653192e-05, "loss": 0.3676, "step": 61038 }, { "epoch": 3.68, "learning_rate": 1.6465227166417695e-05, "loss": 0.2922, "step": 61040 }, { "epoch": 3.68, "learning_rate": 1.6463718256303475e-05, "loss": 0.4181, "step": 61042 }, { "epoch": 3.68, "learning_rate": 1.6462209346189248e-05, "loss": 0.3359, "step": 61044 }, { "epoch": 3.68, "learning_rate": 1.6460700436075024e-05, "loss": 0.3248, "step": 61046 }, { "epoch": 3.68, "learning_rate": 1.6459191525960797e-05, "loss": 0.3012, "step": 61048 }, { "epoch": 3.68, "learning_rate": 1.6457682615846577e-05, "loss": 0.466, "step": 61050 }, { "epoch": 3.68, "learning_rate": 1.645617370573235e-05, "loss": 0.2756, "step": 61052 }, { "epoch": 3.69, "learning_rate": 1.6454664795618127e-05, "loss": 0.4688, "step": 61054 }, { "epoch": 3.69, "learning_rate": 1.64531558855039e-05, "loss": 0.3158, "step": 61056 }, { "epoch": 3.69, "learning_rate": 1.6451646975389677e-05, "loss": 0.408, "step": 61058 }, { "epoch": 3.69, "learning_rate": 1.6450138065275453e-05, "loss": 0.2856, "step": 61060 }, { "epoch": 3.69, "learning_rate": 1.6448629155161226e-05, "loss": 0.3032, "step": 61062 }, { "epoch": 3.69, "learning_rate": 1.6447120245047003e-05, "loss": 0.3554, "step": 61064 }, { "epoch": 3.69, "learning_rate": 1.644561133493278e-05, "loss": 0.3824, "step": 61066 }, { "epoch": 3.69, "learning_rate": 1.6444102424818556e-05, "loss": 0.3089, "step": 61068 }, { "epoch": 3.69, "learning_rate": 1.644259351470433e-05, "loss": 0.3302, "step": 61070 }, { "epoch": 3.69, "learning_rate": 1.6441084604590105e-05, "loss": 0.3348, "step": 61072 }, { "epoch": 3.69, "learning_rate": 1.6439575694475882e-05, "loss": 0.4809, "step": 61074 }, { "epoch": 3.69, "learning_rate": 1.6438066784361655e-05, "loss": 0.3498, "step": 61076 }, { "epoch": 3.69, "learning_rate": 1.643655787424743e-05, "loss": 0.322, "step": 61078 }, { "epoch": 3.69, "learning_rate": 1.6435048964133208e-05, "loss": 0.3466, "step": 61080 }, { "epoch": 3.69, "learning_rate": 1.6433540054018985e-05, "loss": 0.294, "step": 61082 }, { "epoch": 3.69, "learning_rate": 1.6432031143904758e-05, "loss": 0.351, "step": 61084 }, { "epoch": 3.69, "learning_rate": 1.6430522233790534e-05, "loss": 0.2868, "step": 61086 }, { "epoch": 3.69, "learning_rate": 1.642901332367631e-05, "loss": 0.4243, "step": 61088 }, { "epoch": 3.69, "learning_rate": 1.6427504413562084e-05, "loss": 0.3323, "step": 61090 }, { "epoch": 3.69, "learning_rate": 1.642599550344786e-05, "loss": 0.4759, "step": 61092 }, { "epoch": 3.69, "learning_rate": 1.6424486593333634e-05, "loss": 0.5301, "step": 61094 }, { "epoch": 3.69, "learning_rate": 1.6422977683219413e-05, "loss": 0.3781, "step": 61096 }, { "epoch": 3.69, "learning_rate": 1.6421468773105187e-05, "loss": 0.3879, "step": 61098 }, { "epoch": 3.69, "learning_rate": 1.6419959862990963e-05, "loss": 0.4785, "step": 61100 }, { "epoch": 3.69, "learning_rate": 1.6418450952876736e-05, "loss": 0.4084, "step": 61102 }, { "epoch": 3.69, "learning_rate": 1.6416942042762516e-05, "loss": 0.2906, "step": 61104 }, { "epoch": 3.69, "learning_rate": 1.641543313264829e-05, "loss": 0.4319, "step": 61106 }, { "epoch": 3.69, "learning_rate": 1.6413924222534062e-05, "loss": 0.3639, "step": 61108 }, { "epoch": 3.69, "learning_rate": 1.641241531241984e-05, "loss": 0.2484, "step": 61110 }, { "epoch": 3.69, "learning_rate": 1.6410906402305615e-05, "loss": 0.4913, "step": 61112 }, { "epoch": 3.69, "learning_rate": 1.6409397492191392e-05, "loss": 0.3249, "step": 61114 }, { "epoch": 3.69, "learning_rate": 1.6407888582077165e-05, "loss": 0.3598, "step": 61116 }, { "epoch": 3.69, "learning_rate": 1.640637967196294e-05, "loss": 0.2927, "step": 61118 }, { "epoch": 3.69, "learning_rate": 1.6404870761848718e-05, "loss": 0.3271, "step": 61120 }, { "epoch": 3.69, "learning_rate": 1.640336185173449e-05, "loss": 0.3327, "step": 61122 }, { "epoch": 3.69, "learning_rate": 1.6401852941620268e-05, "loss": 0.273, "step": 61124 }, { "epoch": 3.69, "learning_rate": 1.6400344031506044e-05, "loss": 0.3347, "step": 61126 }, { "epoch": 3.69, "learning_rate": 1.639883512139182e-05, "loss": 0.4331, "step": 61128 }, { "epoch": 3.69, "learning_rate": 1.6397326211277594e-05, "loss": 0.3359, "step": 61130 }, { "epoch": 3.69, "learning_rate": 1.639581730116337e-05, "loss": 0.3141, "step": 61132 }, { "epoch": 3.69, "learning_rate": 1.6394308391049147e-05, "loss": 0.398, "step": 61134 }, { "epoch": 3.69, "learning_rate": 1.6392799480934923e-05, "loss": 0.4037, "step": 61136 }, { "epoch": 3.69, "learning_rate": 1.6391290570820697e-05, "loss": 0.5812, "step": 61138 }, { "epoch": 3.69, "learning_rate": 1.638978166070647e-05, "loss": 0.352, "step": 61140 }, { "epoch": 3.69, "learning_rate": 1.638827275059225e-05, "loss": 0.3048, "step": 61142 }, { "epoch": 3.69, "learning_rate": 1.6386763840478023e-05, "loss": 0.3673, "step": 61144 }, { "epoch": 3.69, "learning_rate": 1.63852549303638e-05, "loss": 0.3955, "step": 61146 }, { "epoch": 3.69, "learning_rate": 1.6383746020249576e-05, "loss": 0.2659, "step": 61148 }, { "epoch": 3.69, "learning_rate": 1.6382237110135352e-05, "loss": 0.4132, "step": 61150 }, { "epoch": 3.69, "learning_rate": 1.6380728200021125e-05, "loss": 0.2699, "step": 61152 }, { "epoch": 3.69, "learning_rate": 1.63792192899069e-05, "loss": 0.3294, "step": 61154 }, { "epoch": 3.69, "learning_rate": 1.637771037979268e-05, "loss": 0.4623, "step": 61156 }, { "epoch": 3.69, "learning_rate": 1.637620146967845e-05, "loss": 0.3683, "step": 61158 }, { "epoch": 3.69, "learning_rate": 1.6374692559564228e-05, "loss": 0.4124, "step": 61160 }, { "epoch": 3.69, "learning_rate": 1.637318364945e-05, "loss": 0.275, "step": 61162 }, { "epoch": 3.69, "learning_rate": 1.637167473933578e-05, "loss": 0.3844, "step": 61164 }, { "epoch": 3.69, "learning_rate": 1.6370165829221554e-05, "loss": 0.3816, "step": 61166 }, { "epoch": 3.69, "learning_rate": 1.636865691910733e-05, "loss": 0.2953, "step": 61168 }, { "epoch": 3.69, "learning_rate": 1.6367148008993104e-05, "loss": 0.2892, "step": 61170 }, { "epoch": 3.69, "learning_rate": 1.636563909887888e-05, "loss": 0.5277, "step": 61172 }, { "epoch": 3.69, "learning_rate": 1.6364130188764657e-05, "loss": 0.2888, "step": 61174 }, { "epoch": 3.69, "learning_rate": 1.636262127865043e-05, "loss": 0.2887, "step": 61176 }, { "epoch": 3.69, "learning_rate": 1.6361112368536206e-05, "loss": 0.2393, "step": 61178 }, { "epoch": 3.69, "learning_rate": 1.6359603458421983e-05, "loss": 0.4189, "step": 61180 }, { "epoch": 3.69, "learning_rate": 1.635809454830776e-05, "loss": 0.4319, "step": 61182 }, { "epoch": 3.69, "learning_rate": 1.6356585638193533e-05, "loss": 0.3995, "step": 61184 }, { "epoch": 3.69, "learning_rate": 1.635507672807931e-05, "loss": 0.5435, "step": 61186 }, { "epoch": 3.69, "learning_rate": 1.6353567817965086e-05, "loss": 0.4143, "step": 61188 }, { "epoch": 3.69, "learning_rate": 1.635205890785086e-05, "loss": 0.2727, "step": 61190 }, { "epoch": 3.69, "learning_rate": 1.6350549997736635e-05, "loss": 0.5126, "step": 61192 }, { "epoch": 3.69, "learning_rate": 1.6349041087622412e-05, "loss": 0.4174, "step": 61194 }, { "epoch": 3.69, "learning_rate": 1.634753217750819e-05, "loss": 0.3939, "step": 61196 }, { "epoch": 3.69, "learning_rate": 1.634602326739396e-05, "loss": 0.4719, "step": 61198 }, { "epoch": 3.69, "learning_rate": 1.6344514357279738e-05, "loss": 0.3675, "step": 61200 }, { "epoch": 3.69, "learning_rate": 1.6343005447165514e-05, "loss": 0.2324, "step": 61202 }, { "epoch": 3.69, "learning_rate": 1.6341496537051288e-05, "loss": 0.4459, "step": 61204 }, { "epoch": 3.69, "learning_rate": 1.6339987626937064e-05, "loss": 0.4975, "step": 61206 }, { "epoch": 3.69, "learning_rate": 1.6338478716822837e-05, "loss": 0.565, "step": 61208 }, { "epoch": 3.69, "learning_rate": 1.6336969806708617e-05, "loss": 0.3617, "step": 61210 }, { "epoch": 3.69, "learning_rate": 1.633546089659439e-05, "loss": 0.4688, "step": 61212 }, { "epoch": 3.69, "learning_rate": 1.6333951986480167e-05, "loss": 0.466, "step": 61214 }, { "epoch": 3.69, "learning_rate": 1.633244307636594e-05, "loss": 0.3849, "step": 61216 }, { "epoch": 3.69, "learning_rate": 1.633093416625172e-05, "loss": 0.3786, "step": 61218 }, { "epoch": 3.7, "learning_rate": 1.6329425256137493e-05, "loss": 0.3972, "step": 61220 }, { "epoch": 3.7, "learning_rate": 1.6327916346023266e-05, "loss": 0.4109, "step": 61222 }, { "epoch": 3.7, "learning_rate": 1.6326407435909043e-05, "loss": 0.3073, "step": 61224 }, { "epoch": 3.7, "learning_rate": 1.632489852579482e-05, "loss": 0.4639, "step": 61226 }, { "epoch": 3.7, "learning_rate": 1.6323389615680596e-05, "loss": 0.3647, "step": 61228 }, { "epoch": 3.7, "learning_rate": 1.632188070556637e-05, "loss": 0.3889, "step": 61230 }, { "epoch": 3.7, "learning_rate": 1.6320371795452145e-05, "loss": 0.344, "step": 61232 }, { "epoch": 3.7, "learning_rate": 1.6318862885337922e-05, "loss": 0.2594, "step": 61234 }, { "epoch": 3.7, "learning_rate": 1.6317353975223695e-05, "loss": 0.3703, "step": 61236 }, { "epoch": 3.7, "learning_rate": 1.631584506510947e-05, "loss": 0.3021, "step": 61238 }, { "epoch": 3.7, "learning_rate": 1.6314336154995248e-05, "loss": 0.5265, "step": 61240 }, { "epoch": 3.7, "learning_rate": 1.6312827244881024e-05, "loss": 0.4121, "step": 61242 }, { "epoch": 3.7, "learning_rate": 1.6311318334766798e-05, "loss": 0.4096, "step": 61244 }, { "epoch": 3.7, "learning_rate": 1.6309809424652574e-05, "loss": 0.4085, "step": 61246 }, { "epoch": 3.7, "learning_rate": 1.630830051453835e-05, "loss": 0.3195, "step": 61248 }, { "epoch": 3.7, "learning_rate": 1.6306791604424127e-05, "loss": 0.3272, "step": 61250 }, { "epoch": 3.7, "learning_rate": 1.63052826943099e-05, "loss": 0.401, "step": 61252 }, { "epoch": 3.7, "learning_rate": 1.6303773784195673e-05, "loss": 0.352, "step": 61254 }, { "epoch": 3.7, "learning_rate": 1.6302264874081453e-05, "loss": 0.3082, "step": 61256 }, { "epoch": 3.7, "learning_rate": 1.6300755963967226e-05, "loss": 0.3781, "step": 61258 }, { "epoch": 3.7, "learning_rate": 1.6299247053853003e-05, "loss": 0.513, "step": 61260 }, { "epoch": 3.7, "learning_rate": 1.629773814373878e-05, "loss": 0.477, "step": 61262 }, { "epoch": 3.7, "learning_rate": 1.6296229233624556e-05, "loss": 0.2605, "step": 61264 }, { "epoch": 3.7, "learning_rate": 1.629472032351033e-05, "loss": 0.3345, "step": 61266 }, { "epoch": 3.7, "learning_rate": 1.6293211413396106e-05, "loss": 0.2816, "step": 61268 }, { "epoch": 3.7, "learning_rate": 1.6291702503281882e-05, "loss": 0.4767, "step": 61270 }, { "epoch": 3.7, "learning_rate": 1.6290193593167655e-05, "loss": 0.2895, "step": 61272 }, { "epoch": 3.7, "learning_rate": 1.6288684683053432e-05, "loss": 0.4409, "step": 61274 }, { "epoch": 3.7, "learning_rate": 1.6287175772939205e-05, "loss": 0.4097, "step": 61276 }, { "epoch": 3.7, "learning_rate": 1.6285666862824985e-05, "loss": 0.5299, "step": 61278 }, { "epoch": 3.7, "learning_rate": 1.6284157952710758e-05, "loss": 0.4863, "step": 61280 }, { "epoch": 3.7, "learning_rate": 1.6282649042596534e-05, "loss": 0.5265, "step": 61282 }, { "epoch": 3.7, "learning_rate": 1.6281140132482308e-05, "loss": 0.3101, "step": 61284 }, { "epoch": 3.7, "learning_rate": 1.6279631222368084e-05, "loss": 0.4258, "step": 61286 }, { "epoch": 3.7, "learning_rate": 1.627812231225386e-05, "loss": 0.3078, "step": 61288 }, { "epoch": 3.7, "learning_rate": 1.6276613402139634e-05, "loss": 0.4299, "step": 61290 }, { "epoch": 3.7, "learning_rate": 1.627510449202541e-05, "loss": 0.3051, "step": 61292 }, { "epoch": 3.7, "learning_rate": 1.6273595581911187e-05, "loss": 0.2341, "step": 61294 }, { "epoch": 3.7, "learning_rate": 1.6272086671796963e-05, "loss": 0.2095, "step": 61296 }, { "epoch": 3.7, "learning_rate": 1.6270577761682736e-05, "loss": 0.3901, "step": 61298 }, { "epoch": 3.7, "learning_rate": 1.6269068851568513e-05, "loss": 0.5161, "step": 61300 }, { "epoch": 3.7, "learning_rate": 1.626755994145429e-05, "loss": 0.279, "step": 61302 }, { "epoch": 3.7, "learning_rate": 1.6266051031340062e-05, "loss": 0.4514, "step": 61304 }, { "epoch": 3.7, "learning_rate": 1.626454212122584e-05, "loss": 0.3546, "step": 61306 }, { "epoch": 3.7, "learning_rate": 1.6263033211111616e-05, "loss": 0.3622, "step": 61308 }, { "epoch": 3.7, "learning_rate": 1.6261524300997392e-05, "loss": 0.3798, "step": 61310 }, { "epoch": 3.7, "learning_rate": 1.6260015390883165e-05, "loss": 0.4568, "step": 61312 }, { "epoch": 3.7, "learning_rate": 1.625850648076894e-05, "loss": 0.2434, "step": 61314 }, { "epoch": 3.7, "learning_rate": 1.6256997570654718e-05, "loss": 0.3618, "step": 61316 }, { "epoch": 3.7, "learning_rate": 1.625548866054049e-05, "loss": 0.5126, "step": 61318 }, { "epoch": 3.7, "learning_rate": 1.6253979750426268e-05, "loss": 0.343, "step": 61320 }, { "epoch": 3.7, "learning_rate": 1.625247084031204e-05, "loss": 0.4399, "step": 61322 }, { "epoch": 3.7, "learning_rate": 1.625096193019782e-05, "loss": 0.4452, "step": 61324 }, { "epoch": 3.7, "learning_rate": 1.6249453020083594e-05, "loss": 0.2551, "step": 61326 }, { "epoch": 3.7, "learning_rate": 1.624794410996937e-05, "loss": 0.339, "step": 61328 }, { "epoch": 3.7, "learning_rate": 1.6246435199855144e-05, "loss": 0.2707, "step": 61330 }, { "epoch": 3.7, "learning_rate": 1.6244926289740924e-05, "loss": 0.2976, "step": 61332 }, { "epoch": 3.7, "learning_rate": 1.6243417379626697e-05, "loss": 0.3292, "step": 61334 }, { "epoch": 3.7, "learning_rate": 1.624190846951247e-05, "loss": 0.289, "step": 61336 }, { "epoch": 3.7, "learning_rate": 1.6240399559398246e-05, "loss": 0.3949, "step": 61338 }, { "epoch": 3.7, "learning_rate": 1.6238890649284023e-05, "loss": 0.3068, "step": 61340 }, { "epoch": 3.7, "learning_rate": 1.62373817391698e-05, "loss": 0.4269, "step": 61342 }, { "epoch": 3.7, "learning_rate": 1.6235872829055572e-05, "loss": 0.3338, "step": 61344 }, { "epoch": 3.7, "learning_rate": 1.623436391894135e-05, "loss": 0.4238, "step": 61346 }, { "epoch": 3.7, "learning_rate": 1.6232855008827125e-05, "loss": 0.3682, "step": 61348 }, { "epoch": 3.7, "learning_rate": 1.6231346098712902e-05, "loss": 0.5179, "step": 61350 }, { "epoch": 3.7, "learning_rate": 1.6229837188598675e-05, "loss": 0.3254, "step": 61352 }, { "epoch": 3.7, "learning_rate": 1.622832827848445e-05, "loss": 0.2429, "step": 61354 }, { "epoch": 3.7, "learning_rate": 1.6226819368370228e-05, "loss": 0.5108, "step": 61356 }, { "epoch": 3.7, "learning_rate": 1.6225310458256e-05, "loss": 0.503, "step": 61358 }, { "epoch": 3.7, "learning_rate": 1.6223801548141778e-05, "loss": 0.2573, "step": 61360 }, { "epoch": 3.7, "learning_rate": 1.6222292638027554e-05, "loss": 0.3831, "step": 61362 }, { "epoch": 3.7, "learning_rate": 1.622078372791333e-05, "loss": 0.3128, "step": 61364 }, { "epoch": 3.7, "learning_rate": 1.6219274817799104e-05, "loss": 0.2319, "step": 61366 }, { "epoch": 3.7, "learning_rate": 1.6217765907684877e-05, "loss": 0.3315, "step": 61368 }, { "epoch": 3.7, "learning_rate": 1.6216256997570657e-05, "loss": 0.3697, "step": 61370 }, { "epoch": 3.7, "learning_rate": 1.621474808745643e-05, "loss": 0.381, "step": 61372 }, { "epoch": 3.7, "learning_rate": 1.6213239177342207e-05, "loss": 0.3958, "step": 61374 }, { "epoch": 3.7, "learning_rate": 1.621173026722798e-05, "loss": 0.2767, "step": 61376 }, { "epoch": 3.7, "learning_rate": 1.621022135711376e-05, "loss": 0.4934, "step": 61378 }, { "epoch": 3.7, "learning_rate": 1.6208712446999533e-05, "loss": 0.3562, "step": 61380 }, { "epoch": 3.7, "learning_rate": 1.620720353688531e-05, "loss": 0.396, "step": 61382 }, { "epoch": 3.7, "learning_rate": 1.6205694626771086e-05, "loss": 0.2401, "step": 61384 }, { "epoch": 3.71, "learning_rate": 1.620418571665686e-05, "loss": 0.351, "step": 61386 }, { "epoch": 3.71, "learning_rate": 1.6202676806542635e-05, "loss": 0.3441, "step": 61388 }, { "epoch": 3.71, "learning_rate": 1.620116789642841e-05, "loss": 0.3136, "step": 61390 }, { "epoch": 3.71, "learning_rate": 1.619965898631419e-05, "loss": 0.3316, "step": 61392 }, { "epoch": 3.71, "learning_rate": 1.619815007619996e-05, "loss": 0.3354, "step": 61394 }, { "epoch": 3.71, "learning_rate": 1.6196641166085738e-05, "loss": 0.3423, "step": 61396 }, { "epoch": 3.71, "learning_rate": 1.619513225597151e-05, "loss": 0.3166, "step": 61398 }, { "epoch": 3.71, "learning_rate": 1.6193623345857288e-05, "loss": 0.4836, "step": 61400 }, { "epoch": 3.71, "learning_rate": 1.6192114435743064e-05, "loss": 0.4326, "step": 61402 }, { "epoch": 3.71, "learning_rate": 1.6190605525628837e-05, "loss": 0.2555, "step": 61404 }, { "epoch": 3.71, "learning_rate": 1.6189096615514614e-05, "loss": 0.481, "step": 61406 }, { "epoch": 3.71, "learning_rate": 1.618758770540039e-05, "loss": 0.3713, "step": 61408 }, { "epoch": 3.71, "learning_rate": 1.6186078795286167e-05, "loss": 0.4314, "step": 61410 }, { "epoch": 3.71, "learning_rate": 1.618456988517194e-05, "loss": 0.2968, "step": 61412 }, { "epoch": 3.71, "learning_rate": 1.6183060975057717e-05, "loss": 0.3639, "step": 61414 }, { "epoch": 3.71, "learning_rate": 1.6181552064943493e-05, "loss": 0.2632, "step": 61416 }, { "epoch": 3.71, "learning_rate": 1.6180043154829266e-05, "loss": 0.315, "step": 61418 }, { "epoch": 3.71, "learning_rate": 1.6178534244715043e-05, "loss": 0.2482, "step": 61420 }, { "epoch": 3.71, "learning_rate": 1.617702533460082e-05, "loss": 0.3593, "step": 61422 }, { "epoch": 3.71, "learning_rate": 1.6175516424486596e-05, "loss": 0.4351, "step": 61424 }, { "epoch": 3.71, "learning_rate": 1.617400751437237e-05, "loss": 0.3497, "step": 61426 }, { "epoch": 3.71, "learning_rate": 1.6172498604258145e-05, "loss": 0.2724, "step": 61428 }, { "epoch": 3.71, "learning_rate": 1.6170989694143922e-05, "loss": 0.4273, "step": 61430 }, { "epoch": 3.71, "learning_rate": 1.61694807840297e-05, "loss": 0.417, "step": 61432 }, { "epoch": 3.71, "learning_rate": 1.616797187391547e-05, "loss": 0.3881, "step": 61434 }, { "epoch": 3.71, "learning_rate": 1.6166462963801245e-05, "loss": 0.4744, "step": 61436 }, { "epoch": 3.71, "learning_rate": 1.6164954053687025e-05, "loss": 0.5011, "step": 61438 }, { "epoch": 3.71, "learning_rate": 1.6163445143572798e-05, "loss": 0.4453, "step": 61440 }, { "epoch": 3.71, "learning_rate": 1.6161936233458574e-05, "loss": 0.3582, "step": 61442 }, { "epoch": 3.71, "learning_rate": 1.6160427323344347e-05, "loss": 0.4079, "step": 61444 }, { "epoch": 3.71, "learning_rate": 1.6158918413230127e-05, "loss": 0.4435, "step": 61446 }, { "epoch": 3.71, "learning_rate": 1.61574095031159e-05, "loss": 0.4865, "step": 61448 }, { "epoch": 3.71, "learning_rate": 1.6155900593001673e-05, "loss": 0.4146, "step": 61450 }, { "epoch": 3.71, "learning_rate": 1.615439168288745e-05, "loss": 0.3113, "step": 61452 }, { "epoch": 3.71, "learning_rate": 1.6152882772773226e-05, "loss": 0.361, "step": 61454 }, { "epoch": 3.71, "learning_rate": 1.6151373862659003e-05, "loss": 0.3987, "step": 61456 }, { "epoch": 3.71, "learning_rate": 1.6149864952544776e-05, "loss": 0.4351, "step": 61458 }, { "epoch": 3.71, "learning_rate": 1.6148356042430553e-05, "loss": 0.4116, "step": 61460 }, { "epoch": 3.71, "learning_rate": 1.614684713231633e-05, "loss": 0.3811, "step": 61462 }, { "epoch": 3.71, "learning_rate": 1.6145338222202106e-05, "loss": 0.2812, "step": 61464 }, { "epoch": 3.71, "learning_rate": 1.614382931208788e-05, "loss": 0.4404, "step": 61466 }, { "epoch": 3.71, "learning_rate": 1.6142320401973655e-05, "loss": 0.4354, "step": 61468 }, { "epoch": 3.71, "learning_rate": 1.6140811491859432e-05, "loss": 0.3724, "step": 61470 }, { "epoch": 3.71, "learning_rate": 1.6139302581745205e-05, "loss": 0.3381, "step": 61472 }, { "epoch": 3.71, "learning_rate": 1.613779367163098e-05, "loss": 0.2543, "step": 61474 }, { "epoch": 3.71, "learning_rate": 1.6136284761516758e-05, "loss": 0.3834, "step": 61476 }, { "epoch": 3.71, "learning_rate": 1.6134775851402534e-05, "loss": 0.2966, "step": 61478 }, { "epoch": 3.71, "learning_rate": 1.6133266941288308e-05, "loss": 0.4181, "step": 61480 }, { "epoch": 3.71, "learning_rate": 1.613175803117408e-05, "loss": 0.3876, "step": 61482 }, { "epoch": 3.71, "learning_rate": 1.613024912105986e-05, "loss": 0.3543, "step": 61484 }, { "epoch": 3.71, "learning_rate": 1.6128740210945634e-05, "loss": 0.5198, "step": 61486 }, { "epoch": 3.71, "learning_rate": 1.612723130083141e-05, "loss": 0.3095, "step": 61488 }, { "epoch": 3.71, "learning_rate": 1.6125722390717183e-05, "loss": 0.3602, "step": 61490 }, { "epoch": 3.71, "learning_rate": 1.6124213480602963e-05, "loss": 0.4977, "step": 61492 }, { "epoch": 3.71, "learning_rate": 1.6122704570488736e-05, "loss": 0.4289, "step": 61494 }, { "epoch": 3.71, "learning_rate": 1.6121195660374513e-05, "loss": 0.242, "step": 61496 }, { "epoch": 3.71, "learning_rate": 1.611968675026029e-05, "loss": 0.4204, "step": 61498 }, { "epoch": 3.71, "learning_rate": 1.6118177840146063e-05, "loss": 0.2849, "step": 61500 }, { "epoch": 3.71, "learning_rate": 1.611666893003184e-05, "loss": 0.4392, "step": 61502 }, { "epoch": 3.71, "learning_rate": 1.6115160019917612e-05, "loss": 0.4397, "step": 61504 }, { "epoch": 3.71, "learning_rate": 1.6113651109803392e-05, "loss": 0.3557, "step": 61506 }, { "epoch": 3.71, "learning_rate": 1.6112142199689165e-05, "loss": 0.2482, "step": 61508 }, { "epoch": 3.71, "learning_rate": 1.6110633289574942e-05, "loss": 0.4679, "step": 61510 }, { "epoch": 3.71, "learning_rate": 1.6109124379460715e-05, "loss": 0.3894, "step": 61512 }, { "epoch": 3.71, "learning_rate": 1.6107615469346495e-05, "loss": 0.3482, "step": 61514 }, { "epoch": 3.71, "learning_rate": 1.6106106559232268e-05, "loss": 0.3289, "step": 61516 }, { "epoch": 3.71, "learning_rate": 1.610459764911804e-05, "loss": 0.2913, "step": 61518 }, { "epoch": 3.71, "learning_rate": 1.6103088739003818e-05, "loss": 0.2112, "step": 61520 }, { "epoch": 3.71, "learning_rate": 1.6101579828889594e-05, "loss": 0.3485, "step": 61522 }, { "epoch": 3.71, "learning_rate": 1.610007091877537e-05, "loss": 0.5543, "step": 61524 }, { "epoch": 3.71, "learning_rate": 1.6098562008661144e-05, "loss": 0.5415, "step": 61526 }, { "epoch": 3.71, "learning_rate": 1.609705309854692e-05, "loss": 0.3793, "step": 61528 }, { "epoch": 3.71, "learning_rate": 1.6095544188432697e-05, "loss": 0.3197, "step": 61530 }, { "epoch": 3.71, "learning_rate": 1.609403527831847e-05, "loss": 0.3952, "step": 61532 }, { "epoch": 3.71, "learning_rate": 1.6092526368204246e-05, "loss": 0.5188, "step": 61534 }, { "epoch": 3.71, "learning_rate": 1.6091017458090023e-05, "loss": 0.4107, "step": 61536 }, { "epoch": 3.71, "learning_rate": 1.60895085479758e-05, "loss": 0.4446, "step": 61538 }, { "epoch": 3.71, "learning_rate": 1.6087999637861572e-05, "loss": 0.3186, "step": 61540 }, { "epoch": 3.71, "learning_rate": 1.608649072774735e-05, "loss": 0.4106, "step": 61542 }, { "epoch": 3.71, "learning_rate": 1.6084981817633126e-05, "loss": 0.3305, "step": 61544 }, { "epoch": 3.71, "learning_rate": 1.6083472907518902e-05, "loss": 0.354, "step": 61546 }, { "epoch": 3.71, "learning_rate": 1.6081963997404675e-05, "loss": 0.3013, "step": 61548 }, { "epoch": 3.71, "learning_rate": 1.6080455087290448e-05, "loss": 0.3367, "step": 61550 }, { "epoch": 3.72, "learning_rate": 1.6078946177176228e-05, "loss": 0.6023, "step": 61552 }, { "epoch": 3.72, "learning_rate": 1.6077437267062e-05, "loss": 0.2689, "step": 61554 }, { "epoch": 3.72, "learning_rate": 1.6075928356947778e-05, "loss": 0.3904, "step": 61556 }, { "epoch": 3.72, "learning_rate": 1.607441944683355e-05, "loss": 0.5246, "step": 61558 }, { "epoch": 3.72, "learning_rate": 1.607291053671933e-05, "loss": 0.249, "step": 61560 }, { "epoch": 3.72, "learning_rate": 1.6071401626605104e-05, "loss": 0.2915, "step": 61562 }, { "epoch": 3.72, "learning_rate": 1.606989271649088e-05, "loss": 0.2335, "step": 61564 }, { "epoch": 3.72, "learning_rate": 1.6068383806376654e-05, "loss": 0.3541, "step": 61566 }, { "epoch": 3.72, "learning_rate": 1.606687489626243e-05, "loss": 0.3746, "step": 61568 }, { "epoch": 3.72, "learning_rate": 1.6065365986148207e-05, "loss": 0.4709, "step": 61570 }, { "epoch": 3.72, "learning_rate": 1.606385707603398e-05, "loss": 0.38, "step": 61572 }, { "epoch": 3.72, "learning_rate": 1.6062348165919756e-05, "loss": 0.2797, "step": 61574 }, { "epoch": 3.72, "learning_rate": 1.6060839255805533e-05, "loss": 0.5031, "step": 61576 }, { "epoch": 3.72, "learning_rate": 1.605933034569131e-05, "loss": 0.3425, "step": 61578 }, { "epoch": 3.72, "learning_rate": 1.6057821435577082e-05, "loss": 0.2123, "step": 61580 }, { "epoch": 3.72, "learning_rate": 1.605631252546286e-05, "loss": 0.3739, "step": 61582 }, { "epoch": 3.72, "learning_rate": 1.6054803615348635e-05, "loss": 0.2668, "step": 61584 }, { "epoch": 3.72, "learning_rate": 1.605329470523441e-05, "loss": 0.3838, "step": 61586 }, { "epoch": 3.72, "learning_rate": 1.6051785795120185e-05, "loss": 0.3182, "step": 61588 }, { "epoch": 3.72, "learning_rate": 1.605027688500596e-05, "loss": 0.2992, "step": 61590 }, { "epoch": 3.72, "learning_rate": 1.6048767974891738e-05, "loss": 0.2632, "step": 61592 }, { "epoch": 3.72, "learning_rate": 1.604725906477751e-05, "loss": 0.2866, "step": 61594 }, { "epoch": 3.72, "learning_rate": 1.6045750154663288e-05, "loss": 0.34, "step": 61596 }, { "epoch": 3.72, "learning_rate": 1.6044241244549064e-05, "loss": 0.339, "step": 61598 }, { "epoch": 3.72, "learning_rate": 1.6042732334434837e-05, "loss": 0.2766, "step": 61600 }, { "epoch": 3.72, "learning_rate": 1.6041223424320614e-05, "loss": 0.2405, "step": 61602 }, { "epoch": 3.72, "learning_rate": 1.6039714514206387e-05, "loss": 0.4836, "step": 61604 }, { "epoch": 3.72, "learning_rate": 1.6038205604092167e-05, "loss": 0.2316, "step": 61606 }, { "epoch": 3.72, "learning_rate": 1.603669669397794e-05, "loss": 0.3055, "step": 61608 }, { "epoch": 3.72, "learning_rate": 1.6035187783863717e-05, "loss": 0.2936, "step": 61610 }, { "epoch": 3.72, "learning_rate": 1.603367887374949e-05, "loss": 0.5215, "step": 61612 }, { "epoch": 3.72, "learning_rate": 1.6032169963635266e-05, "loss": 0.4642, "step": 61614 }, { "epoch": 3.72, "learning_rate": 1.6030661053521043e-05, "loss": 0.2912, "step": 61616 }, { "epoch": 3.72, "learning_rate": 1.6029152143406816e-05, "loss": 0.3049, "step": 61618 }, { "epoch": 3.72, "learning_rate": 1.6027643233292596e-05, "loss": 0.3302, "step": 61620 }, { "epoch": 3.72, "learning_rate": 1.602613432317837e-05, "loss": 0.411, "step": 61622 }, { "epoch": 3.72, "learning_rate": 1.6024625413064145e-05, "loss": 0.619, "step": 61624 }, { "epoch": 3.72, "learning_rate": 1.602311650294992e-05, "loss": 0.2813, "step": 61626 }, { "epoch": 3.72, "learning_rate": 1.60216075928357e-05, "loss": 0.3554, "step": 61628 }, { "epoch": 3.72, "learning_rate": 1.602009868272147e-05, "loss": 0.4107, "step": 61630 }, { "epoch": 3.72, "learning_rate": 1.6018589772607245e-05, "loss": 0.3655, "step": 61632 }, { "epoch": 3.72, "learning_rate": 1.601708086249302e-05, "loss": 0.4785, "step": 61634 }, { "epoch": 3.72, "learning_rate": 1.6015571952378798e-05, "loss": 0.4303, "step": 61636 }, { "epoch": 3.72, "learning_rate": 1.6014063042264574e-05, "loss": 0.4691, "step": 61638 }, { "epoch": 3.72, "learning_rate": 1.6012554132150347e-05, "loss": 0.3923, "step": 61640 }, { "epoch": 3.72, "learning_rate": 1.6011045222036124e-05, "loss": 0.2872, "step": 61642 }, { "epoch": 3.72, "learning_rate": 1.60095363119219e-05, "loss": 0.4511, "step": 61644 }, { "epoch": 3.72, "learning_rate": 1.6008027401807677e-05, "loss": 0.4161, "step": 61646 }, { "epoch": 3.72, "learning_rate": 1.600651849169345e-05, "loss": 0.3229, "step": 61648 }, { "epoch": 3.72, "learning_rate": 1.6005009581579227e-05, "loss": 0.4818, "step": 61650 }, { "epoch": 3.72, "learning_rate": 1.6003500671465003e-05, "loss": 0.2664, "step": 61652 }, { "epoch": 3.72, "learning_rate": 1.6001991761350776e-05, "loss": 0.4114, "step": 61654 }, { "epoch": 3.72, "learning_rate": 1.6000482851236553e-05, "loss": 0.2917, "step": 61656 }, { "epoch": 3.72, "learning_rate": 1.599897394112233e-05, "loss": 0.2783, "step": 61658 }, { "epoch": 3.72, "learning_rate": 1.5997465031008106e-05, "loss": 0.3579, "step": 61660 }, { "epoch": 3.72, "learning_rate": 1.599595612089388e-05, "loss": 0.5286, "step": 61662 }, { "epoch": 3.72, "learning_rate": 1.5994447210779652e-05, "loss": 0.2338, "step": 61664 }, { "epoch": 3.72, "learning_rate": 1.5992938300665432e-05, "loss": 0.4452, "step": 61666 }, { "epoch": 3.72, "learning_rate": 1.5991429390551205e-05, "loss": 0.3422, "step": 61668 }, { "epoch": 3.72, "learning_rate": 1.598992048043698e-05, "loss": 0.3544, "step": 61670 }, { "epoch": 3.72, "learning_rate": 1.5988411570322755e-05, "loss": 0.4582, "step": 61672 }, { "epoch": 3.72, "learning_rate": 1.5986902660208535e-05, "loss": 0.3725, "step": 61674 }, { "epoch": 3.72, "learning_rate": 1.5985393750094308e-05, "loss": 0.4429, "step": 61676 }, { "epoch": 3.72, "learning_rate": 1.5983884839980084e-05, "loss": 0.3917, "step": 61678 }, { "epoch": 3.72, "learning_rate": 1.5982375929865857e-05, "loss": 0.3008, "step": 61680 }, { "epoch": 3.72, "learning_rate": 1.5980867019751634e-05, "loss": 0.3363, "step": 61682 }, { "epoch": 3.72, "learning_rate": 1.597935810963741e-05, "loss": 0.2864, "step": 61684 }, { "epoch": 3.72, "learning_rate": 1.5977849199523183e-05, "loss": 0.3836, "step": 61686 }, { "epoch": 3.72, "learning_rate": 1.597634028940896e-05, "loss": 0.2941, "step": 61688 }, { "epoch": 3.72, "learning_rate": 1.5974831379294736e-05, "loss": 0.3912, "step": 61690 }, { "epoch": 3.72, "learning_rate": 1.5973322469180513e-05, "loss": 0.641, "step": 61692 }, { "epoch": 3.72, "learning_rate": 1.5971813559066286e-05, "loss": 0.591, "step": 61694 }, { "epoch": 3.72, "learning_rate": 1.5970304648952063e-05, "loss": 0.4424, "step": 61696 }, { "epoch": 3.72, "learning_rate": 1.596879573883784e-05, "loss": 0.245, "step": 61698 }, { "epoch": 3.72, "learning_rate": 1.5967286828723612e-05, "loss": 0.4694, "step": 61700 }, { "epoch": 3.72, "learning_rate": 1.596577791860939e-05, "loss": 0.2356, "step": 61702 }, { "epoch": 3.72, "learning_rate": 1.5964269008495165e-05, "loss": 0.4191, "step": 61704 }, { "epoch": 3.72, "learning_rate": 1.5962760098380942e-05, "loss": 0.5756, "step": 61706 }, { "epoch": 3.72, "learning_rate": 1.5961251188266715e-05, "loss": 0.4652, "step": 61708 }, { "epoch": 3.72, "learning_rate": 1.595974227815249e-05, "loss": 0.2549, "step": 61710 }, { "epoch": 3.72, "learning_rate": 1.5958233368038268e-05, "loss": 0.2515, "step": 61712 }, { "epoch": 3.72, "learning_rate": 1.595672445792404e-05, "loss": 0.3998, "step": 61714 }, { "epoch": 3.73, "learning_rate": 1.5955215547809818e-05, "loss": 0.3473, "step": 61716 }, { "epoch": 3.73, "learning_rate": 1.595370663769559e-05, "loss": 0.3253, "step": 61718 }, { "epoch": 3.73, "learning_rate": 1.595219772758137e-05, "loss": 0.2285, "step": 61720 }, { "epoch": 3.73, "learning_rate": 1.5950688817467144e-05, "loss": 0.3254, "step": 61722 }, { "epoch": 3.73, "learning_rate": 1.594917990735292e-05, "loss": 0.563, "step": 61724 }, { "epoch": 3.73, "learning_rate": 1.5947670997238693e-05, "loss": 0.3394, "step": 61726 }, { "epoch": 3.73, "learning_rate": 1.5946162087124473e-05, "loss": 0.3487, "step": 61728 }, { "epoch": 3.73, "learning_rate": 1.5944653177010246e-05, "loss": 0.4069, "step": 61730 }, { "epoch": 3.73, "learning_rate": 1.594314426689602e-05, "loss": 0.338, "step": 61732 }, { "epoch": 3.73, "learning_rate": 1.59416353567818e-05, "loss": 0.4742, "step": 61734 }, { "epoch": 3.73, "learning_rate": 1.5940126446667573e-05, "loss": 0.3877, "step": 61736 }, { "epoch": 3.73, "learning_rate": 1.593861753655335e-05, "loss": 0.4553, "step": 61738 }, { "epoch": 3.73, "learning_rate": 1.5937108626439122e-05, "loss": 0.3884, "step": 61740 }, { "epoch": 3.73, "learning_rate": 1.5935599716324902e-05, "loss": 0.421, "step": 61742 }, { "epoch": 3.73, "learning_rate": 1.5934090806210675e-05, "loss": 0.4309, "step": 61744 }, { "epoch": 3.73, "learning_rate": 1.593258189609645e-05, "loss": 0.5895, "step": 61746 }, { "epoch": 3.73, "learning_rate": 1.5931072985982225e-05, "loss": 0.4314, "step": 61748 }, { "epoch": 3.73, "learning_rate": 1.5929564075868e-05, "loss": 0.4385, "step": 61750 }, { "epoch": 3.73, "learning_rate": 1.5928055165753778e-05, "loss": 0.4653, "step": 61752 }, { "epoch": 3.73, "learning_rate": 1.592654625563955e-05, "loss": 0.4821, "step": 61754 }, { "epoch": 3.73, "learning_rate": 1.5925037345525328e-05, "loss": 0.3246, "step": 61756 }, { "epoch": 3.73, "learning_rate": 1.5923528435411104e-05, "loss": 0.4329, "step": 61758 }, { "epoch": 3.73, "learning_rate": 1.592201952529688e-05, "loss": 0.2981, "step": 61760 }, { "epoch": 3.73, "learning_rate": 1.5920510615182654e-05, "loss": 0.3416, "step": 61762 }, { "epoch": 3.73, "learning_rate": 1.591900170506843e-05, "loss": 0.2927, "step": 61764 }, { "epoch": 3.73, "learning_rate": 1.5917492794954207e-05, "loss": 0.4557, "step": 61766 }, { "epoch": 3.73, "learning_rate": 1.591598388483998e-05, "loss": 0.4124, "step": 61768 }, { "epoch": 3.73, "learning_rate": 1.5914474974725756e-05, "loss": 0.4269, "step": 61770 }, { "epoch": 3.73, "learning_rate": 1.5912966064611533e-05, "loss": 0.4275, "step": 61772 }, { "epoch": 3.73, "learning_rate": 1.591145715449731e-05, "loss": 0.2998, "step": 61774 }, { "epoch": 3.73, "learning_rate": 1.5909948244383083e-05, "loss": 0.2027, "step": 61776 }, { "epoch": 3.73, "learning_rate": 1.5908439334268856e-05, "loss": 0.2822, "step": 61778 }, { "epoch": 3.73, "learning_rate": 1.5906930424154636e-05, "loss": 0.361, "step": 61780 }, { "epoch": 3.73, "learning_rate": 1.590542151404041e-05, "loss": 0.2697, "step": 61782 }, { "epoch": 3.73, "learning_rate": 1.5903912603926185e-05, "loss": 0.5347, "step": 61784 }, { "epoch": 3.73, "learning_rate": 1.590240369381196e-05, "loss": 0.3704, "step": 61786 }, { "epoch": 3.73, "learning_rate": 1.5900894783697738e-05, "loss": 0.3349, "step": 61788 }, { "epoch": 3.73, "learning_rate": 1.589938587358351e-05, "loss": 0.5749, "step": 61790 }, { "epoch": 3.73, "learning_rate": 1.5897876963469288e-05, "loss": 0.324, "step": 61792 }, { "epoch": 3.73, "learning_rate": 1.589636805335506e-05, "loss": 0.4515, "step": 61794 }, { "epoch": 3.73, "learning_rate": 1.5894859143240837e-05, "loss": 0.3263, "step": 61796 }, { "epoch": 3.73, "learning_rate": 1.5893350233126614e-05, "loss": 0.2773, "step": 61798 }, { "epoch": 3.73, "learning_rate": 1.5891841323012387e-05, "loss": 0.2553, "step": 61800 }, { "epoch": 3.73, "learning_rate": 1.5890332412898164e-05, "loss": 0.2702, "step": 61802 }, { "epoch": 3.73, "learning_rate": 1.588882350278394e-05, "loss": 0.2548, "step": 61804 }, { "epoch": 3.73, "learning_rate": 1.5887314592669717e-05, "loss": 0.353, "step": 61806 }, { "epoch": 3.73, "learning_rate": 1.588580568255549e-05, "loss": 0.4978, "step": 61808 }, { "epoch": 3.73, "learning_rate": 1.5884296772441266e-05, "loss": 0.3908, "step": 61810 }, { "epoch": 3.73, "learning_rate": 1.5882787862327043e-05, "loss": 0.3167, "step": 61812 }, { "epoch": 3.73, "learning_rate": 1.5881278952212816e-05, "loss": 0.4852, "step": 61814 }, { "epoch": 3.73, "learning_rate": 1.5879770042098592e-05, "loss": 0.3884, "step": 61816 }, { "epoch": 3.73, "learning_rate": 1.587826113198437e-05, "loss": 0.2094, "step": 61818 }, { "epoch": 3.73, "learning_rate": 1.5876752221870145e-05, "loss": 0.3364, "step": 61820 }, { "epoch": 3.73, "learning_rate": 1.587524331175592e-05, "loss": 0.3863, "step": 61822 }, { "epoch": 3.73, "learning_rate": 1.5873734401641695e-05, "loss": 0.3471, "step": 61824 }, { "epoch": 3.73, "learning_rate": 1.587222549152747e-05, "loss": 0.326, "step": 61826 }, { "epoch": 3.73, "learning_rate": 1.5870716581413245e-05, "loss": 0.3375, "step": 61828 }, { "epoch": 3.73, "learning_rate": 1.586920767129902e-05, "loss": 0.4279, "step": 61830 }, { "epoch": 3.73, "learning_rate": 1.5867698761184794e-05, "loss": 0.3366, "step": 61832 }, { "epoch": 3.73, "learning_rate": 1.5866189851070574e-05, "loss": 0.2999, "step": 61834 }, { "epoch": 3.73, "learning_rate": 1.5864680940956347e-05, "loss": 0.1742, "step": 61836 }, { "epoch": 3.73, "learning_rate": 1.5863172030842124e-05, "loss": 0.3983, "step": 61838 }, { "epoch": 3.73, "learning_rate": 1.5861663120727897e-05, "loss": 0.4752, "step": 61840 }, { "epoch": 3.73, "learning_rate": 1.5860154210613677e-05, "loss": 0.3535, "step": 61842 }, { "epoch": 3.73, "learning_rate": 1.585864530049945e-05, "loss": 0.2168, "step": 61844 }, { "epoch": 3.73, "learning_rate": 1.5857136390385223e-05, "loss": 0.3855, "step": 61846 }, { "epoch": 3.73, "learning_rate": 1.5855627480271e-05, "loss": 0.3524, "step": 61848 }, { "epoch": 3.73, "learning_rate": 1.5854118570156776e-05, "loss": 0.3843, "step": 61850 }, { "epoch": 3.73, "learning_rate": 1.5852609660042553e-05, "loss": 0.2677, "step": 61852 }, { "epoch": 3.73, "learning_rate": 1.5851100749928326e-05, "loss": 0.3351, "step": 61854 }, { "epoch": 3.73, "learning_rate": 1.5849591839814106e-05, "loss": 0.3986, "step": 61856 }, { "epoch": 3.73, "learning_rate": 1.584808292969988e-05, "loss": 0.4483, "step": 61858 }, { "epoch": 3.73, "learning_rate": 1.5846574019585652e-05, "loss": 0.3165, "step": 61860 }, { "epoch": 3.73, "learning_rate": 1.584506510947143e-05, "loss": 0.2741, "step": 61862 }, { "epoch": 3.73, "learning_rate": 1.5843556199357205e-05, "loss": 0.3809, "step": 61864 }, { "epoch": 3.73, "learning_rate": 1.584204728924298e-05, "loss": 0.2498, "step": 61866 }, { "epoch": 3.73, "learning_rate": 1.5840538379128755e-05, "loss": 0.3599, "step": 61868 }, { "epoch": 3.73, "learning_rate": 1.583902946901453e-05, "loss": 0.5999, "step": 61870 }, { "epoch": 3.73, "learning_rate": 1.5837520558900308e-05, "loss": 0.4667, "step": 61872 }, { "epoch": 3.73, "learning_rate": 1.5836011648786084e-05, "loss": 0.2621, "step": 61874 }, { "epoch": 3.73, "learning_rate": 1.5834502738671857e-05, "loss": 0.4937, "step": 61876 }, { "epoch": 3.73, "learning_rate": 1.5832993828557634e-05, "loss": 0.3577, "step": 61878 }, { "epoch": 3.73, "learning_rate": 1.583148491844341e-05, "loss": 0.2695, "step": 61880 }, { "epoch": 3.74, "learning_rate": 1.5829976008329184e-05, "loss": 0.3205, "step": 61882 }, { "epoch": 3.74, "learning_rate": 1.582846709821496e-05, "loss": 0.4505, "step": 61884 }, { "epoch": 3.74, "learning_rate": 1.5826958188100737e-05, "loss": 0.3995, "step": 61886 }, { "epoch": 3.74, "learning_rate": 1.5825449277986513e-05, "loss": 0.4542, "step": 61888 }, { "epoch": 3.74, "learning_rate": 1.5823940367872286e-05, "loss": 0.5313, "step": 61890 }, { "epoch": 3.74, "learning_rate": 1.5822431457758063e-05, "loss": 0.4607, "step": 61892 }, { "epoch": 3.74, "learning_rate": 1.582092254764384e-05, "loss": 0.4333, "step": 61894 }, { "epoch": 3.74, "learning_rate": 1.5819413637529612e-05, "loss": 0.4456, "step": 61896 }, { "epoch": 3.74, "learning_rate": 1.581790472741539e-05, "loss": 0.3449, "step": 61898 }, { "epoch": 3.74, "learning_rate": 1.5816395817301162e-05, "loss": 0.2877, "step": 61900 }, { "epoch": 3.74, "learning_rate": 1.5814886907186942e-05, "loss": 0.4104, "step": 61902 }, { "epoch": 3.74, "learning_rate": 1.5813377997072715e-05, "loss": 0.5102, "step": 61904 }, { "epoch": 3.74, "learning_rate": 1.581186908695849e-05, "loss": 0.3446, "step": 61906 }, { "epoch": 3.74, "learning_rate": 1.5810360176844265e-05, "loss": 0.4044, "step": 61908 }, { "epoch": 3.74, "learning_rate": 1.580885126673004e-05, "loss": 0.173, "step": 61910 }, { "epoch": 3.74, "learning_rate": 1.5807342356615818e-05, "loss": 0.3362, "step": 61912 }, { "epoch": 3.74, "learning_rate": 1.580583344650159e-05, "loss": 0.351, "step": 61914 }, { "epoch": 3.74, "learning_rate": 1.5804324536387367e-05, "loss": 0.4651, "step": 61916 }, { "epoch": 3.74, "learning_rate": 1.5802815626273144e-05, "loss": 0.3688, "step": 61918 }, { "epoch": 3.74, "learning_rate": 1.580130671615892e-05, "loss": 0.4639, "step": 61920 }, { "epoch": 3.74, "learning_rate": 1.5799797806044693e-05, "loss": 0.3568, "step": 61922 }, { "epoch": 3.74, "learning_rate": 1.579828889593047e-05, "loss": 0.2693, "step": 61924 }, { "epoch": 3.74, "learning_rate": 1.5796779985816247e-05, "loss": 0.3557, "step": 61926 }, { "epoch": 3.74, "learning_rate": 1.579527107570202e-05, "loss": 0.2949, "step": 61928 }, { "epoch": 3.74, "learning_rate": 1.5793762165587796e-05, "loss": 0.3484, "step": 61930 }, { "epoch": 3.74, "learning_rate": 1.5792253255473573e-05, "loss": 0.4271, "step": 61932 }, { "epoch": 3.74, "learning_rate": 1.579074434535935e-05, "loss": 0.2705, "step": 61934 }, { "epoch": 3.74, "learning_rate": 1.5789235435245122e-05, "loss": 0.2606, "step": 61936 }, { "epoch": 3.74, "learning_rate": 1.57877265251309e-05, "loss": 0.6411, "step": 61938 }, { "epoch": 3.74, "learning_rate": 1.5786217615016675e-05, "loss": 0.3728, "step": 61940 }, { "epoch": 3.74, "learning_rate": 1.578470870490245e-05, "loss": 0.4377, "step": 61942 }, { "epoch": 3.74, "learning_rate": 1.5783199794788225e-05, "loss": 0.2601, "step": 61944 }, { "epoch": 3.74, "learning_rate": 1.5781690884673998e-05, "loss": 0.317, "step": 61946 }, { "epoch": 3.74, "learning_rate": 1.5780181974559778e-05, "loss": 0.4054, "step": 61948 }, { "epoch": 3.74, "learning_rate": 1.577867306444555e-05, "loss": 0.4139, "step": 61950 }, { "epoch": 3.74, "learning_rate": 1.5777164154331328e-05, "loss": 0.4393, "step": 61952 }, { "epoch": 3.74, "learning_rate": 1.57756552442171e-05, "loss": 0.3745, "step": 61954 }, { "epoch": 3.74, "learning_rate": 1.577414633410288e-05, "loss": 0.2932, "step": 61956 }, { "epoch": 3.74, "learning_rate": 1.5772637423988654e-05, "loss": 0.2626, "step": 61958 }, { "epoch": 3.74, "learning_rate": 1.5771128513874427e-05, "loss": 0.4116, "step": 61960 }, { "epoch": 3.74, "learning_rate": 1.5769619603760203e-05, "loss": 0.2794, "step": 61962 }, { "epoch": 3.74, "learning_rate": 1.576811069364598e-05, "loss": 0.3987, "step": 61964 }, { "epoch": 3.74, "learning_rate": 1.5766601783531756e-05, "loss": 0.2768, "step": 61966 }, { "epoch": 3.74, "learning_rate": 1.576509287341753e-05, "loss": 0.5198, "step": 61968 }, { "epoch": 3.74, "learning_rate": 1.5763583963303306e-05, "loss": 0.419, "step": 61970 }, { "epoch": 3.74, "learning_rate": 1.5762075053189083e-05, "loss": 0.3349, "step": 61972 }, { "epoch": 3.74, "learning_rate": 1.576056614307486e-05, "loss": 0.3401, "step": 61974 }, { "epoch": 3.74, "learning_rate": 1.5759057232960632e-05, "loss": 0.3818, "step": 61976 }, { "epoch": 3.74, "learning_rate": 1.575754832284641e-05, "loss": 0.3533, "step": 61978 }, { "epoch": 3.74, "learning_rate": 1.5756039412732185e-05, "loss": 0.2977, "step": 61980 }, { "epoch": 3.74, "learning_rate": 1.575453050261796e-05, "loss": 0.3994, "step": 61982 }, { "epoch": 3.74, "learning_rate": 1.5753021592503735e-05, "loss": 0.3454, "step": 61984 }, { "epoch": 3.74, "learning_rate": 1.575151268238951e-05, "loss": 0.413, "step": 61986 }, { "epoch": 3.74, "learning_rate": 1.5750003772275288e-05, "loss": 0.2589, "step": 61988 }, { "epoch": 3.74, "learning_rate": 1.574849486216106e-05, "loss": 0.416, "step": 61990 }, { "epoch": 3.74, "learning_rate": 1.5746985952046838e-05, "loss": 0.4245, "step": 61992 }, { "epoch": 3.74, "learning_rate": 1.5745477041932614e-05, "loss": 0.3789, "step": 61994 }, { "epoch": 3.74, "learning_rate": 1.5743968131818387e-05, "loss": 0.259, "step": 61996 }, { "epoch": 3.74, "learning_rate": 1.5742459221704164e-05, "loss": 0.3091, "step": 61998 }, { "epoch": 3.74, "learning_rate": 1.574095031158994e-05, "loss": 0.333, "step": 62000 }, { "epoch": 3.74, "learning_rate": 1.5739441401475717e-05, "loss": 0.2789, "step": 62002 }, { "epoch": 3.74, "learning_rate": 1.573793249136149e-05, "loss": 0.333, "step": 62004 }, { "epoch": 3.74, "learning_rate": 1.5736423581247266e-05, "loss": 0.297, "step": 62006 }, { "epoch": 3.74, "learning_rate": 1.5734914671133043e-05, "loss": 0.4641, "step": 62008 }, { "epoch": 3.74, "learning_rate": 1.5733405761018816e-05, "loss": 0.5584, "step": 62010 }, { "epoch": 3.74, "learning_rate": 1.5731896850904593e-05, "loss": 0.3692, "step": 62012 }, { "epoch": 3.74, "learning_rate": 1.5730387940790366e-05, "loss": 0.5131, "step": 62014 }, { "epoch": 3.74, "learning_rate": 1.5728879030676146e-05, "loss": 0.2886, "step": 62016 }, { "epoch": 3.74, "learning_rate": 1.572737012056192e-05, "loss": 0.4184, "step": 62018 }, { "epoch": 3.74, "learning_rate": 1.5725861210447695e-05, "loss": 0.217, "step": 62020 }, { "epoch": 3.74, "learning_rate": 1.572435230033347e-05, "loss": 0.4803, "step": 62022 }, { "epoch": 3.74, "learning_rate": 1.5722843390219245e-05, "loss": 0.4382, "step": 62024 }, { "epoch": 3.74, "learning_rate": 1.572133448010502e-05, "loss": 0.3614, "step": 62026 }, { "epoch": 3.74, "learning_rate": 1.5719825569990794e-05, "loss": 0.3627, "step": 62028 }, { "epoch": 3.74, "learning_rate": 1.571831665987657e-05, "loss": 0.3795, "step": 62030 }, { "epoch": 3.74, "learning_rate": 1.5716807749762348e-05, "loss": 0.255, "step": 62032 }, { "epoch": 3.74, "learning_rate": 1.5715298839648124e-05, "loss": 0.3347, "step": 62034 }, { "epoch": 3.74, "learning_rate": 1.5713789929533897e-05, "loss": 0.3047, "step": 62036 }, { "epoch": 3.74, "learning_rate": 1.5712281019419674e-05, "loss": 0.393, "step": 62038 }, { "epoch": 3.74, "learning_rate": 1.571077210930545e-05, "loss": 0.3076, "step": 62040 }, { "epoch": 3.74, "learning_rate": 1.5709263199191223e-05, "loss": 0.3733, "step": 62042 }, { "epoch": 3.74, "learning_rate": 1.5707754289077e-05, "loss": 0.3258, "step": 62044 }, { "epoch": 3.74, "learning_rate": 1.5706245378962776e-05, "loss": 0.3599, "step": 62046 }, { "epoch": 3.75, "learning_rate": 1.5704736468848553e-05, "loss": 0.3432, "step": 62048 }, { "epoch": 3.75, "learning_rate": 1.5703227558734326e-05, "loss": 0.3203, "step": 62050 }, { "epoch": 3.75, "learning_rate": 1.5701718648620102e-05, "loss": 0.3643, "step": 62052 }, { "epoch": 3.75, "learning_rate": 1.570020973850588e-05, "loss": 0.2784, "step": 62054 }, { "epoch": 3.75, "learning_rate": 1.5698700828391656e-05, "loss": 0.435, "step": 62056 }, { "epoch": 3.75, "learning_rate": 1.569719191827743e-05, "loss": 0.2875, "step": 62058 }, { "epoch": 3.75, "learning_rate": 1.5695683008163202e-05, "loss": 0.3335, "step": 62060 }, { "epoch": 3.75, "learning_rate": 1.569417409804898e-05, "loss": 0.367, "step": 62062 }, { "epoch": 3.75, "learning_rate": 1.5692665187934755e-05, "loss": 0.1606, "step": 62064 }, { "epoch": 3.75, "learning_rate": 1.569115627782053e-05, "loss": 0.3913, "step": 62066 }, { "epoch": 3.75, "learning_rate": 1.5689647367706304e-05, "loss": 0.5262, "step": 62068 }, { "epoch": 3.75, "learning_rate": 1.5688138457592084e-05, "loss": 0.4219, "step": 62070 }, { "epoch": 3.75, "learning_rate": 1.5686629547477857e-05, "loss": 0.4704, "step": 62072 }, { "epoch": 3.75, "learning_rate": 1.568512063736363e-05, "loss": 0.3039, "step": 62074 }, { "epoch": 3.75, "learning_rate": 1.5683611727249407e-05, "loss": 0.4948, "step": 62076 }, { "epoch": 3.75, "learning_rate": 1.5682102817135184e-05, "loss": 0.4098, "step": 62078 }, { "epoch": 3.75, "learning_rate": 1.568059390702096e-05, "loss": 0.3926, "step": 62080 }, { "epoch": 3.75, "learning_rate": 1.5679084996906733e-05, "loss": 0.2733, "step": 62082 }, { "epoch": 3.75, "learning_rate": 1.567757608679251e-05, "loss": 0.3548, "step": 62084 }, { "epoch": 3.75, "learning_rate": 1.5676067176678286e-05, "loss": 0.4244, "step": 62086 }, { "epoch": 3.75, "learning_rate": 1.5674558266564063e-05, "loss": 0.4551, "step": 62088 }, { "epoch": 3.75, "learning_rate": 1.5673049356449836e-05, "loss": 0.312, "step": 62090 }, { "epoch": 3.75, "learning_rate": 1.5671540446335612e-05, "loss": 0.3859, "step": 62092 }, { "epoch": 3.75, "learning_rate": 1.567003153622139e-05, "loss": 0.3837, "step": 62094 }, { "epoch": 3.75, "learning_rate": 1.5668522626107162e-05, "loss": 0.3997, "step": 62096 }, { "epoch": 3.75, "learning_rate": 1.566701371599294e-05, "loss": 0.3099, "step": 62098 }, { "epoch": 3.75, "learning_rate": 1.5665504805878715e-05, "loss": 0.5283, "step": 62100 }, { "epoch": 3.75, "learning_rate": 1.566399589576449e-05, "loss": 0.4589, "step": 62102 }, { "epoch": 3.75, "learning_rate": 1.5662486985650265e-05, "loss": 0.4408, "step": 62104 }, { "epoch": 3.75, "learning_rate": 1.566097807553604e-05, "loss": 0.4415, "step": 62106 }, { "epoch": 3.75, "learning_rate": 1.5659469165421818e-05, "loss": 0.3715, "step": 62108 }, { "epoch": 3.75, "learning_rate": 1.565796025530759e-05, "loss": 0.2571, "step": 62110 }, { "epoch": 3.75, "learning_rate": 1.5656451345193367e-05, "loss": 0.3036, "step": 62112 }, { "epoch": 3.75, "learning_rate": 1.5654942435079144e-05, "loss": 0.4908, "step": 62114 }, { "epoch": 3.75, "learning_rate": 1.565343352496492e-05, "loss": 0.241, "step": 62116 }, { "epoch": 3.75, "learning_rate": 1.5651924614850694e-05, "loss": 0.3013, "step": 62118 }, { "epoch": 3.75, "learning_rate": 1.565041570473647e-05, "loss": 0.361, "step": 62120 }, { "epoch": 3.75, "learning_rate": 1.5648906794622247e-05, "loss": 0.4342, "step": 62122 }, { "epoch": 3.75, "learning_rate": 1.564739788450802e-05, "loss": 0.2325, "step": 62124 }, { "epoch": 3.75, "learning_rate": 1.5645888974393796e-05, "loss": 0.2253, "step": 62126 }, { "epoch": 3.75, "learning_rate": 1.564438006427957e-05, "loss": 0.3786, "step": 62128 }, { "epoch": 3.75, "learning_rate": 1.564287115416535e-05, "loss": 0.2674, "step": 62130 }, { "epoch": 3.75, "learning_rate": 1.5641362244051122e-05, "loss": 0.6622, "step": 62132 }, { "epoch": 3.75, "learning_rate": 1.56398533339369e-05, "loss": 0.5308, "step": 62134 }, { "epoch": 3.75, "learning_rate": 1.5638344423822672e-05, "loss": 0.275, "step": 62136 }, { "epoch": 3.75, "learning_rate": 1.5636835513708452e-05, "loss": 0.2799, "step": 62138 }, { "epoch": 3.75, "learning_rate": 1.5635326603594225e-05, "loss": 0.4763, "step": 62140 }, { "epoch": 3.75, "learning_rate": 1.5633817693479998e-05, "loss": 0.35, "step": 62142 }, { "epoch": 3.75, "learning_rate": 1.5632308783365775e-05, "loss": 0.3972, "step": 62144 }, { "epoch": 3.75, "learning_rate": 1.563079987325155e-05, "loss": 0.4567, "step": 62146 }, { "epoch": 3.75, "learning_rate": 1.5629290963137328e-05, "loss": 0.3778, "step": 62148 }, { "epoch": 3.75, "learning_rate": 1.56277820530231e-05, "loss": 0.4537, "step": 62150 }, { "epoch": 3.75, "learning_rate": 1.5626273142908877e-05, "loss": 0.3875, "step": 62152 }, { "epoch": 3.75, "learning_rate": 1.5624764232794654e-05, "loss": 0.3951, "step": 62154 }, { "epoch": 3.75, "learning_rate": 1.5623255322680427e-05, "loss": 0.415, "step": 62156 }, { "epoch": 3.75, "learning_rate": 1.5621746412566203e-05, "loss": 0.4018, "step": 62158 }, { "epoch": 3.75, "learning_rate": 1.562023750245198e-05, "loss": 0.3067, "step": 62160 }, { "epoch": 3.75, "learning_rate": 1.5618728592337757e-05, "loss": 0.5127, "step": 62162 }, { "epoch": 3.75, "learning_rate": 1.561721968222353e-05, "loss": 0.4758, "step": 62164 }, { "epoch": 3.75, "learning_rate": 1.5615710772109306e-05, "loss": 0.381, "step": 62166 }, { "epoch": 3.75, "learning_rate": 1.5614201861995083e-05, "loss": 0.189, "step": 62168 }, { "epoch": 3.75, "learning_rate": 1.561269295188086e-05, "loss": 0.3327, "step": 62170 }, { "epoch": 3.75, "learning_rate": 1.5611184041766632e-05, "loss": 0.5905, "step": 62172 }, { "epoch": 3.75, "learning_rate": 1.5609675131652405e-05, "loss": 0.3176, "step": 62174 }, { "epoch": 3.75, "learning_rate": 1.5608166221538185e-05, "loss": 0.3706, "step": 62176 }, { "epoch": 3.75, "learning_rate": 1.560665731142396e-05, "loss": 0.4171, "step": 62178 }, { "epoch": 3.75, "learning_rate": 1.5605148401309735e-05, "loss": 0.2621, "step": 62180 }, { "epoch": 3.75, "learning_rate": 1.5603639491195508e-05, "loss": 0.4491, "step": 62182 }, { "epoch": 3.75, "learning_rate": 1.5602130581081288e-05, "loss": 0.2667, "step": 62184 }, { "epoch": 3.75, "learning_rate": 1.560062167096706e-05, "loss": 0.2022, "step": 62186 }, { "epoch": 3.75, "learning_rate": 1.5599112760852838e-05, "loss": 0.4832, "step": 62188 }, { "epoch": 3.75, "learning_rate": 1.559760385073861e-05, "loss": 0.2923, "step": 62190 }, { "epoch": 3.75, "learning_rate": 1.5596094940624387e-05, "loss": 0.279, "step": 62192 }, { "epoch": 3.75, "learning_rate": 1.5594586030510164e-05, "loss": 0.2865, "step": 62194 }, { "epoch": 3.75, "learning_rate": 1.5593077120395937e-05, "loss": 0.3803, "step": 62196 }, { "epoch": 3.75, "learning_rate": 1.5591568210281713e-05, "loss": 0.2368, "step": 62198 }, { "epoch": 3.75, "learning_rate": 1.559005930016749e-05, "loss": 0.2653, "step": 62200 }, { "epoch": 3.75, "learning_rate": 1.5588550390053266e-05, "loss": 0.2362, "step": 62202 }, { "epoch": 3.75, "learning_rate": 1.558704147993904e-05, "loss": 0.4449, "step": 62204 }, { "epoch": 3.75, "learning_rate": 1.5585532569824816e-05, "loss": 0.3241, "step": 62206 }, { "epoch": 3.75, "learning_rate": 1.5584023659710593e-05, "loss": 0.2602, "step": 62208 }, { "epoch": 3.75, "learning_rate": 1.5582514749596366e-05, "loss": 0.3529, "step": 62210 }, { "epoch": 3.75, "learning_rate": 1.5581005839482142e-05, "loss": 0.2004, "step": 62212 }, { "epoch": 3.76, "learning_rate": 1.557949692936792e-05, "loss": 0.3456, "step": 62214 }, { "epoch": 3.76, "learning_rate": 1.5577988019253695e-05, "loss": 0.3946, "step": 62216 }, { "epoch": 3.76, "learning_rate": 1.557647910913947e-05, "loss": 0.1855, "step": 62218 }, { "epoch": 3.76, "learning_rate": 1.5574970199025245e-05, "loss": 0.5109, "step": 62220 }, { "epoch": 3.76, "learning_rate": 1.557346128891102e-05, "loss": 0.2683, "step": 62222 }, { "epoch": 3.76, "learning_rate": 1.5571952378796795e-05, "loss": 0.3274, "step": 62224 }, { "epoch": 3.76, "learning_rate": 1.557044346868257e-05, "loss": 0.3999, "step": 62226 }, { "epoch": 3.76, "learning_rate": 1.5568934558568348e-05, "loss": 0.3485, "step": 62228 }, { "epoch": 3.76, "learning_rate": 1.5567425648454124e-05, "loss": 0.3702, "step": 62230 }, { "epoch": 3.76, "learning_rate": 1.5565916738339897e-05, "loss": 0.4313, "step": 62232 }, { "epoch": 3.76, "learning_rate": 1.5564407828225674e-05, "loss": 0.4362, "step": 62234 }, { "epoch": 3.76, "learning_rate": 1.556289891811145e-05, "loss": 0.3852, "step": 62236 }, { "epoch": 3.76, "learning_rate": 1.5561390007997223e-05, "loss": 0.2731, "step": 62238 }, { "epoch": 3.76, "learning_rate": 1.5559881097883e-05, "loss": 0.2633, "step": 62240 }, { "epoch": 3.76, "learning_rate": 1.5558372187768773e-05, "loss": 0.3134, "step": 62242 }, { "epoch": 3.76, "learning_rate": 1.5556863277654553e-05, "loss": 0.3462, "step": 62244 }, { "epoch": 3.76, "learning_rate": 1.5555354367540326e-05, "loss": 0.2354, "step": 62246 }, { "epoch": 3.76, "learning_rate": 1.5553845457426103e-05, "loss": 0.3433, "step": 62248 }, { "epoch": 3.76, "learning_rate": 1.5552336547311876e-05, "loss": 0.4147, "step": 62250 }, { "epoch": 3.76, "learning_rate": 1.5550827637197656e-05, "loss": 0.3854, "step": 62252 }, { "epoch": 3.76, "learning_rate": 1.554931872708343e-05, "loss": 0.3016, "step": 62254 }, { "epoch": 3.76, "learning_rate": 1.5547809816969202e-05, "loss": 0.3343, "step": 62256 }, { "epoch": 3.76, "learning_rate": 1.554630090685498e-05, "loss": 0.3166, "step": 62258 }, { "epoch": 3.76, "learning_rate": 1.5544791996740755e-05, "loss": 0.4021, "step": 62260 }, { "epoch": 3.76, "learning_rate": 1.554328308662653e-05, "loss": 0.3899, "step": 62262 }, { "epoch": 3.76, "learning_rate": 1.5541774176512305e-05, "loss": 0.4165, "step": 62264 }, { "epoch": 3.76, "learning_rate": 1.554026526639808e-05, "loss": 0.3215, "step": 62266 }, { "epoch": 3.76, "learning_rate": 1.5538756356283858e-05, "loss": 0.2688, "step": 62268 }, { "epoch": 3.76, "learning_rate": 1.5537247446169634e-05, "loss": 0.2594, "step": 62270 }, { "epoch": 3.76, "learning_rate": 1.5535738536055407e-05, "loss": 0.2489, "step": 62272 }, { "epoch": 3.76, "learning_rate": 1.5534229625941184e-05, "loss": 0.3913, "step": 62274 }, { "epoch": 3.76, "learning_rate": 1.553272071582696e-05, "loss": 0.3014, "step": 62276 }, { "epoch": 3.76, "learning_rate": 1.5531211805712733e-05, "loss": 0.3856, "step": 62278 }, { "epoch": 3.76, "learning_rate": 1.552970289559851e-05, "loss": 0.4028, "step": 62280 }, { "epoch": 3.76, "learning_rate": 1.5528193985484286e-05, "loss": 0.4576, "step": 62282 }, { "epoch": 3.76, "learning_rate": 1.5526685075370063e-05, "loss": 0.3063, "step": 62284 }, { "epoch": 3.76, "learning_rate": 1.5525176165255836e-05, "loss": 0.2081, "step": 62286 }, { "epoch": 3.76, "learning_rate": 1.552366725514161e-05, "loss": 0.6035, "step": 62288 }, { "epoch": 3.76, "learning_rate": 1.552215834502739e-05, "loss": 0.4576, "step": 62290 }, { "epoch": 3.76, "learning_rate": 1.5520649434913162e-05, "loss": 0.58, "step": 62292 }, { "epoch": 3.76, "learning_rate": 1.551914052479894e-05, "loss": 0.3992, "step": 62294 }, { "epoch": 3.76, "learning_rate": 1.5517631614684712e-05, "loss": 0.3821, "step": 62296 }, { "epoch": 3.76, "learning_rate": 1.551612270457049e-05, "loss": 0.442, "step": 62298 }, { "epoch": 3.76, "learning_rate": 1.5514613794456265e-05, "loss": 0.3077, "step": 62300 }, { "epoch": 3.76, "learning_rate": 1.551310488434204e-05, "loss": 0.3611, "step": 62302 }, { "epoch": 3.76, "learning_rate": 1.5511595974227814e-05, "loss": 0.3149, "step": 62304 }, { "epoch": 3.76, "learning_rate": 1.551008706411359e-05, "loss": 0.2944, "step": 62306 }, { "epoch": 3.76, "learning_rate": 1.5508578153999367e-05, "loss": 0.2708, "step": 62308 }, { "epoch": 3.76, "learning_rate": 1.550706924388514e-05, "loss": 0.516, "step": 62310 }, { "epoch": 3.76, "learning_rate": 1.5505560333770917e-05, "loss": 0.3814, "step": 62312 }, { "epoch": 3.76, "learning_rate": 1.5504051423656694e-05, "loss": 0.2719, "step": 62314 }, { "epoch": 3.76, "learning_rate": 1.550254251354247e-05, "loss": 0.3751, "step": 62316 }, { "epoch": 3.76, "learning_rate": 1.5501033603428243e-05, "loss": 0.3442, "step": 62318 }, { "epoch": 3.76, "learning_rate": 1.549952469331402e-05, "loss": 0.3004, "step": 62320 }, { "epoch": 3.76, "learning_rate": 1.5498015783199796e-05, "loss": 0.2557, "step": 62322 }, { "epoch": 3.76, "learning_rate": 1.549650687308557e-05, "loss": 0.4409, "step": 62324 }, { "epoch": 3.76, "learning_rate": 1.5494997962971346e-05, "loss": 0.2402, "step": 62326 }, { "epoch": 3.76, "learning_rate": 1.5493489052857122e-05, "loss": 0.2422, "step": 62328 }, { "epoch": 3.76, "learning_rate": 1.54919801427429e-05, "loss": 0.3747, "step": 62330 }, { "epoch": 3.76, "learning_rate": 1.5490471232628672e-05, "loss": 0.2623, "step": 62332 }, { "epoch": 3.76, "learning_rate": 1.548896232251445e-05, "loss": 0.2169, "step": 62334 }, { "epoch": 3.76, "learning_rate": 1.5487453412400225e-05, "loss": 0.2952, "step": 62336 }, { "epoch": 3.76, "learning_rate": 1.5485944502285998e-05, "loss": 0.4129, "step": 62338 }, { "epoch": 3.76, "learning_rate": 1.5484435592171775e-05, "loss": 0.4207, "step": 62340 }, { "epoch": 3.76, "learning_rate": 1.548292668205755e-05, "loss": 0.4178, "step": 62342 }, { "epoch": 3.76, "learning_rate": 1.5481417771943328e-05, "loss": 0.3417, "step": 62344 }, { "epoch": 3.76, "learning_rate": 1.54799088618291e-05, "loss": 0.4239, "step": 62346 }, { "epoch": 3.76, "learning_rate": 1.5478399951714877e-05, "loss": 0.4096, "step": 62348 }, { "epoch": 3.76, "learning_rate": 1.5476891041600654e-05, "loss": 0.3899, "step": 62350 }, { "epoch": 3.76, "learning_rate": 1.547538213148643e-05, "loss": 0.272, "step": 62352 }, { "epoch": 3.76, "learning_rate": 1.5473873221372204e-05, "loss": 0.3833, "step": 62354 }, { "epoch": 3.76, "learning_rate": 1.5472364311257977e-05, "loss": 0.3483, "step": 62356 }, { "epoch": 3.76, "learning_rate": 1.5470855401143757e-05, "loss": 0.3011, "step": 62358 }, { "epoch": 3.76, "learning_rate": 1.546934649102953e-05, "loss": 0.3589, "step": 62360 }, { "epoch": 3.76, "learning_rate": 1.5467837580915306e-05, "loss": 0.4571, "step": 62362 }, { "epoch": 3.76, "learning_rate": 1.546632867080108e-05, "loss": 0.475, "step": 62364 }, { "epoch": 3.76, "learning_rate": 1.546481976068686e-05, "loss": 0.557, "step": 62366 }, { "epoch": 3.76, "learning_rate": 1.5463310850572632e-05, "loss": 0.3351, "step": 62368 }, { "epoch": 3.76, "learning_rate": 1.5461801940458406e-05, "loss": 0.2923, "step": 62370 }, { "epoch": 3.76, "learning_rate": 1.5460293030344182e-05, "loss": 0.3082, "step": 62372 }, { "epoch": 3.76, "learning_rate": 1.545878412022996e-05, "loss": 0.3229, "step": 62374 }, { "epoch": 3.76, "learning_rate": 1.5457275210115735e-05, "loss": 0.3539, "step": 62376 }, { "epoch": 3.76, "learning_rate": 1.5455766300001508e-05, "loss": 0.4583, "step": 62378 }, { "epoch": 3.77, "learning_rate": 1.5454257389887285e-05, "loss": 0.3459, "step": 62380 }, { "epoch": 3.77, "learning_rate": 1.545274847977306e-05, "loss": 0.5721, "step": 62382 }, { "epoch": 3.77, "learning_rate": 1.5451239569658838e-05, "loss": 0.4111, "step": 62384 }, { "epoch": 3.77, "learning_rate": 1.544973065954461e-05, "loss": 0.3815, "step": 62386 }, { "epoch": 3.77, "learning_rate": 1.5448221749430387e-05, "loss": 0.498, "step": 62388 }, { "epoch": 3.77, "learning_rate": 1.5446712839316164e-05, "loss": 0.3164, "step": 62390 }, { "epoch": 3.77, "learning_rate": 1.5445203929201937e-05, "loss": 0.3971, "step": 62392 }, { "epoch": 3.77, "learning_rate": 1.5443695019087714e-05, "loss": 0.3764, "step": 62394 }, { "epoch": 3.77, "learning_rate": 1.544218610897349e-05, "loss": 0.3247, "step": 62396 }, { "epoch": 3.77, "learning_rate": 1.5440677198859267e-05, "loss": 0.3591, "step": 62398 }, { "epoch": 3.77, "learning_rate": 1.543916828874504e-05, "loss": 0.4329, "step": 62400 }, { "epoch": 3.77, "learning_rate": 1.5437659378630813e-05, "loss": 0.3985, "step": 62402 }, { "epoch": 3.77, "learning_rate": 1.5436150468516593e-05, "loss": 0.293, "step": 62404 }, { "epoch": 3.77, "learning_rate": 1.5434641558402366e-05, "loss": 0.3901, "step": 62406 }, { "epoch": 3.77, "learning_rate": 1.5433132648288142e-05, "loss": 0.3796, "step": 62408 }, { "epoch": 3.77, "learning_rate": 1.5431623738173915e-05, "loss": 0.2994, "step": 62410 }, { "epoch": 3.77, "learning_rate": 1.5430114828059695e-05, "loss": 0.3571, "step": 62412 }, { "epoch": 3.77, "learning_rate": 1.542860591794547e-05, "loss": 0.4043, "step": 62414 }, { "epoch": 3.77, "learning_rate": 1.5427097007831245e-05, "loss": 0.3816, "step": 62416 }, { "epoch": 3.77, "learning_rate": 1.5425588097717018e-05, "loss": 0.3926, "step": 62418 }, { "epoch": 3.77, "learning_rate": 1.5424079187602795e-05, "loss": 0.2412, "step": 62420 }, { "epoch": 3.77, "learning_rate": 1.542257027748857e-05, "loss": 0.3857, "step": 62422 }, { "epoch": 3.77, "learning_rate": 1.5421061367374344e-05, "loss": 0.3566, "step": 62424 }, { "epoch": 3.77, "learning_rate": 1.541955245726012e-05, "loss": 0.4481, "step": 62426 }, { "epoch": 3.77, "learning_rate": 1.5418043547145897e-05, "loss": 0.3315, "step": 62428 }, { "epoch": 3.77, "learning_rate": 1.5416534637031674e-05, "loss": 0.2353, "step": 62430 }, { "epoch": 3.77, "learning_rate": 1.5415025726917447e-05, "loss": 0.5096, "step": 62432 }, { "epoch": 3.77, "learning_rate": 1.5413516816803223e-05, "loss": 0.4241, "step": 62434 }, { "epoch": 3.77, "learning_rate": 1.5412007906689e-05, "loss": 0.322, "step": 62436 }, { "epoch": 3.77, "learning_rate": 1.5410498996574773e-05, "loss": 0.4078, "step": 62438 }, { "epoch": 3.77, "learning_rate": 1.540899008646055e-05, "loss": 0.2892, "step": 62440 }, { "epoch": 3.77, "learning_rate": 1.5407481176346326e-05, "loss": 0.383, "step": 62442 }, { "epoch": 3.77, "learning_rate": 1.5405972266232103e-05, "loss": 0.3019, "step": 62444 }, { "epoch": 3.77, "learning_rate": 1.5404463356117876e-05, "loss": 0.2861, "step": 62446 }, { "epoch": 3.77, "learning_rate": 1.5402954446003652e-05, "loss": 0.2991, "step": 62448 }, { "epoch": 3.77, "learning_rate": 1.540144553588943e-05, "loss": 0.4044, "step": 62450 }, { "epoch": 3.77, "learning_rate": 1.5399936625775202e-05, "loss": 0.2379, "step": 62452 }, { "epoch": 3.77, "learning_rate": 1.539842771566098e-05, "loss": 0.5519, "step": 62454 }, { "epoch": 3.77, "learning_rate": 1.5396918805546755e-05, "loss": 0.2634, "step": 62456 }, { "epoch": 3.77, "learning_rate": 1.539540989543253e-05, "loss": 0.5127, "step": 62458 }, { "epoch": 3.77, "learning_rate": 1.5393900985318305e-05, "loss": 0.3521, "step": 62460 }, { "epoch": 3.77, "learning_rate": 1.539239207520408e-05, "loss": 0.4432, "step": 62462 }, { "epoch": 3.77, "learning_rate": 1.5390883165089858e-05, "loss": 0.3312, "step": 62464 }, { "epoch": 3.77, "learning_rate": 1.5389374254975634e-05, "loss": 0.3241, "step": 62466 }, { "epoch": 3.77, "learning_rate": 1.5387865344861407e-05, "loss": 0.606, "step": 62468 }, { "epoch": 3.77, "learning_rate": 1.538635643474718e-05, "loss": 0.4527, "step": 62470 }, { "epoch": 3.77, "learning_rate": 1.538484752463296e-05, "loss": 0.2189, "step": 62472 }, { "epoch": 3.77, "learning_rate": 1.5383338614518733e-05, "loss": 0.3441, "step": 62474 }, { "epoch": 3.77, "learning_rate": 1.538182970440451e-05, "loss": 0.3565, "step": 62476 }, { "epoch": 3.77, "learning_rate": 1.5380320794290283e-05, "loss": 0.26, "step": 62478 }, { "epoch": 3.77, "learning_rate": 1.5378811884176063e-05, "loss": 0.4216, "step": 62480 }, { "epoch": 3.77, "learning_rate": 1.5377302974061836e-05, "loss": 0.3758, "step": 62482 }, { "epoch": 3.77, "learning_rate": 1.537579406394761e-05, "loss": 0.3446, "step": 62484 }, { "epoch": 3.77, "learning_rate": 1.5374285153833386e-05, "loss": 0.3327, "step": 62486 }, { "epoch": 3.77, "learning_rate": 1.5372776243719162e-05, "loss": 0.3281, "step": 62488 }, { "epoch": 3.77, "learning_rate": 1.537126733360494e-05, "loss": 0.4759, "step": 62490 }, { "epoch": 3.77, "learning_rate": 1.5369758423490712e-05, "loss": 0.4132, "step": 62492 }, { "epoch": 3.77, "learning_rate": 1.536824951337649e-05, "loss": 0.5623, "step": 62494 }, { "epoch": 3.77, "learning_rate": 1.5366740603262265e-05, "loss": 0.2807, "step": 62496 }, { "epoch": 3.77, "learning_rate": 1.536523169314804e-05, "loss": 0.381, "step": 62498 }, { "epoch": 3.77, "learning_rate": 1.5363722783033815e-05, "loss": 0.4027, "step": 62500 }, { "epoch": 3.77, "learning_rate": 1.536221387291959e-05, "loss": 0.2517, "step": 62502 }, { "epoch": 3.77, "learning_rate": 1.5360704962805368e-05, "loss": 0.2762, "step": 62504 }, { "epoch": 3.77, "learning_rate": 1.535919605269114e-05, "loss": 0.3287, "step": 62506 }, { "epoch": 3.77, "learning_rate": 1.5357687142576917e-05, "loss": 0.4582, "step": 62508 }, { "epoch": 3.77, "learning_rate": 1.5356178232462694e-05, "loss": 0.2816, "step": 62510 }, { "epoch": 3.77, "learning_rate": 1.535466932234847e-05, "loss": 0.3319, "step": 62512 }, { "epoch": 3.77, "learning_rate": 1.5353160412234243e-05, "loss": 0.4107, "step": 62514 }, { "epoch": 3.77, "learning_rate": 1.535165150212002e-05, "loss": 0.3711, "step": 62516 }, { "epoch": 3.77, "learning_rate": 1.5350142592005796e-05, "loss": 0.29, "step": 62518 }, { "epoch": 3.77, "learning_rate": 1.534863368189157e-05, "loss": 0.2606, "step": 62520 }, { "epoch": 3.77, "learning_rate": 1.5347124771777346e-05, "loss": 0.4727, "step": 62522 }, { "epoch": 3.77, "learning_rate": 1.534561586166312e-05, "loss": 0.3984, "step": 62524 }, { "epoch": 3.77, "learning_rate": 1.53441069515489e-05, "loss": 0.3313, "step": 62526 }, { "epoch": 3.77, "learning_rate": 1.5342598041434672e-05, "loss": 0.4527, "step": 62528 }, { "epoch": 3.77, "learning_rate": 1.534108913132045e-05, "loss": 0.3507, "step": 62530 }, { "epoch": 3.77, "learning_rate": 1.5339580221206222e-05, "loss": 0.2594, "step": 62532 }, { "epoch": 3.77, "learning_rate": 1.5338071311092e-05, "loss": 0.3683, "step": 62534 }, { "epoch": 3.77, "learning_rate": 1.5336562400977775e-05, "loss": 0.329, "step": 62536 }, { "epoch": 3.77, "learning_rate": 1.5335053490863548e-05, "loss": 0.3839, "step": 62538 }, { "epoch": 3.77, "learning_rate": 1.5333544580749324e-05, "loss": 0.3757, "step": 62540 }, { "epoch": 3.77, "learning_rate": 1.53320356706351e-05, "loss": 0.4045, "step": 62542 }, { "epoch": 3.77, "learning_rate": 1.5330526760520878e-05, "loss": 0.3429, "step": 62544 }, { "epoch": 3.78, "learning_rate": 1.532901785040665e-05, "loss": 0.2236, "step": 62546 }, { "epoch": 3.78, "learning_rate": 1.5327508940292427e-05, "loss": 0.2793, "step": 62548 }, { "epoch": 3.78, "learning_rate": 1.5326000030178204e-05, "loss": 0.352, "step": 62550 }, { "epoch": 3.78, "learning_rate": 1.5324491120063977e-05, "loss": 0.5069, "step": 62552 }, { "epoch": 3.78, "learning_rate": 1.5322982209949753e-05, "loss": 0.5666, "step": 62554 }, { "epoch": 3.78, "learning_rate": 1.532147329983553e-05, "loss": 0.4359, "step": 62556 }, { "epoch": 3.78, "learning_rate": 1.5319964389721306e-05, "loss": 0.2514, "step": 62558 }, { "epoch": 3.78, "learning_rate": 1.531845547960708e-05, "loss": 0.2797, "step": 62560 }, { "epoch": 3.78, "learning_rate": 1.5316946569492856e-05, "loss": 0.3941, "step": 62562 }, { "epoch": 3.78, "learning_rate": 1.5315437659378632e-05, "loss": 0.2288, "step": 62564 }, { "epoch": 3.78, "learning_rate": 1.5313928749264406e-05, "loss": 0.3513, "step": 62566 }, { "epoch": 3.78, "learning_rate": 1.5312419839150182e-05, "loss": 0.283, "step": 62568 }, { "epoch": 3.78, "learning_rate": 1.531091092903596e-05, "loss": 0.4211, "step": 62570 }, { "epoch": 3.78, "learning_rate": 1.5309402018921735e-05, "loss": 0.4478, "step": 62572 }, { "epoch": 3.78, "learning_rate": 1.5307893108807508e-05, "loss": 0.3044, "step": 62574 }, { "epoch": 3.78, "learning_rate": 1.5306384198693285e-05, "loss": 0.3294, "step": 62576 }, { "epoch": 3.78, "learning_rate": 1.530487528857906e-05, "loss": 0.1929, "step": 62578 }, { "epoch": 3.78, "learning_rate": 1.5303366378464838e-05, "loss": 0.3567, "step": 62580 }, { "epoch": 3.78, "learning_rate": 1.530185746835061e-05, "loss": 0.233, "step": 62582 }, { "epoch": 3.78, "learning_rate": 1.5300348558236384e-05, "loss": 0.2976, "step": 62584 }, { "epoch": 3.78, "learning_rate": 1.5298839648122164e-05, "loss": 0.2678, "step": 62586 }, { "epoch": 3.78, "learning_rate": 1.5297330738007937e-05, "loss": 0.371, "step": 62588 }, { "epoch": 3.78, "learning_rate": 1.5295821827893714e-05, "loss": 0.2932, "step": 62590 }, { "epoch": 3.78, "learning_rate": 1.5294312917779487e-05, "loss": 0.2794, "step": 62592 }, { "epoch": 3.78, "learning_rate": 1.5292804007665267e-05, "loss": 0.3609, "step": 62594 }, { "epoch": 3.78, "learning_rate": 1.529129509755104e-05, "loss": 0.3109, "step": 62596 }, { "epoch": 3.78, "learning_rate": 1.5289786187436816e-05, "loss": 0.3695, "step": 62598 }, { "epoch": 3.78, "learning_rate": 1.528827727732259e-05, "loss": 0.3382, "step": 62600 }, { "epoch": 3.78, "learning_rate": 1.5286768367208366e-05, "loss": 0.4001, "step": 62602 }, { "epoch": 3.78, "learning_rate": 1.5285259457094142e-05, "loss": 0.3983, "step": 62604 }, { "epoch": 3.78, "learning_rate": 1.5283750546979916e-05, "loss": 0.2713, "step": 62606 }, { "epoch": 3.78, "learning_rate": 1.5282241636865692e-05, "loss": 0.3174, "step": 62608 }, { "epoch": 3.78, "learning_rate": 1.528073272675147e-05, "loss": 0.3803, "step": 62610 }, { "epoch": 3.78, "learning_rate": 1.5279223816637245e-05, "loss": 0.4696, "step": 62612 }, { "epoch": 3.78, "learning_rate": 1.5277714906523018e-05, "loss": 0.432, "step": 62614 }, { "epoch": 3.78, "learning_rate": 1.5276205996408795e-05, "loss": 0.3372, "step": 62616 }, { "epoch": 3.78, "learning_rate": 1.527469708629457e-05, "loss": 0.3943, "step": 62618 }, { "epoch": 3.78, "learning_rate": 1.5273188176180344e-05, "loss": 0.418, "step": 62620 }, { "epoch": 3.78, "learning_rate": 1.527167926606612e-05, "loss": 0.4363, "step": 62622 }, { "epoch": 3.78, "learning_rate": 1.5270170355951897e-05, "loss": 0.2627, "step": 62624 }, { "epoch": 3.78, "learning_rate": 1.5268661445837674e-05, "loss": 0.4145, "step": 62626 }, { "epoch": 3.78, "learning_rate": 1.5267152535723447e-05, "loss": 0.3343, "step": 62628 }, { "epoch": 3.78, "learning_rate": 1.5265643625609224e-05, "loss": 0.3374, "step": 62630 }, { "epoch": 3.78, "learning_rate": 1.5264134715495e-05, "loss": 0.2589, "step": 62632 }, { "epoch": 3.78, "learning_rate": 1.5262625805380773e-05, "loss": 0.3439, "step": 62634 }, { "epoch": 3.78, "learning_rate": 1.526111689526655e-05, "loss": 0.3044, "step": 62636 }, { "epoch": 3.78, "learning_rate": 1.5259607985152323e-05, "loss": 0.4605, "step": 62638 }, { "epoch": 3.78, "learning_rate": 1.5258099075038101e-05, "loss": 0.3533, "step": 62640 }, { "epoch": 3.78, "learning_rate": 1.5256590164923876e-05, "loss": 0.4408, "step": 62642 }, { "epoch": 3.78, "learning_rate": 1.5255081254809652e-05, "loss": 0.2892, "step": 62644 }, { "epoch": 3.78, "learning_rate": 1.5253572344695427e-05, "loss": 0.5385, "step": 62646 }, { "epoch": 3.78, "learning_rate": 1.5252063434581202e-05, "loss": 0.3084, "step": 62648 }, { "epoch": 3.78, "learning_rate": 1.5250554524466979e-05, "loss": 0.4606, "step": 62650 }, { "epoch": 3.78, "learning_rate": 1.5249045614352753e-05, "loss": 0.4235, "step": 62652 }, { "epoch": 3.78, "learning_rate": 1.524753670423853e-05, "loss": 0.393, "step": 62654 }, { "epoch": 3.78, "learning_rate": 1.5246027794124305e-05, "loss": 0.375, "step": 62656 }, { "epoch": 3.78, "learning_rate": 1.5244518884010081e-05, "loss": 0.3199, "step": 62658 }, { "epoch": 3.78, "learning_rate": 1.5243009973895856e-05, "loss": 0.5124, "step": 62660 }, { "epoch": 3.78, "learning_rate": 1.5241501063781633e-05, "loss": 0.2214, "step": 62662 }, { "epoch": 3.78, "learning_rate": 1.5239992153667407e-05, "loss": 0.2106, "step": 62664 }, { "epoch": 3.78, "learning_rate": 1.523848324355318e-05, "loss": 0.3253, "step": 62666 }, { "epoch": 3.78, "learning_rate": 1.5236974333438959e-05, "loss": 0.4432, "step": 62668 }, { "epoch": 3.78, "learning_rate": 1.5235465423324732e-05, "loss": 0.2736, "step": 62670 }, { "epoch": 3.78, "learning_rate": 1.523395651321051e-05, "loss": 0.4595, "step": 62672 }, { "epoch": 3.78, "learning_rate": 1.5232447603096283e-05, "loss": 0.4876, "step": 62674 }, { "epoch": 3.78, "learning_rate": 1.5230938692982061e-05, "loss": 0.3794, "step": 62676 }, { "epoch": 3.78, "learning_rate": 1.5229429782867834e-05, "loss": 0.3451, "step": 62678 }, { "epoch": 3.78, "learning_rate": 1.5227920872753613e-05, "loss": 0.2916, "step": 62680 }, { "epoch": 3.78, "learning_rate": 1.5226411962639386e-05, "loss": 0.2228, "step": 62682 }, { "epoch": 3.78, "learning_rate": 1.522490305252516e-05, "loss": 0.3745, "step": 62684 }, { "epoch": 3.78, "learning_rate": 1.5223394142410937e-05, "loss": 0.2798, "step": 62686 }, { "epoch": 3.78, "learning_rate": 1.5221885232296712e-05, "loss": 0.2279, "step": 62688 }, { "epoch": 3.78, "learning_rate": 1.5220376322182488e-05, "loss": 0.4532, "step": 62690 }, { "epoch": 3.78, "learning_rate": 1.5218867412068263e-05, "loss": 0.3787, "step": 62692 }, { "epoch": 3.78, "learning_rate": 1.521735850195404e-05, "loss": 0.2574, "step": 62694 }, { "epoch": 3.78, "learning_rate": 1.5215849591839815e-05, "loss": 0.3435, "step": 62696 }, { "epoch": 3.78, "learning_rate": 1.521434068172559e-05, "loss": 0.4013, "step": 62698 }, { "epoch": 3.78, "learning_rate": 1.5212831771611366e-05, "loss": 0.386, "step": 62700 }, { "epoch": 3.78, "learning_rate": 1.521132286149714e-05, "loss": 0.3084, "step": 62702 }, { "epoch": 3.78, "learning_rate": 1.5209813951382917e-05, "loss": 0.3301, "step": 62704 }, { "epoch": 3.78, "learning_rate": 1.5208305041268692e-05, "loss": 0.4326, "step": 62706 }, { "epoch": 3.78, "learning_rate": 1.5206796131154469e-05, "loss": 0.448, "step": 62708 }, { "epoch": 3.79, "learning_rate": 1.5205287221040243e-05, "loss": 0.2977, "step": 62710 }, { "epoch": 3.79, "learning_rate": 1.520377831092602e-05, "loss": 0.4019, "step": 62712 }, { "epoch": 3.79, "learning_rate": 1.5202269400811795e-05, "loss": 0.3517, "step": 62714 }, { "epoch": 3.79, "learning_rate": 1.5200760490697568e-05, "loss": 0.3603, "step": 62716 }, { "epoch": 3.79, "learning_rate": 1.5199251580583346e-05, "loss": 0.2472, "step": 62718 }, { "epoch": 3.79, "learning_rate": 1.519774267046912e-05, "loss": 0.2879, "step": 62720 }, { "epoch": 3.79, "learning_rate": 1.5196233760354897e-05, "loss": 0.3695, "step": 62722 }, { "epoch": 3.79, "learning_rate": 1.519472485024067e-05, "loss": 0.3255, "step": 62724 }, { "epoch": 3.79, "learning_rate": 1.5193215940126449e-05, "loss": 0.3735, "step": 62726 }, { "epoch": 3.79, "learning_rate": 1.5191707030012222e-05, "loss": 0.4334, "step": 62728 }, { "epoch": 3.79, "learning_rate": 1.5190198119898e-05, "loss": 0.3609, "step": 62730 }, { "epoch": 3.79, "learning_rate": 1.5188689209783775e-05, "loss": 0.4794, "step": 62732 }, { "epoch": 3.79, "learning_rate": 1.5187180299669548e-05, "loss": 0.3333, "step": 62734 }, { "epoch": 3.79, "learning_rate": 1.5185671389555326e-05, "loss": 0.4127, "step": 62736 }, { "epoch": 3.79, "learning_rate": 1.51841624794411e-05, "loss": 0.2788, "step": 62738 }, { "epoch": 3.79, "learning_rate": 1.5182653569326878e-05, "loss": 0.3462, "step": 62740 }, { "epoch": 3.79, "learning_rate": 1.518114465921265e-05, "loss": 0.398, "step": 62742 }, { "epoch": 3.79, "learning_rate": 1.5179635749098429e-05, "loss": 0.2825, "step": 62744 }, { "epoch": 3.79, "learning_rate": 1.5178126838984202e-05, "loss": 0.3644, "step": 62746 }, { "epoch": 3.79, "learning_rate": 1.5176617928869977e-05, "loss": 0.2636, "step": 62748 }, { "epoch": 3.79, "learning_rate": 1.5175109018755753e-05, "loss": 0.2898, "step": 62750 }, { "epoch": 3.79, "learning_rate": 1.5173600108641528e-05, "loss": 0.409, "step": 62752 }, { "epoch": 3.79, "learning_rate": 1.5172091198527305e-05, "loss": 0.2551, "step": 62754 }, { "epoch": 3.79, "learning_rate": 1.517058228841308e-05, "loss": 0.4024, "step": 62756 }, { "epoch": 3.79, "learning_rate": 1.5169073378298856e-05, "loss": 0.3755, "step": 62758 }, { "epoch": 3.79, "learning_rate": 1.5167564468184631e-05, "loss": 0.2435, "step": 62760 }, { "epoch": 3.79, "learning_rate": 1.5166055558070407e-05, "loss": 0.3221, "step": 62762 }, { "epoch": 3.79, "learning_rate": 1.5164546647956182e-05, "loss": 0.2292, "step": 62764 }, { "epoch": 3.79, "learning_rate": 1.5163037737841957e-05, "loss": 0.3048, "step": 62766 }, { "epoch": 3.79, "learning_rate": 1.5161528827727734e-05, "loss": 0.3291, "step": 62768 }, { "epoch": 3.79, "learning_rate": 1.5160019917613508e-05, "loss": 0.2211, "step": 62770 }, { "epoch": 3.79, "learning_rate": 1.5158511007499285e-05, "loss": 0.3376, "step": 62772 }, { "epoch": 3.79, "learning_rate": 1.515700209738506e-05, "loss": 0.2514, "step": 62774 }, { "epoch": 3.79, "learning_rate": 1.5155493187270836e-05, "loss": 0.292, "step": 62776 }, { "epoch": 3.79, "learning_rate": 1.5153984277156611e-05, "loss": 0.4036, "step": 62778 }, { "epoch": 3.79, "learning_rate": 1.5152475367042384e-05, "loss": 0.4728, "step": 62780 }, { "epoch": 3.79, "learning_rate": 1.5150966456928162e-05, "loss": 0.4215, "step": 62782 }, { "epoch": 3.79, "learning_rate": 1.5149457546813935e-05, "loss": 0.3631, "step": 62784 }, { "epoch": 3.79, "learning_rate": 1.5147948636699714e-05, "loss": 0.2535, "step": 62786 }, { "epoch": 3.79, "learning_rate": 1.5146439726585487e-05, "loss": 0.3925, "step": 62788 }, { "epoch": 3.79, "learning_rate": 1.5144930816471265e-05, "loss": 0.1626, "step": 62790 }, { "epoch": 3.79, "learning_rate": 1.5143421906357038e-05, "loss": 0.4504, "step": 62792 }, { "epoch": 3.79, "learning_rate": 1.5141912996242816e-05, "loss": 0.5947, "step": 62794 }, { "epoch": 3.79, "learning_rate": 1.514040408612859e-05, "loss": 0.2647, "step": 62796 }, { "epoch": 3.79, "learning_rate": 1.5138895176014364e-05, "loss": 0.2737, "step": 62798 }, { "epoch": 3.79, "learning_rate": 1.513738626590014e-05, "loss": 0.4544, "step": 62800 }, { "epoch": 3.79, "learning_rate": 1.5135877355785916e-05, "loss": 0.6219, "step": 62802 }, { "epoch": 3.79, "learning_rate": 1.5134368445671692e-05, "loss": 0.3921, "step": 62804 }, { "epoch": 3.79, "learning_rate": 1.5132859535557467e-05, "loss": 0.513, "step": 62806 }, { "epoch": 3.79, "learning_rate": 1.5131350625443243e-05, "loss": 0.2881, "step": 62808 }, { "epoch": 3.79, "learning_rate": 1.5129841715329018e-05, "loss": 0.4359, "step": 62810 }, { "epoch": 3.79, "learning_rate": 1.5128332805214795e-05, "loss": 0.3913, "step": 62812 }, { "epoch": 3.79, "learning_rate": 1.512682389510057e-05, "loss": 0.4544, "step": 62814 }, { "epoch": 3.79, "learning_rate": 1.5125314984986344e-05, "loss": 0.2716, "step": 62816 }, { "epoch": 3.79, "learning_rate": 1.5123806074872121e-05, "loss": 0.2638, "step": 62818 }, { "epoch": 3.79, "learning_rate": 1.5122297164757896e-05, "loss": 0.3468, "step": 62820 }, { "epoch": 3.79, "learning_rate": 1.5120788254643672e-05, "loss": 0.2457, "step": 62822 }, { "epoch": 3.79, "learning_rate": 1.5119279344529447e-05, "loss": 0.3348, "step": 62824 }, { "epoch": 3.79, "learning_rate": 1.5117770434415224e-05, "loss": 0.2958, "step": 62826 }, { "epoch": 3.79, "learning_rate": 1.5116261524300998e-05, "loss": 0.3679, "step": 62828 }, { "epoch": 3.79, "learning_rate": 1.5114752614186772e-05, "loss": 0.5249, "step": 62830 }, { "epoch": 3.79, "learning_rate": 1.511324370407255e-05, "loss": 0.4137, "step": 62832 }, { "epoch": 3.79, "learning_rate": 1.5111734793958323e-05, "loss": 0.3822, "step": 62834 }, { "epoch": 3.79, "learning_rate": 1.5110225883844101e-05, "loss": 0.3347, "step": 62836 }, { "epoch": 3.79, "learning_rate": 1.5108716973729874e-05, "loss": 0.4172, "step": 62838 }, { "epoch": 3.79, "learning_rate": 1.5107208063615652e-05, "loss": 0.463, "step": 62840 }, { "epoch": 3.79, "learning_rate": 1.5105699153501426e-05, "loss": 0.2767, "step": 62842 }, { "epoch": 3.79, "learning_rate": 1.5104190243387204e-05, "loss": 0.4839, "step": 62844 }, { "epoch": 3.79, "learning_rate": 1.5102681333272977e-05, "loss": 0.3614, "step": 62846 }, { "epoch": 3.79, "learning_rate": 1.5101926878215867e-05, "loss": 0.4501, "step": 62848 }, { "epoch": 3.79, "learning_rate": 1.510041796810164e-05, "loss": 0.3355, "step": 62850 }, { "epoch": 3.79, "learning_rate": 1.5098909057987418e-05, "loss": 0.3892, "step": 62852 }, { "epoch": 3.79, "learning_rate": 1.5097400147873191e-05, "loss": 0.2693, "step": 62854 }, { "epoch": 3.79, "learning_rate": 1.5095891237758966e-05, "loss": 0.2852, "step": 62856 }, { "epoch": 3.79, "learning_rate": 1.5094382327644743e-05, "loss": 0.5192, "step": 62858 }, { "epoch": 3.79, "learning_rate": 1.5092873417530517e-05, "loss": 0.2248, "step": 62860 }, { "epoch": 3.79, "learning_rate": 1.5091364507416294e-05, "loss": 0.2962, "step": 62862 }, { "epoch": 3.79, "learning_rate": 1.5089855597302069e-05, "loss": 0.3495, "step": 62864 }, { "epoch": 3.79, "learning_rate": 1.5088346687187845e-05, "loss": 0.3783, "step": 62866 }, { "epoch": 3.79, "learning_rate": 1.508683777707362e-05, "loss": 0.3687, "step": 62868 }, { "epoch": 3.79, "learning_rate": 1.5085328866959397e-05, "loss": 0.5961, "step": 62870 }, { "epoch": 3.79, "learning_rate": 1.5083819956845171e-05, "loss": 0.3778, "step": 62872 }, { "epoch": 3.79, "learning_rate": 1.5082311046730946e-05, "loss": 0.4674, "step": 62874 }, { "epoch": 3.8, "learning_rate": 1.5080802136616723e-05, "loss": 0.3917, "step": 62876 }, { "epoch": 3.8, "learning_rate": 1.5079293226502498e-05, "loss": 0.3459, "step": 62878 }, { "epoch": 3.8, "learning_rate": 1.5077784316388274e-05, "loss": 0.4079, "step": 62880 }, { "epoch": 3.8, "learning_rate": 1.5076275406274049e-05, "loss": 0.4799, "step": 62882 }, { "epoch": 3.8, "learning_rate": 1.5074766496159825e-05, "loss": 0.5443, "step": 62884 }, { "epoch": 3.8, "learning_rate": 1.50732575860456e-05, "loss": 0.3877, "step": 62886 }, { "epoch": 3.8, "learning_rate": 1.5071748675931373e-05, "loss": 0.3548, "step": 62888 }, { "epoch": 3.8, "learning_rate": 1.5070239765817152e-05, "loss": 0.2738, "step": 62890 }, { "epoch": 3.8, "learning_rate": 1.5068730855702925e-05, "loss": 0.3667, "step": 62892 }, { "epoch": 3.8, "learning_rate": 1.5067221945588703e-05, "loss": 0.3917, "step": 62894 }, { "epoch": 3.8, "learning_rate": 1.5065713035474476e-05, "loss": 0.3357, "step": 62896 }, { "epoch": 3.8, "learning_rate": 1.5064204125360254e-05, "loss": 0.4151, "step": 62898 }, { "epoch": 3.8, "learning_rate": 1.5062695215246027e-05, "loss": 0.3241, "step": 62900 }, { "epoch": 3.8, "learning_rate": 1.5061186305131806e-05, "loss": 0.4283, "step": 62902 }, { "epoch": 3.8, "learning_rate": 1.5059677395017579e-05, "loss": 0.3455, "step": 62904 }, { "epoch": 3.8, "learning_rate": 1.5058168484903354e-05, "loss": 0.3457, "step": 62906 }, { "epoch": 3.8, "learning_rate": 1.505665957478913e-05, "loss": 0.31, "step": 62908 }, { "epoch": 3.8, "learning_rate": 1.5055150664674905e-05, "loss": 0.4037, "step": 62910 }, { "epoch": 3.8, "learning_rate": 1.5053641754560683e-05, "loss": 0.4518, "step": 62912 }, { "epoch": 3.8, "learning_rate": 1.5052132844446456e-05, "loss": 0.3044, "step": 62914 }, { "epoch": 3.8, "learning_rate": 1.5050623934332234e-05, "loss": 0.3053, "step": 62916 }, { "epoch": 3.8, "learning_rate": 1.5049115024218008e-05, "loss": 0.4404, "step": 62918 }, { "epoch": 3.8, "learning_rate": 1.5047606114103782e-05, "loss": 0.4249, "step": 62920 }, { "epoch": 3.8, "learning_rate": 1.5046097203989559e-05, "loss": 0.3845, "step": 62922 }, { "epoch": 3.8, "learning_rate": 1.5044588293875334e-05, "loss": 0.4015, "step": 62924 }, { "epoch": 3.8, "learning_rate": 1.504307938376111e-05, "loss": 0.3335, "step": 62926 }, { "epoch": 3.8, "learning_rate": 1.5041570473646885e-05, "loss": 0.3781, "step": 62928 }, { "epoch": 3.8, "learning_rate": 1.5040061563532662e-05, "loss": 0.2738, "step": 62930 }, { "epoch": 3.8, "learning_rate": 1.5038552653418436e-05, "loss": 0.3725, "step": 62932 }, { "epoch": 3.8, "learning_rate": 1.5037043743304213e-05, "loss": 0.3115, "step": 62934 }, { "epoch": 3.8, "learning_rate": 1.5035534833189988e-05, "loss": 0.3039, "step": 62936 }, { "epoch": 3.8, "learning_rate": 1.5034025923075763e-05, "loss": 0.3446, "step": 62938 }, { "epoch": 3.8, "learning_rate": 1.5032517012961539e-05, "loss": 0.2656, "step": 62940 }, { "epoch": 3.8, "learning_rate": 1.5031008102847314e-05, "loss": 0.3461, "step": 62942 }, { "epoch": 3.8, "learning_rate": 1.502949919273309e-05, "loss": 0.4059, "step": 62944 }, { "epoch": 3.8, "learning_rate": 1.5027990282618865e-05, "loss": 0.3719, "step": 62946 }, { "epoch": 3.8, "learning_rate": 1.5026481372504642e-05, "loss": 0.4509, "step": 62948 }, { "epoch": 3.8, "learning_rate": 1.5024972462390417e-05, "loss": 0.2593, "step": 62950 }, { "epoch": 3.8, "learning_rate": 1.5023463552276193e-05, "loss": 0.2694, "step": 62952 }, { "epoch": 3.8, "learning_rate": 1.5021954642161968e-05, "loss": 0.4308, "step": 62954 }, { "epoch": 3.8, "learning_rate": 1.5020445732047741e-05, "loss": 0.2796, "step": 62956 }, { "epoch": 3.8, "learning_rate": 1.501893682193352e-05, "loss": 0.2777, "step": 62958 }, { "epoch": 3.8, "learning_rate": 1.5017427911819292e-05, "loss": 0.4569, "step": 62960 }, { "epoch": 3.8, "learning_rate": 1.501591900170507e-05, "loss": 0.419, "step": 62962 }, { "epoch": 3.8, "learning_rate": 1.5014410091590844e-05, "loss": 0.3937, "step": 62964 }, { "epoch": 3.8, "learning_rate": 1.5012901181476622e-05, "loss": 0.3301, "step": 62966 }, { "epoch": 3.8, "learning_rate": 1.5011392271362395e-05, "loss": 0.3602, "step": 62968 }, { "epoch": 3.8, "learning_rate": 1.500988336124817e-05, "loss": 0.3752, "step": 62970 }, { "epoch": 3.8, "learning_rate": 1.5008374451133946e-05, "loss": 0.3602, "step": 62972 }, { "epoch": 3.8, "learning_rate": 1.5006865541019721e-05, "loss": 0.444, "step": 62974 }, { "epoch": 3.8, "learning_rate": 1.5005356630905498e-05, "loss": 0.3775, "step": 62976 }, { "epoch": 3.8, "learning_rate": 1.5003847720791272e-05, "loss": 0.3066, "step": 62978 }, { "epoch": 3.8, "learning_rate": 1.5002338810677049e-05, "loss": 0.3259, "step": 62980 }, { "epoch": 3.8, "learning_rate": 1.5000829900562824e-05, "loss": 0.4226, "step": 62982 }, { "epoch": 3.8, "learning_rate": 1.49993209904486e-05, "loss": 0.3814, "step": 62984 }, { "epoch": 3.8, "learning_rate": 1.4997812080334375e-05, "loss": 0.5039, "step": 62986 }, { "epoch": 3.8, "learning_rate": 1.499630317022015e-05, "loss": 0.2437, "step": 62988 }, { "epoch": 3.8, "learning_rate": 1.4994794260105926e-05, "loss": 0.4087, "step": 62990 }, { "epoch": 3.8, "learning_rate": 1.4993285349991701e-05, "loss": 0.4136, "step": 62992 }, { "epoch": 3.8, "learning_rate": 1.4991776439877478e-05, "loss": 0.3685, "step": 62994 }, { "epoch": 3.8, "learning_rate": 1.4990267529763253e-05, "loss": 0.323, "step": 62996 }, { "epoch": 3.8, "learning_rate": 1.4988758619649029e-05, "loss": 0.3841, "step": 62998 }, { "epoch": 3.8, "learning_rate": 1.4987249709534804e-05, "loss": 0.3166, "step": 63000 }, { "epoch": 3.8, "learning_rate": 1.4985740799420577e-05, "loss": 0.3236, "step": 63002 }, { "epoch": 3.8, "learning_rate": 1.4984231889306355e-05, "loss": 0.3636, "step": 63004 }, { "epoch": 3.8, "learning_rate": 1.4982722979192128e-05, "loss": 0.2924, "step": 63006 }, { "epoch": 3.8, "learning_rate": 1.4981214069077907e-05, "loss": 0.4804, "step": 63008 }, { "epoch": 3.8, "learning_rate": 1.497970515896368e-05, "loss": 0.4036, "step": 63010 }, { "epoch": 3.8, "learning_rate": 1.4978196248849458e-05, "loss": 0.365, "step": 63012 }, { "epoch": 3.8, "learning_rate": 1.4976687338735231e-05, "loss": 0.455, "step": 63014 }, { "epoch": 3.8, "learning_rate": 1.497517842862101e-05, "loss": 0.2679, "step": 63016 }, { "epoch": 3.8, "learning_rate": 1.4973669518506782e-05, "loss": 0.3727, "step": 63018 }, { "epoch": 3.8, "learning_rate": 1.4972160608392557e-05, "loss": 0.5627, "step": 63020 }, { "epoch": 3.8, "learning_rate": 1.4970651698278334e-05, "loss": 0.351, "step": 63022 }, { "epoch": 3.8, "learning_rate": 1.4969142788164109e-05, "loss": 0.3805, "step": 63024 }, { "epoch": 3.8, "learning_rate": 1.4967633878049885e-05, "loss": 0.4301, "step": 63026 }, { "epoch": 3.8, "learning_rate": 1.496612496793566e-05, "loss": 0.2683, "step": 63028 }, { "epoch": 3.8, "learning_rate": 1.4964616057821438e-05, "loss": 0.3529, "step": 63030 }, { "epoch": 3.8, "learning_rate": 1.4963107147707211e-05, "loss": 0.2788, "step": 63032 }, { "epoch": 3.8, "learning_rate": 1.496159823759299e-05, "loss": 0.2373, "step": 63034 }, { "epoch": 3.8, "learning_rate": 1.4960089327478763e-05, "loss": 0.3416, "step": 63036 }, { "epoch": 3.8, "learning_rate": 1.4958580417364537e-05, "loss": 0.3017, "step": 63038 }, { "epoch": 3.8, "learning_rate": 1.4957071507250314e-05, "loss": 0.3376, "step": 63040 }, { "epoch": 3.81, "learning_rate": 1.4955562597136089e-05, "loss": 0.2825, "step": 63042 }, { "epoch": 3.81, "learning_rate": 1.4954053687021865e-05, "loss": 0.3087, "step": 63044 }, { "epoch": 3.81, "learning_rate": 1.495254477690764e-05, "loss": 0.382, "step": 63046 }, { "epoch": 3.81, "learning_rate": 1.4951035866793417e-05, "loss": 0.5689, "step": 63048 }, { "epoch": 3.81, "learning_rate": 1.4949526956679191e-05, "loss": 0.3743, "step": 63050 }, { "epoch": 3.81, "learning_rate": 1.4948018046564966e-05, "loss": 0.2184, "step": 63052 }, { "epoch": 3.81, "learning_rate": 1.4946509136450743e-05, "loss": 0.4902, "step": 63054 }, { "epoch": 3.81, "learning_rate": 1.4945000226336518e-05, "loss": 0.3561, "step": 63056 }, { "epoch": 3.81, "learning_rate": 1.4943491316222294e-05, "loss": 0.3784, "step": 63058 }, { "epoch": 3.81, "learning_rate": 1.4941982406108069e-05, "loss": 0.4479, "step": 63060 }, { "epoch": 3.81, "learning_rate": 1.4940473495993845e-05, "loss": 0.3677, "step": 63062 }, { "epoch": 3.81, "learning_rate": 1.493896458587962e-05, "loss": 0.2547, "step": 63064 }, { "epoch": 3.81, "learning_rate": 1.4937455675765397e-05, "loss": 0.528, "step": 63066 }, { "epoch": 3.81, "learning_rate": 1.4935946765651172e-05, "loss": 0.4236, "step": 63068 }, { "epoch": 3.81, "learning_rate": 1.4934437855536945e-05, "loss": 0.2416, "step": 63070 }, { "epoch": 3.81, "learning_rate": 1.4932928945422723e-05, "loss": 0.2493, "step": 63072 }, { "epoch": 3.81, "learning_rate": 1.4931420035308496e-05, "loss": 0.4999, "step": 63074 }, { "epoch": 3.81, "learning_rate": 1.4929911125194274e-05, "loss": 0.3744, "step": 63076 }, { "epoch": 3.81, "learning_rate": 1.4928402215080047e-05, "loss": 0.3948, "step": 63078 }, { "epoch": 3.81, "learning_rate": 1.4926893304965826e-05, "loss": 0.3266, "step": 63080 }, { "epoch": 3.81, "learning_rate": 1.4925384394851599e-05, "loss": 0.3523, "step": 63082 }, { "epoch": 3.81, "learning_rate": 1.4923875484737377e-05, "loss": 0.4821, "step": 63084 }, { "epoch": 3.81, "learning_rate": 1.492236657462315e-05, "loss": 0.4006, "step": 63086 }, { "epoch": 3.81, "learning_rate": 1.4920857664508925e-05, "loss": 0.3406, "step": 63088 }, { "epoch": 3.81, "learning_rate": 1.4919348754394701e-05, "loss": 0.2932, "step": 63090 }, { "epoch": 3.81, "learning_rate": 1.4917839844280476e-05, "loss": 0.2785, "step": 63092 }, { "epoch": 3.81, "learning_rate": 1.4916330934166253e-05, "loss": 0.2982, "step": 63094 }, { "epoch": 3.81, "learning_rate": 1.4914822024052027e-05, "loss": 0.3213, "step": 63096 }, { "epoch": 3.81, "learning_rate": 1.4913313113937804e-05, "loss": 0.2589, "step": 63098 }, { "epoch": 3.81, "learning_rate": 1.4911804203823579e-05, "loss": 0.5055, "step": 63100 }, { "epoch": 3.81, "learning_rate": 1.4910295293709354e-05, "loss": 0.5981, "step": 63102 }, { "epoch": 3.81, "learning_rate": 1.490878638359513e-05, "loss": 0.3743, "step": 63104 }, { "epoch": 3.81, "learning_rate": 1.4907277473480905e-05, "loss": 0.3943, "step": 63106 }, { "epoch": 3.81, "learning_rate": 1.4905768563366681e-05, "loss": 0.4675, "step": 63108 }, { "epoch": 3.81, "learning_rate": 1.4904259653252456e-05, "loss": 0.4532, "step": 63110 }, { "epoch": 3.81, "learning_rate": 1.4902750743138233e-05, "loss": 0.3283, "step": 63112 }, { "epoch": 3.81, "learning_rate": 1.4901241833024008e-05, "loss": 0.2324, "step": 63114 }, { "epoch": 3.81, "learning_rate": 1.4899732922909784e-05, "loss": 0.3764, "step": 63116 }, { "epoch": 3.81, "learning_rate": 1.4898224012795559e-05, "loss": 0.3146, "step": 63118 }, { "epoch": 3.81, "learning_rate": 1.4896715102681332e-05, "loss": 0.4351, "step": 63120 }, { "epoch": 3.81, "learning_rate": 1.489520619256711e-05, "loss": 0.4039, "step": 63122 }, { "epoch": 3.81, "learning_rate": 1.4893697282452883e-05, "loss": 0.283, "step": 63124 }, { "epoch": 3.81, "learning_rate": 1.4892188372338662e-05, "loss": 0.3132, "step": 63126 }, { "epoch": 3.81, "learning_rate": 1.4890679462224435e-05, "loss": 0.5108, "step": 63128 }, { "epoch": 3.81, "learning_rate": 1.4889170552110213e-05, "loss": 0.3973, "step": 63130 }, { "epoch": 3.81, "learning_rate": 1.4887661641995986e-05, "loss": 0.3626, "step": 63132 }, { "epoch": 3.81, "learning_rate": 1.4886152731881761e-05, "loss": 0.4684, "step": 63134 }, { "epoch": 3.81, "learning_rate": 1.4884643821767537e-05, "loss": 0.2989, "step": 63136 }, { "epoch": 3.81, "learning_rate": 1.4883134911653312e-05, "loss": 0.3437, "step": 63138 }, { "epoch": 3.81, "learning_rate": 1.4881626001539089e-05, "loss": 0.3578, "step": 63140 }, { "epoch": 3.81, "learning_rate": 1.4880117091424864e-05, "loss": 0.3522, "step": 63142 }, { "epoch": 3.81, "learning_rate": 1.487860818131064e-05, "loss": 0.2307, "step": 63144 }, { "epoch": 3.81, "learning_rate": 1.4877099271196415e-05, "loss": 0.2998, "step": 63146 }, { "epoch": 3.81, "learning_rate": 1.4875590361082193e-05, "loss": 0.3594, "step": 63148 }, { "epoch": 3.81, "learning_rate": 1.4874081450967966e-05, "loss": 0.3687, "step": 63150 }, { "epoch": 3.81, "learning_rate": 1.4872572540853741e-05, "loss": 0.3847, "step": 63152 }, { "epoch": 3.81, "learning_rate": 1.4871063630739518e-05, "loss": 0.25, "step": 63154 }, { "epoch": 3.81, "learning_rate": 1.4869554720625292e-05, "loss": 0.6096, "step": 63156 }, { "epoch": 3.81, "learning_rate": 1.4868045810511069e-05, "loss": 0.3162, "step": 63158 }, { "epoch": 3.81, "learning_rate": 1.4866536900396844e-05, "loss": 0.3541, "step": 63160 }, { "epoch": 3.81, "learning_rate": 1.486502799028262e-05, "loss": 0.466, "step": 63162 }, { "epoch": 3.81, "learning_rate": 1.4863519080168395e-05, "loss": 0.3861, "step": 63164 }, { "epoch": 3.81, "learning_rate": 1.4862010170054172e-05, "loss": 0.3618, "step": 63166 }, { "epoch": 3.81, "learning_rate": 1.4860501259939946e-05, "loss": 0.3176, "step": 63168 }, { "epoch": 3.81, "learning_rate": 1.4858992349825721e-05, "loss": 0.4414, "step": 63170 }, { "epoch": 3.81, "learning_rate": 1.4857483439711498e-05, "loss": 0.4126, "step": 63172 }, { "epoch": 3.81, "learning_rate": 1.4855974529597273e-05, "loss": 0.3965, "step": 63174 }, { "epoch": 3.81, "learning_rate": 1.4854465619483049e-05, "loss": 0.2776, "step": 63176 }, { "epoch": 3.81, "learning_rate": 1.4852956709368824e-05, "loss": 0.3964, "step": 63178 }, { "epoch": 3.81, "learning_rate": 1.48514477992546e-05, "loss": 0.3272, "step": 63180 }, { "epoch": 3.81, "learning_rate": 1.4849938889140375e-05, "loss": 0.4065, "step": 63182 }, { "epoch": 3.81, "learning_rate": 1.4848429979026148e-05, "loss": 0.2258, "step": 63184 }, { "epoch": 3.81, "learning_rate": 1.4846921068911927e-05, "loss": 0.395, "step": 63186 }, { "epoch": 3.81, "learning_rate": 1.48454121587977e-05, "loss": 0.2536, "step": 63188 }, { "epoch": 3.81, "learning_rate": 1.4843903248683478e-05, "loss": 0.301, "step": 63190 }, { "epoch": 3.81, "learning_rate": 1.4842394338569251e-05, "loss": 0.2951, "step": 63192 }, { "epoch": 3.81, "learning_rate": 1.484088542845503e-05, "loss": 0.2293, "step": 63194 }, { "epoch": 3.81, "learning_rate": 1.4839376518340802e-05, "loss": 0.2817, "step": 63196 }, { "epoch": 3.81, "learning_rate": 1.483786760822658e-05, "loss": 0.2174, "step": 63198 }, { "epoch": 3.81, "learning_rate": 1.4836358698112354e-05, "loss": 0.3396, "step": 63200 }, { "epoch": 3.81, "learning_rate": 1.4834849787998128e-05, "loss": 0.3982, "step": 63202 }, { "epoch": 3.81, "learning_rate": 1.4833340877883905e-05, "loss": 0.3681, "step": 63204 }, { "epoch": 3.81, "learning_rate": 1.483183196776968e-05, "loss": 0.3574, "step": 63206 }, { "epoch": 3.82, "learning_rate": 1.4830323057655456e-05, "loss": 0.2116, "step": 63208 }, { "epoch": 3.82, "learning_rate": 1.4828814147541231e-05, "loss": 0.3896, "step": 63210 }, { "epoch": 3.82, "learning_rate": 1.4827305237427008e-05, "loss": 0.3369, "step": 63212 }, { "epoch": 3.82, "learning_rate": 1.4825796327312782e-05, "loss": 0.341, "step": 63214 }, { "epoch": 3.82, "learning_rate": 1.4824287417198557e-05, "loss": 0.3156, "step": 63216 }, { "epoch": 3.82, "learning_rate": 1.4822778507084334e-05, "loss": 0.3594, "step": 63218 }, { "epoch": 3.82, "learning_rate": 1.4821269596970109e-05, "loss": 0.2962, "step": 63220 }, { "epoch": 3.82, "learning_rate": 1.4819760686855885e-05, "loss": 0.2998, "step": 63222 }, { "epoch": 3.82, "learning_rate": 1.481825177674166e-05, "loss": 0.6832, "step": 63224 }, { "epoch": 3.82, "learning_rate": 1.4816742866627436e-05, "loss": 0.3579, "step": 63226 }, { "epoch": 3.82, "learning_rate": 1.4815233956513211e-05, "loss": 0.3782, "step": 63228 }, { "epoch": 3.82, "learning_rate": 1.4813725046398988e-05, "loss": 0.391, "step": 63230 }, { "epoch": 3.82, "learning_rate": 1.4812216136284763e-05, "loss": 0.3578, "step": 63232 }, { "epoch": 3.82, "learning_rate": 1.4810707226170536e-05, "loss": 0.3345, "step": 63234 }, { "epoch": 3.82, "learning_rate": 1.4809198316056314e-05, "loss": 0.388, "step": 63236 }, { "epoch": 3.82, "learning_rate": 1.4807689405942087e-05, "loss": 0.4504, "step": 63238 }, { "epoch": 3.82, "learning_rate": 1.4806180495827865e-05, "loss": 0.3053, "step": 63240 }, { "epoch": 3.82, "learning_rate": 1.4804671585713638e-05, "loss": 0.3009, "step": 63242 }, { "epoch": 3.82, "learning_rate": 1.4803162675599417e-05, "loss": 0.3353, "step": 63244 }, { "epoch": 3.82, "learning_rate": 1.480165376548519e-05, "loss": 0.3128, "step": 63246 }, { "epoch": 3.82, "learning_rate": 1.4800144855370968e-05, "loss": 0.3828, "step": 63248 }, { "epoch": 3.82, "learning_rate": 1.4798635945256741e-05, "loss": 0.5043, "step": 63250 }, { "epoch": 3.82, "learning_rate": 1.4797127035142516e-05, "loss": 0.3553, "step": 63252 }, { "epoch": 3.82, "learning_rate": 1.4795618125028292e-05, "loss": 0.3016, "step": 63254 }, { "epoch": 3.82, "learning_rate": 1.4794109214914067e-05, "loss": 0.4611, "step": 63256 }, { "epoch": 3.82, "learning_rate": 1.4792600304799844e-05, "loss": 0.3872, "step": 63258 }, { "epoch": 3.82, "learning_rate": 1.4791091394685619e-05, "loss": 0.4687, "step": 63260 }, { "epoch": 3.82, "learning_rate": 1.4789582484571395e-05, "loss": 0.4125, "step": 63262 }, { "epoch": 3.82, "learning_rate": 1.478807357445717e-05, "loss": 0.427, "step": 63264 }, { "epoch": 3.82, "learning_rate": 1.4786564664342945e-05, "loss": 0.3335, "step": 63266 }, { "epoch": 3.82, "learning_rate": 1.4785055754228721e-05, "loss": 0.3184, "step": 63268 }, { "epoch": 3.82, "learning_rate": 1.4783546844114496e-05, "loss": 0.3881, "step": 63270 }, { "epoch": 3.82, "learning_rate": 1.4782037934000273e-05, "loss": 0.4386, "step": 63272 }, { "epoch": 3.82, "learning_rate": 1.4780529023886047e-05, "loss": 0.3272, "step": 63274 }, { "epoch": 3.82, "learning_rate": 1.4779020113771824e-05, "loss": 0.2749, "step": 63276 }, { "epoch": 3.82, "learning_rate": 1.4777511203657599e-05, "loss": 0.326, "step": 63278 }, { "epoch": 3.82, "learning_rate": 1.4776002293543375e-05, "loss": 0.3715, "step": 63280 }, { "epoch": 3.82, "learning_rate": 1.477449338342915e-05, "loss": 0.3309, "step": 63282 }, { "epoch": 3.82, "learning_rate": 1.4772984473314925e-05, "loss": 0.4596, "step": 63284 }, { "epoch": 3.82, "learning_rate": 1.4771475563200701e-05, "loss": 0.3049, "step": 63286 }, { "epoch": 3.82, "learning_rate": 1.4769966653086476e-05, "loss": 0.3363, "step": 63288 }, { "epoch": 3.82, "learning_rate": 1.4768457742972253e-05, "loss": 0.2578, "step": 63290 }, { "epoch": 3.82, "learning_rate": 1.4766948832858028e-05, "loss": 0.2626, "step": 63292 }, { "epoch": 3.82, "learning_rate": 1.4765439922743804e-05, "loss": 0.2749, "step": 63294 }, { "epoch": 3.82, "learning_rate": 1.4763931012629579e-05, "loss": 0.4253, "step": 63296 }, { "epoch": 3.82, "learning_rate": 1.4762422102515352e-05, "loss": 0.4031, "step": 63298 }, { "epoch": 3.82, "learning_rate": 1.476091319240113e-05, "loss": 0.2582, "step": 63300 }, { "epoch": 3.82, "learning_rate": 1.4759404282286903e-05, "loss": 0.2777, "step": 63302 }, { "epoch": 3.82, "learning_rate": 1.4757895372172682e-05, "loss": 0.404, "step": 63304 }, { "epoch": 3.82, "learning_rate": 1.4756386462058455e-05, "loss": 0.3088, "step": 63306 }, { "epoch": 3.82, "learning_rate": 1.4754877551944233e-05, "loss": 0.2063, "step": 63308 }, { "epoch": 3.82, "learning_rate": 1.4753368641830006e-05, "loss": 0.5121, "step": 63310 }, { "epoch": 3.82, "learning_rate": 1.4751859731715784e-05, "loss": 0.3656, "step": 63312 }, { "epoch": 3.82, "learning_rate": 1.4750350821601557e-05, "loss": 0.317, "step": 63314 }, { "epoch": 3.82, "learning_rate": 1.4748841911487332e-05, "loss": 0.319, "step": 63316 }, { "epoch": 3.82, "learning_rate": 1.4747333001373109e-05, "loss": 0.3835, "step": 63318 }, { "epoch": 3.82, "learning_rate": 1.4745824091258884e-05, "loss": 0.2852, "step": 63320 }, { "epoch": 3.82, "learning_rate": 1.474431518114466e-05, "loss": 0.2877, "step": 63322 }, { "epoch": 3.82, "learning_rate": 1.4742806271030435e-05, "loss": 0.248, "step": 63324 }, { "epoch": 3.82, "learning_rate": 1.4741297360916211e-05, "loss": 0.3615, "step": 63326 }, { "epoch": 3.82, "learning_rate": 1.4739788450801986e-05, "loss": 0.2785, "step": 63328 }, { "epoch": 3.82, "learning_rate": 1.4738279540687763e-05, "loss": 0.378, "step": 63330 }, { "epoch": 3.82, "learning_rate": 1.4736770630573538e-05, "loss": 0.273, "step": 63332 }, { "epoch": 3.82, "learning_rate": 1.4735261720459312e-05, "loss": 0.2754, "step": 63334 }, { "epoch": 3.82, "learning_rate": 1.4733752810345089e-05, "loss": 0.4192, "step": 63336 }, { "epoch": 3.82, "learning_rate": 1.4732243900230864e-05, "loss": 0.2923, "step": 63338 }, { "epoch": 3.82, "learning_rate": 1.473073499011664e-05, "loss": 0.3046, "step": 63340 }, { "epoch": 3.82, "learning_rate": 1.4729226080002415e-05, "loss": 0.2926, "step": 63342 }, { "epoch": 3.82, "learning_rate": 1.4727717169888192e-05, "loss": 0.355, "step": 63344 }, { "epoch": 3.82, "learning_rate": 1.4726208259773966e-05, "loss": 0.4052, "step": 63346 }, { "epoch": 3.82, "learning_rate": 1.472469934965974e-05, "loss": 0.3217, "step": 63348 }, { "epoch": 3.82, "learning_rate": 1.4723190439545518e-05, "loss": 0.3983, "step": 63350 }, { "epoch": 3.82, "learning_rate": 1.472168152943129e-05, "loss": 0.3529, "step": 63352 }, { "epoch": 3.82, "learning_rate": 1.4720172619317069e-05, "loss": 0.3124, "step": 63354 }, { "epoch": 3.82, "learning_rate": 1.4718663709202842e-05, "loss": 0.2634, "step": 63356 }, { "epoch": 3.82, "learning_rate": 1.471715479908862e-05, "loss": 0.5576, "step": 63358 }, { "epoch": 3.82, "learning_rate": 1.4715645888974393e-05, "loss": 0.2779, "step": 63360 }, { "epoch": 3.82, "learning_rate": 1.4714136978860172e-05, "loss": 0.4716, "step": 63362 }, { "epoch": 3.82, "learning_rate": 1.4712628068745945e-05, "loss": 0.401, "step": 63364 }, { "epoch": 3.82, "learning_rate": 1.471111915863172e-05, "loss": 0.2676, "step": 63366 }, { "epoch": 3.82, "learning_rate": 1.4709610248517496e-05, "loss": 0.2824, "step": 63368 }, { "epoch": 3.82, "learning_rate": 1.4708101338403271e-05, "loss": 0.2377, "step": 63370 }, { "epoch": 3.82, "learning_rate": 1.4706592428289047e-05, "loss": 0.3209, "step": 63372 }, { "epoch": 3.83, "learning_rate": 1.4705083518174822e-05, "loss": 0.4031, "step": 63374 }, { "epoch": 3.83, "learning_rate": 1.4703574608060599e-05, "loss": 0.4236, "step": 63376 }, { "epoch": 3.83, "learning_rate": 1.4702065697946374e-05, "loss": 0.427, "step": 63378 }, { "epoch": 3.83, "learning_rate": 1.4700556787832148e-05, "loss": 0.402, "step": 63380 }, { "epoch": 3.83, "learning_rate": 1.4699047877717925e-05, "loss": 0.4331, "step": 63382 }, { "epoch": 3.83, "learning_rate": 1.46975389676037e-05, "loss": 0.4454, "step": 63384 }, { "epoch": 3.83, "learning_rate": 1.4696030057489476e-05, "loss": 0.2976, "step": 63386 }, { "epoch": 3.83, "learning_rate": 1.4694521147375251e-05, "loss": 0.3056, "step": 63388 }, { "epoch": 3.83, "learning_rate": 1.4693012237261028e-05, "loss": 0.2889, "step": 63390 }, { "epoch": 3.83, "learning_rate": 1.4691503327146802e-05, "loss": 0.4218, "step": 63392 }, { "epoch": 3.83, "learning_rate": 1.4689994417032579e-05, "loss": 0.3499, "step": 63394 }, { "epoch": 3.83, "learning_rate": 1.4688485506918354e-05, "loss": 0.3003, "step": 63396 }, { "epoch": 3.83, "learning_rate": 1.4686976596804129e-05, "loss": 0.5029, "step": 63398 }, { "epoch": 3.83, "learning_rate": 1.4685467686689905e-05, "loss": 0.2718, "step": 63400 }, { "epoch": 3.83, "learning_rate": 1.468395877657568e-05, "loss": 0.2969, "step": 63402 }, { "epoch": 3.83, "learning_rate": 1.4682449866461456e-05, "loss": 0.2746, "step": 63404 }, { "epoch": 3.83, "learning_rate": 1.4680940956347231e-05, "loss": 0.2723, "step": 63406 }, { "epoch": 3.83, "learning_rate": 1.4679432046233008e-05, "loss": 0.3304, "step": 63408 }, { "epoch": 3.83, "learning_rate": 1.4677923136118783e-05, "loss": 0.5063, "step": 63410 }, { "epoch": 3.83, "learning_rate": 1.4676414226004559e-05, "loss": 0.4163, "step": 63412 }, { "epoch": 3.83, "learning_rate": 1.4674905315890334e-05, "loss": 0.3912, "step": 63414 }, { "epoch": 3.83, "learning_rate": 1.4673396405776107e-05, "loss": 0.4049, "step": 63416 }, { "epoch": 3.83, "learning_rate": 1.4671887495661885e-05, "loss": 0.48, "step": 63418 }, { "epoch": 3.83, "learning_rate": 1.4670378585547658e-05, "loss": 0.4926, "step": 63420 }, { "epoch": 3.83, "learning_rate": 1.4668869675433437e-05, "loss": 0.2878, "step": 63422 }, { "epoch": 3.83, "learning_rate": 1.466736076531921e-05, "loss": 0.3392, "step": 63424 }, { "epoch": 3.83, "learning_rate": 1.4665851855204988e-05, "loss": 0.4006, "step": 63426 }, { "epoch": 3.83, "learning_rate": 1.4664342945090761e-05, "loss": 0.4489, "step": 63428 }, { "epoch": 3.83, "learning_rate": 1.4662834034976536e-05, "loss": 0.3984, "step": 63430 }, { "epoch": 3.83, "learning_rate": 1.4661325124862312e-05, "loss": 0.3974, "step": 63432 }, { "epoch": 3.83, "learning_rate": 1.4659816214748087e-05, "loss": 0.3572, "step": 63434 }, { "epoch": 3.83, "learning_rate": 1.4658307304633864e-05, "loss": 0.361, "step": 63436 }, { "epoch": 3.83, "learning_rate": 1.4656798394519639e-05, "loss": 0.3162, "step": 63438 }, { "epoch": 3.83, "learning_rate": 1.4655289484405415e-05, "loss": 0.2676, "step": 63440 }, { "epoch": 3.83, "learning_rate": 1.465378057429119e-05, "loss": 0.3312, "step": 63442 }, { "epoch": 3.83, "learning_rate": 1.4652271664176966e-05, "loss": 0.2298, "step": 63444 }, { "epoch": 3.83, "learning_rate": 1.4650762754062741e-05, "loss": 0.3445, "step": 63446 }, { "epoch": 3.83, "learning_rate": 1.4649253843948516e-05, "loss": 0.2913, "step": 63448 }, { "epoch": 3.83, "learning_rate": 1.4647744933834293e-05, "loss": 0.3504, "step": 63450 }, { "epoch": 3.83, "learning_rate": 1.4646236023720067e-05, "loss": 0.4707, "step": 63452 }, { "epoch": 3.83, "learning_rate": 1.4644727113605844e-05, "loss": 0.3824, "step": 63454 }, { "epoch": 3.83, "learning_rate": 1.4643218203491619e-05, "loss": 0.4066, "step": 63456 }, { "epoch": 3.83, "learning_rate": 1.4641709293377395e-05, "loss": 0.2901, "step": 63458 }, { "epoch": 3.83, "learning_rate": 1.464020038326317e-05, "loss": 0.4202, "step": 63460 }, { "epoch": 3.83, "learning_rate": 1.4638691473148943e-05, "loss": 0.3247, "step": 63462 }, { "epoch": 3.83, "learning_rate": 1.4637182563034721e-05, "loss": 0.3749, "step": 63464 }, { "epoch": 3.83, "learning_rate": 1.4635673652920494e-05, "loss": 0.366, "step": 63466 }, { "epoch": 3.83, "learning_rate": 1.4634164742806273e-05, "loss": 0.3782, "step": 63468 }, { "epoch": 3.83, "learning_rate": 1.4632655832692046e-05, "loss": 0.2426, "step": 63470 }, { "epoch": 3.83, "learning_rate": 1.4631146922577824e-05, "loss": 0.3012, "step": 63472 }, { "epoch": 3.83, "learning_rate": 1.4629638012463597e-05, "loss": 0.4134, "step": 63474 }, { "epoch": 3.83, "learning_rate": 1.4628129102349375e-05, "loss": 0.4498, "step": 63476 }, { "epoch": 3.83, "learning_rate": 1.4626620192235148e-05, "loss": 0.3488, "step": 63478 }, { "epoch": 3.83, "learning_rate": 1.4625111282120923e-05, "loss": 0.2342, "step": 63480 }, { "epoch": 3.83, "learning_rate": 1.46236023720067e-05, "loss": 0.3463, "step": 63482 }, { "epoch": 3.83, "learning_rate": 1.4622093461892475e-05, "loss": 0.33, "step": 63484 }, { "epoch": 3.83, "learning_rate": 1.4620584551778251e-05, "loss": 0.2686, "step": 63486 }, { "epoch": 3.83, "learning_rate": 1.4619075641664026e-05, "loss": 0.3254, "step": 63488 }, { "epoch": 3.83, "learning_rate": 1.4617566731549802e-05, "loss": 0.4448, "step": 63490 }, { "epoch": 3.83, "learning_rate": 1.4616057821435577e-05, "loss": 0.3704, "step": 63492 }, { "epoch": 3.83, "learning_rate": 1.4614548911321354e-05, "loss": 0.3207, "step": 63494 }, { "epoch": 3.83, "learning_rate": 1.4613040001207129e-05, "loss": 0.2151, "step": 63496 }, { "epoch": 3.83, "learning_rate": 1.4611531091092903e-05, "loss": 0.333, "step": 63498 }, { "epoch": 3.83, "learning_rate": 1.461002218097868e-05, "loss": 0.3382, "step": 63500 }, { "epoch": 3.83, "learning_rate": 1.4608513270864455e-05, "loss": 0.6255, "step": 63502 }, { "epoch": 3.83, "learning_rate": 1.4607004360750231e-05, "loss": 0.2996, "step": 63504 }, { "epoch": 3.83, "learning_rate": 1.4605495450636006e-05, "loss": 0.4207, "step": 63506 }, { "epoch": 3.83, "learning_rate": 1.4603986540521783e-05, "loss": 0.4179, "step": 63508 }, { "epoch": 3.83, "learning_rate": 1.4602477630407557e-05, "loss": 0.3501, "step": 63510 }, { "epoch": 3.83, "learning_rate": 1.4600968720293332e-05, "loss": 0.4032, "step": 63512 }, { "epoch": 3.83, "learning_rate": 1.4599459810179109e-05, "loss": 0.3904, "step": 63514 }, { "epoch": 3.83, "learning_rate": 1.4597950900064884e-05, "loss": 0.4757, "step": 63516 }, { "epoch": 3.83, "learning_rate": 1.459644198995066e-05, "loss": 0.3626, "step": 63518 }, { "epoch": 3.83, "learning_rate": 1.4594933079836435e-05, "loss": 0.3124, "step": 63520 }, { "epoch": 3.83, "learning_rate": 1.4593424169722211e-05, "loss": 0.4107, "step": 63522 }, { "epoch": 3.83, "learning_rate": 1.4591915259607986e-05, "loss": 0.3904, "step": 63524 }, { "epoch": 3.83, "learning_rate": 1.4590406349493763e-05, "loss": 0.3578, "step": 63526 }, { "epoch": 3.83, "learning_rate": 1.4588897439379538e-05, "loss": 0.4218, "step": 63528 }, { "epoch": 3.83, "learning_rate": 1.458738852926531e-05, "loss": 0.2876, "step": 63530 }, { "epoch": 3.83, "learning_rate": 1.4585879619151089e-05, "loss": 0.2534, "step": 63532 }, { "epoch": 3.83, "learning_rate": 1.4584370709036862e-05, "loss": 0.4108, "step": 63534 }, { "epoch": 3.83, "learning_rate": 1.458286179892264e-05, "loss": 0.3868, "step": 63536 }, { "epoch": 3.83, "learning_rate": 1.4581352888808413e-05, "loss": 0.4864, "step": 63538 }, { "epoch": 3.84, "learning_rate": 1.4579843978694192e-05, "loss": 0.4991, "step": 63540 }, { "epoch": 3.84, "learning_rate": 1.4578335068579965e-05, "loss": 0.2418, "step": 63542 }, { "epoch": 3.84, "learning_rate": 1.457682615846574e-05, "loss": 0.3215, "step": 63544 }, { "epoch": 3.84, "learning_rate": 1.4575317248351516e-05, "loss": 0.3812, "step": 63546 }, { "epoch": 3.84, "learning_rate": 1.4573808338237291e-05, "loss": 0.2917, "step": 63548 }, { "epoch": 3.84, "learning_rate": 1.4572299428123067e-05, "loss": 0.5139, "step": 63550 }, { "epoch": 3.84, "learning_rate": 1.4570790518008842e-05, "loss": 0.3379, "step": 63552 }, { "epoch": 3.84, "learning_rate": 1.4569281607894619e-05, "loss": 0.3597, "step": 63554 }, { "epoch": 3.84, "learning_rate": 1.4567772697780394e-05, "loss": 0.3224, "step": 63556 }, { "epoch": 3.84, "learning_rate": 1.456626378766617e-05, "loss": 0.3678, "step": 63558 }, { "epoch": 3.84, "learning_rate": 1.4564754877551945e-05, "loss": 0.2461, "step": 63560 }, { "epoch": 3.84, "learning_rate": 1.456324596743772e-05, "loss": 0.489, "step": 63562 }, { "epoch": 3.84, "learning_rate": 1.4561737057323496e-05, "loss": 0.3736, "step": 63564 }, { "epoch": 3.84, "learning_rate": 1.4560228147209271e-05, "loss": 0.3757, "step": 63566 }, { "epoch": 3.84, "learning_rate": 1.4558719237095048e-05, "loss": 0.3695, "step": 63568 }, { "epoch": 3.84, "learning_rate": 1.4557210326980822e-05, "loss": 0.4891, "step": 63570 }, { "epoch": 3.84, "learning_rate": 1.4555701416866599e-05, "loss": 0.5419, "step": 63572 }, { "epoch": 3.84, "learning_rate": 1.4554192506752374e-05, "loss": 0.2621, "step": 63574 }, { "epoch": 3.84, "learning_rate": 1.455268359663815e-05, "loss": 0.333, "step": 63576 }, { "epoch": 3.84, "learning_rate": 1.4551174686523925e-05, "loss": 0.3765, "step": 63578 }, { "epoch": 3.84, "learning_rate": 1.4549665776409698e-05, "loss": 0.3849, "step": 63580 }, { "epoch": 3.84, "learning_rate": 1.4548156866295476e-05, "loss": 0.4399, "step": 63582 }, { "epoch": 3.84, "learning_rate": 1.454664795618125e-05, "loss": 0.4918, "step": 63584 }, { "epoch": 3.84, "learning_rate": 1.4545139046067028e-05, "loss": 0.2907, "step": 63586 }, { "epoch": 3.84, "learning_rate": 1.45436301359528e-05, "loss": 0.4389, "step": 63588 }, { "epoch": 3.84, "learning_rate": 1.4542121225838579e-05, "loss": 0.46, "step": 63590 }, { "epoch": 3.84, "learning_rate": 1.4540612315724352e-05, "loss": 0.3565, "step": 63592 }, { "epoch": 3.84, "learning_rate": 1.4539103405610127e-05, "loss": 0.3146, "step": 63594 }, { "epoch": 3.84, "learning_rate": 1.4537594495495903e-05, "loss": 0.2666, "step": 63596 }, { "epoch": 3.84, "learning_rate": 1.4536085585381678e-05, "loss": 0.4551, "step": 63598 }, { "epoch": 3.84, "learning_rate": 1.4534576675267455e-05, "loss": 0.2543, "step": 63600 }, { "epoch": 3.84, "learning_rate": 1.453306776515323e-05, "loss": 0.4188, "step": 63602 }, { "epoch": 3.84, "learning_rate": 1.4531558855039006e-05, "loss": 0.5197, "step": 63604 }, { "epoch": 3.84, "learning_rate": 1.4530049944924781e-05, "loss": 0.301, "step": 63606 }, { "epoch": 3.84, "learning_rate": 1.4528541034810557e-05, "loss": 0.2788, "step": 63608 }, { "epoch": 3.84, "learning_rate": 1.4527032124696332e-05, "loss": 0.3217, "step": 63610 }, { "epoch": 3.84, "learning_rate": 1.4525523214582107e-05, "loss": 0.3368, "step": 63612 }, { "epoch": 3.84, "learning_rate": 1.4524014304467884e-05, "loss": 0.4132, "step": 63614 }, { "epoch": 3.84, "learning_rate": 1.4522505394353658e-05, "loss": 0.4068, "step": 63616 }, { "epoch": 3.84, "learning_rate": 1.4520996484239435e-05, "loss": 0.385, "step": 63618 }, { "epoch": 3.84, "learning_rate": 1.451948757412521e-05, "loss": 0.3971, "step": 63620 }, { "epoch": 3.84, "learning_rate": 1.4517978664010986e-05, "loss": 0.5365, "step": 63622 }, { "epoch": 3.84, "learning_rate": 1.4516469753896761e-05, "loss": 0.3913, "step": 63624 }, { "epoch": 3.84, "learning_rate": 1.4514960843782538e-05, "loss": 0.3031, "step": 63626 }, { "epoch": 3.84, "learning_rate": 1.4513451933668312e-05, "loss": 0.3011, "step": 63628 }, { "epoch": 3.84, "learning_rate": 1.4511943023554087e-05, "loss": 0.3732, "step": 63630 }, { "epoch": 3.84, "learning_rate": 1.4510434113439864e-05, "loss": 0.3794, "step": 63632 }, { "epoch": 3.84, "learning_rate": 1.4508925203325639e-05, "loss": 0.2937, "step": 63634 }, { "epoch": 3.84, "learning_rate": 1.4507416293211415e-05, "loss": 0.3678, "step": 63636 }, { "epoch": 3.84, "learning_rate": 1.450590738309719e-05, "loss": 0.3928, "step": 63638 }, { "epoch": 3.84, "learning_rate": 1.4504398472982966e-05, "loss": 0.4103, "step": 63640 }, { "epoch": 3.84, "learning_rate": 1.4502889562868741e-05, "loss": 0.3734, "step": 63642 }, { "epoch": 3.84, "learning_rate": 1.4501380652754514e-05, "loss": 0.2731, "step": 63644 }, { "epoch": 3.84, "learning_rate": 1.4499871742640293e-05, "loss": 0.4703, "step": 63646 }, { "epoch": 3.84, "learning_rate": 1.4498362832526066e-05, "loss": 0.3408, "step": 63648 }, { "epoch": 3.84, "learning_rate": 1.4496853922411844e-05, "loss": 0.4614, "step": 63650 }, { "epoch": 3.84, "learning_rate": 1.4495345012297617e-05, "loss": 0.4466, "step": 63652 }, { "epoch": 3.84, "learning_rate": 1.4493836102183395e-05, "loss": 0.3555, "step": 63654 }, { "epoch": 3.84, "learning_rate": 1.4492327192069168e-05, "loss": 0.2721, "step": 63656 }, { "epoch": 3.84, "learning_rate": 1.4490818281954947e-05, "loss": 0.3556, "step": 63658 }, { "epoch": 3.84, "learning_rate": 1.448930937184072e-05, "loss": 0.3546, "step": 63660 }, { "epoch": 3.84, "learning_rate": 1.4487800461726495e-05, "loss": 0.4798, "step": 63662 }, { "epoch": 3.84, "learning_rate": 1.4486291551612271e-05, "loss": 0.5061, "step": 63664 }, { "epoch": 3.84, "learning_rate": 1.4484782641498046e-05, "loss": 0.3159, "step": 63666 }, { "epoch": 3.84, "learning_rate": 1.4483273731383822e-05, "loss": 0.4031, "step": 63668 }, { "epoch": 3.84, "learning_rate": 1.4481764821269597e-05, "loss": 0.351, "step": 63670 }, { "epoch": 3.84, "learning_rate": 1.4480255911155374e-05, "loss": 0.3504, "step": 63672 }, { "epoch": 3.84, "learning_rate": 1.4478747001041149e-05, "loss": 0.3087, "step": 63674 }, { "epoch": 3.84, "learning_rate": 1.4477238090926923e-05, "loss": 0.383, "step": 63676 }, { "epoch": 3.84, "learning_rate": 1.44757291808127e-05, "loss": 0.3605, "step": 63678 }, { "epoch": 3.84, "learning_rate": 1.4474220270698475e-05, "loss": 0.2784, "step": 63680 }, { "epoch": 3.84, "learning_rate": 1.4472711360584251e-05, "loss": 0.3677, "step": 63682 }, { "epoch": 3.84, "learning_rate": 1.4471202450470026e-05, "loss": 0.5709, "step": 63684 }, { "epoch": 3.84, "learning_rate": 1.4469693540355803e-05, "loss": 0.4618, "step": 63686 }, { "epoch": 3.84, "learning_rate": 1.4468184630241577e-05, "loss": 0.4315, "step": 63688 }, { "epoch": 3.84, "learning_rate": 1.4466675720127354e-05, "loss": 0.3727, "step": 63690 }, { "epoch": 3.84, "learning_rate": 1.4465166810013129e-05, "loss": 0.2153, "step": 63692 }, { "epoch": 3.84, "learning_rate": 1.4463657899898902e-05, "loss": 0.2485, "step": 63694 }, { "epoch": 3.84, "learning_rate": 1.446214898978468e-05, "loss": 0.3659, "step": 63696 }, { "epoch": 3.84, "learning_rate": 1.4460640079670453e-05, "loss": 0.406, "step": 63698 }, { "epoch": 3.84, "learning_rate": 1.4459131169556231e-05, "loss": 0.4085, "step": 63700 }, { "epoch": 3.84, "learning_rate": 1.4457622259442004e-05, "loss": 0.4182, "step": 63702 }, { "epoch": 3.85, "learning_rate": 1.4456113349327783e-05, "loss": 0.2919, "step": 63704 }, { "epoch": 3.85, "learning_rate": 1.4454604439213556e-05, "loss": 0.3138, "step": 63706 }, { "epoch": 3.85, "learning_rate": 1.4453095529099334e-05, "loss": 0.2919, "step": 63708 }, { "epoch": 3.85, "learning_rate": 1.4451586618985107e-05, "loss": 0.4956, "step": 63710 }, { "epoch": 3.85, "learning_rate": 1.4450077708870882e-05, "loss": 0.2781, "step": 63712 }, { "epoch": 3.85, "learning_rate": 1.4448568798756658e-05, "loss": 0.4641, "step": 63714 }, { "epoch": 3.85, "learning_rate": 1.4447059888642433e-05, "loss": 0.3241, "step": 63716 }, { "epoch": 3.85, "learning_rate": 1.444555097852821e-05, "loss": 0.579, "step": 63718 }, { "epoch": 3.85, "learning_rate": 1.4444042068413985e-05, "loss": 0.2898, "step": 63720 }, { "epoch": 3.85, "learning_rate": 1.4442533158299761e-05, "loss": 0.3683, "step": 63722 }, { "epoch": 3.85, "learning_rate": 1.4441024248185536e-05, "loss": 0.2601, "step": 63724 }, { "epoch": 3.85, "learning_rate": 1.443951533807131e-05, "loss": 0.2682, "step": 63726 }, { "epoch": 3.85, "learning_rate": 1.4438006427957087e-05, "loss": 0.2393, "step": 63728 }, { "epoch": 3.85, "learning_rate": 1.4436497517842862e-05, "loss": 0.2424, "step": 63730 }, { "epoch": 3.85, "learning_rate": 1.4434988607728639e-05, "loss": 0.5434, "step": 63732 }, { "epoch": 3.85, "learning_rate": 1.4433479697614413e-05, "loss": 0.2749, "step": 63734 }, { "epoch": 3.85, "learning_rate": 1.443197078750019e-05, "loss": 0.274, "step": 63736 }, { "epoch": 3.85, "learning_rate": 1.4430461877385965e-05, "loss": 0.3139, "step": 63738 }, { "epoch": 3.85, "learning_rate": 1.4428952967271741e-05, "loss": 0.2087, "step": 63740 }, { "epoch": 3.85, "learning_rate": 1.4427444057157516e-05, "loss": 0.3829, "step": 63742 }, { "epoch": 3.85, "learning_rate": 1.4425935147043291e-05, "loss": 0.2432, "step": 63744 }, { "epoch": 3.85, "learning_rate": 1.4424426236929067e-05, "loss": 0.5656, "step": 63746 }, { "epoch": 3.85, "learning_rate": 1.4422917326814842e-05, "loss": 0.4466, "step": 63748 }, { "epoch": 3.85, "learning_rate": 1.4421408416700619e-05, "loss": 0.262, "step": 63750 }, { "epoch": 3.85, "learning_rate": 1.4419899506586394e-05, "loss": 0.2661, "step": 63752 }, { "epoch": 3.85, "learning_rate": 1.441839059647217e-05, "loss": 0.4079, "step": 63754 }, { "epoch": 3.85, "learning_rate": 1.4416881686357945e-05, "loss": 0.5184, "step": 63756 }, { "epoch": 3.85, "learning_rate": 1.4415372776243718e-05, "loss": 0.357, "step": 63758 }, { "epoch": 3.85, "learning_rate": 1.4413863866129496e-05, "loss": 0.3318, "step": 63760 }, { "epoch": 3.85, "learning_rate": 1.441235495601527e-05, "loss": 0.1824, "step": 63762 }, { "epoch": 3.85, "learning_rate": 1.4410846045901048e-05, "loss": 0.3096, "step": 63764 }, { "epoch": 3.85, "learning_rate": 1.440933713578682e-05, "loss": 0.4955, "step": 63766 }, { "epoch": 3.85, "learning_rate": 1.4407828225672599e-05, "loss": 0.3803, "step": 63768 }, { "epoch": 3.85, "learning_rate": 1.4406319315558372e-05, "loss": 0.3693, "step": 63770 }, { "epoch": 3.85, "learning_rate": 1.440481040544415e-05, "loss": 0.3123, "step": 63772 }, { "epoch": 3.85, "learning_rate": 1.4403301495329923e-05, "loss": 0.2666, "step": 63774 }, { "epoch": 3.85, "learning_rate": 1.4401792585215698e-05, "loss": 0.5377, "step": 63776 }, { "epoch": 3.85, "learning_rate": 1.4400283675101475e-05, "loss": 0.2855, "step": 63778 }, { "epoch": 3.85, "learning_rate": 1.439877476498725e-05, "loss": 0.3489, "step": 63780 }, { "epoch": 3.85, "learning_rate": 1.4397265854873026e-05, "loss": 0.3239, "step": 63782 }, { "epoch": 3.85, "learning_rate": 1.4395756944758801e-05, "loss": 0.3394, "step": 63784 }, { "epoch": 3.85, "learning_rate": 1.4394248034644577e-05, "loss": 0.2441, "step": 63786 }, { "epoch": 3.85, "learning_rate": 1.4392739124530352e-05, "loss": 0.3136, "step": 63788 }, { "epoch": 3.85, "learning_rate": 1.4391230214416129e-05, "loss": 0.2914, "step": 63790 }, { "epoch": 3.85, "learning_rate": 1.4389721304301904e-05, "loss": 0.2488, "step": 63792 }, { "epoch": 3.85, "learning_rate": 1.4388212394187678e-05, "loss": 0.2123, "step": 63794 }, { "epoch": 3.85, "learning_rate": 1.4386703484073455e-05, "loss": 0.3816, "step": 63796 }, { "epoch": 3.85, "learning_rate": 1.438519457395923e-05, "loss": 0.4398, "step": 63798 }, { "epoch": 3.85, "learning_rate": 1.4383685663845006e-05, "loss": 0.4456, "step": 63800 }, { "epoch": 3.85, "learning_rate": 1.4382176753730781e-05, "loss": 0.4074, "step": 63802 }, { "epoch": 3.85, "learning_rate": 1.4380667843616558e-05, "loss": 0.4112, "step": 63804 }, { "epoch": 3.85, "learning_rate": 1.4379158933502332e-05, "loss": 0.3656, "step": 63806 }, { "epoch": 3.85, "learning_rate": 1.4377650023388105e-05, "loss": 0.1798, "step": 63808 }, { "epoch": 3.85, "learning_rate": 1.4376141113273884e-05, "loss": 0.2804, "step": 63810 }, { "epoch": 3.85, "learning_rate": 1.4374632203159657e-05, "loss": 0.5493, "step": 63812 }, { "epoch": 3.85, "learning_rate": 1.4373123293045435e-05, "loss": 0.3175, "step": 63814 }, { "epoch": 3.85, "learning_rate": 1.4371614382931208e-05, "loss": 0.4306, "step": 63816 }, { "epoch": 3.85, "learning_rate": 1.4370105472816986e-05, "loss": 0.2664, "step": 63818 }, { "epoch": 3.85, "learning_rate": 1.436859656270276e-05, "loss": 0.2957, "step": 63820 }, { "epoch": 3.85, "learning_rate": 1.4367087652588538e-05, "loss": 0.2353, "step": 63822 }, { "epoch": 3.85, "learning_rate": 1.436557874247431e-05, "loss": 0.2747, "step": 63824 }, { "epoch": 3.85, "learning_rate": 1.4364069832360086e-05, "loss": 0.2818, "step": 63826 }, { "epoch": 3.85, "learning_rate": 1.4362560922245862e-05, "loss": 0.4388, "step": 63828 }, { "epoch": 3.85, "learning_rate": 1.4361052012131637e-05, "loss": 0.5596, "step": 63830 }, { "epoch": 3.85, "learning_rate": 1.4359543102017413e-05, "loss": 0.4456, "step": 63832 }, { "epoch": 3.85, "learning_rate": 1.4358034191903188e-05, "loss": 0.3438, "step": 63834 }, { "epoch": 3.85, "learning_rate": 1.4356525281788965e-05, "loss": 0.4187, "step": 63836 }, { "epoch": 3.85, "learning_rate": 1.435501637167474e-05, "loss": 0.2373, "step": 63838 }, { "epoch": 3.85, "learning_rate": 1.4353507461560514e-05, "loss": 0.3568, "step": 63840 }, { "epoch": 3.85, "learning_rate": 1.4351998551446291e-05, "loss": 0.3901, "step": 63842 }, { "epoch": 3.85, "learning_rate": 1.4350489641332066e-05, "loss": 0.2166, "step": 63844 }, { "epoch": 3.85, "learning_rate": 1.4348980731217842e-05, "loss": 0.4814, "step": 63846 }, { "epoch": 3.85, "learning_rate": 1.4347471821103617e-05, "loss": 0.2988, "step": 63848 }, { "epoch": 3.85, "learning_rate": 1.4345962910989394e-05, "loss": 0.3902, "step": 63850 }, { "epoch": 3.85, "learning_rate": 1.4344454000875168e-05, "loss": 0.3903, "step": 63852 }, { "epoch": 3.85, "learning_rate": 1.4342945090760945e-05, "loss": 0.6109, "step": 63854 }, { "epoch": 3.85, "learning_rate": 1.434143618064672e-05, "loss": 0.3866, "step": 63856 }, { "epoch": 3.85, "learning_rate": 1.4339927270532495e-05, "loss": 0.3204, "step": 63858 }, { "epoch": 3.85, "learning_rate": 1.4338418360418271e-05, "loss": 0.4026, "step": 63860 }, { "epoch": 3.85, "learning_rate": 1.4336909450304046e-05, "loss": 0.379, "step": 63862 }, { "epoch": 3.85, "learning_rate": 1.4335400540189822e-05, "loss": 0.3806, "step": 63864 }, { "epoch": 3.85, "learning_rate": 1.4333891630075597e-05, "loss": 0.3151, "step": 63866 }, { "epoch": 3.85, "learning_rate": 1.4332382719961374e-05, "loss": 0.4427, "step": 63868 }, { "epoch": 3.86, "learning_rate": 1.4330873809847149e-05, "loss": 0.4772, "step": 63870 }, { "epoch": 3.86, "learning_rate": 1.4329364899732925e-05, "loss": 0.3097, "step": 63872 }, { "epoch": 3.86, "learning_rate": 1.43278559896187e-05, "loss": 0.3353, "step": 63874 }, { "epoch": 3.86, "learning_rate": 1.4326347079504473e-05, "loss": 0.3486, "step": 63876 }, { "epoch": 3.86, "learning_rate": 1.4324838169390251e-05, "loss": 0.4454, "step": 63878 }, { "epoch": 3.86, "learning_rate": 1.4323329259276024e-05, "loss": 0.4107, "step": 63880 }, { "epoch": 3.86, "learning_rate": 1.4321820349161803e-05, "loss": 0.2781, "step": 63882 }, { "epoch": 3.86, "learning_rate": 1.4320311439047576e-05, "loss": 0.3493, "step": 63884 }, { "epoch": 3.86, "learning_rate": 1.4318802528933354e-05, "loss": 0.2314, "step": 63886 }, { "epoch": 3.86, "learning_rate": 1.4317293618819127e-05, "loss": 0.2481, "step": 63888 }, { "epoch": 3.86, "learning_rate": 1.4315784708704902e-05, "loss": 0.4174, "step": 63890 }, { "epoch": 3.86, "learning_rate": 1.4314275798590678e-05, "loss": 0.4622, "step": 63892 }, { "epoch": 3.86, "learning_rate": 1.4312766888476453e-05, "loss": 0.3726, "step": 63894 }, { "epoch": 3.86, "learning_rate": 1.431125797836223e-05, "loss": 0.446, "step": 63896 }, { "epoch": 3.86, "learning_rate": 1.4309749068248005e-05, "loss": 0.471, "step": 63898 }, { "epoch": 3.86, "learning_rate": 1.4308240158133781e-05, "loss": 0.4129, "step": 63900 }, { "epoch": 3.86, "learning_rate": 1.4306731248019556e-05, "loss": 0.348, "step": 63902 }, { "epoch": 3.86, "learning_rate": 1.4305222337905332e-05, "loss": 0.3437, "step": 63904 }, { "epoch": 3.86, "learning_rate": 1.4303713427791107e-05, "loss": 0.3629, "step": 63906 }, { "epoch": 3.86, "learning_rate": 1.4302204517676882e-05, "loss": 0.2698, "step": 63908 }, { "epoch": 3.86, "learning_rate": 1.4300695607562659e-05, "loss": 0.4077, "step": 63910 }, { "epoch": 3.86, "learning_rate": 1.4299186697448433e-05, "loss": 0.446, "step": 63912 }, { "epoch": 3.86, "learning_rate": 1.429767778733421e-05, "loss": 0.4441, "step": 63914 }, { "epoch": 3.86, "learning_rate": 1.4296168877219985e-05, "loss": 0.3037, "step": 63916 }, { "epoch": 3.86, "learning_rate": 1.4294659967105761e-05, "loss": 0.2969, "step": 63918 }, { "epoch": 3.86, "learning_rate": 1.4293151056991536e-05, "loss": 0.2727, "step": 63920 }, { "epoch": 3.86, "learning_rate": 1.429164214687731e-05, "loss": 0.2111, "step": 63922 }, { "epoch": 3.86, "learning_rate": 1.4290133236763087e-05, "loss": 0.2608, "step": 63924 }, { "epoch": 3.86, "learning_rate": 1.428862432664886e-05, "loss": 0.3878, "step": 63926 }, { "epoch": 3.86, "learning_rate": 1.4287115416534639e-05, "loss": 0.2883, "step": 63928 }, { "epoch": 3.86, "learning_rate": 1.4285606506420412e-05, "loss": 0.2711, "step": 63930 }, { "epoch": 3.86, "learning_rate": 1.428409759630619e-05, "loss": 0.3978, "step": 63932 }, { "epoch": 3.86, "learning_rate": 1.4282588686191963e-05, "loss": 0.4158, "step": 63934 }, { "epoch": 3.86, "learning_rate": 1.4281079776077741e-05, "loss": 0.3767, "step": 63936 }, { "epoch": 3.86, "learning_rate": 1.4279570865963515e-05, "loss": 0.2499, "step": 63938 }, { "epoch": 3.86, "learning_rate": 1.427806195584929e-05, "loss": 0.4119, "step": 63940 }, { "epoch": 3.86, "learning_rate": 1.4276553045735066e-05, "loss": 0.3677, "step": 63942 }, { "epoch": 3.86, "learning_rate": 1.427504413562084e-05, "loss": 0.4205, "step": 63944 }, { "epoch": 3.86, "learning_rate": 1.4273535225506617e-05, "loss": 0.459, "step": 63946 }, { "epoch": 3.86, "learning_rate": 1.4272026315392392e-05, "loss": 0.4127, "step": 63948 }, { "epoch": 3.86, "learning_rate": 1.4270517405278169e-05, "loss": 0.5211, "step": 63950 }, { "epoch": 3.86, "learning_rate": 1.4269008495163943e-05, "loss": 0.4735, "step": 63952 }, { "epoch": 3.86, "learning_rate": 1.426749958504972e-05, "loss": 0.3435, "step": 63954 }, { "epoch": 3.86, "learning_rate": 1.4265990674935495e-05, "loss": 0.2911, "step": 63956 }, { "epoch": 3.86, "learning_rate": 1.426448176482127e-05, "loss": 0.3093, "step": 63958 }, { "epoch": 3.86, "learning_rate": 1.4262972854707046e-05, "loss": 0.364, "step": 63960 }, { "epoch": 3.86, "learning_rate": 1.426146394459282e-05, "loss": 0.3635, "step": 63962 }, { "epoch": 3.86, "learning_rate": 1.4259955034478597e-05, "loss": 0.2726, "step": 63964 }, { "epoch": 3.86, "learning_rate": 1.4258446124364372e-05, "loss": 0.3885, "step": 63966 }, { "epoch": 3.86, "learning_rate": 1.4256937214250149e-05, "loss": 0.3726, "step": 63968 }, { "epoch": 3.86, "learning_rate": 1.4255428304135923e-05, "loss": 0.3041, "step": 63970 }, { "epoch": 3.86, "learning_rate": 1.4253919394021698e-05, "loss": 0.3299, "step": 63972 }, { "epoch": 3.86, "learning_rate": 1.4252410483907475e-05, "loss": 0.3161, "step": 63974 }, { "epoch": 3.86, "learning_rate": 1.425090157379325e-05, "loss": 0.435, "step": 63976 }, { "epoch": 3.86, "learning_rate": 1.4249392663679026e-05, "loss": 0.3744, "step": 63978 }, { "epoch": 3.86, "learning_rate": 1.4247883753564801e-05, "loss": 0.3867, "step": 63980 }, { "epoch": 3.86, "learning_rate": 1.4246374843450577e-05, "loss": 0.3078, "step": 63982 }, { "epoch": 3.86, "learning_rate": 1.4244865933336352e-05, "loss": 0.4076, "step": 63984 }, { "epoch": 3.86, "learning_rate": 1.4243357023222129e-05, "loss": 0.4308, "step": 63986 }, { "epoch": 3.86, "learning_rate": 1.4241848113107904e-05, "loss": 0.2825, "step": 63988 }, { "epoch": 3.86, "learning_rate": 1.4240339202993677e-05, "loss": 0.2953, "step": 63990 }, { "epoch": 3.86, "learning_rate": 1.4238830292879455e-05, "loss": 0.3455, "step": 63992 }, { "epoch": 3.86, "learning_rate": 1.4237321382765228e-05, "loss": 0.3197, "step": 63994 }, { "epoch": 3.86, "learning_rate": 1.4235812472651006e-05, "loss": 0.365, "step": 63996 }, { "epoch": 3.86, "learning_rate": 1.423430356253678e-05, "loss": 0.3285, "step": 63998 }, { "epoch": 3.86, "learning_rate": 1.4232794652422558e-05, "loss": 0.4187, "step": 64000 }, { "epoch": 3.86, "learning_rate": 1.423128574230833e-05, "loss": 0.5016, "step": 64002 }, { "epoch": 3.86, "learning_rate": 1.4229776832194106e-05, "loss": 0.6123, "step": 64004 }, { "epoch": 3.86, "learning_rate": 1.4228267922079882e-05, "loss": 0.2763, "step": 64006 }, { "epoch": 3.86, "learning_rate": 1.4226759011965657e-05, "loss": 0.5017, "step": 64008 }, { "epoch": 3.86, "learning_rate": 1.4225250101851433e-05, "loss": 0.244, "step": 64010 }, { "epoch": 3.86, "learning_rate": 1.4223741191737208e-05, "loss": 0.3204, "step": 64012 }, { "epoch": 3.86, "learning_rate": 1.4222232281622985e-05, "loss": 0.3548, "step": 64014 }, { "epoch": 3.86, "learning_rate": 1.422072337150876e-05, "loss": 0.4713, "step": 64016 }, { "epoch": 3.86, "learning_rate": 1.4219214461394536e-05, "loss": 0.3333, "step": 64018 }, { "epoch": 3.86, "learning_rate": 1.4217705551280311e-05, "loss": 0.3417, "step": 64020 }, { "epoch": 3.86, "learning_rate": 1.4216196641166086e-05, "loss": 0.3467, "step": 64022 }, { "epoch": 3.86, "learning_rate": 1.4214687731051862e-05, "loss": 0.3074, "step": 64024 }, { "epoch": 3.86, "learning_rate": 1.4213178820937637e-05, "loss": 0.2491, "step": 64026 }, { "epoch": 3.86, "learning_rate": 1.4211669910823414e-05, "loss": 0.5639, "step": 64028 }, { "epoch": 3.86, "learning_rate": 1.4210161000709188e-05, "loss": 0.5051, "step": 64030 }, { "epoch": 3.86, "learning_rate": 1.4208652090594965e-05, "loss": 0.4579, "step": 64032 }, { "epoch": 3.86, "learning_rate": 1.420714318048074e-05, "loss": 0.3381, "step": 64034 }, { "epoch": 3.87, "learning_rate": 1.4205634270366516e-05, "loss": 0.3541, "step": 64036 }, { "epoch": 3.87, "learning_rate": 1.4204125360252291e-05, "loss": 0.3652, "step": 64038 }, { "epoch": 3.87, "learning_rate": 1.4202616450138064e-05, "loss": 0.3999, "step": 64040 }, { "epoch": 3.87, "learning_rate": 1.4201107540023842e-05, "loss": 0.6056, "step": 64042 }, { "epoch": 3.87, "learning_rate": 1.4199598629909616e-05, "loss": 0.297, "step": 64044 }, { "epoch": 3.87, "learning_rate": 1.4198089719795394e-05, "loss": 0.4032, "step": 64046 }, { "epoch": 3.87, "learning_rate": 1.4196580809681167e-05, "loss": 0.2368, "step": 64048 }, { "epoch": 3.87, "learning_rate": 1.4195071899566945e-05, "loss": 0.4677, "step": 64050 }, { "epoch": 3.87, "learning_rate": 1.4193562989452718e-05, "loss": 0.4636, "step": 64052 }, { "epoch": 3.87, "learning_rate": 1.4192054079338493e-05, "loss": 0.3171, "step": 64054 }, { "epoch": 3.87, "learning_rate": 1.419054516922427e-05, "loss": 0.3428, "step": 64056 }, { "epoch": 3.87, "learning_rate": 1.4189036259110044e-05, "loss": 0.2866, "step": 64058 }, { "epoch": 3.87, "learning_rate": 1.4187527348995821e-05, "loss": 0.4183, "step": 64060 }, { "epoch": 3.87, "learning_rate": 1.4186018438881596e-05, "loss": 0.3383, "step": 64062 }, { "epoch": 3.87, "learning_rate": 1.4184509528767372e-05, "loss": 0.5162, "step": 64064 }, { "epoch": 3.87, "learning_rate": 1.4183000618653147e-05, "loss": 0.1591, "step": 64066 }, { "epoch": 3.87, "learning_rate": 1.4181491708538924e-05, "loss": 0.3059, "step": 64068 }, { "epoch": 3.87, "learning_rate": 1.4179982798424698e-05, "loss": 0.4009, "step": 64070 }, { "epoch": 3.87, "learning_rate": 1.4178473888310473e-05, "loss": 0.4139, "step": 64072 }, { "epoch": 3.87, "learning_rate": 1.417696497819625e-05, "loss": 0.1603, "step": 64074 }, { "epoch": 3.87, "learning_rate": 1.4175456068082024e-05, "loss": 0.3405, "step": 64076 }, { "epoch": 3.87, "learning_rate": 1.4173947157967801e-05, "loss": 0.3082, "step": 64078 }, { "epoch": 3.87, "learning_rate": 1.4172438247853576e-05, "loss": 0.4037, "step": 64080 }, { "epoch": 3.87, "learning_rate": 1.4170929337739352e-05, "loss": 0.3969, "step": 64082 }, { "epoch": 3.87, "learning_rate": 1.4169420427625127e-05, "loss": 0.3743, "step": 64084 }, { "epoch": 3.87, "learning_rate": 1.4167911517510902e-05, "loss": 0.3697, "step": 64086 }, { "epoch": 3.87, "learning_rate": 1.4166402607396678e-05, "loss": 0.2913, "step": 64088 }, { "epoch": 3.87, "learning_rate": 1.4164893697282453e-05, "loss": 0.5648, "step": 64090 }, { "epoch": 3.87, "learning_rate": 1.416338478716823e-05, "loss": 0.2988, "step": 64092 }, { "epoch": 3.87, "learning_rate": 1.4161875877054005e-05, "loss": 0.3579, "step": 64094 }, { "epoch": 3.87, "learning_rate": 1.4160366966939781e-05, "loss": 0.4101, "step": 64096 }, { "epoch": 3.87, "learning_rate": 1.4158858056825556e-05, "loss": 0.4034, "step": 64098 }, { "epoch": 3.87, "learning_rate": 1.4157349146711332e-05, "loss": 0.2698, "step": 64100 }, { "epoch": 3.87, "learning_rate": 1.4155840236597107e-05, "loss": 0.4193, "step": 64102 }, { "epoch": 3.87, "learning_rate": 1.415433132648288e-05, "loss": 0.3248, "step": 64104 }, { "epoch": 3.87, "learning_rate": 1.4152822416368659e-05, "loss": 0.4246, "step": 64106 }, { "epoch": 3.87, "learning_rate": 1.4151313506254432e-05, "loss": 0.29, "step": 64108 }, { "epoch": 3.87, "learning_rate": 1.414980459614021e-05, "loss": 0.5105, "step": 64110 }, { "epoch": 3.87, "learning_rate": 1.4148295686025983e-05, "loss": 0.2952, "step": 64112 }, { "epoch": 3.87, "learning_rate": 1.4146786775911761e-05, "loss": 0.2634, "step": 64114 }, { "epoch": 3.87, "learning_rate": 1.4145277865797534e-05, "loss": 0.4306, "step": 64116 }, { "epoch": 3.87, "learning_rate": 1.4143768955683313e-05, "loss": 0.2283, "step": 64118 }, { "epoch": 3.87, "learning_rate": 1.4142260045569086e-05, "loss": 0.358, "step": 64120 }, { "epoch": 3.87, "learning_rate": 1.414075113545486e-05, "loss": 0.3775, "step": 64122 }, { "epoch": 3.87, "learning_rate": 1.4139242225340637e-05, "loss": 0.2056, "step": 64124 }, { "epoch": 3.87, "learning_rate": 1.4137733315226412e-05, "loss": 0.3998, "step": 64126 }, { "epoch": 3.87, "learning_rate": 1.4136224405112188e-05, "loss": 0.378, "step": 64128 }, { "epoch": 3.87, "learning_rate": 1.4134715494997963e-05, "loss": 0.2976, "step": 64130 }, { "epoch": 3.87, "learning_rate": 1.413320658488374e-05, "loss": 0.4564, "step": 64132 }, { "epoch": 3.87, "learning_rate": 1.4131697674769515e-05, "loss": 0.319, "step": 64134 }, { "epoch": 3.87, "learning_rate": 1.413018876465529e-05, "loss": 0.4184, "step": 64136 }, { "epoch": 3.87, "learning_rate": 1.4128679854541066e-05, "loss": 0.2557, "step": 64138 }, { "epoch": 3.87, "learning_rate": 1.412717094442684e-05, "loss": 0.4263, "step": 64140 }, { "epoch": 3.87, "learning_rate": 1.4125662034312617e-05, "loss": 0.2698, "step": 64142 }, { "epoch": 3.87, "learning_rate": 1.4124153124198392e-05, "loss": 0.342, "step": 64144 }, { "epoch": 3.87, "learning_rate": 1.4122644214084169e-05, "loss": 0.2641, "step": 64146 }, { "epoch": 3.87, "learning_rate": 1.4121135303969943e-05, "loss": 0.2966, "step": 64148 }, { "epoch": 3.87, "learning_rate": 1.411962639385572e-05, "loss": 0.436, "step": 64150 }, { "epoch": 3.87, "learning_rate": 1.4118117483741495e-05, "loss": 0.347, "step": 64152 }, { "epoch": 3.87, "learning_rate": 1.4116608573627268e-05, "loss": 0.3515, "step": 64154 }, { "epoch": 3.87, "learning_rate": 1.4115099663513046e-05, "loss": 0.4323, "step": 64156 }, { "epoch": 3.87, "learning_rate": 1.411359075339882e-05, "loss": 0.3138, "step": 64158 }, { "epoch": 3.87, "learning_rate": 1.4112081843284597e-05, "loss": 0.3222, "step": 64160 }, { "epoch": 3.87, "learning_rate": 1.411057293317037e-05, "loss": 0.4536, "step": 64162 }, { "epoch": 3.87, "learning_rate": 1.4109064023056149e-05, "loss": 0.2821, "step": 64164 }, { "epoch": 3.87, "learning_rate": 1.4107555112941922e-05, "loss": 0.2769, "step": 64166 }, { "epoch": 3.87, "learning_rate": 1.4106046202827697e-05, "loss": 0.3462, "step": 64168 }, { "epoch": 3.87, "learning_rate": 1.4104537292713473e-05, "loss": 0.2635, "step": 64170 }, { "epoch": 3.87, "learning_rate": 1.4103028382599248e-05, "loss": 0.271, "step": 64172 }, { "epoch": 3.87, "learning_rate": 1.4101519472485025e-05, "loss": 0.4101, "step": 64174 }, { "epoch": 3.87, "learning_rate": 1.41000105623708e-05, "loss": 0.3564, "step": 64176 }, { "epoch": 3.87, "learning_rate": 1.4098501652256576e-05, "loss": 0.5223, "step": 64178 }, { "epoch": 3.87, "learning_rate": 1.409699274214235e-05, "loss": 0.3753, "step": 64180 }, { "epoch": 3.87, "learning_rate": 1.4095483832028127e-05, "loss": 0.3449, "step": 64182 }, { "epoch": 3.87, "learning_rate": 1.4093974921913902e-05, "loss": 0.3683, "step": 64184 }, { "epoch": 3.87, "learning_rate": 1.4092466011799677e-05, "loss": 0.5439, "step": 64186 }, { "epoch": 3.87, "learning_rate": 1.4090957101685453e-05, "loss": 0.5356, "step": 64188 }, { "epoch": 3.87, "learning_rate": 1.4089448191571228e-05, "loss": 0.3612, "step": 64190 }, { "epoch": 3.87, "learning_rate": 1.4087939281457005e-05, "loss": 0.33, "step": 64192 }, { "epoch": 3.87, "learning_rate": 1.408643037134278e-05, "loss": 0.3682, "step": 64194 }, { "epoch": 3.87, "learning_rate": 1.4084921461228556e-05, "loss": 0.4225, "step": 64196 }, { "epoch": 3.87, "learning_rate": 1.408341255111433e-05, "loss": 0.3657, "step": 64198 }, { "epoch": 3.87, "learning_rate": 1.4081903641000107e-05, "loss": 0.2447, "step": 64200 }, { "epoch": 3.88, "learning_rate": 1.4080394730885882e-05, "loss": 0.4513, "step": 64202 }, { "epoch": 3.88, "learning_rate": 1.4078885820771657e-05, "loss": 0.3778, "step": 64204 }, { "epoch": 3.88, "learning_rate": 1.4077376910657434e-05, "loss": 0.419, "step": 64206 }, { "epoch": 3.88, "learning_rate": 1.4075868000543208e-05, "loss": 0.3812, "step": 64208 }, { "epoch": 3.88, "learning_rate": 1.4074359090428985e-05, "loss": 0.4276, "step": 64210 }, { "epoch": 3.88, "learning_rate": 1.407285018031476e-05, "loss": 0.2751, "step": 64212 }, { "epoch": 3.88, "learning_rate": 1.4071341270200536e-05, "loss": 0.3423, "step": 64214 }, { "epoch": 3.88, "learning_rate": 1.4069832360086311e-05, "loss": 0.2977, "step": 64216 }, { "epoch": 3.88, "learning_rate": 1.4068323449972084e-05, "loss": 0.296, "step": 64218 }, { "epoch": 3.88, "learning_rate": 1.4066814539857862e-05, "loss": 0.3977, "step": 64220 }, { "epoch": 3.88, "learning_rate": 1.4065305629743635e-05, "loss": 0.2726, "step": 64222 }, { "epoch": 3.88, "learning_rate": 1.4063796719629414e-05, "loss": 0.4356, "step": 64224 }, { "epoch": 3.88, "learning_rate": 1.4062287809515187e-05, "loss": 0.3469, "step": 64226 }, { "epoch": 3.88, "learning_rate": 1.4060778899400965e-05, "loss": 0.2404, "step": 64228 }, { "epoch": 3.88, "learning_rate": 1.4059269989286738e-05, "loss": 0.4106, "step": 64230 }, { "epoch": 3.88, "learning_rate": 1.4057761079172516e-05, "loss": 0.461, "step": 64232 }, { "epoch": 3.88, "learning_rate": 1.405625216905829e-05, "loss": 0.3483, "step": 64234 }, { "epoch": 3.88, "learning_rate": 1.4054743258944064e-05, "loss": 0.3673, "step": 64236 }, { "epoch": 3.88, "learning_rate": 1.405323434882984e-05, "loss": 0.4939, "step": 64238 }, { "epoch": 3.88, "learning_rate": 1.4051725438715616e-05, "loss": 0.3233, "step": 64240 }, { "epoch": 3.88, "learning_rate": 1.4050216528601392e-05, "loss": 0.2484, "step": 64242 }, { "epoch": 3.88, "learning_rate": 1.4048707618487167e-05, "loss": 0.5766, "step": 64244 }, { "epoch": 3.88, "learning_rate": 1.4047198708372943e-05, "loss": 0.3996, "step": 64246 }, { "epoch": 3.88, "learning_rate": 1.4045689798258718e-05, "loss": 0.4259, "step": 64248 }, { "epoch": 3.88, "learning_rate": 1.4044180888144495e-05, "loss": 0.418, "step": 64250 }, { "epoch": 3.88, "learning_rate": 1.404267197803027e-05, "loss": 0.5011, "step": 64252 }, { "epoch": 3.88, "learning_rate": 1.4041163067916044e-05, "loss": 0.235, "step": 64254 }, { "epoch": 3.88, "learning_rate": 1.4039654157801821e-05, "loss": 0.3724, "step": 64256 }, { "epoch": 3.88, "learning_rate": 1.4038145247687596e-05, "loss": 0.3727, "step": 64258 }, { "epoch": 3.88, "learning_rate": 1.4036636337573372e-05, "loss": 0.2867, "step": 64260 }, { "epoch": 3.88, "learning_rate": 1.4035127427459147e-05, "loss": 0.3455, "step": 64262 }, { "epoch": 3.88, "learning_rate": 1.4033618517344924e-05, "loss": 0.2283, "step": 64264 }, { "epoch": 3.88, "learning_rate": 1.4032109607230698e-05, "loss": 0.5782, "step": 64266 }, { "epoch": 3.88, "learning_rate": 1.4030600697116472e-05, "loss": 0.335, "step": 64268 }, { "epoch": 3.88, "learning_rate": 1.402909178700225e-05, "loss": 0.4093, "step": 64270 }, { "epoch": 3.88, "learning_rate": 1.4027582876888023e-05, "loss": 0.3453, "step": 64272 }, { "epoch": 3.88, "learning_rate": 1.4026073966773801e-05, "loss": 0.3173, "step": 64274 }, { "epoch": 3.88, "learning_rate": 1.4024565056659574e-05, "loss": 0.4711, "step": 64276 }, { "epoch": 3.88, "learning_rate": 1.4023056146545352e-05, "loss": 0.4474, "step": 64278 }, { "epoch": 3.88, "learning_rate": 1.4021547236431126e-05, "loss": 0.309, "step": 64280 }, { "epoch": 3.88, "learning_rate": 1.4020038326316904e-05, "loss": 0.3144, "step": 64282 }, { "epoch": 3.88, "learning_rate": 1.4018529416202677e-05, "loss": 0.3625, "step": 64284 }, { "epoch": 3.88, "learning_rate": 1.4017020506088452e-05, "loss": 0.2916, "step": 64286 }, { "epoch": 3.88, "learning_rate": 1.4015511595974228e-05, "loss": 0.3807, "step": 64288 }, { "epoch": 3.88, "learning_rate": 1.4014002685860003e-05, "loss": 0.3162, "step": 64290 }, { "epoch": 3.88, "learning_rate": 1.401249377574578e-05, "loss": 0.3798, "step": 64292 }, { "epoch": 3.88, "learning_rate": 1.4010984865631554e-05, "loss": 0.469, "step": 64294 }, { "epoch": 3.88, "learning_rate": 1.4009475955517331e-05, "loss": 0.3219, "step": 64296 }, { "epoch": 3.88, "learning_rate": 1.4007967045403106e-05, "loss": 0.4161, "step": 64298 }, { "epoch": 3.88, "learning_rate": 1.400645813528888e-05, "loss": 0.4319, "step": 64300 }, { "epoch": 3.88, "learning_rate": 1.4004949225174657e-05, "loss": 0.3058, "step": 64302 }, { "epoch": 3.88, "learning_rate": 1.4003440315060432e-05, "loss": 0.4184, "step": 64304 }, { "epoch": 3.88, "learning_rate": 1.4001931404946208e-05, "loss": 0.3858, "step": 64306 }, { "epoch": 3.88, "learning_rate": 1.4000422494831983e-05, "loss": 0.3309, "step": 64308 }, { "epoch": 3.88, "learning_rate": 1.399891358471776e-05, "loss": 0.192, "step": 64310 }, { "epoch": 3.88, "learning_rate": 1.3997404674603535e-05, "loss": 0.4595, "step": 64312 }, { "epoch": 3.88, "learning_rate": 1.3995895764489311e-05, "loss": 0.4179, "step": 64314 }, { "epoch": 3.88, "learning_rate": 1.3994386854375086e-05, "loss": 0.3682, "step": 64316 }, { "epoch": 3.88, "learning_rate": 1.399287794426086e-05, "loss": 0.2265, "step": 64318 }, { "epoch": 3.88, "learning_rate": 1.3991369034146637e-05, "loss": 0.3711, "step": 64320 }, { "epoch": 3.88, "learning_rate": 1.3989860124032412e-05, "loss": 0.2404, "step": 64322 }, { "epoch": 3.88, "learning_rate": 1.3988351213918189e-05, "loss": 0.278, "step": 64324 }, { "epoch": 3.88, "learning_rate": 1.3986842303803963e-05, "loss": 0.3504, "step": 64326 }, { "epoch": 3.88, "learning_rate": 1.398533339368974e-05, "loss": 0.3771, "step": 64328 }, { "epoch": 3.88, "learning_rate": 1.3983824483575515e-05, "loss": 0.3229, "step": 64330 }, { "epoch": 3.88, "learning_rate": 1.3982315573461291e-05, "loss": 0.3709, "step": 64332 }, { "epoch": 3.88, "learning_rate": 1.3980806663347066e-05, "loss": 0.3053, "step": 64334 }, { "epoch": 3.88, "learning_rate": 1.3979297753232839e-05, "loss": 0.249, "step": 64336 }, { "epoch": 3.88, "learning_rate": 1.3977788843118617e-05, "loss": 0.5153, "step": 64338 }, { "epoch": 3.88, "learning_rate": 1.397627993300439e-05, "loss": 0.3231, "step": 64340 }, { "epoch": 3.88, "learning_rate": 1.3974771022890169e-05, "loss": 0.5233, "step": 64342 }, { "epoch": 3.88, "learning_rate": 1.3973262112775942e-05, "loss": 0.3248, "step": 64344 }, { "epoch": 3.88, "learning_rate": 1.397175320266172e-05, "loss": 0.2836, "step": 64346 }, { "epoch": 3.88, "learning_rate": 1.3970244292547493e-05, "loss": 0.2879, "step": 64348 }, { "epoch": 3.88, "learning_rate": 1.3968735382433268e-05, "loss": 0.4448, "step": 64350 }, { "epoch": 3.88, "learning_rate": 1.3967226472319044e-05, "loss": 0.4901, "step": 64352 }, { "epoch": 3.88, "learning_rate": 1.396571756220482e-05, "loss": 0.3552, "step": 64354 }, { "epoch": 3.88, "learning_rate": 1.3964208652090596e-05, "loss": 0.4575, "step": 64356 }, { "epoch": 3.88, "learning_rate": 1.396269974197637e-05, "loss": 0.4162, "step": 64358 }, { "epoch": 3.88, "learning_rate": 1.3961190831862147e-05, "loss": 0.4065, "step": 64360 }, { "epoch": 3.88, "learning_rate": 1.3959681921747922e-05, "loss": 0.2389, "step": 64362 }, { "epoch": 3.88, "learning_rate": 1.3958173011633698e-05, "loss": 0.2903, "step": 64364 }, { "epoch": 3.88, "learning_rate": 1.3956664101519473e-05, "loss": 0.5208, "step": 64366 }, { "epoch": 3.89, "learning_rate": 1.3955155191405248e-05, "loss": 0.3616, "step": 64368 }, { "epoch": 3.89, "learning_rate": 1.3953646281291025e-05, "loss": 0.3821, "step": 64370 }, { "epoch": 3.89, "learning_rate": 1.39521373711768e-05, "loss": 0.3131, "step": 64372 }, { "epoch": 3.89, "learning_rate": 1.3950628461062576e-05, "loss": 0.3204, "step": 64374 }, { "epoch": 3.89, "learning_rate": 1.394911955094835e-05, "loss": 0.4317, "step": 64376 }, { "epoch": 3.89, "learning_rate": 1.3947610640834127e-05, "loss": 0.2883, "step": 64378 }, { "epoch": 3.89, "learning_rate": 1.3946101730719902e-05, "loss": 0.345, "step": 64380 }, { "epoch": 3.89, "learning_rate": 1.3944592820605675e-05, "loss": 0.4059, "step": 64382 }, { "epoch": 3.89, "learning_rate": 1.3943083910491453e-05, "loss": 0.4089, "step": 64384 }, { "epoch": 3.89, "learning_rate": 1.3941575000377227e-05, "loss": 0.3251, "step": 64386 }, { "epoch": 3.89, "learning_rate": 1.3940066090263005e-05, "loss": 0.2365, "step": 64388 }, { "epoch": 3.89, "learning_rate": 1.3938557180148778e-05, "loss": 0.2059, "step": 64390 }, { "epoch": 3.89, "learning_rate": 1.3937048270034556e-05, "loss": 0.3404, "step": 64392 }, { "epoch": 3.89, "learning_rate": 1.393553935992033e-05, "loss": 0.2185, "step": 64394 }, { "epoch": 3.89, "learning_rate": 1.3934030449806107e-05, "loss": 0.4552, "step": 64396 }, { "epoch": 3.89, "learning_rate": 1.393252153969188e-05, "loss": 0.4315, "step": 64398 }, { "epoch": 3.89, "learning_rate": 1.3931012629577655e-05, "loss": 0.2358, "step": 64400 }, { "epoch": 3.89, "learning_rate": 1.3929503719463432e-05, "loss": 0.3064, "step": 64402 }, { "epoch": 3.89, "learning_rate": 1.3927994809349207e-05, "loss": 0.3983, "step": 64404 }, { "epoch": 3.89, "learning_rate": 1.3926485899234983e-05, "loss": 0.3559, "step": 64406 }, { "epoch": 3.89, "learning_rate": 1.3924976989120758e-05, "loss": 0.1885, "step": 64408 }, { "epoch": 3.89, "learning_rate": 1.3923468079006535e-05, "loss": 0.2411, "step": 64410 }, { "epoch": 3.89, "learning_rate": 1.392195916889231e-05, "loss": 0.3877, "step": 64412 }, { "epoch": 3.89, "learning_rate": 1.3920450258778086e-05, "loss": 0.326, "step": 64414 }, { "epoch": 3.89, "learning_rate": 1.391894134866386e-05, "loss": 0.3529, "step": 64416 }, { "epoch": 3.89, "learning_rate": 1.3917432438549636e-05, "loss": 0.2087, "step": 64418 }, { "epoch": 3.89, "learning_rate": 1.3915923528435412e-05, "loss": 0.3041, "step": 64420 }, { "epoch": 3.89, "learning_rate": 1.3914414618321187e-05, "loss": 0.3608, "step": 64422 }, { "epoch": 3.89, "learning_rate": 1.3912905708206963e-05, "loss": 0.35, "step": 64424 }, { "epoch": 3.89, "learning_rate": 1.3911396798092738e-05, "loss": 0.2783, "step": 64426 }, { "epoch": 3.89, "learning_rate": 1.3909887887978515e-05, "loss": 0.3528, "step": 64428 }, { "epoch": 3.89, "learning_rate": 1.390837897786429e-05, "loss": 0.3327, "step": 64430 }, { "epoch": 3.89, "learning_rate": 1.3906870067750064e-05, "loss": 0.4746, "step": 64432 }, { "epoch": 3.89, "learning_rate": 1.3905361157635841e-05, "loss": 0.3702, "step": 64434 }, { "epoch": 3.89, "learning_rate": 1.3903852247521616e-05, "loss": 0.4261, "step": 64436 }, { "epoch": 3.89, "learning_rate": 1.3902343337407392e-05, "loss": 0.2084, "step": 64438 }, { "epoch": 3.89, "learning_rate": 1.3900834427293167e-05, "loss": 0.1757, "step": 64440 }, { "epoch": 3.89, "learning_rate": 1.3899325517178944e-05, "loss": 0.2823, "step": 64442 }, { "epoch": 3.89, "learning_rate": 1.3897816607064718e-05, "loss": 0.3345, "step": 64444 }, { "epoch": 3.89, "learning_rate": 1.3896307696950495e-05, "loss": 0.3634, "step": 64446 }, { "epoch": 3.89, "learning_rate": 1.389479878683627e-05, "loss": 0.4286, "step": 64448 }, { "epoch": 3.89, "learning_rate": 1.3893289876722043e-05, "loss": 0.2557, "step": 64450 }, { "epoch": 3.89, "learning_rate": 1.3891780966607821e-05, "loss": 0.3369, "step": 64452 }, { "epoch": 3.89, "learning_rate": 1.3890272056493594e-05, "loss": 0.3469, "step": 64454 }, { "epoch": 3.89, "learning_rate": 1.3888763146379372e-05, "loss": 0.3128, "step": 64456 }, { "epoch": 3.89, "learning_rate": 1.3887254236265145e-05, "loss": 0.3348, "step": 64458 }, { "epoch": 3.89, "learning_rate": 1.3885745326150924e-05, "loss": 0.3872, "step": 64460 }, { "epoch": 3.89, "learning_rate": 1.3884236416036697e-05, "loss": 0.3612, "step": 64462 }, { "epoch": 3.89, "learning_rate": 1.3882727505922472e-05, "loss": 0.508, "step": 64464 }, { "epoch": 3.89, "learning_rate": 1.3881218595808248e-05, "loss": 0.4001, "step": 64466 }, { "epoch": 3.89, "learning_rate": 1.3879709685694023e-05, "loss": 0.3777, "step": 64468 }, { "epoch": 3.89, "learning_rate": 1.38782007755798e-05, "loss": 0.4107, "step": 64470 }, { "epoch": 3.89, "learning_rate": 1.3876691865465574e-05, "loss": 0.4244, "step": 64472 }, { "epoch": 3.89, "learning_rate": 1.387518295535135e-05, "loss": 0.4413, "step": 64474 }, { "epoch": 3.89, "learning_rate": 1.3873674045237126e-05, "loss": 0.4556, "step": 64476 }, { "epoch": 3.89, "learning_rate": 1.3872165135122902e-05, "loss": 0.3721, "step": 64478 }, { "epoch": 3.89, "learning_rate": 1.3870656225008677e-05, "loss": 0.5194, "step": 64480 }, { "epoch": 3.89, "learning_rate": 1.3869147314894452e-05, "loss": 0.4265, "step": 64482 }, { "epoch": 3.89, "learning_rate": 1.3867638404780228e-05, "loss": 0.2364, "step": 64484 }, { "epoch": 3.89, "learning_rate": 1.3866129494666003e-05, "loss": 0.5194, "step": 64486 }, { "epoch": 3.89, "learning_rate": 1.386462058455178e-05, "loss": 0.2535, "step": 64488 }, { "epoch": 3.89, "learning_rate": 1.3863111674437554e-05, "loss": 0.3629, "step": 64490 }, { "epoch": 3.89, "learning_rate": 1.3861602764323331e-05, "loss": 0.3304, "step": 64492 }, { "epoch": 3.89, "learning_rate": 1.3860093854209106e-05, "loss": 0.4435, "step": 64494 }, { "epoch": 3.89, "learning_rate": 1.3858584944094882e-05, "loss": 0.2587, "step": 64496 }, { "epoch": 3.89, "learning_rate": 1.3857076033980657e-05, "loss": 0.2609, "step": 64498 }, { "epoch": 3.89, "learning_rate": 1.385556712386643e-05, "loss": 0.3598, "step": 64500 }, { "epoch": 3.89, "learning_rate": 1.3854058213752208e-05, "loss": 0.3806, "step": 64502 }, { "epoch": 3.89, "learning_rate": 1.3852549303637982e-05, "loss": 0.3424, "step": 64504 }, { "epoch": 3.89, "learning_rate": 1.385104039352376e-05, "loss": 0.273, "step": 64506 }, { "epoch": 3.89, "learning_rate": 1.3849531483409533e-05, "loss": 0.2895, "step": 64508 }, { "epoch": 3.89, "learning_rate": 1.3848022573295311e-05, "loss": 0.2395, "step": 64510 }, { "epoch": 3.89, "learning_rate": 1.3846513663181084e-05, "loss": 0.4381, "step": 64512 }, { "epoch": 3.89, "learning_rate": 1.3845004753066859e-05, "loss": 0.3962, "step": 64514 }, { "epoch": 3.89, "learning_rate": 1.3843495842952636e-05, "loss": 0.3404, "step": 64516 }, { "epoch": 3.89, "learning_rate": 1.384198693283841e-05, "loss": 0.3082, "step": 64518 }, { "epoch": 3.89, "learning_rate": 1.3840478022724187e-05, "loss": 0.39, "step": 64520 }, { "epoch": 3.89, "learning_rate": 1.3838969112609962e-05, "loss": 0.3006, "step": 64522 }, { "epoch": 3.89, "learning_rate": 1.3837460202495738e-05, "loss": 0.398, "step": 64524 }, { "epoch": 3.89, "learning_rate": 1.3835951292381513e-05, "loss": 0.3574, "step": 64526 }, { "epoch": 3.89, "learning_rate": 1.383444238226729e-05, "loss": 0.4977, "step": 64528 }, { "epoch": 3.89, "learning_rate": 1.3832933472153064e-05, "loss": 0.3556, "step": 64530 }, { "epoch": 3.89, "learning_rate": 1.383142456203884e-05, "loss": 0.2242, "step": 64532 }, { "epoch": 3.9, "learning_rate": 1.3829915651924616e-05, "loss": 0.3057, "step": 64534 }, { "epoch": 3.9, "learning_rate": 1.382840674181039e-05, "loss": 0.4794, "step": 64536 }, { "epoch": 3.9, "learning_rate": 1.3826897831696167e-05, "loss": 0.3452, "step": 64538 }, { "epoch": 3.9, "learning_rate": 1.3825388921581942e-05, "loss": 0.3156, "step": 64540 }, { "epoch": 3.9, "learning_rate": 1.3823880011467718e-05, "loss": 0.5161, "step": 64542 }, { "epoch": 3.9, "learning_rate": 1.3822371101353493e-05, "loss": 0.2849, "step": 64544 }, { "epoch": 3.9, "learning_rate": 1.3820862191239268e-05, "loss": 0.3726, "step": 64546 }, { "epoch": 3.9, "learning_rate": 1.3819353281125045e-05, "loss": 0.4289, "step": 64548 }, { "epoch": 3.9, "learning_rate": 1.381784437101082e-05, "loss": 0.3568, "step": 64550 }, { "epoch": 3.9, "learning_rate": 1.3816335460896596e-05, "loss": 0.3093, "step": 64552 }, { "epoch": 3.9, "learning_rate": 1.381482655078237e-05, "loss": 0.3326, "step": 64554 }, { "epoch": 3.9, "learning_rate": 1.3813317640668147e-05, "loss": 0.4327, "step": 64556 }, { "epoch": 3.9, "learning_rate": 1.3811808730553922e-05, "loss": 0.3695, "step": 64558 }, { "epoch": 3.9, "learning_rate": 1.3810299820439699e-05, "loss": 0.428, "step": 64560 }, { "epoch": 3.9, "learning_rate": 1.3808790910325473e-05, "loss": 0.3035, "step": 64562 }, { "epoch": 3.9, "learning_rate": 1.3807282000211246e-05, "loss": 0.2694, "step": 64564 }, { "epoch": 3.9, "learning_rate": 1.3805773090097025e-05, "loss": 0.3035, "step": 64566 }, { "epoch": 3.9, "learning_rate": 1.3804264179982798e-05, "loss": 0.3515, "step": 64568 }, { "epoch": 3.9, "learning_rate": 1.3802755269868576e-05, "loss": 0.289, "step": 64570 }, { "epoch": 3.9, "learning_rate": 1.3801246359754349e-05, "loss": 0.2702, "step": 64572 }, { "epoch": 3.9, "learning_rate": 1.3799737449640127e-05, "loss": 0.2546, "step": 64574 }, { "epoch": 3.9, "learning_rate": 1.37982285395259e-05, "loss": 0.4297, "step": 64576 }, { "epoch": 3.9, "learning_rate": 1.3796719629411679e-05, "loss": 0.3753, "step": 64578 }, { "epoch": 3.9, "learning_rate": 1.3795210719297452e-05, "loss": 0.4462, "step": 64580 }, { "epoch": 3.9, "learning_rate": 1.3793701809183227e-05, "loss": 0.2734, "step": 64582 }, { "epoch": 3.9, "learning_rate": 1.3792192899069003e-05, "loss": 0.3703, "step": 64584 }, { "epoch": 3.9, "learning_rate": 1.3790683988954778e-05, "loss": 0.43, "step": 64586 }, { "epoch": 3.9, "learning_rate": 1.3789175078840554e-05, "loss": 0.3486, "step": 64588 }, { "epoch": 3.9, "learning_rate": 1.378766616872633e-05, "loss": 0.2307, "step": 64590 }, { "epoch": 3.9, "learning_rate": 1.3786157258612106e-05, "loss": 0.4426, "step": 64592 }, { "epoch": 3.9, "learning_rate": 1.378464834849788e-05, "loss": 0.4611, "step": 64594 }, { "epoch": 3.9, "learning_rate": 1.3783139438383655e-05, "loss": 0.3166, "step": 64596 }, { "epoch": 3.9, "learning_rate": 1.3781630528269432e-05, "loss": 0.4846, "step": 64598 }, { "epoch": 3.9, "learning_rate": 1.3780121618155207e-05, "loss": 0.2703, "step": 64600 }, { "epoch": 3.9, "learning_rate": 1.3778612708040983e-05, "loss": 0.4587, "step": 64602 }, { "epoch": 3.9, "learning_rate": 1.3777103797926758e-05, "loss": 0.3514, "step": 64604 }, { "epoch": 3.9, "learning_rate": 1.3775594887812535e-05, "loss": 0.4332, "step": 64606 }, { "epoch": 3.9, "learning_rate": 1.377408597769831e-05, "loss": 0.243, "step": 64608 }, { "epoch": 3.9, "learning_rate": 1.3772577067584086e-05, "loss": 0.3234, "step": 64610 }, { "epoch": 3.9, "learning_rate": 1.377106815746986e-05, "loss": 0.4287, "step": 64612 }, { "epoch": 3.9, "learning_rate": 1.3769559247355634e-05, "loss": 0.326, "step": 64614 }, { "epoch": 3.9, "learning_rate": 1.3768050337241412e-05, "loss": 0.2682, "step": 64616 }, { "epoch": 3.9, "learning_rate": 1.3766541427127185e-05, "loss": 0.3152, "step": 64618 }, { "epoch": 3.9, "learning_rate": 1.3765032517012963e-05, "loss": 0.2736, "step": 64620 }, { "epoch": 3.9, "learning_rate": 1.3763523606898737e-05, "loss": 0.3033, "step": 64622 }, { "epoch": 3.9, "learning_rate": 1.3762014696784515e-05, "loss": 0.2661, "step": 64624 }, { "epoch": 3.9, "learning_rate": 1.3760505786670288e-05, "loss": 0.3923, "step": 64626 }, { "epoch": 3.9, "learning_rate": 1.3758996876556063e-05, "loss": 0.3628, "step": 64628 }, { "epoch": 3.9, "learning_rate": 1.375748796644184e-05, "loss": 0.1977, "step": 64630 }, { "epoch": 3.9, "learning_rate": 1.3755979056327614e-05, "loss": 0.2781, "step": 64632 }, { "epoch": 3.9, "learning_rate": 1.375447014621339e-05, "loss": 0.2952, "step": 64634 }, { "epoch": 3.9, "learning_rate": 1.3752961236099165e-05, "loss": 0.3894, "step": 64636 }, { "epoch": 3.9, "learning_rate": 1.3751452325984942e-05, "loss": 0.4842, "step": 64638 }, { "epoch": 3.9, "learning_rate": 1.3749943415870717e-05, "loss": 0.3474, "step": 64640 }, { "epoch": 3.9, "learning_rate": 1.3748434505756493e-05, "loss": 0.3263, "step": 64642 }, { "epoch": 3.9, "learning_rate": 1.3746925595642268e-05, "loss": 0.3147, "step": 64644 }, { "epoch": 3.9, "learning_rate": 1.3745416685528043e-05, "loss": 0.1984, "step": 64646 }, { "epoch": 3.9, "learning_rate": 1.374390777541382e-05, "loss": 0.392, "step": 64648 }, { "epoch": 3.9, "learning_rate": 1.3742398865299594e-05, "loss": 0.3087, "step": 64650 }, { "epoch": 3.9, "learning_rate": 1.374088995518537e-05, "loss": 0.5521, "step": 64652 }, { "epoch": 3.9, "learning_rate": 1.3739381045071146e-05, "loss": 0.3937, "step": 64654 }, { "epoch": 3.9, "learning_rate": 1.3737872134956922e-05, "loss": 0.5558, "step": 64656 }, { "epoch": 3.9, "learning_rate": 1.3736363224842697e-05, "loss": 0.3445, "step": 64658 }, { "epoch": 3.9, "learning_rate": 1.3734854314728473e-05, "loss": 0.3234, "step": 64660 }, { "epoch": 3.9, "learning_rate": 1.3733345404614248e-05, "loss": 0.4076, "step": 64662 }, { "epoch": 3.9, "learning_rate": 1.3731836494500023e-05, "loss": 0.3867, "step": 64664 }, { "epoch": 3.9, "learning_rate": 1.37303275843858e-05, "loss": 0.3287, "step": 64666 }, { "epoch": 3.9, "learning_rate": 1.3728818674271574e-05, "loss": 0.3635, "step": 64668 }, { "epoch": 3.9, "learning_rate": 1.3727309764157351e-05, "loss": 0.2993, "step": 64670 }, { "epoch": 3.9, "learning_rate": 1.3725800854043126e-05, "loss": 0.4725, "step": 64672 }, { "epoch": 3.9, "learning_rate": 1.3724291943928902e-05, "loss": 0.3079, "step": 64674 }, { "epoch": 3.9, "learning_rate": 1.3722783033814677e-05, "loss": 0.3755, "step": 64676 }, { "epoch": 3.9, "learning_rate": 1.372127412370045e-05, "loss": 0.3291, "step": 64678 }, { "epoch": 3.9, "learning_rate": 1.3719765213586228e-05, "loss": 0.2965, "step": 64680 }, { "epoch": 3.9, "learning_rate": 1.3718256303472001e-05, "loss": 0.3345, "step": 64682 }, { "epoch": 3.9, "learning_rate": 1.371674739335778e-05, "loss": 0.2259, "step": 64684 }, { "epoch": 3.9, "learning_rate": 1.3715238483243553e-05, "loss": 0.3233, "step": 64686 }, { "epoch": 3.9, "learning_rate": 1.3713729573129331e-05, "loss": 0.2586, "step": 64688 }, { "epoch": 3.9, "learning_rate": 1.3712220663015104e-05, "loss": 0.4045, "step": 64690 }, { "epoch": 3.9, "learning_rate": 1.3710711752900882e-05, "loss": 0.3004, "step": 64692 }, { "epoch": 3.9, "learning_rate": 1.3709202842786655e-05, "loss": 0.3343, "step": 64694 }, { "epoch": 3.9, "learning_rate": 1.370769393267243e-05, "loss": 0.1781, "step": 64696 }, { "epoch": 3.9, "learning_rate": 1.3706185022558207e-05, "loss": 0.3443, "step": 64698 }, { "epoch": 3.91, "learning_rate": 1.3704676112443982e-05, "loss": 0.2993, "step": 64700 }, { "epoch": 3.91, "learning_rate": 1.3703167202329758e-05, "loss": 0.3096, "step": 64702 }, { "epoch": 3.91, "learning_rate": 1.3701658292215533e-05, "loss": 0.3197, "step": 64704 }, { "epoch": 3.91, "learning_rate": 1.370014938210131e-05, "loss": 0.3275, "step": 64706 }, { "epoch": 3.91, "learning_rate": 1.3698640471987084e-05, "loss": 0.3175, "step": 64708 }, { "epoch": 3.91, "learning_rate": 1.3697131561872859e-05, "loss": 0.4543, "step": 64710 }, { "epoch": 3.91, "learning_rate": 1.3695622651758636e-05, "loss": 0.4818, "step": 64712 }, { "epoch": 3.91, "learning_rate": 1.369411374164441e-05, "loss": 0.3397, "step": 64714 }, { "epoch": 3.91, "learning_rate": 1.3692604831530187e-05, "loss": 0.3064, "step": 64716 }, { "epoch": 3.91, "learning_rate": 1.3691095921415962e-05, "loss": 0.3642, "step": 64718 }, { "epoch": 3.91, "learning_rate": 1.3689587011301738e-05, "loss": 0.409, "step": 64720 }, { "epoch": 3.91, "learning_rate": 1.3688078101187513e-05, "loss": 0.3915, "step": 64722 }, { "epoch": 3.91, "learning_rate": 1.368656919107329e-05, "loss": 0.2865, "step": 64724 }, { "epoch": 3.91, "learning_rate": 1.3685060280959064e-05, "loss": 0.256, "step": 64726 }, { "epoch": 3.91, "learning_rate": 1.3683551370844838e-05, "loss": 0.3178, "step": 64728 }, { "epoch": 3.91, "learning_rate": 1.3682042460730616e-05, "loss": 0.4187, "step": 64730 }, { "epoch": 3.91, "learning_rate": 1.3680533550616389e-05, "loss": 0.2911, "step": 64732 }, { "epoch": 3.91, "learning_rate": 1.3679024640502167e-05, "loss": 0.3113, "step": 64734 }, { "epoch": 3.91, "learning_rate": 1.367751573038794e-05, "loss": 0.2798, "step": 64736 }, { "epoch": 3.91, "learning_rate": 1.3676006820273718e-05, "loss": 0.607, "step": 64738 }, { "epoch": 3.91, "learning_rate": 1.3674497910159492e-05, "loss": 0.3427, "step": 64740 }, { "epoch": 3.91, "learning_rate": 1.367298900004527e-05, "loss": 0.3518, "step": 64742 }, { "epoch": 3.91, "learning_rate": 1.3671480089931043e-05, "loss": 0.3161, "step": 64744 }, { "epoch": 3.91, "learning_rate": 1.3669971179816818e-05, "loss": 0.2534, "step": 64746 }, { "epoch": 3.91, "learning_rate": 1.3668462269702594e-05, "loss": 0.2991, "step": 64748 }, { "epoch": 3.91, "learning_rate": 1.3666953359588369e-05, "loss": 0.3051, "step": 64750 }, { "epoch": 3.91, "learning_rate": 1.3665444449474146e-05, "loss": 0.284, "step": 64752 }, { "epoch": 3.91, "learning_rate": 1.366393553935992e-05, "loss": 0.2619, "step": 64754 }, { "epoch": 3.91, "learning_rate": 1.3662426629245697e-05, "loss": 0.3917, "step": 64756 }, { "epoch": 3.91, "learning_rate": 1.3660917719131472e-05, "loss": 0.3371, "step": 64758 }, { "epoch": 3.91, "learning_rate": 1.3659408809017247e-05, "loss": 0.4082, "step": 64760 }, { "epoch": 3.91, "learning_rate": 1.3657899898903023e-05, "loss": 0.472, "step": 64762 }, { "epoch": 3.91, "learning_rate": 1.3656390988788798e-05, "loss": 0.3974, "step": 64764 }, { "epoch": 3.91, "learning_rate": 1.3654882078674574e-05, "loss": 0.3731, "step": 64766 }, { "epoch": 3.91, "learning_rate": 1.365337316856035e-05, "loss": 0.2341, "step": 64768 }, { "epoch": 3.91, "learning_rate": 1.3651864258446126e-05, "loss": 0.2673, "step": 64770 }, { "epoch": 3.91, "learning_rate": 1.36503553483319e-05, "loss": 0.291, "step": 64772 }, { "epoch": 3.91, "learning_rate": 1.3648846438217677e-05, "loss": 0.4038, "step": 64774 }, { "epoch": 3.91, "learning_rate": 1.3647337528103452e-05, "loss": 0.2123, "step": 64776 }, { "epoch": 3.91, "learning_rate": 1.3645828617989225e-05, "loss": 0.1963, "step": 64778 }, { "epoch": 3.91, "learning_rate": 1.3644319707875003e-05, "loss": 0.5296, "step": 64780 }, { "epoch": 3.91, "learning_rate": 1.3642810797760776e-05, "loss": 0.4483, "step": 64782 }, { "epoch": 3.91, "learning_rate": 1.3641301887646555e-05, "loss": 0.3019, "step": 64784 }, { "epoch": 3.91, "learning_rate": 1.363979297753233e-05, "loss": 0.3506, "step": 64786 }, { "epoch": 3.91, "learning_rate": 1.3638284067418106e-05, "loss": 0.2206, "step": 64788 }, { "epoch": 3.91, "learning_rate": 1.363677515730388e-05, "loss": 0.5837, "step": 64790 }, { "epoch": 3.91, "learning_rate": 1.3635266247189657e-05, "loss": 0.2239, "step": 64792 }, { "epoch": 3.91, "learning_rate": 1.3633757337075432e-05, "loss": 0.3806, "step": 64794 }, { "epoch": 3.91, "learning_rate": 1.3632248426961205e-05, "loss": 0.5302, "step": 64796 }, { "epoch": 3.91, "learning_rate": 1.3630739516846983e-05, "loss": 0.6057, "step": 64798 }, { "epoch": 3.91, "learning_rate": 1.3629230606732757e-05, "loss": 0.5393, "step": 64800 }, { "epoch": 3.91, "learning_rate": 1.3627721696618535e-05, "loss": 0.5381, "step": 64802 }, { "epoch": 3.91, "learning_rate": 1.3626212786504308e-05, "loss": 0.2635, "step": 64804 }, { "epoch": 3.91, "learning_rate": 1.3624703876390086e-05, "loss": 0.3609, "step": 64806 }, { "epoch": 3.91, "learning_rate": 1.362319496627586e-05, "loss": 0.4233, "step": 64808 }, { "epoch": 3.91, "learning_rate": 1.3621686056161634e-05, "loss": 0.5188, "step": 64810 }, { "epoch": 3.91, "learning_rate": 1.362017714604741e-05, "loss": 0.2882, "step": 64812 }, { "epoch": 3.91, "learning_rate": 1.3618668235933185e-05, "loss": 0.2659, "step": 64814 }, { "epoch": 3.91, "learning_rate": 1.3617159325818962e-05, "loss": 0.2893, "step": 64816 }, { "epoch": 3.91, "learning_rate": 1.3615650415704737e-05, "loss": 0.4561, "step": 64818 }, { "epoch": 3.91, "learning_rate": 1.3614141505590513e-05, "loss": 0.4095, "step": 64820 }, { "epoch": 3.91, "learning_rate": 1.3612632595476288e-05, "loss": 0.488, "step": 64822 }, { "epoch": 3.91, "learning_rate": 1.3611123685362065e-05, "loss": 0.3638, "step": 64824 }, { "epoch": 3.91, "learning_rate": 1.360961477524784e-05, "loss": 0.3283, "step": 64826 }, { "epoch": 3.91, "learning_rate": 1.3608105865133614e-05, "loss": 0.411, "step": 64828 }, { "epoch": 3.91, "learning_rate": 1.360659695501939e-05, "loss": 0.3808, "step": 64830 }, { "epoch": 3.91, "learning_rate": 1.3605088044905165e-05, "loss": 0.3219, "step": 64832 }, { "epoch": 3.91, "learning_rate": 1.3603579134790942e-05, "loss": 0.3737, "step": 64834 }, { "epoch": 3.91, "learning_rate": 1.3602070224676717e-05, "loss": 0.3767, "step": 64836 }, { "epoch": 3.91, "learning_rate": 1.3600561314562493e-05, "loss": 0.3333, "step": 64838 }, { "epoch": 3.91, "learning_rate": 1.3599052404448268e-05, "loss": 0.3899, "step": 64840 }, { "epoch": 3.91, "learning_rate": 1.3597543494334041e-05, "loss": 0.2563, "step": 64842 }, { "epoch": 3.91, "learning_rate": 1.359603458421982e-05, "loss": 0.279, "step": 64844 }, { "epoch": 3.91, "learning_rate": 1.3594525674105593e-05, "loss": 0.2897, "step": 64846 }, { "epoch": 3.91, "learning_rate": 1.359301676399137e-05, "loss": 0.5006, "step": 64848 }, { "epoch": 3.91, "learning_rate": 1.3591507853877144e-05, "loss": 0.4524, "step": 64850 }, { "epoch": 3.91, "learning_rate": 1.3589998943762922e-05, "loss": 0.3904, "step": 64852 }, { "epoch": 3.91, "learning_rate": 1.3588490033648695e-05, "loss": 0.3157, "step": 64854 }, { "epoch": 3.91, "learning_rate": 1.3586981123534473e-05, "loss": 0.4167, "step": 64856 }, { "epoch": 3.91, "learning_rate": 1.3585472213420247e-05, "loss": 0.3711, "step": 64858 }, { "epoch": 3.91, "learning_rate": 1.3583963303306021e-05, "loss": 0.2102, "step": 64860 }, { "epoch": 3.91, "learning_rate": 1.3582454393191798e-05, "loss": 0.4824, "step": 64862 }, { "epoch": 3.92, "learning_rate": 1.3580945483077573e-05, "loss": 0.5728, "step": 64864 }, { "epoch": 3.92, "learning_rate": 1.357943657296335e-05, "loss": 0.3264, "step": 64866 }, { "epoch": 3.92, "learning_rate": 1.3577927662849124e-05, "loss": 0.2132, "step": 64868 }, { "epoch": 3.92, "learning_rate": 1.35764187527349e-05, "loss": 0.3792, "step": 64870 }, { "epoch": 3.92, "learning_rate": 1.3574909842620675e-05, "loss": 0.2498, "step": 64872 }, { "epoch": 3.92, "learning_rate": 1.3573400932506452e-05, "loss": 0.3168, "step": 64874 }, { "epoch": 3.92, "learning_rate": 1.3571892022392227e-05, "loss": 0.3685, "step": 64876 }, { "epoch": 3.92, "learning_rate": 1.3570383112278002e-05, "loss": 0.482, "step": 64878 }, { "epoch": 3.92, "learning_rate": 1.3568874202163778e-05, "loss": 0.3001, "step": 64880 }, { "epoch": 3.92, "learning_rate": 1.3567365292049553e-05, "loss": 0.3577, "step": 64882 }, { "epoch": 3.92, "learning_rate": 1.356585638193533e-05, "loss": 0.3277, "step": 64884 }, { "epoch": 3.92, "learning_rate": 1.3564347471821104e-05, "loss": 0.4925, "step": 64886 }, { "epoch": 3.92, "learning_rate": 1.356283856170688e-05, "loss": 0.3935, "step": 64888 }, { "epoch": 3.92, "learning_rate": 1.3561329651592656e-05, "loss": 0.3827, "step": 64890 }, { "epoch": 3.92, "learning_rate": 1.3559820741478429e-05, "loss": 0.31, "step": 64892 }, { "epoch": 3.92, "learning_rate": 1.3558311831364207e-05, "loss": 0.3775, "step": 64894 }, { "epoch": 3.92, "learning_rate": 1.355680292124998e-05, "loss": 0.3076, "step": 64896 }, { "epoch": 3.92, "learning_rate": 1.3555294011135758e-05, "loss": 0.4869, "step": 64898 }, { "epoch": 3.92, "learning_rate": 1.3553785101021531e-05, "loss": 0.3101, "step": 64900 }, { "epoch": 3.92, "learning_rate": 1.355227619090731e-05, "loss": 0.4437, "step": 64902 }, { "epoch": 3.92, "learning_rate": 1.3550767280793084e-05, "loss": 0.3521, "step": 64904 }, { "epoch": 3.92, "learning_rate": 1.3549258370678861e-05, "loss": 0.3848, "step": 64906 }, { "epoch": 3.92, "learning_rate": 1.3547749460564636e-05, "loss": 0.3135, "step": 64908 }, { "epoch": 3.92, "learning_rate": 1.3546240550450409e-05, "loss": 0.3546, "step": 64910 }, { "epoch": 3.92, "learning_rate": 1.3544731640336187e-05, "loss": 0.308, "step": 64912 }, { "epoch": 3.92, "learning_rate": 1.354322273022196e-05, "loss": 0.5354, "step": 64914 }, { "epoch": 3.92, "learning_rate": 1.3541713820107738e-05, "loss": 0.3477, "step": 64916 }, { "epoch": 3.92, "learning_rate": 1.3540204909993512e-05, "loss": 0.4627, "step": 64918 }, { "epoch": 3.92, "learning_rate": 1.353869599987929e-05, "loss": 0.3059, "step": 64920 }, { "epoch": 3.92, "learning_rate": 1.3537187089765063e-05, "loss": 0.4358, "step": 64922 }, { "epoch": 3.92, "learning_rate": 1.3535678179650838e-05, "loss": 0.3212, "step": 64924 }, { "epoch": 3.92, "learning_rate": 1.3534169269536614e-05, "loss": 0.3817, "step": 64926 }, { "epoch": 3.92, "learning_rate": 1.3532660359422389e-05, "loss": 0.416, "step": 64928 }, { "epoch": 3.92, "learning_rate": 1.3531151449308166e-05, "loss": 0.2177, "step": 64930 }, { "epoch": 3.92, "learning_rate": 1.352964253919394e-05, "loss": 0.3423, "step": 64932 }, { "epoch": 3.92, "learning_rate": 1.3528133629079717e-05, "loss": 0.3722, "step": 64934 }, { "epoch": 3.92, "learning_rate": 1.3526624718965492e-05, "loss": 0.3337, "step": 64936 }, { "epoch": 3.92, "learning_rate": 1.3525115808851268e-05, "loss": 0.3737, "step": 64938 }, { "epoch": 3.92, "learning_rate": 1.3523606898737043e-05, "loss": 0.3117, "step": 64940 }, { "epoch": 3.92, "learning_rate": 1.3522097988622818e-05, "loss": 0.3338, "step": 64942 }, { "epoch": 3.92, "learning_rate": 1.3520589078508594e-05, "loss": 0.2914, "step": 64944 }, { "epoch": 3.92, "learning_rate": 1.3519080168394369e-05, "loss": 0.3247, "step": 64946 }, { "epoch": 3.92, "learning_rate": 1.3517571258280146e-05, "loss": 0.271, "step": 64948 }, { "epoch": 3.92, "learning_rate": 1.351606234816592e-05, "loss": 0.3787, "step": 64950 }, { "epoch": 3.92, "learning_rate": 1.3514553438051697e-05, "loss": 0.4153, "step": 64952 }, { "epoch": 3.92, "learning_rate": 1.3513044527937472e-05, "loss": 0.2732, "step": 64954 }, { "epoch": 3.92, "learning_rate": 1.3511535617823248e-05, "loss": 0.3779, "step": 64956 }, { "epoch": 3.92, "learning_rate": 1.3510026707709023e-05, "loss": 0.3854, "step": 64958 }, { "epoch": 3.92, "learning_rate": 1.3508517797594796e-05, "loss": 0.4799, "step": 64960 }, { "epoch": 3.92, "learning_rate": 1.3507008887480574e-05, "loss": 0.5294, "step": 64962 }, { "epoch": 3.92, "learning_rate": 1.3505499977366348e-05, "loss": 0.2244, "step": 64964 }, { "epoch": 3.92, "learning_rate": 1.3503991067252126e-05, "loss": 0.325, "step": 64966 }, { "epoch": 3.92, "learning_rate": 1.3502482157137899e-05, "loss": 0.3208, "step": 64968 }, { "epoch": 3.92, "learning_rate": 1.3500973247023677e-05, "loss": 0.3321, "step": 64970 }, { "epoch": 3.92, "learning_rate": 1.349946433690945e-05, "loss": 0.2561, "step": 64972 }, { "epoch": 3.92, "learning_rate": 1.3497955426795225e-05, "loss": 0.4048, "step": 64974 }, { "epoch": 3.92, "learning_rate": 1.3496446516681002e-05, "loss": 0.345, "step": 64976 }, { "epoch": 3.92, "learning_rate": 1.3494937606566776e-05, "loss": 0.2377, "step": 64978 }, { "epoch": 3.92, "learning_rate": 1.3493428696452553e-05, "loss": 0.4128, "step": 64980 }, { "epoch": 3.92, "learning_rate": 1.3491919786338328e-05, "loss": 0.2779, "step": 64982 }, { "epoch": 3.92, "learning_rate": 1.3490410876224104e-05, "loss": 0.3516, "step": 64984 }, { "epoch": 3.92, "learning_rate": 1.3488901966109879e-05, "loss": 0.454, "step": 64986 }, { "epoch": 3.92, "learning_rate": 1.3487393055995656e-05, "loss": 0.3012, "step": 64988 }, { "epoch": 3.92, "learning_rate": 1.348588414588143e-05, "loss": 0.3125, "step": 64990 }, { "epoch": 3.92, "learning_rate": 1.3484375235767205e-05, "loss": 0.246, "step": 64992 }, { "epoch": 3.92, "learning_rate": 1.3482866325652982e-05, "loss": 0.5139, "step": 64994 }, { "epoch": 3.92, "learning_rate": 1.3481357415538757e-05, "loss": 0.4071, "step": 64996 }, { "epoch": 3.92, "learning_rate": 1.3479848505424533e-05, "loss": 0.3502, "step": 64998 }, { "epoch": 3.92, "learning_rate": 1.3478339595310308e-05, "loss": 0.3552, "step": 65000 }, { "epoch": 3.92, "learning_rate": 1.3476830685196084e-05, "loss": 0.4401, "step": 65002 }, { "epoch": 3.92, "learning_rate": 1.347532177508186e-05, "loss": 0.3153, "step": 65004 }, { "epoch": 3.92, "learning_rate": 1.3473812864967632e-05, "loss": 0.2381, "step": 65006 }, { "epoch": 3.92, "learning_rate": 1.347230395485341e-05, "loss": 0.3342, "step": 65008 }, { "epoch": 3.92, "learning_rate": 1.3470795044739184e-05, "loss": 0.5184, "step": 65010 }, { "epoch": 3.92, "learning_rate": 1.3469286134624962e-05, "loss": 0.3692, "step": 65012 }, { "epoch": 3.92, "learning_rate": 1.3467777224510735e-05, "loss": 0.3014, "step": 65014 }, { "epoch": 3.92, "learning_rate": 1.3466268314396513e-05, "loss": 0.3386, "step": 65016 }, { "epoch": 3.92, "learning_rate": 1.3464759404282286e-05, "loss": 0.4506, "step": 65018 }, { "epoch": 3.92, "learning_rate": 1.3463250494168065e-05, "loss": 0.3891, "step": 65020 }, { "epoch": 3.92, "learning_rate": 1.346174158405384e-05, "loss": 0.3316, "step": 65022 }, { "epoch": 3.92, "learning_rate": 1.3460232673939613e-05, "loss": 0.3537, "step": 65024 }, { "epoch": 3.92, "learning_rate": 1.345872376382539e-05, "loss": 0.4256, "step": 65026 }, { "epoch": 3.92, "learning_rate": 1.3457214853711164e-05, "loss": 0.4393, "step": 65028 }, { "epoch": 3.93, "learning_rate": 1.3455705943596942e-05, "loss": 0.4403, "step": 65030 }, { "epoch": 3.93, "learning_rate": 1.3454197033482715e-05, "loss": 0.346, "step": 65032 }, { "epoch": 3.93, "learning_rate": 1.3452688123368493e-05, "loss": 0.3171, "step": 65034 }, { "epoch": 3.93, "learning_rate": 1.3451179213254267e-05, "loss": 0.3389, "step": 65036 }, { "epoch": 3.93, "learning_rate": 1.3449670303140045e-05, "loss": 0.2155, "step": 65038 }, { "epoch": 3.93, "learning_rate": 1.3448161393025818e-05, "loss": 0.2995, "step": 65040 }, { "epoch": 3.93, "learning_rate": 1.3446652482911593e-05, "loss": 0.4739, "step": 65042 }, { "epoch": 3.93, "learning_rate": 1.344514357279737e-05, "loss": 0.2727, "step": 65044 }, { "epoch": 3.93, "learning_rate": 1.3443634662683144e-05, "loss": 0.274, "step": 65046 }, { "epoch": 3.93, "learning_rate": 1.344212575256892e-05, "loss": 0.3852, "step": 65048 }, { "epoch": 3.93, "learning_rate": 1.3440616842454695e-05, "loss": 0.4928, "step": 65050 }, { "epoch": 3.93, "learning_rate": 1.3439107932340472e-05, "loss": 0.2781, "step": 65052 }, { "epoch": 3.93, "learning_rate": 1.3437599022226247e-05, "loss": 0.3363, "step": 65054 }, { "epoch": 3.93, "learning_rate": 1.3436090112112021e-05, "loss": 0.395, "step": 65056 }, { "epoch": 3.93, "learning_rate": 1.3434581201997798e-05, "loss": 0.4912, "step": 65058 }, { "epoch": 3.93, "learning_rate": 1.3433072291883573e-05, "loss": 0.4212, "step": 65060 }, { "epoch": 3.93, "learning_rate": 1.343156338176935e-05, "loss": 0.2461, "step": 65062 }, { "epoch": 3.93, "learning_rate": 1.3430054471655124e-05, "loss": 0.2306, "step": 65064 }, { "epoch": 3.93, "learning_rate": 1.34285455615409e-05, "loss": 0.2761, "step": 65066 }, { "epoch": 3.93, "learning_rate": 1.3427036651426675e-05, "loss": 0.4772, "step": 65068 }, { "epoch": 3.93, "learning_rate": 1.3425527741312452e-05, "loss": 0.5139, "step": 65070 }, { "epoch": 3.93, "learning_rate": 1.3424018831198227e-05, "loss": 0.3479, "step": 65072 }, { "epoch": 3.93, "learning_rate": 1.3422509921084e-05, "loss": 0.3913, "step": 65074 }, { "epoch": 3.93, "learning_rate": 1.3421001010969778e-05, "loss": 0.3936, "step": 65076 }, { "epoch": 3.93, "learning_rate": 1.3419492100855551e-05, "loss": 0.4473, "step": 65078 }, { "epoch": 3.93, "learning_rate": 1.341798319074133e-05, "loss": 0.353, "step": 65080 }, { "epoch": 3.93, "learning_rate": 1.3416474280627103e-05, "loss": 0.4707, "step": 65082 }, { "epoch": 3.93, "learning_rate": 1.341496537051288e-05, "loss": 0.4439, "step": 65084 }, { "epoch": 3.93, "learning_rate": 1.3413456460398654e-05, "loss": 0.3317, "step": 65086 }, { "epoch": 3.93, "learning_rate": 1.3411947550284429e-05, "loss": 0.2651, "step": 65088 }, { "epoch": 3.93, "learning_rate": 1.3410438640170205e-05, "loss": 0.3168, "step": 65090 }, { "epoch": 3.93, "learning_rate": 1.340892973005598e-05, "loss": 0.4787, "step": 65092 }, { "epoch": 3.93, "learning_rate": 1.3407420819941757e-05, "loss": 0.3322, "step": 65094 }, { "epoch": 3.93, "learning_rate": 1.3405911909827531e-05, "loss": 0.4032, "step": 65096 }, { "epoch": 3.93, "learning_rate": 1.3404402999713308e-05, "loss": 0.3674, "step": 65098 }, { "epoch": 3.93, "learning_rate": 1.3402894089599083e-05, "loss": 0.3265, "step": 65100 }, { "epoch": 3.93, "learning_rate": 1.340138517948486e-05, "loss": 0.3509, "step": 65102 }, { "epoch": 3.93, "learning_rate": 1.3399876269370634e-05, "loss": 0.4406, "step": 65104 }, { "epoch": 3.93, "learning_rate": 1.3398367359256409e-05, "loss": 0.2307, "step": 65106 }, { "epoch": 3.93, "learning_rate": 1.3396858449142185e-05, "loss": 0.3018, "step": 65108 }, { "epoch": 3.93, "learning_rate": 1.339534953902796e-05, "loss": 0.3644, "step": 65110 }, { "epoch": 3.93, "learning_rate": 1.3393840628913737e-05, "loss": 0.4851, "step": 65112 }, { "epoch": 3.93, "learning_rate": 1.3392331718799512e-05, "loss": 0.3146, "step": 65114 }, { "epoch": 3.93, "learning_rate": 1.3390822808685288e-05, "loss": 0.4861, "step": 65116 }, { "epoch": 3.93, "learning_rate": 1.3389313898571063e-05, "loss": 0.39, "step": 65118 }, { "epoch": 3.93, "learning_rate": 1.338780498845684e-05, "loss": 0.3836, "step": 65120 }, { "epoch": 3.93, "learning_rate": 1.3386296078342614e-05, "loss": 0.3499, "step": 65122 }, { "epoch": 3.93, "learning_rate": 1.3384787168228387e-05, "loss": 0.2276, "step": 65124 }, { "epoch": 3.93, "learning_rate": 1.3383278258114166e-05, "loss": 0.4399, "step": 65126 }, { "epoch": 3.93, "learning_rate": 1.3381769347999939e-05, "loss": 0.3433, "step": 65128 }, { "epoch": 3.93, "learning_rate": 1.3380260437885717e-05, "loss": 0.3087, "step": 65130 }, { "epoch": 3.93, "learning_rate": 1.337875152777149e-05, "loss": 0.3908, "step": 65132 }, { "epoch": 3.93, "learning_rate": 1.3377242617657268e-05, "loss": 0.3642, "step": 65134 }, { "epoch": 3.93, "learning_rate": 1.3375733707543041e-05, "loss": 0.277, "step": 65136 }, { "epoch": 3.93, "learning_rate": 1.3374224797428816e-05, "loss": 0.2824, "step": 65138 }, { "epoch": 3.93, "learning_rate": 1.3372715887314594e-05, "loss": 0.2877, "step": 65140 }, { "epoch": 3.93, "learning_rate": 1.3371206977200368e-05, "loss": 0.3445, "step": 65142 }, { "epoch": 3.93, "learning_rate": 1.3369698067086146e-05, "loss": 0.3974, "step": 65144 }, { "epoch": 3.93, "learning_rate": 1.3368189156971919e-05, "loss": 0.2951, "step": 65146 }, { "epoch": 3.93, "learning_rate": 1.3366680246857697e-05, "loss": 0.4015, "step": 65148 }, { "epoch": 3.93, "learning_rate": 1.336517133674347e-05, "loss": 0.3534, "step": 65150 }, { "epoch": 3.93, "learning_rate": 1.3363662426629248e-05, "loss": 0.2944, "step": 65152 }, { "epoch": 3.93, "learning_rate": 1.3362153516515022e-05, "loss": 0.3686, "step": 65154 }, { "epoch": 3.93, "learning_rate": 1.3360644606400796e-05, "loss": 0.1946, "step": 65156 }, { "epoch": 3.93, "learning_rate": 1.3359135696286573e-05, "loss": 0.2525, "step": 65158 }, { "epoch": 3.93, "learning_rate": 1.3357626786172348e-05, "loss": 0.2406, "step": 65160 }, { "epoch": 3.93, "learning_rate": 1.3356117876058124e-05, "loss": 0.277, "step": 65162 }, { "epoch": 3.93, "learning_rate": 1.3354608965943899e-05, "loss": 0.2954, "step": 65164 }, { "epoch": 3.93, "learning_rate": 1.3353100055829676e-05, "loss": 0.4383, "step": 65166 }, { "epoch": 3.93, "learning_rate": 1.335159114571545e-05, "loss": 0.2423, "step": 65168 }, { "epoch": 3.93, "learning_rate": 1.3350082235601225e-05, "loss": 0.4958, "step": 65170 }, { "epoch": 3.93, "learning_rate": 1.3348573325487002e-05, "loss": 0.585, "step": 65172 }, { "epoch": 3.93, "learning_rate": 1.3347064415372777e-05, "loss": 0.3246, "step": 65174 }, { "epoch": 3.93, "learning_rate": 1.3345555505258553e-05, "loss": 0.3368, "step": 65176 }, { "epoch": 3.93, "learning_rate": 1.3344046595144328e-05, "loss": 0.3715, "step": 65178 }, { "epoch": 3.93, "learning_rate": 1.3342537685030104e-05, "loss": 0.4128, "step": 65180 }, { "epoch": 3.93, "learning_rate": 1.334102877491588e-05, "loss": 0.4119, "step": 65182 }, { "epoch": 3.93, "learning_rate": 1.3339519864801656e-05, "loss": 0.459, "step": 65184 }, { "epoch": 3.93, "learning_rate": 1.333801095468743e-05, "loss": 0.3609, "step": 65186 }, { "epoch": 3.93, "learning_rate": 1.3336502044573204e-05, "loss": 0.2603, "step": 65188 }, { "epoch": 3.93, "learning_rate": 1.3334993134458982e-05, "loss": 0.3287, "step": 65190 }, { "epoch": 3.93, "learning_rate": 1.3333484224344755e-05, "loss": 0.3311, "step": 65192 }, { "epoch": 3.93, "learning_rate": 1.3331975314230533e-05, "loss": 0.3757, "step": 65194 }, { "epoch": 3.94, "learning_rate": 1.3330466404116306e-05, "loss": 0.4555, "step": 65196 }, { "epoch": 3.94, "learning_rate": 1.3328957494002085e-05, "loss": 0.3519, "step": 65198 }, { "epoch": 3.94, "learning_rate": 1.3327448583887858e-05, "loss": 0.2619, "step": 65200 }, { "epoch": 3.94, "learning_rate": 1.3325939673773636e-05, "loss": 0.3383, "step": 65202 }, { "epoch": 3.94, "learning_rate": 1.3324430763659409e-05, "loss": 0.2457, "step": 65204 }, { "epoch": 3.94, "learning_rate": 1.3322921853545184e-05, "loss": 0.4854, "step": 65206 }, { "epoch": 3.94, "learning_rate": 1.332141294343096e-05, "loss": 0.452, "step": 65208 }, { "epoch": 3.94, "learning_rate": 1.3319904033316735e-05, "loss": 0.2606, "step": 65210 }, { "epoch": 3.94, "learning_rate": 1.3318395123202512e-05, "loss": 0.3493, "step": 65212 }, { "epoch": 3.94, "learning_rate": 1.3316886213088286e-05, "loss": 0.52, "step": 65214 }, { "epoch": 3.94, "learning_rate": 1.3315377302974063e-05, "loss": 0.4373, "step": 65216 }, { "epoch": 3.94, "learning_rate": 1.3313868392859838e-05, "loss": 0.3253, "step": 65218 }, { "epoch": 3.94, "learning_rate": 1.3312359482745613e-05, "loss": 0.3455, "step": 65220 }, { "epoch": 3.94, "learning_rate": 1.3310850572631389e-05, "loss": 0.4413, "step": 65222 }, { "epoch": 3.94, "learning_rate": 1.3309341662517164e-05, "loss": 0.3277, "step": 65224 }, { "epoch": 3.94, "learning_rate": 1.330783275240294e-05, "loss": 0.4052, "step": 65226 }, { "epoch": 3.94, "learning_rate": 1.3306323842288715e-05, "loss": 0.2845, "step": 65228 }, { "epoch": 3.94, "learning_rate": 1.3304814932174492e-05, "loss": 0.432, "step": 65230 }, { "epoch": 3.94, "learning_rate": 1.3303306022060267e-05, "loss": 0.4091, "step": 65232 }, { "epoch": 3.94, "learning_rate": 1.3301797111946043e-05, "loss": 0.3779, "step": 65234 }, { "epoch": 3.94, "learning_rate": 1.3300288201831818e-05, "loss": 0.3515, "step": 65236 }, { "epoch": 3.94, "learning_rate": 1.3298779291717591e-05, "loss": 0.3592, "step": 65238 }, { "epoch": 3.94, "learning_rate": 1.329727038160337e-05, "loss": 0.3054, "step": 65240 }, { "epoch": 3.94, "learning_rate": 1.3295761471489142e-05, "loss": 0.267, "step": 65242 }, { "epoch": 3.94, "learning_rate": 1.329425256137492e-05, "loss": 0.4087, "step": 65244 }, { "epoch": 3.94, "learning_rate": 1.3292743651260694e-05, "loss": 0.1982, "step": 65246 }, { "epoch": 3.94, "learning_rate": 1.3291234741146472e-05, "loss": 0.2736, "step": 65248 }, { "epoch": 3.94, "learning_rate": 1.3289725831032245e-05, "loss": 0.5297, "step": 65250 }, { "epoch": 3.94, "learning_rate": 1.328821692091802e-05, "loss": 0.2, "step": 65252 }, { "epoch": 3.94, "learning_rate": 1.3286708010803796e-05, "loss": 0.3962, "step": 65254 }, { "epoch": 3.94, "learning_rate": 1.3285199100689571e-05, "loss": 0.3777, "step": 65256 }, { "epoch": 3.94, "learning_rate": 1.328369019057535e-05, "loss": 0.4637, "step": 65258 }, { "epoch": 3.94, "learning_rate": 1.3282181280461123e-05, "loss": 0.3582, "step": 65260 }, { "epoch": 3.94, "learning_rate": 1.32806723703469e-05, "loss": 0.3085, "step": 65262 }, { "epoch": 3.94, "learning_rate": 1.3279163460232674e-05, "loss": 0.3081, "step": 65264 }, { "epoch": 3.94, "learning_rate": 1.3277654550118452e-05, "loss": 0.4429, "step": 65266 }, { "epoch": 3.94, "learning_rate": 1.3276145640004225e-05, "loss": 0.3231, "step": 65268 }, { "epoch": 3.94, "learning_rate": 1.327463672989e-05, "loss": 0.5387, "step": 65270 }, { "epoch": 3.94, "learning_rate": 1.3273127819775777e-05, "loss": 0.4731, "step": 65272 }, { "epoch": 3.94, "learning_rate": 1.3271618909661551e-05, "loss": 0.3812, "step": 65274 }, { "epoch": 3.94, "learning_rate": 1.3270109999547328e-05, "loss": 0.3521, "step": 65276 }, { "epoch": 3.94, "learning_rate": 1.3268601089433103e-05, "loss": 0.3796, "step": 65278 }, { "epoch": 3.94, "learning_rate": 1.326709217931888e-05, "loss": 0.5078, "step": 65280 }, { "epoch": 3.94, "learning_rate": 1.3265583269204654e-05, "loss": 0.3603, "step": 65282 }, { "epoch": 3.94, "learning_rate": 1.326407435909043e-05, "loss": 0.4577, "step": 65284 }, { "epoch": 3.94, "learning_rate": 1.3262565448976205e-05, "loss": 0.3119, "step": 65286 }, { "epoch": 3.94, "learning_rate": 1.326105653886198e-05, "loss": 0.4337, "step": 65288 }, { "epoch": 3.94, "learning_rate": 1.3259547628747757e-05, "loss": 0.302, "step": 65290 }, { "epoch": 3.94, "learning_rate": 1.3258038718633532e-05, "loss": 0.4264, "step": 65292 }, { "epoch": 3.94, "learning_rate": 1.3256529808519308e-05, "loss": 0.2725, "step": 65294 }, { "epoch": 3.94, "learning_rate": 1.3255020898405083e-05, "loss": 0.3532, "step": 65296 }, { "epoch": 3.94, "learning_rate": 1.325351198829086e-05, "loss": 0.395, "step": 65298 }, { "epoch": 3.94, "learning_rate": 1.3252003078176634e-05, "loss": 0.3084, "step": 65300 }, { "epoch": 3.94, "learning_rate": 1.3250494168062407e-05, "loss": 0.4867, "step": 65302 }, { "epoch": 3.94, "learning_rate": 1.3248985257948186e-05, "loss": 0.2759, "step": 65304 }, { "epoch": 3.94, "learning_rate": 1.3247476347833959e-05, "loss": 0.3604, "step": 65306 }, { "epoch": 3.94, "learning_rate": 1.3245967437719737e-05, "loss": 0.3563, "step": 65308 }, { "epoch": 3.94, "learning_rate": 1.324445852760551e-05, "loss": 0.4028, "step": 65310 }, { "epoch": 3.94, "learning_rate": 1.3242949617491288e-05, "loss": 0.411, "step": 65312 }, { "epoch": 3.94, "learning_rate": 1.3241440707377061e-05, "loss": 0.2631, "step": 65314 }, { "epoch": 3.94, "learning_rate": 1.323993179726284e-05, "loss": 0.2772, "step": 65316 }, { "epoch": 3.94, "learning_rate": 1.3238422887148613e-05, "loss": 0.3058, "step": 65318 }, { "epoch": 3.94, "learning_rate": 1.3236913977034387e-05, "loss": 0.3971, "step": 65320 }, { "epoch": 3.94, "learning_rate": 1.3235405066920164e-05, "loss": 0.2723, "step": 65322 }, { "epoch": 3.94, "learning_rate": 1.3233896156805939e-05, "loss": 0.3576, "step": 65324 }, { "epoch": 3.94, "learning_rate": 1.3232387246691715e-05, "loss": 0.4589, "step": 65326 }, { "epoch": 3.94, "learning_rate": 1.323087833657749e-05, "loss": 0.3765, "step": 65328 }, { "epoch": 3.94, "learning_rate": 1.3229369426463267e-05, "loss": 0.2382, "step": 65330 }, { "epoch": 3.94, "learning_rate": 1.3227860516349041e-05, "loss": 0.3625, "step": 65332 }, { "epoch": 3.94, "learning_rate": 1.3226351606234816e-05, "loss": 0.2329, "step": 65334 }, { "epoch": 3.94, "learning_rate": 1.3224842696120593e-05, "loss": 0.3317, "step": 65336 }, { "epoch": 3.94, "learning_rate": 1.3223333786006368e-05, "loss": 0.4449, "step": 65338 }, { "epoch": 3.94, "learning_rate": 1.3221824875892144e-05, "loss": 0.4335, "step": 65340 }, { "epoch": 3.94, "learning_rate": 1.3220315965777919e-05, "loss": 0.3112, "step": 65342 }, { "epoch": 3.94, "learning_rate": 1.3218807055663695e-05, "loss": 0.3149, "step": 65344 }, { "epoch": 3.94, "learning_rate": 1.321729814554947e-05, "loss": 0.4361, "step": 65346 }, { "epoch": 3.94, "learning_rate": 1.3215789235435247e-05, "loss": 0.4199, "step": 65348 }, { "epoch": 3.94, "learning_rate": 1.3214280325321022e-05, "loss": 0.3566, "step": 65350 }, { "epoch": 3.94, "learning_rate": 1.3212771415206795e-05, "loss": 0.3198, "step": 65352 }, { "epoch": 3.94, "learning_rate": 1.3211262505092573e-05, "loss": 0.2492, "step": 65354 }, { "epoch": 3.94, "learning_rate": 1.3209753594978346e-05, "loss": 0.4057, "step": 65356 }, { "epoch": 3.94, "learning_rate": 1.3208244684864124e-05, "loss": 0.3865, "step": 65358 }, { "epoch": 3.94, "learning_rate": 1.3206735774749897e-05, "loss": 0.3525, "step": 65360 }, { "epoch": 3.95, "learning_rate": 1.3205226864635676e-05, "loss": 0.2702, "step": 65362 }, { "epoch": 3.95, "learning_rate": 1.3203717954521449e-05, "loss": 0.3389, "step": 65364 }, { "epoch": 3.95, "learning_rate": 1.3202209044407227e-05, "loss": 0.346, "step": 65366 }, { "epoch": 3.95, "learning_rate": 1.3200700134293e-05, "loss": 0.3497, "step": 65368 }, { "epoch": 3.95, "learning_rate": 1.3199191224178775e-05, "loss": 0.4248, "step": 65370 }, { "epoch": 3.95, "learning_rate": 1.3197682314064551e-05, "loss": 0.2229, "step": 65372 }, { "epoch": 3.95, "learning_rate": 1.3196173403950326e-05, "loss": 0.3543, "step": 65374 }, { "epoch": 3.95, "learning_rate": 1.3194664493836104e-05, "loss": 0.2162, "step": 65376 }, { "epoch": 3.95, "learning_rate": 1.3193155583721878e-05, "loss": 0.246, "step": 65378 }, { "epoch": 3.95, "learning_rate": 1.3191646673607656e-05, "loss": 0.4295, "step": 65380 }, { "epoch": 3.95, "learning_rate": 1.3190137763493429e-05, "loss": 0.2656, "step": 65382 }, { "epoch": 3.95, "learning_rate": 1.3188628853379204e-05, "loss": 0.4041, "step": 65384 }, { "epoch": 3.95, "learning_rate": 1.318711994326498e-05, "loss": 0.4332, "step": 65386 }, { "epoch": 3.95, "learning_rate": 1.3185611033150755e-05, "loss": 0.5911, "step": 65388 }, { "epoch": 3.95, "learning_rate": 1.3184102123036532e-05, "loss": 0.2938, "step": 65390 }, { "epoch": 3.95, "learning_rate": 1.3182593212922306e-05, "loss": 0.3677, "step": 65392 }, { "epoch": 3.95, "learning_rate": 1.3181084302808083e-05, "loss": 0.3439, "step": 65394 }, { "epoch": 3.95, "learning_rate": 1.3179575392693858e-05, "loss": 0.3342, "step": 65396 }, { "epoch": 3.95, "learning_rate": 1.3178066482579634e-05, "loss": 0.386, "step": 65398 }, { "epoch": 3.95, "learning_rate": 1.3176557572465409e-05, "loss": 0.3076, "step": 65400 }, { "epoch": 3.95, "learning_rate": 1.3175048662351184e-05, "loss": 0.348, "step": 65402 }, { "epoch": 3.95, "learning_rate": 1.317353975223696e-05, "loss": 0.314, "step": 65404 }, { "epoch": 3.95, "learning_rate": 1.3172030842122735e-05, "loss": 0.1645, "step": 65406 }, { "epoch": 3.95, "learning_rate": 1.3170521932008512e-05, "loss": 0.4409, "step": 65408 }, { "epoch": 3.95, "learning_rate": 1.3169013021894287e-05, "loss": 0.2567, "step": 65410 }, { "epoch": 3.95, "learning_rate": 1.3167504111780063e-05, "loss": 0.2671, "step": 65412 }, { "epoch": 3.95, "learning_rate": 1.3165995201665838e-05, "loss": 0.2803, "step": 65414 }, { "epoch": 3.95, "learning_rate": 1.3164486291551614e-05, "loss": 0.2568, "step": 65416 }, { "epoch": 3.95, "learning_rate": 1.316297738143739e-05, "loss": 0.3048, "step": 65418 }, { "epoch": 3.95, "learning_rate": 1.3161468471323162e-05, "loss": 0.4709, "step": 65420 }, { "epoch": 3.95, "learning_rate": 1.315995956120894e-05, "loss": 0.2458, "step": 65422 }, { "epoch": 3.95, "learning_rate": 1.3158450651094714e-05, "loss": 0.3399, "step": 65424 }, { "epoch": 3.95, "learning_rate": 1.3156941740980492e-05, "loss": 0.5884, "step": 65426 }, { "epoch": 3.95, "learning_rate": 1.3155432830866265e-05, "loss": 0.3076, "step": 65428 }, { "epoch": 3.95, "learning_rate": 1.3153923920752043e-05, "loss": 0.2916, "step": 65430 }, { "epoch": 3.95, "learning_rate": 1.3152415010637816e-05, "loss": 0.2489, "step": 65432 }, { "epoch": 3.95, "learning_rate": 1.3150906100523591e-05, "loss": 0.215, "step": 65434 }, { "epoch": 3.95, "learning_rate": 1.3149397190409368e-05, "loss": 0.3631, "step": 65436 }, { "epoch": 3.95, "learning_rate": 1.3147888280295142e-05, "loss": 0.268, "step": 65438 }, { "epoch": 3.95, "learning_rate": 1.3146379370180919e-05, "loss": 0.1934, "step": 65440 }, { "epoch": 3.95, "learning_rate": 1.3144870460066694e-05, "loss": 0.2727, "step": 65442 }, { "epoch": 3.95, "learning_rate": 1.314336154995247e-05, "loss": 0.351, "step": 65444 }, { "epoch": 3.95, "learning_rate": 1.3141852639838245e-05, "loss": 0.3486, "step": 65446 }, { "epoch": 3.95, "learning_rate": 1.3140343729724022e-05, "loss": 0.397, "step": 65448 }, { "epoch": 3.95, "learning_rate": 1.3138834819609796e-05, "loss": 0.3138, "step": 65450 }, { "epoch": 3.95, "learning_rate": 1.3137325909495571e-05, "loss": 0.5191, "step": 65452 }, { "epoch": 3.95, "learning_rate": 1.3135816999381348e-05, "loss": 0.3095, "step": 65454 }, { "epoch": 3.95, "learning_rate": 1.3134308089267123e-05, "loss": 0.3713, "step": 65456 }, { "epoch": 3.95, "learning_rate": 1.3132799179152899e-05, "loss": 0.2346, "step": 65458 }, { "epoch": 3.95, "learning_rate": 1.3131290269038674e-05, "loss": 0.4224, "step": 65460 }, { "epoch": 3.95, "learning_rate": 1.312978135892445e-05, "loss": 0.346, "step": 65462 }, { "epoch": 3.95, "learning_rate": 1.3128272448810225e-05, "loss": 0.3075, "step": 65464 }, { "epoch": 3.95, "learning_rate": 1.3126763538695998e-05, "loss": 0.531, "step": 65466 }, { "epoch": 3.95, "learning_rate": 1.3125254628581777e-05, "loss": 0.517, "step": 65468 }, { "epoch": 3.95, "learning_rate": 1.312374571846755e-05, "loss": 0.2276, "step": 65470 }, { "epoch": 3.95, "learning_rate": 1.3122236808353328e-05, "loss": 0.3632, "step": 65472 }, { "epoch": 3.95, "learning_rate": 1.3120727898239101e-05, "loss": 0.2549, "step": 65474 }, { "epoch": 3.95, "learning_rate": 1.311921898812488e-05, "loss": 0.4306, "step": 65476 }, { "epoch": 3.95, "learning_rate": 1.3117710078010652e-05, "loss": 0.3196, "step": 65478 }, { "epoch": 3.95, "learning_rate": 1.311620116789643e-05, "loss": 0.2505, "step": 65480 }, { "epoch": 3.95, "learning_rate": 1.3114692257782204e-05, "loss": 0.4764, "step": 65482 }, { "epoch": 3.95, "learning_rate": 1.3113183347667979e-05, "loss": 0.3601, "step": 65484 }, { "epoch": 3.95, "learning_rate": 1.3111674437553755e-05, "loss": 0.324, "step": 65486 }, { "epoch": 3.95, "learning_rate": 1.311016552743953e-05, "loss": 0.4541, "step": 65488 }, { "epoch": 3.95, "learning_rate": 1.3108656617325306e-05, "loss": 0.3674, "step": 65490 }, { "epoch": 3.95, "learning_rate": 1.3107147707211081e-05, "loss": 0.3055, "step": 65492 }, { "epoch": 3.95, "learning_rate": 1.310563879709686e-05, "loss": 0.3436, "step": 65494 }, { "epoch": 3.95, "learning_rate": 1.3104129886982633e-05, "loss": 0.3157, "step": 65496 }, { "epoch": 3.95, "learning_rate": 1.310262097686841e-05, "loss": 0.6329, "step": 65498 }, { "epoch": 3.95, "learning_rate": 1.3101112066754184e-05, "loss": 0.3528, "step": 65500 }, { "epoch": 3.95, "learning_rate": 1.3099603156639959e-05, "loss": 0.3019, "step": 65502 }, { "epoch": 3.95, "learning_rate": 1.3098094246525735e-05, "loss": 0.3957, "step": 65504 }, { "epoch": 3.95, "learning_rate": 1.309658533641151e-05, "loss": 0.2812, "step": 65506 }, { "epoch": 3.95, "learning_rate": 1.3095076426297287e-05, "loss": 0.4271, "step": 65508 }, { "epoch": 3.95, "learning_rate": 1.3093567516183061e-05, "loss": 0.3491, "step": 65510 }, { "epoch": 3.95, "learning_rate": 1.3092058606068838e-05, "loss": 0.3732, "step": 65512 }, { "epoch": 3.95, "learning_rate": 1.3090549695954613e-05, "loss": 0.3168, "step": 65514 }, { "epoch": 3.95, "learning_rate": 1.3089040785840388e-05, "loss": 0.4774, "step": 65516 }, { "epoch": 3.95, "learning_rate": 1.3087531875726164e-05, "loss": 0.3913, "step": 65518 }, { "epoch": 3.95, "learning_rate": 1.3086022965611939e-05, "loss": 0.3252, "step": 65520 }, { "epoch": 3.95, "learning_rate": 1.3084514055497715e-05, "loss": 0.3791, "step": 65522 }, { "epoch": 3.95, "learning_rate": 1.308300514538349e-05, "loss": 0.3149, "step": 65524 }, { "epoch": 3.95, "learning_rate": 1.3081496235269267e-05, "loss": 0.3243, "step": 65526 }, { "epoch": 3.96, "learning_rate": 1.3079987325155042e-05, "loss": 0.4311, "step": 65528 }, { "epoch": 3.96, "learning_rate": 1.3078478415040818e-05, "loss": 0.3837, "step": 65530 }, { "epoch": 3.96, "learning_rate": 1.3076969504926593e-05, "loss": 0.4091, "step": 65532 }, { "epoch": 3.96, "learning_rate": 1.3075460594812366e-05, "loss": 0.37, "step": 65534 }, { "epoch": 3.96, "learning_rate": 1.3073951684698144e-05, "loss": 0.4239, "step": 65536 }, { "epoch": 3.96, "learning_rate": 1.3072442774583917e-05, "loss": 0.2599, "step": 65538 }, { "epoch": 3.96, "learning_rate": 1.3070933864469696e-05, "loss": 0.3361, "step": 65540 }, { "epoch": 3.96, "learning_rate": 1.3069424954355469e-05, "loss": 0.4337, "step": 65542 }, { "epoch": 3.96, "learning_rate": 1.3067916044241247e-05, "loss": 0.3706, "step": 65544 }, { "epoch": 3.96, "learning_rate": 1.306640713412702e-05, "loss": 0.4446, "step": 65546 }, { "epoch": 3.96, "learning_rate": 1.3064898224012795e-05, "loss": 0.4478, "step": 65548 }, { "epoch": 3.96, "learning_rate": 1.3063389313898571e-05, "loss": 0.3961, "step": 65550 }, { "epoch": 3.96, "learning_rate": 1.3061880403784346e-05, "loss": 0.5183, "step": 65552 }, { "epoch": 3.96, "learning_rate": 1.3060371493670123e-05, "loss": 0.2902, "step": 65554 }, { "epoch": 3.96, "learning_rate": 1.3058862583555897e-05, "loss": 0.4693, "step": 65556 }, { "epoch": 3.96, "learning_rate": 1.3057353673441674e-05, "loss": 0.212, "step": 65558 }, { "epoch": 3.96, "learning_rate": 1.3055844763327449e-05, "loss": 0.4135, "step": 65560 }, { "epoch": 3.96, "learning_rate": 1.3054335853213225e-05, "loss": 0.3061, "step": 65562 }, { "epoch": 3.96, "learning_rate": 1.3052826943099e-05, "loss": 0.2237, "step": 65564 }, { "epoch": 3.96, "learning_rate": 1.3051318032984775e-05, "loss": 0.3687, "step": 65566 }, { "epoch": 3.96, "learning_rate": 1.3049809122870551e-05, "loss": 0.2355, "step": 65568 }, { "epoch": 3.96, "learning_rate": 1.3048300212756326e-05, "loss": 0.3337, "step": 65570 }, { "epoch": 3.96, "learning_rate": 1.3046791302642103e-05, "loss": 0.2211, "step": 65572 }, { "epoch": 3.96, "learning_rate": 1.3045282392527878e-05, "loss": 0.5918, "step": 65574 }, { "epoch": 3.96, "learning_rate": 1.3043773482413654e-05, "loss": 0.2603, "step": 65576 }, { "epoch": 3.96, "learning_rate": 1.3042264572299429e-05, "loss": 0.1957, "step": 65578 }, { "epoch": 3.96, "learning_rate": 1.3040755662185205e-05, "loss": 0.2863, "step": 65580 }, { "epoch": 3.96, "learning_rate": 1.303924675207098e-05, "loss": 0.2873, "step": 65582 }, { "epoch": 3.96, "learning_rate": 1.3037737841956753e-05, "loss": 0.2703, "step": 65584 }, { "epoch": 3.96, "learning_rate": 1.3036228931842532e-05, "loss": 0.2494, "step": 65586 }, { "epoch": 3.96, "learning_rate": 1.3034720021728305e-05, "loss": 0.5108, "step": 65588 }, { "epoch": 3.96, "learning_rate": 1.3033211111614083e-05, "loss": 0.3718, "step": 65590 }, { "epoch": 3.96, "learning_rate": 1.3031702201499856e-05, "loss": 0.2566, "step": 65592 }, { "epoch": 3.96, "learning_rate": 1.3030193291385634e-05, "loss": 0.4463, "step": 65594 }, { "epoch": 3.96, "learning_rate": 1.3028684381271407e-05, "loss": 0.3204, "step": 65596 }, { "epoch": 3.96, "learning_rate": 1.3027175471157182e-05, "loss": 0.3659, "step": 65598 }, { "epoch": 3.96, "learning_rate": 1.3025666561042959e-05, "loss": 0.3407, "step": 65600 }, { "epoch": 3.96, "learning_rate": 1.3024157650928734e-05, "loss": 0.3101, "step": 65602 }, { "epoch": 3.96, "learning_rate": 1.302264874081451e-05, "loss": 0.3991, "step": 65604 }, { "epoch": 3.96, "learning_rate": 1.3021139830700285e-05, "loss": 0.4329, "step": 65606 }, { "epoch": 3.96, "learning_rate": 1.3019630920586061e-05, "loss": 0.2889, "step": 65608 }, { "epoch": 3.96, "learning_rate": 1.3018122010471836e-05, "loss": 0.2026, "step": 65610 }, { "epoch": 3.96, "learning_rate": 1.3016613100357614e-05, "loss": 0.4783, "step": 65612 }, { "epoch": 3.96, "learning_rate": 1.3015104190243388e-05, "loss": 0.3581, "step": 65614 }, { "epoch": 3.96, "learning_rate": 1.3013595280129162e-05, "loss": 0.352, "step": 65616 }, { "epoch": 3.96, "learning_rate": 1.3012086370014939e-05, "loss": 0.3042, "step": 65618 }, { "epoch": 3.96, "learning_rate": 1.3010577459900714e-05, "loss": 0.3075, "step": 65620 }, { "epoch": 3.96, "learning_rate": 1.300906854978649e-05, "loss": 0.2519, "step": 65622 }, { "epoch": 3.96, "learning_rate": 1.3007559639672265e-05, "loss": 0.2085, "step": 65624 }, { "epoch": 3.96, "learning_rate": 1.3006050729558042e-05, "loss": 0.4041, "step": 65626 }, { "epoch": 3.96, "learning_rate": 1.3004541819443816e-05, "loss": 0.4328, "step": 65628 }, { "epoch": 3.96, "learning_rate": 1.3003032909329591e-05, "loss": 0.3315, "step": 65630 }, { "epoch": 3.96, "learning_rate": 1.3001523999215368e-05, "loss": 0.2659, "step": 65632 }, { "epoch": 3.96, "learning_rate": 1.3000015089101143e-05, "loss": 0.2769, "step": 65634 }, { "epoch": 3.96, "learning_rate": 1.2998506178986919e-05, "loss": 0.2876, "step": 65636 }, { "epoch": 3.96, "learning_rate": 1.2996997268872694e-05, "loss": 0.2628, "step": 65638 }, { "epoch": 3.96, "learning_rate": 1.299548835875847e-05, "loss": 0.3529, "step": 65640 }, { "epoch": 3.96, "learning_rate": 1.2993979448644245e-05, "loss": 0.3982, "step": 65642 }, { "epoch": 3.96, "learning_rate": 1.2992470538530022e-05, "loss": 0.3717, "step": 65644 }, { "epoch": 3.96, "learning_rate": 1.2990961628415797e-05, "loss": 0.3738, "step": 65646 }, { "epoch": 3.96, "learning_rate": 1.298945271830157e-05, "loss": 0.3628, "step": 65648 }, { "epoch": 3.96, "learning_rate": 1.2987943808187348e-05, "loss": 0.2536, "step": 65650 }, { "epoch": 3.96, "learning_rate": 1.2986434898073121e-05, "loss": 0.2256, "step": 65652 }, { "epoch": 3.96, "learning_rate": 1.29849259879589e-05, "loss": 0.4467, "step": 65654 }, { "epoch": 3.96, "learning_rate": 1.2983417077844672e-05, "loss": 0.4029, "step": 65656 }, { "epoch": 3.96, "learning_rate": 1.298190816773045e-05, "loss": 0.3566, "step": 65658 }, { "epoch": 3.96, "learning_rate": 1.2980399257616224e-05, "loss": 0.2997, "step": 65660 }, { "epoch": 3.96, "learning_rate": 1.2978890347502002e-05, "loss": 0.2219, "step": 65662 }, { "epoch": 3.96, "learning_rate": 1.2977381437387775e-05, "loss": 0.3183, "step": 65664 }, { "epoch": 3.96, "learning_rate": 1.297587252727355e-05, "loss": 0.4213, "step": 65666 }, { "epoch": 3.96, "learning_rate": 1.2974363617159326e-05, "loss": 0.2492, "step": 65668 }, { "epoch": 3.96, "learning_rate": 1.2972854707045101e-05, "loss": 0.4091, "step": 65670 }, { "epoch": 3.96, "learning_rate": 1.2971345796930878e-05, "loss": 0.3393, "step": 65672 }, { "epoch": 3.96, "learning_rate": 1.2969836886816652e-05, "loss": 0.3311, "step": 65674 }, { "epoch": 3.96, "learning_rate": 1.2968327976702429e-05, "loss": 0.4008, "step": 65676 }, { "epoch": 3.96, "learning_rate": 1.2966819066588204e-05, "loss": 0.4034, "step": 65678 }, { "epoch": 3.96, "learning_rate": 1.2965310156473979e-05, "loss": 0.3608, "step": 65680 }, { "epoch": 3.96, "learning_rate": 1.2963801246359755e-05, "loss": 0.4573, "step": 65682 }, { "epoch": 3.96, "learning_rate": 1.296229233624553e-05, "loss": 0.1781, "step": 65684 }, { "epoch": 3.96, "learning_rate": 1.2960783426131306e-05, "loss": 0.27, "step": 65686 }, { "epoch": 3.96, "learning_rate": 1.2959274516017081e-05, "loss": 0.3935, "step": 65688 }, { "epoch": 3.96, "learning_rate": 1.2957765605902858e-05, "loss": 0.3631, "step": 65690 }, { "epoch": 3.96, "learning_rate": 1.2956256695788633e-05, "loss": 0.3625, "step": 65692 }, { "epoch": 3.97, "learning_rate": 1.295474778567441e-05, "loss": 0.4961, "step": 65694 }, { "epoch": 3.97, "learning_rate": 1.2953238875560184e-05, "loss": 0.2873, "step": 65696 }, { "epoch": 3.97, "learning_rate": 1.2951729965445957e-05, "loss": 0.3635, "step": 65698 }, { "epoch": 3.97, "learning_rate": 1.2950221055331735e-05, "loss": 0.2381, "step": 65700 }, { "epoch": 3.97, "learning_rate": 1.2948712145217508e-05, "loss": 0.3466, "step": 65702 }, { "epoch": 3.97, "learning_rate": 1.2947203235103287e-05, "loss": 0.4263, "step": 65704 }, { "epoch": 3.97, "learning_rate": 1.294569432498906e-05, "loss": 0.3736, "step": 65706 }, { "epoch": 3.97, "learning_rate": 1.2944185414874838e-05, "loss": 0.2815, "step": 65708 }, { "epoch": 3.97, "learning_rate": 1.2942676504760611e-05, "loss": 0.4026, "step": 65710 }, { "epoch": 3.97, "learning_rate": 1.2941167594646386e-05, "loss": 0.3289, "step": 65712 }, { "epoch": 3.97, "learning_rate": 1.2939658684532162e-05, "loss": 0.1929, "step": 65714 }, { "epoch": 3.97, "learning_rate": 1.2938149774417937e-05, "loss": 0.558, "step": 65716 }, { "epoch": 3.97, "learning_rate": 1.2936640864303714e-05, "loss": 0.2962, "step": 65718 }, { "epoch": 3.97, "learning_rate": 1.2935131954189489e-05, "loss": 0.4444, "step": 65720 }, { "epoch": 3.97, "learning_rate": 1.2933623044075265e-05, "loss": 0.2773, "step": 65722 }, { "epoch": 3.97, "learning_rate": 1.293211413396104e-05, "loss": 0.4716, "step": 65724 }, { "epoch": 3.97, "learning_rate": 1.2930605223846816e-05, "loss": 0.3942, "step": 65726 }, { "epoch": 3.97, "learning_rate": 1.2929096313732591e-05, "loss": 0.3539, "step": 65728 }, { "epoch": 3.97, "learning_rate": 1.2927587403618366e-05, "loss": 0.346, "step": 65730 }, { "epoch": 3.97, "learning_rate": 1.2926078493504143e-05, "loss": 0.3325, "step": 65732 }, { "epoch": 3.97, "learning_rate": 1.2924569583389917e-05, "loss": 0.2999, "step": 65734 }, { "epoch": 3.97, "learning_rate": 1.2923060673275694e-05, "loss": 0.3533, "step": 65736 }, { "epoch": 3.97, "learning_rate": 1.2921551763161469e-05, "loss": 0.4155, "step": 65738 }, { "epoch": 3.97, "learning_rate": 1.2920042853047245e-05, "loss": 0.2972, "step": 65740 }, { "epoch": 3.97, "learning_rate": 1.291853394293302e-05, "loss": 0.4363, "step": 65742 }, { "epoch": 3.97, "learning_rate": 1.2917025032818797e-05, "loss": 0.377, "step": 65744 }, { "epoch": 3.97, "learning_rate": 1.2915516122704571e-05, "loss": 0.1989, "step": 65746 }, { "epoch": 3.97, "learning_rate": 1.2914007212590346e-05, "loss": 0.385, "step": 65748 }, { "epoch": 3.97, "learning_rate": 1.2912498302476123e-05, "loss": 0.3881, "step": 65750 }, { "epoch": 3.97, "learning_rate": 1.2910989392361898e-05, "loss": 0.3214, "step": 65752 }, { "epoch": 3.97, "learning_rate": 1.2909480482247674e-05, "loss": 0.4903, "step": 65754 }, { "epoch": 3.97, "learning_rate": 1.2907971572133449e-05, "loss": 0.4964, "step": 65756 }, { "epoch": 3.97, "learning_rate": 1.2906462662019225e-05, "loss": 0.3075, "step": 65758 }, { "epoch": 3.97, "learning_rate": 1.2904953751905e-05, "loss": 0.2919, "step": 65760 }, { "epoch": 3.97, "learning_rate": 1.2903444841790773e-05, "loss": 0.2634, "step": 65762 }, { "epoch": 3.97, "learning_rate": 1.2901935931676552e-05, "loss": 0.2787, "step": 65764 }, { "epoch": 3.97, "learning_rate": 1.2900427021562325e-05, "loss": 0.2467, "step": 65766 }, { "epoch": 3.97, "learning_rate": 1.2898918111448103e-05, "loss": 0.1702, "step": 65768 }, { "epoch": 3.97, "learning_rate": 1.2897409201333876e-05, "loss": 0.4075, "step": 65770 }, { "epoch": 3.97, "learning_rate": 1.2895900291219654e-05, "loss": 0.4865, "step": 65772 }, { "epoch": 3.97, "learning_rate": 1.2894391381105427e-05, "loss": 0.2707, "step": 65774 }, { "epoch": 3.97, "learning_rate": 1.2892882470991206e-05, "loss": 0.2933, "step": 65776 }, { "epoch": 3.97, "learning_rate": 1.2891373560876979e-05, "loss": 0.3359, "step": 65778 }, { "epoch": 3.97, "learning_rate": 1.2889864650762754e-05, "loss": 0.3741, "step": 65780 }, { "epoch": 3.97, "learning_rate": 1.288835574064853e-05, "loss": 0.3932, "step": 65782 }, { "epoch": 3.97, "learning_rate": 1.2886846830534305e-05, "loss": 0.3087, "step": 65784 }, { "epoch": 3.97, "learning_rate": 1.2885337920420081e-05, "loss": 0.1659, "step": 65786 }, { "epoch": 3.97, "learning_rate": 1.2883829010305856e-05, "loss": 0.5346, "step": 65788 }, { "epoch": 3.97, "learning_rate": 1.2882320100191633e-05, "loss": 0.3557, "step": 65790 }, { "epoch": 3.97, "learning_rate": 1.2880811190077408e-05, "loss": 0.3012, "step": 65792 }, { "epoch": 3.97, "learning_rate": 1.2879302279963182e-05, "loss": 0.1673, "step": 65794 }, { "epoch": 3.97, "learning_rate": 1.2877793369848959e-05, "loss": 0.4416, "step": 65796 }, { "epoch": 3.97, "learning_rate": 1.2876284459734734e-05, "loss": 0.3413, "step": 65798 }, { "epoch": 3.97, "learning_rate": 1.287477554962051e-05, "loss": 0.4761, "step": 65800 }, { "epoch": 3.97, "learning_rate": 1.2873266639506285e-05, "loss": 0.4061, "step": 65802 }, { "epoch": 3.97, "learning_rate": 1.2871757729392062e-05, "loss": 0.5022, "step": 65804 }, { "epoch": 3.97, "learning_rate": 1.2870248819277836e-05, "loss": 0.4299, "step": 65806 }, { "epoch": 3.97, "learning_rate": 1.2868739909163613e-05, "loss": 0.2876, "step": 65808 }, { "epoch": 3.97, "learning_rate": 1.2867230999049388e-05, "loss": 0.3168, "step": 65810 }, { "epoch": 3.97, "learning_rate": 1.286572208893516e-05, "loss": 0.2453, "step": 65812 }, { "epoch": 3.97, "learning_rate": 1.2864213178820939e-05, "loss": 0.4002, "step": 65814 }, { "epoch": 3.97, "learning_rate": 1.2862704268706712e-05, "loss": 0.2675, "step": 65816 }, { "epoch": 3.97, "learning_rate": 1.286119535859249e-05, "loss": 0.2976, "step": 65818 }, { "epoch": 3.97, "learning_rate": 1.2859686448478263e-05, "loss": 0.3479, "step": 65820 }, { "epoch": 3.97, "learning_rate": 1.2858177538364042e-05, "loss": 0.2227, "step": 65822 }, { "epoch": 3.97, "learning_rate": 1.2856668628249815e-05, "loss": 0.2887, "step": 65824 }, { "epoch": 3.97, "learning_rate": 1.2855159718135593e-05, "loss": 0.3788, "step": 65826 }, { "epoch": 3.97, "learning_rate": 1.2853650808021366e-05, "loss": 0.278, "step": 65828 }, { "epoch": 3.97, "learning_rate": 1.2852141897907141e-05, "loss": 0.2624, "step": 65830 }, { "epoch": 3.97, "learning_rate": 1.2850632987792917e-05, "loss": 0.1806, "step": 65832 }, { "epoch": 3.97, "learning_rate": 1.2849124077678692e-05, "loss": 0.4419, "step": 65834 }, { "epoch": 3.97, "learning_rate": 1.2847615167564469e-05, "loss": 0.3303, "step": 65836 }, { "epoch": 3.97, "learning_rate": 1.2846106257450244e-05, "loss": 0.2642, "step": 65838 }, { "epoch": 3.97, "learning_rate": 1.284459734733602e-05, "loss": 0.1753, "step": 65840 }, { "epoch": 3.97, "learning_rate": 1.2843088437221795e-05, "loss": 0.1977, "step": 65842 }, { "epoch": 3.97, "learning_rate": 1.284157952710757e-05, "loss": 0.306, "step": 65844 }, { "epoch": 3.97, "learning_rate": 1.2840070616993346e-05, "loss": 0.2312, "step": 65846 }, { "epoch": 3.97, "learning_rate": 1.2838561706879121e-05, "loss": 0.3161, "step": 65848 }, { "epoch": 3.97, "learning_rate": 1.2837052796764898e-05, "loss": 0.2618, "step": 65850 }, { "epoch": 3.97, "learning_rate": 1.2835543886650672e-05, "loss": 0.3685, "step": 65852 }, { "epoch": 3.97, "learning_rate": 1.2834034976536449e-05, "loss": 0.2809, "step": 65854 }, { "epoch": 3.97, "learning_rate": 1.2832526066422224e-05, "loss": 0.3276, "step": 65856 }, { "epoch": 3.98, "learning_rate": 1.2831017156308e-05, "loss": 0.2738, "step": 65858 }, { "epoch": 3.98, "learning_rate": 1.2829508246193775e-05, "loss": 0.3928, "step": 65860 }, { "epoch": 3.98, "learning_rate": 1.282799933607955e-05, "loss": 0.3808, "step": 65862 }, { "epoch": 3.98, "learning_rate": 1.2826490425965326e-05, "loss": 0.2168, "step": 65864 }, { "epoch": 3.98, "learning_rate": 1.2824981515851101e-05, "loss": 0.3516, "step": 65866 }, { "epoch": 3.98, "learning_rate": 1.2823472605736878e-05, "loss": 0.3815, "step": 65868 }, { "epoch": 3.98, "learning_rate": 1.2821963695622653e-05, "loss": 0.338, "step": 65870 }, { "epoch": 3.98, "learning_rate": 1.2820454785508429e-05, "loss": 0.3165, "step": 65872 }, { "epoch": 3.98, "learning_rate": 1.2818945875394204e-05, "loss": 0.2563, "step": 65874 }, { "epoch": 3.98, "learning_rate": 1.2817436965279977e-05, "loss": 0.4596, "step": 65876 }, { "epoch": 3.98, "learning_rate": 1.2815928055165755e-05, "loss": 0.2511, "step": 65878 }, { "epoch": 3.98, "learning_rate": 1.2814419145051528e-05, "loss": 0.3316, "step": 65880 }, { "epoch": 3.98, "learning_rate": 1.2812910234937307e-05, "loss": 0.2733, "step": 65882 }, { "epoch": 3.98, "learning_rate": 1.281140132482308e-05, "loss": 0.4674, "step": 65884 }, { "epoch": 3.98, "learning_rate": 1.2809892414708858e-05, "loss": 0.4419, "step": 65886 }, { "epoch": 3.98, "learning_rate": 1.2808383504594631e-05, "loss": 0.2606, "step": 65888 }, { "epoch": 3.98, "learning_rate": 1.280687459448041e-05, "loss": 0.3854, "step": 65890 }, { "epoch": 3.98, "learning_rate": 1.2805365684366182e-05, "loss": 0.2864, "step": 65892 }, { "epoch": 3.98, "learning_rate": 1.2803856774251957e-05, "loss": 0.3722, "step": 65894 }, { "epoch": 3.98, "learning_rate": 1.2802347864137734e-05, "loss": 0.391, "step": 65896 }, { "epoch": 3.98, "learning_rate": 1.2800838954023509e-05, "loss": 0.501, "step": 65898 }, { "epoch": 3.98, "learning_rate": 1.2799330043909285e-05, "loss": 0.5103, "step": 65900 }, { "epoch": 3.98, "learning_rate": 1.279782113379506e-05, "loss": 0.3981, "step": 65902 }, { "epoch": 3.98, "learning_rate": 1.2796312223680836e-05, "loss": 0.4644, "step": 65904 }, { "epoch": 3.98, "learning_rate": 1.2794803313566611e-05, "loss": 0.38, "step": 65906 }, { "epoch": 3.98, "learning_rate": 1.2793294403452388e-05, "loss": 0.3092, "step": 65908 }, { "epoch": 3.98, "learning_rate": 1.2791785493338163e-05, "loss": 0.2929, "step": 65910 }, { "epoch": 3.98, "learning_rate": 1.2790276583223937e-05, "loss": 0.4128, "step": 65912 }, { "epoch": 3.98, "learning_rate": 1.2788767673109714e-05, "loss": 0.3195, "step": 65914 }, { "epoch": 3.98, "learning_rate": 1.2787258762995489e-05, "loss": 0.3487, "step": 65916 }, { "epoch": 3.98, "learning_rate": 1.2785749852881265e-05, "loss": 0.2656, "step": 65918 }, { "epoch": 3.98, "learning_rate": 1.278424094276704e-05, "loss": 0.3842, "step": 65920 }, { "epoch": 3.98, "learning_rate": 1.2782732032652817e-05, "loss": 0.277, "step": 65922 }, { "epoch": 3.98, "learning_rate": 1.2781223122538591e-05, "loss": 0.3013, "step": 65924 }, { "epoch": 3.98, "learning_rate": 1.2779714212424364e-05, "loss": 0.3283, "step": 65926 }, { "epoch": 3.98, "learning_rate": 1.2778205302310143e-05, "loss": 0.2668, "step": 65928 }, { "epoch": 3.98, "learning_rate": 1.2776696392195916e-05, "loss": 0.3312, "step": 65930 }, { "epoch": 3.98, "learning_rate": 1.2775187482081694e-05, "loss": 0.2534, "step": 65932 }, { "epoch": 3.98, "learning_rate": 1.2773678571967467e-05, "loss": 0.2852, "step": 65934 }, { "epoch": 3.98, "learning_rate": 1.2772169661853245e-05, "loss": 0.4922, "step": 65936 }, { "epoch": 3.98, "learning_rate": 1.2770660751739018e-05, "loss": 0.4048, "step": 65938 }, { "epoch": 3.98, "learning_rate": 1.2769151841624797e-05, "loss": 0.3381, "step": 65940 }, { "epoch": 3.98, "learning_rate": 1.276764293151057e-05, "loss": 0.3316, "step": 65942 }, { "epoch": 3.98, "learning_rate": 1.2766134021396345e-05, "loss": 0.3978, "step": 65944 }, { "epoch": 3.98, "learning_rate": 1.2764625111282121e-05, "loss": 0.305, "step": 65946 }, { "epoch": 3.98, "learning_rate": 1.2763116201167896e-05, "loss": 0.2719, "step": 65948 }, { "epoch": 3.98, "learning_rate": 1.2761607291053672e-05, "loss": 0.2665, "step": 65950 }, { "epoch": 3.98, "learning_rate": 1.2760098380939447e-05, "loss": 0.3118, "step": 65952 }, { "epoch": 3.98, "learning_rate": 1.2758589470825224e-05, "loss": 0.3254, "step": 65954 }, { "epoch": 3.98, "learning_rate": 1.2757080560710999e-05, "loss": 0.342, "step": 65956 }, { "epoch": 3.98, "learning_rate": 1.2755571650596775e-05, "loss": 0.241, "step": 65958 }, { "epoch": 3.98, "learning_rate": 1.275406274048255e-05, "loss": 0.3484, "step": 65960 }, { "epoch": 3.98, "learning_rate": 1.2752553830368325e-05, "loss": 0.2155, "step": 65962 }, { "epoch": 3.98, "learning_rate": 1.2751044920254101e-05, "loss": 0.324, "step": 65964 }, { "epoch": 3.98, "learning_rate": 1.2749536010139876e-05, "loss": 0.2616, "step": 65966 }, { "epoch": 3.98, "learning_rate": 1.2748027100025653e-05, "loss": 0.4679, "step": 65968 }, { "epoch": 3.98, "learning_rate": 1.2746518189911427e-05, "loss": 0.3897, "step": 65970 }, { "epoch": 3.98, "learning_rate": 1.2745009279797204e-05, "loss": 0.4334, "step": 65972 }, { "epoch": 3.98, "learning_rate": 1.2743500369682979e-05, "loss": 0.2406, "step": 65974 }, { "epoch": 3.98, "learning_rate": 1.2741991459568754e-05, "loss": 0.3435, "step": 65976 }, { "epoch": 3.98, "learning_rate": 1.274048254945453e-05, "loss": 0.3119, "step": 65978 }, { "epoch": 3.98, "learning_rate": 1.2738973639340305e-05, "loss": 0.4506, "step": 65980 }, { "epoch": 3.98, "learning_rate": 1.2737464729226081e-05, "loss": 0.3133, "step": 65982 }, { "epoch": 3.98, "learning_rate": 1.2735955819111856e-05, "loss": 0.3703, "step": 65984 }, { "epoch": 3.98, "learning_rate": 1.2734446908997633e-05, "loss": 0.3407, "step": 65986 }, { "epoch": 3.98, "learning_rate": 1.2732937998883408e-05, "loss": 0.4518, "step": 65988 }, { "epoch": 3.98, "learning_rate": 1.2731429088769184e-05, "loss": 0.3659, "step": 65990 }, { "epoch": 3.98, "learning_rate": 1.2729920178654959e-05, "loss": 0.2713, "step": 65992 }, { "epoch": 3.98, "learning_rate": 1.2728411268540732e-05, "loss": 0.4346, "step": 65994 }, { "epoch": 3.98, "learning_rate": 1.272690235842651e-05, "loss": 0.3838, "step": 65996 }, { "epoch": 3.98, "learning_rate": 1.2725393448312283e-05, "loss": 0.3138, "step": 65998 }, { "epoch": 3.98, "learning_rate": 1.2723884538198062e-05, "loss": 0.2261, "step": 66000 }, { "epoch": 3.98, "learning_rate": 1.2722375628083835e-05, "loss": 0.5016, "step": 66002 }, { "epoch": 3.98, "learning_rate": 1.2720866717969613e-05, "loss": 0.2986, "step": 66004 }, { "epoch": 3.98, "learning_rate": 1.2719357807855386e-05, "loss": 0.3632, "step": 66006 }, { "epoch": 3.98, "learning_rate": 1.2717848897741161e-05, "loss": 0.3897, "step": 66008 }, { "epoch": 3.98, "learning_rate": 1.2716339987626937e-05, "loss": 0.4011, "step": 66010 }, { "epoch": 3.98, "learning_rate": 1.2714831077512712e-05, "loss": 0.3585, "step": 66012 }, { "epoch": 3.98, "learning_rate": 1.2713322167398489e-05, "loss": 0.4109, "step": 66014 }, { "epoch": 3.98, "learning_rate": 1.2711813257284264e-05, "loss": 0.3896, "step": 66016 }, { "epoch": 3.98, "learning_rate": 1.271030434717004e-05, "loss": 0.2918, "step": 66018 }, { "epoch": 3.98, "learning_rate": 1.2708795437055815e-05, "loss": 0.3853, "step": 66020 }, { "epoch": 3.98, "learning_rate": 1.2707286526941591e-05, "loss": 0.392, "step": 66022 }, { "epoch": 3.99, "learning_rate": 1.2705777616827366e-05, "loss": 0.2439, "step": 66024 }, { "epoch": 3.99, "learning_rate": 1.2704268706713141e-05, "loss": 0.3071, "step": 66026 }, { "epoch": 3.99, "learning_rate": 1.2702759796598918e-05, "loss": 0.2364, "step": 66028 }, { "epoch": 3.99, "learning_rate": 1.2701250886484692e-05, "loss": 0.2621, "step": 66030 }, { "epoch": 3.99, "learning_rate": 1.2699741976370469e-05, "loss": 0.4278, "step": 66032 }, { "epoch": 3.99, "learning_rate": 1.2698233066256244e-05, "loss": 0.4465, "step": 66034 }, { "epoch": 3.99, "learning_rate": 1.269672415614202e-05, "loss": 0.2595, "step": 66036 }, { "epoch": 3.99, "learning_rate": 1.2695215246027795e-05, "loss": 0.4522, "step": 66038 }, { "epoch": 3.99, "learning_rate": 1.2693706335913572e-05, "loss": 0.4802, "step": 66040 }, { "epoch": 3.99, "learning_rate": 1.2692197425799346e-05, "loss": 0.2806, "step": 66042 }, { "epoch": 3.99, "learning_rate": 1.269068851568512e-05, "loss": 0.3743, "step": 66044 }, { "epoch": 3.99, "learning_rate": 1.2689179605570898e-05, "loss": 0.3391, "step": 66046 }, { "epoch": 3.99, "learning_rate": 1.268767069545667e-05, "loss": 0.2222, "step": 66048 }, { "epoch": 3.99, "learning_rate": 1.2686161785342449e-05, "loss": 0.3556, "step": 66050 }, { "epoch": 3.99, "learning_rate": 1.2684652875228222e-05, "loss": 0.3434, "step": 66052 }, { "epoch": 3.99, "learning_rate": 1.2683143965114e-05, "loss": 0.314, "step": 66054 }, { "epoch": 3.99, "learning_rate": 1.2681635054999773e-05, "loss": 0.3265, "step": 66056 }, { "epoch": 3.99, "learning_rate": 1.2680126144885548e-05, "loss": 0.3643, "step": 66058 }, { "epoch": 3.99, "learning_rate": 1.2678617234771325e-05, "loss": 0.3718, "step": 66060 }, { "epoch": 3.99, "learning_rate": 1.26771083246571e-05, "loss": 0.3632, "step": 66062 }, { "epoch": 3.99, "learning_rate": 1.2675599414542876e-05, "loss": 0.3363, "step": 66064 }, { "epoch": 3.99, "learning_rate": 1.2674090504428651e-05, "loss": 0.4449, "step": 66066 }, { "epoch": 3.99, "learning_rate": 1.2672581594314427e-05, "loss": 0.2926, "step": 66068 }, { "epoch": 3.99, "learning_rate": 1.2671072684200202e-05, "loss": 0.4994, "step": 66070 }, { "epoch": 3.99, "learning_rate": 1.2669563774085979e-05, "loss": 0.3471, "step": 66072 }, { "epoch": 3.99, "learning_rate": 1.2668054863971754e-05, "loss": 0.3238, "step": 66074 }, { "epoch": 3.99, "learning_rate": 1.2666545953857528e-05, "loss": 0.4905, "step": 66076 }, { "epoch": 3.99, "learning_rate": 1.2665037043743305e-05, "loss": 0.3281, "step": 66078 }, { "epoch": 3.99, "learning_rate": 1.266352813362908e-05, "loss": 0.2642, "step": 66080 }, { "epoch": 3.99, "learning_rate": 1.2662019223514856e-05, "loss": 0.2935, "step": 66082 }, { "epoch": 3.99, "learning_rate": 1.2660510313400631e-05, "loss": 0.4426, "step": 66084 }, { "epoch": 3.99, "learning_rate": 1.2659001403286408e-05, "loss": 0.244, "step": 66086 }, { "epoch": 3.99, "learning_rate": 1.2657492493172182e-05, "loss": 0.3854, "step": 66088 }, { "epoch": 3.99, "learning_rate": 1.2655983583057957e-05, "loss": 0.3621, "step": 66090 }, { "epoch": 3.99, "learning_rate": 1.2654474672943734e-05, "loss": 0.2591, "step": 66092 }, { "epoch": 3.99, "learning_rate": 1.2652965762829509e-05, "loss": 0.2117, "step": 66094 }, { "epoch": 3.99, "learning_rate": 1.2651456852715285e-05, "loss": 0.2845, "step": 66096 }, { "epoch": 3.99, "learning_rate": 1.264994794260106e-05, "loss": 0.3407, "step": 66098 }, { "epoch": 3.99, "learning_rate": 1.2648439032486836e-05, "loss": 0.4075, "step": 66100 }, { "epoch": 3.99, "learning_rate": 1.2646930122372611e-05, "loss": 0.3755, "step": 66102 }, { "epoch": 3.99, "learning_rate": 1.2645421212258388e-05, "loss": 0.4264, "step": 66104 }, { "epoch": 3.99, "learning_rate": 1.2643912302144163e-05, "loss": 0.3123, "step": 66106 }, { "epoch": 3.99, "learning_rate": 1.2642403392029936e-05, "loss": 0.247, "step": 66108 }, { "epoch": 3.99, "learning_rate": 1.2640894481915714e-05, "loss": 0.2214, "step": 66110 }, { "epoch": 3.99, "learning_rate": 1.2639385571801487e-05, "loss": 0.2412, "step": 66112 }, { "epoch": 3.99, "learning_rate": 1.2637876661687265e-05, "loss": 0.3164, "step": 66114 }, { "epoch": 3.99, "learning_rate": 1.2636367751573038e-05, "loss": 0.275, "step": 66116 }, { "epoch": 3.99, "learning_rate": 1.2634858841458817e-05, "loss": 0.2642, "step": 66118 }, { "epoch": 3.99, "learning_rate": 1.263334993134459e-05, "loss": 0.2402, "step": 66120 }, { "epoch": 3.99, "learning_rate": 1.2631841021230368e-05, "loss": 0.4937, "step": 66122 }, { "epoch": 3.99, "learning_rate": 1.2630332111116141e-05, "loss": 0.437, "step": 66124 }, { "epoch": 3.99, "learning_rate": 1.2628823201001916e-05, "loss": 0.4208, "step": 66126 }, { "epoch": 3.99, "learning_rate": 1.2627314290887692e-05, "loss": 0.2331, "step": 66128 }, { "epoch": 3.99, "learning_rate": 1.2625805380773467e-05, "loss": 0.2606, "step": 66130 }, { "epoch": 3.99, "learning_rate": 1.2624296470659244e-05, "loss": 0.2751, "step": 66132 }, { "epoch": 3.99, "learning_rate": 1.2622787560545019e-05, "loss": 0.3974, "step": 66134 }, { "epoch": 3.99, "learning_rate": 1.2621278650430795e-05, "loss": 0.2982, "step": 66136 }, { "epoch": 3.99, "learning_rate": 1.261976974031657e-05, "loss": 0.3683, "step": 66138 }, { "epoch": 3.99, "learning_rate": 1.2618260830202345e-05, "loss": 0.2522, "step": 66140 }, { "epoch": 3.99, "learning_rate": 1.2616751920088121e-05, "loss": 0.3904, "step": 66142 }, { "epoch": 3.99, "learning_rate": 1.2615243009973896e-05, "loss": 0.4834, "step": 66144 }, { "epoch": 3.99, "learning_rate": 1.2613734099859673e-05, "loss": 0.3799, "step": 66146 }, { "epoch": 3.99, "learning_rate": 1.2612225189745447e-05, "loss": 0.2727, "step": 66148 }, { "epoch": 3.99, "learning_rate": 1.2610716279631224e-05, "loss": 0.315, "step": 66150 }, { "epoch": 3.99, "learning_rate": 1.2609207369516999e-05, "loss": 0.3741, "step": 66152 }, { "epoch": 3.99, "learning_rate": 1.2607698459402775e-05, "loss": 0.2986, "step": 66154 }, { "epoch": 3.99, "learning_rate": 1.260618954928855e-05, "loss": 0.3991, "step": 66156 }, { "epoch": 3.99, "learning_rate": 1.2604680639174323e-05, "loss": 0.2943, "step": 66158 }, { "epoch": 3.99, "learning_rate": 1.2603171729060101e-05, "loss": 0.2027, "step": 66160 }, { "epoch": 3.99, "learning_rate": 1.2601662818945874e-05, "loss": 0.3912, "step": 66162 }, { "epoch": 3.99, "learning_rate": 1.2600153908831653e-05, "loss": 0.4073, "step": 66164 }, { "epoch": 3.99, "learning_rate": 1.2598644998717426e-05, "loss": 0.3596, "step": 66166 }, { "epoch": 3.99, "learning_rate": 1.2597136088603204e-05, "loss": 0.4116, "step": 66168 }, { "epoch": 3.99, "learning_rate": 1.2595627178488977e-05, "loss": 0.2699, "step": 66170 }, { "epoch": 3.99, "learning_rate": 1.2594118268374752e-05, "loss": 0.2209, "step": 66172 }, { "epoch": 3.99, "learning_rate": 1.2592609358260528e-05, "loss": 0.3613, "step": 66174 }, { "epoch": 3.99, "learning_rate": 1.2591100448146303e-05, "loss": 0.4204, "step": 66176 }, { "epoch": 3.99, "learning_rate": 1.258959153803208e-05, "loss": 0.3182, "step": 66178 }, { "epoch": 3.99, "learning_rate": 1.2588082627917855e-05, "loss": 0.2503, "step": 66180 }, { "epoch": 3.99, "learning_rate": 1.2586573717803631e-05, "loss": 0.4823, "step": 66182 }, { "epoch": 3.99, "learning_rate": 1.2585064807689406e-05, "loss": 0.366, "step": 66184 }, { "epoch": 3.99, "learning_rate": 1.2583555897575182e-05, "loss": 0.5175, "step": 66186 }, { "epoch": 3.99, "learning_rate": 1.2582046987460957e-05, "loss": 0.5479, "step": 66188 }, { "epoch": 4.0, "learning_rate": 1.2580538077346732e-05, "loss": 0.4209, "step": 66190 }, { "epoch": 4.0, "learning_rate": 1.2579029167232509e-05, "loss": 0.2979, "step": 66192 }, { "epoch": 4.0, "learning_rate": 1.2577520257118283e-05, "loss": 0.362, "step": 66194 }, { "epoch": 4.0, "learning_rate": 1.257601134700406e-05, "loss": 0.2742, "step": 66196 }, { "epoch": 4.0, "learning_rate": 1.2574502436889835e-05, "loss": 0.3903, "step": 66198 }, { "epoch": 4.0, "learning_rate": 1.2572993526775611e-05, "loss": 0.3003, "step": 66200 }, { "epoch": 4.0, "learning_rate": 1.2571484616661386e-05, "loss": 0.4393, "step": 66202 }, { "epoch": 4.0, "learning_rate": 1.2569975706547163e-05, "loss": 0.234, "step": 66204 }, { "epoch": 4.0, "learning_rate": 1.2568466796432937e-05, "loss": 0.3823, "step": 66206 }, { "epoch": 4.0, "learning_rate": 1.2566957886318712e-05, "loss": 0.3538, "step": 66208 }, { "epoch": 4.0, "learning_rate": 1.2565448976204489e-05, "loss": 0.1861, "step": 66210 }, { "epoch": 4.0, "learning_rate": 1.2563940066090264e-05, "loss": 0.2434, "step": 66212 }, { "epoch": 4.0, "learning_rate": 1.256243115597604e-05, "loss": 0.292, "step": 66214 }, { "epoch": 4.0, "learning_rate": 1.2560922245861815e-05, "loss": 0.3106, "step": 66216 }, { "epoch": 4.0, "learning_rate": 1.2559413335747591e-05, "loss": 0.3789, "step": 66218 }, { "epoch": 4.0, "learning_rate": 1.2557904425633366e-05, "loss": 0.3116, "step": 66220 }, { "epoch": 4.0, "learning_rate": 1.255639551551914e-05, "loss": 0.2589, "step": 66222 }, { "epoch": 4.0, "learning_rate": 1.2554886605404918e-05, "loss": 0.4849, "step": 66224 }, { "epoch": 4.0, "learning_rate": 1.255337769529069e-05, "loss": 0.3996, "step": 66226 }, { "epoch": 4.0, "learning_rate": 1.2551868785176469e-05, "loss": 0.2384, "step": 66228 }, { "epoch": 4.0, "learning_rate": 1.2550359875062242e-05, "loss": 0.383, "step": 66230 }, { "epoch": 4.0, "learning_rate": 1.254885096494802e-05, "loss": 0.3305, "step": 66232 }, { "epoch": 4.0, "learning_rate": 1.2547342054833793e-05, "loss": 0.2506, "step": 66234 }, { "epoch": 4.0, "learning_rate": 1.2545833144719572e-05, "loss": 0.3231, "step": 66236 }, { "epoch": 4.0, "learning_rate": 1.2544324234605345e-05, "loss": 0.2699, "step": 66238 }, { "epoch": 4.0, "learning_rate": 1.254281532449112e-05, "loss": 0.4059, "step": 66240 }, { "epoch": 4.0, "learning_rate": 1.2541306414376896e-05, "loss": 0.5378, "step": 66242 }, { "epoch": 4.0, "learning_rate": 1.2539797504262671e-05, "loss": 0.5071, "step": 66244 }, { "epoch": 4.0, "learning_rate": 1.2538288594148447e-05, "loss": 0.4379, "step": 66246 }, { "epoch": 4.0, "learning_rate": 1.2536779684034222e-05, "loss": 0.4669, "step": 66248 }, { "epoch": 4.0, "learning_rate": 1.2535270773919999e-05, "loss": 0.2591, "step": 66250 }, { "epoch": 4.0, "learning_rate": 1.2533761863805774e-05, "loss": 0.3226, "step": 66252 }, { "epoch": 4.0, "learning_rate": 1.2532252953691548e-05, "loss": 0.2469, "step": 66254 }, { "epoch": 4.0, "learning_rate": 1.2530744043577325e-05, "loss": 0.323, "step": 66256 }, { "epoch": 4.0, "learning_rate": 1.25292351334631e-05, "loss": 0.3579, "step": 66258 }, { "epoch": 4.0, "learning_rate": 1.2527726223348876e-05, "loss": 0.3229, "step": 66260 }, { "epoch": 4.0, "learning_rate": 1.2526217313234651e-05, "loss": 0.448, "step": 66262 }, { "epoch": 4.0, "learning_rate": 1.2524708403120428e-05, "loss": 0.3571, "step": 66264 }, { "epoch": 4.0, "learning_rate": 1.2523199493006202e-05, "loss": 0.2598, "step": 66266 }, { "epoch": 4.0, "learning_rate": 1.2521690582891979e-05, "loss": 0.2837, "step": 66268 }, { "epoch": 4.0, "eval_cer": 0.07859486213120523, "eval_loss": 0.485178679227829, "eval_runtime": 9385.8963, "eval_samples_per_second": 1.765, "eval_steps_per_second": 0.221, "step": 66268 }, { "epoch": 4.0, "learning_rate": 1.2520181672777754e-05, "loss": 0.2667, "step": 66270 }, { "epoch": 4.0, "learning_rate": 1.2518672762663527e-05, "loss": 0.4274, "step": 66272 }, { "epoch": 4.0, "learning_rate": 1.2517163852549305e-05, "loss": 0.2492, "step": 66274 }, { "epoch": 4.0, "learning_rate": 1.2515654942435078e-05, "loss": 0.2241, "step": 66276 }, { "epoch": 4.0, "learning_rate": 1.2514146032320856e-05, "loss": 0.1538, "step": 66278 }, { "epoch": 4.0, "learning_rate": 1.251263712220663e-05, "loss": 0.255, "step": 66280 }, { "epoch": 4.0, "learning_rate": 1.2511128212092408e-05, "loss": 0.2271, "step": 66282 }, { "epoch": 4.0, "learning_rate": 1.250961930197818e-05, "loss": 0.2394, "step": 66284 }, { "epoch": 4.0, "learning_rate": 1.2508110391863959e-05, "loss": 0.3183, "step": 66286 }, { "epoch": 4.0, "learning_rate": 1.2506601481749732e-05, "loss": 0.1336, "step": 66288 }, { "epoch": 4.0, "learning_rate": 1.2505092571635507e-05, "loss": 0.2022, "step": 66290 }, { "epoch": 4.0, "learning_rate": 1.2503583661521283e-05, "loss": 0.443, "step": 66292 }, { "epoch": 4.0, "learning_rate": 1.2502074751407058e-05, "loss": 0.1548, "step": 66294 }, { "epoch": 4.0, "learning_rate": 1.2500565841292835e-05, "loss": 0.2309, "step": 66296 }, { "epoch": 4.0, "learning_rate": 1.249905693117861e-05, "loss": 0.2351, "step": 66298 }, { "epoch": 4.0, "learning_rate": 1.2497548021064386e-05, "loss": 0.1969, "step": 66300 }, { "epoch": 4.0, "learning_rate": 1.2496039110950161e-05, "loss": 0.2534, "step": 66302 }, { "epoch": 4.0, "learning_rate": 1.2494530200835937e-05, "loss": 0.3149, "step": 66304 }, { "epoch": 4.0, "learning_rate": 1.2493021290721712e-05, "loss": 0.2629, "step": 66306 }, { "epoch": 4.0, "learning_rate": 1.2491512380607489e-05, "loss": 0.2687, "step": 66308 }, { "epoch": 4.0, "learning_rate": 1.2490003470493264e-05, "loss": 0.1806, "step": 66310 }, { "epoch": 4.0, "learning_rate": 1.2488494560379038e-05, "loss": 0.1961, "step": 66312 }, { "epoch": 4.0, "learning_rate": 1.2486985650264813e-05, "loss": 0.2405, "step": 66314 }, { "epoch": 4.0, "learning_rate": 1.248547674015059e-05, "loss": 0.2489, "step": 66316 }, { "epoch": 4.0, "learning_rate": 1.2483967830036365e-05, "loss": 0.2229, "step": 66318 }, { "epoch": 4.0, "learning_rate": 1.2482458919922141e-05, "loss": 0.3203, "step": 66320 }, { "epoch": 4.0, "learning_rate": 1.2480950009807916e-05, "loss": 0.2039, "step": 66322 }, { "epoch": 4.0, "learning_rate": 1.2479441099693692e-05, "loss": 0.2549, "step": 66324 }, { "epoch": 4.0, "learning_rate": 1.2477932189579467e-05, "loss": 0.3291, "step": 66326 }, { "epoch": 4.0, "learning_rate": 1.2476423279465242e-05, "loss": 0.2326, "step": 66328 }, { "epoch": 4.0, "learning_rate": 1.2474914369351019e-05, "loss": 0.3476, "step": 66330 }, { "epoch": 4.0, "learning_rate": 1.2473405459236793e-05, "loss": 0.1868, "step": 66332 }, { "epoch": 4.0, "learning_rate": 1.247189654912257e-05, "loss": 0.3544, "step": 66334 }, { "epoch": 4.0, "learning_rate": 1.2470387639008345e-05, "loss": 0.198, "step": 66336 }, { "epoch": 4.0, "learning_rate": 1.2468878728894121e-05, "loss": 0.2466, "step": 66338 }, { "epoch": 4.0, "learning_rate": 1.2467369818779896e-05, "loss": 0.2807, "step": 66340 }, { "epoch": 4.0, "learning_rate": 1.2465860908665673e-05, "loss": 0.3027, "step": 66342 }, { "epoch": 4.0, "learning_rate": 1.2464351998551446e-05, "loss": 0.2378, "step": 66344 }, { "epoch": 4.0, "learning_rate": 1.2462843088437222e-05, "loss": 0.1952, "step": 66346 }, { "epoch": 4.0, "learning_rate": 1.2461334178322997e-05, "loss": 0.2225, "step": 66348 }, { "epoch": 4.0, "learning_rate": 1.2459825268208774e-05, "loss": 0.2583, "step": 66350 }, { "epoch": 4.0, "learning_rate": 1.2458316358094548e-05, "loss": 0.2394, "step": 66352 }, { "epoch": 4.0, "learning_rate": 1.2456807447980325e-05, "loss": 0.193, "step": 66354 }, { "epoch": 4.01, "learning_rate": 1.24552985378661e-05, "loss": 0.1916, "step": 66356 }, { "epoch": 4.01, "learning_rate": 1.2453789627751876e-05, "loss": 0.1358, "step": 66358 }, { "epoch": 4.01, "learning_rate": 1.2452280717637651e-05, "loss": 0.1602, "step": 66360 }, { "epoch": 4.01, "learning_rate": 1.2450771807523426e-05, "loss": 0.2604, "step": 66362 }, { "epoch": 4.01, "learning_rate": 1.24492628974092e-05, "loss": 0.2916, "step": 66364 }, { "epoch": 4.01, "learning_rate": 1.2447753987294977e-05, "loss": 0.1776, "step": 66366 }, { "epoch": 4.01, "learning_rate": 1.2446245077180752e-05, "loss": 0.186, "step": 66368 }, { "epoch": 4.01, "learning_rate": 1.2444736167066529e-05, "loss": 0.3799, "step": 66370 }, { "epoch": 4.01, "learning_rate": 1.2443227256952305e-05, "loss": 0.2972, "step": 66372 }, { "epoch": 4.01, "learning_rate": 1.244171834683808e-05, "loss": 0.1615, "step": 66374 }, { "epoch": 4.01, "learning_rate": 1.2440209436723856e-05, "loss": 0.1105, "step": 66376 }, { "epoch": 4.01, "learning_rate": 1.243870052660963e-05, "loss": 0.1936, "step": 66378 }, { "epoch": 4.01, "learning_rate": 1.2437191616495406e-05, "loss": 0.2435, "step": 66380 }, { "epoch": 4.01, "learning_rate": 1.2435682706381181e-05, "loss": 0.1643, "step": 66382 }, { "epoch": 4.01, "learning_rate": 1.2434173796266957e-05, "loss": 0.2516, "step": 66384 }, { "epoch": 4.01, "learning_rate": 1.2432664886152732e-05, "loss": 0.3345, "step": 66386 }, { "epoch": 4.01, "learning_rate": 1.2431155976038509e-05, "loss": 0.1981, "step": 66388 }, { "epoch": 4.01, "learning_rate": 1.2429647065924284e-05, "loss": 0.1946, "step": 66390 }, { "epoch": 4.01, "learning_rate": 1.242813815581006e-05, "loss": 0.091, "step": 66392 }, { "epoch": 4.01, "learning_rate": 1.2426629245695833e-05, "loss": 0.2651, "step": 66394 }, { "epoch": 4.01, "learning_rate": 1.242512033558161e-05, "loss": 0.2343, "step": 66396 }, { "epoch": 4.01, "learning_rate": 1.2423611425467385e-05, "loss": 0.1617, "step": 66398 }, { "epoch": 4.01, "learning_rate": 1.2422102515353161e-05, "loss": 0.3003, "step": 66400 }, { "epoch": 4.01, "learning_rate": 1.2420593605238936e-05, "loss": 0.218, "step": 66402 }, { "epoch": 4.01, "learning_rate": 1.2419084695124712e-05, "loss": 0.2015, "step": 66404 }, { "epoch": 4.01, "learning_rate": 1.2417575785010487e-05, "loss": 0.2802, "step": 66406 }, { "epoch": 4.01, "learning_rate": 1.2416066874896264e-05, "loss": 0.2712, "step": 66408 }, { "epoch": 4.01, "learning_rate": 1.2414557964782039e-05, "loss": 0.1866, "step": 66410 }, { "epoch": 4.01, "learning_rate": 1.2413049054667813e-05, "loss": 0.2495, "step": 66412 }, { "epoch": 4.01, "learning_rate": 1.241154014455359e-05, "loss": 0.2094, "step": 66414 }, { "epoch": 4.01, "learning_rate": 1.2410031234439365e-05, "loss": 0.3248, "step": 66416 }, { "epoch": 4.01, "learning_rate": 1.2408522324325141e-05, "loss": 0.2772, "step": 66418 }, { "epoch": 4.01, "learning_rate": 1.2407013414210916e-05, "loss": 0.3287, "step": 66420 }, { "epoch": 4.01, "learning_rate": 1.2405504504096693e-05, "loss": 0.2157, "step": 66422 }, { "epoch": 4.01, "learning_rate": 1.2403995593982467e-05, "loss": 0.2763, "step": 66424 }, { "epoch": 4.01, "learning_rate": 1.2402486683868244e-05, "loss": 0.2175, "step": 66426 }, { "epoch": 4.01, "learning_rate": 1.2400977773754017e-05, "loss": 0.2399, "step": 66428 }, { "epoch": 4.01, "learning_rate": 1.2399468863639793e-05, "loss": 0.3443, "step": 66430 }, { "epoch": 4.01, "learning_rate": 1.2397959953525568e-05, "loss": 0.2416, "step": 66432 }, { "epoch": 4.01, "learning_rate": 1.2396451043411345e-05, "loss": 0.1736, "step": 66434 }, { "epoch": 4.01, "learning_rate": 1.239494213329712e-05, "loss": 0.2113, "step": 66436 }, { "epoch": 4.01, "learning_rate": 1.2393433223182896e-05, "loss": 0.2975, "step": 66438 }, { "epoch": 4.01, "learning_rate": 1.2391924313068671e-05, "loss": 0.2014, "step": 66440 }, { "epoch": 4.01, "learning_rate": 1.2390415402954447e-05, "loss": 0.2618, "step": 66442 }, { "epoch": 4.01, "learning_rate": 1.2388906492840222e-05, "loss": 0.2804, "step": 66444 }, { "epoch": 4.01, "learning_rate": 1.2387397582725997e-05, "loss": 0.2067, "step": 66446 }, { "epoch": 4.01, "learning_rate": 1.2385888672611774e-05, "loss": 0.3607, "step": 66448 }, { "epoch": 4.01, "learning_rate": 1.2384379762497548e-05, "loss": 0.2387, "step": 66450 }, { "epoch": 4.01, "learning_rate": 1.2382870852383325e-05, "loss": 0.409, "step": 66452 }, { "epoch": 4.01, "learning_rate": 1.23813619422691e-05, "loss": 0.3541, "step": 66454 }, { "epoch": 4.01, "learning_rate": 1.2379853032154876e-05, "loss": 0.1443, "step": 66456 }, { "epoch": 4.01, "learning_rate": 1.2378344122040651e-05, "loss": 0.3189, "step": 66458 }, { "epoch": 4.01, "learning_rate": 1.2376835211926426e-05, "loss": 0.2814, "step": 66460 }, { "epoch": 4.01, "learning_rate": 1.23753263018122e-05, "loss": 0.1427, "step": 66462 }, { "epoch": 4.01, "learning_rate": 1.2373817391697977e-05, "loss": 0.176, "step": 66464 }, { "epoch": 4.01, "learning_rate": 1.2372308481583752e-05, "loss": 0.2571, "step": 66466 }, { "epoch": 4.01, "learning_rate": 1.2370799571469529e-05, "loss": 0.2982, "step": 66468 }, { "epoch": 4.01, "learning_rate": 1.2369290661355303e-05, "loss": 0.1833, "step": 66470 }, { "epoch": 4.01, "learning_rate": 1.236778175124108e-05, "loss": 0.3061, "step": 66472 }, { "epoch": 4.01, "learning_rate": 1.2366272841126855e-05, "loss": 0.1755, "step": 66474 }, { "epoch": 4.01, "learning_rate": 1.236476393101263e-05, "loss": 0.2618, "step": 66476 }, { "epoch": 4.01, "learning_rate": 1.2363255020898404e-05, "loss": 0.3026, "step": 66478 }, { "epoch": 4.01, "learning_rate": 1.2361746110784181e-05, "loss": 0.27, "step": 66480 }, { "epoch": 4.01, "learning_rate": 1.2360237200669956e-05, "loss": 0.2836, "step": 66482 }, { "epoch": 4.01, "learning_rate": 1.2358728290555732e-05, "loss": 0.243, "step": 66484 }, { "epoch": 4.01, "learning_rate": 1.2357219380441507e-05, "loss": 0.2824, "step": 66486 }, { "epoch": 4.01, "learning_rate": 1.2355710470327284e-05, "loss": 0.1933, "step": 66488 }, { "epoch": 4.01, "learning_rate": 1.235420156021306e-05, "loss": 0.158, "step": 66490 }, { "epoch": 4.01, "learning_rate": 1.2352692650098833e-05, "loss": 0.2307, "step": 66492 }, { "epoch": 4.01, "learning_rate": 1.235118373998461e-05, "loss": 0.3706, "step": 66494 }, { "epoch": 4.01, "learning_rate": 1.2349674829870385e-05, "loss": 0.2641, "step": 66496 }, { "epoch": 4.01, "learning_rate": 1.2348165919756161e-05, "loss": 0.3068, "step": 66498 }, { "epoch": 4.01, "learning_rate": 1.2346657009641936e-05, "loss": 0.2028, "step": 66500 }, { "epoch": 4.01, "learning_rate": 1.2345148099527712e-05, "loss": 0.2633, "step": 66502 }, { "epoch": 4.01, "learning_rate": 1.2343639189413487e-05, "loss": 0.2773, "step": 66504 }, { "epoch": 4.01, "learning_rate": 1.2342130279299264e-05, "loss": 0.3796, "step": 66506 }, { "epoch": 4.01, "learning_rate": 1.2340621369185039e-05, "loss": 0.2032, "step": 66508 }, { "epoch": 4.01, "learning_rate": 1.2339112459070813e-05, "loss": 0.2214, "step": 66510 }, { "epoch": 4.01, "learning_rate": 1.2337603548956588e-05, "loss": 0.2308, "step": 66512 }, { "epoch": 4.01, "learning_rate": 1.2336094638842365e-05, "loss": 0.1201, "step": 66514 }, { "epoch": 4.01, "learning_rate": 1.233458572872814e-05, "loss": 0.2761, "step": 66516 }, { "epoch": 4.01, "learning_rate": 1.2333076818613916e-05, "loss": 0.2157, "step": 66518 }, { "epoch": 4.01, "learning_rate": 1.2331567908499691e-05, "loss": 0.1526, "step": 66520 }, { "epoch": 4.02, "learning_rate": 1.2330058998385467e-05, "loss": 0.2322, "step": 66522 }, { "epoch": 4.02, "learning_rate": 1.2328550088271242e-05, "loss": 0.3867, "step": 66524 }, { "epoch": 4.02, "learning_rate": 1.2327041178157017e-05, "loss": 0.2547, "step": 66526 }, { "epoch": 4.02, "learning_rate": 1.2325532268042794e-05, "loss": 0.2086, "step": 66528 }, { "epoch": 4.02, "learning_rate": 1.2324023357928568e-05, "loss": 0.2564, "step": 66530 }, { "epoch": 4.02, "learning_rate": 1.2322514447814345e-05, "loss": 0.306, "step": 66532 }, { "epoch": 4.02, "learning_rate": 1.232100553770012e-05, "loss": 0.2757, "step": 66534 }, { "epoch": 4.02, "learning_rate": 1.2319496627585896e-05, "loss": 0.2053, "step": 66536 }, { "epoch": 4.02, "learning_rate": 1.2317987717471671e-05, "loss": 0.3339, "step": 66538 }, { "epoch": 4.02, "learning_rate": 1.2316478807357448e-05, "loss": 0.3161, "step": 66540 }, { "epoch": 4.02, "learning_rate": 1.231496989724322e-05, "loss": 0.1976, "step": 66542 }, { "epoch": 4.02, "learning_rate": 1.2313460987128997e-05, "loss": 0.1815, "step": 66544 }, { "epoch": 4.02, "learning_rate": 1.2311952077014772e-05, "loss": 0.26, "step": 66546 }, { "epoch": 4.02, "learning_rate": 1.2310443166900548e-05, "loss": 0.3132, "step": 66548 }, { "epoch": 4.02, "learning_rate": 1.2308934256786323e-05, "loss": 0.2426, "step": 66550 }, { "epoch": 4.02, "learning_rate": 1.23074253466721e-05, "loss": 0.2828, "step": 66552 }, { "epoch": 4.02, "learning_rate": 1.2305916436557875e-05, "loss": 0.2241, "step": 66554 }, { "epoch": 4.02, "learning_rate": 1.2304407526443651e-05, "loss": 0.1925, "step": 66556 }, { "epoch": 4.02, "learning_rate": 1.2302898616329426e-05, "loss": 0.341, "step": 66558 }, { "epoch": 4.02, "learning_rate": 1.23013897062152e-05, "loss": 0.2629, "step": 66560 }, { "epoch": 4.02, "learning_rate": 1.2299880796100977e-05, "loss": 0.2475, "step": 66562 }, { "epoch": 4.02, "learning_rate": 1.2298371885986752e-05, "loss": 0.1837, "step": 66564 }, { "epoch": 4.02, "learning_rate": 1.2296862975872529e-05, "loss": 0.2629, "step": 66566 }, { "epoch": 4.02, "learning_rate": 1.2295354065758303e-05, "loss": 0.2135, "step": 66568 }, { "epoch": 4.02, "learning_rate": 1.229384515564408e-05, "loss": 0.188, "step": 66570 }, { "epoch": 4.02, "learning_rate": 1.2292336245529855e-05, "loss": 0.1937, "step": 66572 }, { "epoch": 4.02, "learning_rate": 1.229082733541563e-05, "loss": 0.3108, "step": 66574 }, { "epoch": 4.02, "learning_rate": 1.2289318425301404e-05, "loss": 0.3433, "step": 66576 }, { "epoch": 4.02, "learning_rate": 1.2287809515187181e-05, "loss": 0.2109, "step": 66578 }, { "epoch": 4.02, "learning_rate": 1.2286300605072956e-05, "loss": 0.2839, "step": 66580 }, { "epoch": 4.02, "learning_rate": 1.2284791694958732e-05, "loss": 0.2584, "step": 66582 }, { "epoch": 4.02, "learning_rate": 1.2283282784844507e-05, "loss": 0.2419, "step": 66584 }, { "epoch": 4.02, "learning_rate": 1.2281773874730284e-05, "loss": 0.1999, "step": 66586 }, { "epoch": 4.02, "learning_rate": 1.2280264964616058e-05, "loss": 0.2175, "step": 66588 }, { "epoch": 4.02, "learning_rate": 1.2278756054501835e-05, "loss": 0.1581, "step": 66590 }, { "epoch": 4.02, "learning_rate": 1.2277247144387608e-05, "loss": 0.2596, "step": 66592 }, { "epoch": 4.02, "learning_rate": 1.2275738234273385e-05, "loss": 0.2772, "step": 66594 }, { "epoch": 4.02, "learning_rate": 1.227422932415916e-05, "loss": 0.3652, "step": 66596 }, { "epoch": 4.02, "learning_rate": 1.2272720414044936e-05, "loss": 0.2004, "step": 66598 }, { "epoch": 4.02, "learning_rate": 1.227121150393071e-05, "loss": 0.2077, "step": 66600 }, { "epoch": 4.02, "learning_rate": 1.2269702593816487e-05, "loss": 0.3324, "step": 66602 }, { "epoch": 4.02, "learning_rate": 1.2268193683702262e-05, "loss": 0.2634, "step": 66604 }, { "epoch": 4.02, "learning_rate": 1.2266684773588039e-05, "loss": 0.2261, "step": 66606 }, { "epoch": 4.02, "learning_rate": 1.2265175863473813e-05, "loss": 0.2144, "step": 66608 }, { "epoch": 4.02, "learning_rate": 1.2263666953359588e-05, "loss": 0.2751, "step": 66610 }, { "epoch": 4.02, "learning_rate": 1.2262158043245365e-05, "loss": 0.2362, "step": 66612 }, { "epoch": 4.02, "learning_rate": 1.226064913313114e-05, "loss": 0.3287, "step": 66614 }, { "epoch": 4.02, "learning_rate": 1.2259140223016916e-05, "loss": 0.1569, "step": 66616 }, { "epoch": 4.02, "learning_rate": 1.2257631312902691e-05, "loss": 0.3144, "step": 66618 }, { "epoch": 4.02, "learning_rate": 1.2256122402788467e-05, "loss": 0.1607, "step": 66620 }, { "epoch": 4.02, "learning_rate": 1.2254613492674242e-05, "loss": 0.3483, "step": 66622 }, { "epoch": 4.02, "learning_rate": 1.2253104582560017e-05, "loss": 0.2454, "step": 66624 }, { "epoch": 4.02, "learning_rate": 1.2251595672445792e-05, "loss": 0.241, "step": 66626 }, { "epoch": 4.02, "learning_rate": 1.2250086762331568e-05, "loss": 0.1906, "step": 66628 }, { "epoch": 4.02, "learning_rate": 1.2248577852217343e-05, "loss": 0.1985, "step": 66630 }, { "epoch": 4.02, "learning_rate": 1.224706894210312e-05, "loss": 0.2163, "step": 66632 }, { "epoch": 4.02, "learning_rate": 1.2245560031988895e-05, "loss": 0.2493, "step": 66634 }, { "epoch": 4.02, "learning_rate": 1.2244051121874671e-05, "loss": 0.2142, "step": 66636 }, { "epoch": 4.02, "learning_rate": 1.2242542211760446e-05, "loss": 0.2844, "step": 66638 }, { "epoch": 4.02, "learning_rate": 1.224103330164622e-05, "loss": 0.3005, "step": 66640 }, { "epoch": 4.02, "learning_rate": 1.2239524391531997e-05, "loss": 0.1803, "step": 66642 }, { "epoch": 4.02, "learning_rate": 1.2238015481417772e-05, "loss": 0.2474, "step": 66644 }, { "epoch": 4.02, "learning_rate": 1.2236506571303549e-05, "loss": 0.2359, "step": 66646 }, { "epoch": 4.02, "learning_rate": 1.2234997661189323e-05, "loss": 0.1917, "step": 66648 }, { "epoch": 4.02, "learning_rate": 1.22334887510751e-05, "loss": 0.2029, "step": 66650 }, { "epoch": 4.02, "learning_rate": 1.2231979840960875e-05, "loss": 0.3361, "step": 66652 }, { "epoch": 4.02, "learning_rate": 1.2230470930846651e-05, "loss": 0.2637, "step": 66654 }, { "epoch": 4.02, "learning_rate": 1.2228962020732424e-05, "loss": 0.2462, "step": 66656 }, { "epoch": 4.02, "learning_rate": 1.22274531106182e-05, "loss": 0.245, "step": 66658 }, { "epoch": 4.02, "learning_rate": 1.2225944200503976e-05, "loss": 0.2864, "step": 66660 }, { "epoch": 4.02, "learning_rate": 1.2224435290389752e-05, "loss": 0.3252, "step": 66662 }, { "epoch": 4.02, "learning_rate": 1.2222926380275527e-05, "loss": 0.1871, "step": 66664 }, { "epoch": 4.02, "learning_rate": 1.2221417470161304e-05, "loss": 0.264, "step": 66666 }, { "epoch": 4.02, "learning_rate": 1.2219908560047078e-05, "loss": 0.2116, "step": 66668 }, { "epoch": 4.02, "learning_rate": 1.2218399649932855e-05, "loss": 0.1478, "step": 66670 }, { "epoch": 4.02, "learning_rate": 1.221689073981863e-05, "loss": 0.2281, "step": 66672 }, { "epoch": 4.02, "learning_rate": 1.2215381829704404e-05, "loss": 0.2817, "step": 66674 }, { "epoch": 4.02, "learning_rate": 1.2213872919590181e-05, "loss": 0.2101, "step": 66676 }, { "epoch": 4.02, "learning_rate": 1.2212364009475956e-05, "loss": 0.184, "step": 66678 }, { "epoch": 4.02, "learning_rate": 1.2210855099361732e-05, "loss": 0.1437, "step": 66680 }, { "epoch": 4.02, "learning_rate": 1.2209346189247507e-05, "loss": 0.2157, "step": 66682 }, { "epoch": 4.02, "learning_rate": 1.2207837279133284e-05, "loss": 0.241, "step": 66684 }, { "epoch": 4.02, "learning_rate": 1.2206328369019058e-05, "loss": 0.2168, "step": 66686 }, { "epoch": 4.03, "learning_rate": 1.2204819458904835e-05, "loss": 0.1376, "step": 66688 }, { "epoch": 4.03, "learning_rate": 1.2203310548790608e-05, "loss": 0.1597, "step": 66690 }, { "epoch": 4.03, "learning_rate": 1.2201801638676385e-05, "loss": 0.1997, "step": 66692 }, { "epoch": 4.03, "learning_rate": 1.220029272856216e-05, "loss": 0.4212, "step": 66694 }, { "epoch": 4.03, "learning_rate": 1.2198783818447936e-05, "loss": 0.2604, "step": 66696 }, { "epoch": 4.03, "learning_rate": 1.219727490833371e-05, "loss": 0.2803, "step": 66698 }, { "epoch": 4.03, "learning_rate": 1.2195765998219487e-05, "loss": 0.1841, "step": 66700 }, { "epoch": 4.03, "learning_rate": 1.2194257088105262e-05, "loss": 0.2701, "step": 66702 }, { "epoch": 4.03, "learning_rate": 1.2192748177991039e-05, "loss": 0.235, "step": 66704 }, { "epoch": 4.03, "learning_rate": 1.2191239267876812e-05, "loss": 0.2248, "step": 66706 }, { "epoch": 4.03, "learning_rate": 1.2189730357762588e-05, "loss": 0.257, "step": 66708 }, { "epoch": 4.03, "learning_rate": 1.2188221447648363e-05, "loss": 0.2558, "step": 66710 }, { "epoch": 4.03, "learning_rate": 1.218671253753414e-05, "loss": 0.1802, "step": 66712 }, { "epoch": 4.03, "learning_rate": 1.2185203627419914e-05, "loss": 0.3332, "step": 66714 }, { "epoch": 4.03, "learning_rate": 1.2183694717305691e-05, "loss": 0.1992, "step": 66716 }, { "epoch": 4.03, "learning_rate": 1.2182185807191466e-05, "loss": 0.1732, "step": 66718 }, { "epoch": 4.03, "learning_rate": 1.2180676897077242e-05, "loss": 0.2443, "step": 66720 }, { "epoch": 4.03, "learning_rate": 1.2179167986963017e-05, "loss": 0.3838, "step": 66722 }, { "epoch": 4.03, "learning_rate": 1.2177659076848792e-05, "loss": 0.1654, "step": 66724 }, { "epoch": 4.03, "learning_rate": 1.2176150166734568e-05, "loss": 0.2198, "step": 66726 }, { "epoch": 4.03, "learning_rate": 1.2174641256620343e-05, "loss": 0.376, "step": 66728 }, { "epoch": 4.03, "learning_rate": 1.217313234650612e-05, "loss": 0.1764, "step": 66730 }, { "epoch": 4.03, "learning_rate": 1.2171623436391895e-05, "loss": 0.1697, "step": 66732 }, { "epoch": 4.03, "learning_rate": 1.2170114526277671e-05, "loss": 0.1531, "step": 66734 }, { "epoch": 4.03, "learning_rate": 1.2168605616163446e-05, "loss": 0.2441, "step": 66736 }, { "epoch": 4.03, "learning_rate": 1.2167096706049222e-05, "loss": 0.1991, "step": 66738 }, { "epoch": 4.03, "learning_rate": 1.2165587795934996e-05, "loss": 0.402, "step": 66740 }, { "epoch": 4.03, "learning_rate": 1.2164078885820772e-05, "loss": 0.2239, "step": 66742 }, { "epoch": 4.03, "learning_rate": 1.2162569975706547e-05, "loss": 0.216, "step": 66744 }, { "epoch": 4.03, "learning_rate": 1.2161061065592323e-05, "loss": 0.1772, "step": 66746 }, { "epoch": 4.03, "learning_rate": 1.2159552155478098e-05, "loss": 0.1843, "step": 66748 }, { "epoch": 4.03, "learning_rate": 1.2158043245363875e-05, "loss": 0.251, "step": 66750 }, { "epoch": 4.03, "learning_rate": 1.215653433524965e-05, "loss": 0.1878, "step": 66752 }, { "epoch": 4.03, "learning_rate": 1.2155025425135426e-05, "loss": 0.3178, "step": 66754 }, { "epoch": 4.03, "learning_rate": 1.2153516515021201e-05, "loss": 0.2913, "step": 66756 }, { "epoch": 4.03, "learning_rate": 1.2152007604906976e-05, "loss": 0.2388, "step": 66758 }, { "epoch": 4.03, "learning_rate": 1.2150498694792752e-05, "loss": 0.1702, "step": 66760 }, { "epoch": 4.03, "learning_rate": 1.2148989784678527e-05, "loss": 0.167, "step": 66762 }, { "epoch": 4.03, "learning_rate": 1.2147480874564304e-05, "loss": 0.3212, "step": 66764 }, { "epoch": 4.03, "learning_rate": 1.2145971964450078e-05, "loss": 0.1985, "step": 66766 }, { "epoch": 4.03, "learning_rate": 1.2144463054335855e-05, "loss": 0.2207, "step": 66768 }, { "epoch": 4.03, "learning_rate": 1.214295414422163e-05, "loss": 0.3123, "step": 66770 }, { "epoch": 4.03, "learning_rate": 1.2141445234107405e-05, "loss": 0.2761, "step": 66772 }, { "epoch": 4.03, "learning_rate": 1.213993632399318e-05, "loss": 0.2144, "step": 66774 }, { "epoch": 4.03, "learning_rate": 1.2138427413878956e-05, "loss": 0.2509, "step": 66776 }, { "epoch": 4.03, "learning_rate": 1.213691850376473e-05, "loss": 0.1925, "step": 66778 }, { "epoch": 4.03, "learning_rate": 1.2135409593650507e-05, "loss": 0.3128, "step": 66780 }, { "epoch": 4.03, "learning_rate": 1.2133900683536282e-05, "loss": 0.1623, "step": 66782 }, { "epoch": 4.03, "learning_rate": 1.2132391773422059e-05, "loss": 0.3106, "step": 66784 }, { "epoch": 4.03, "learning_rate": 1.2130882863307833e-05, "loss": 0.092, "step": 66786 }, { "epoch": 4.03, "learning_rate": 1.2129373953193608e-05, "loss": 0.1872, "step": 66788 }, { "epoch": 4.03, "learning_rate": 1.2127865043079385e-05, "loss": 0.1731, "step": 66790 }, { "epoch": 4.03, "learning_rate": 1.212635613296516e-05, "loss": 0.248, "step": 66792 }, { "epoch": 4.03, "learning_rate": 1.2124847222850936e-05, "loss": 0.2257, "step": 66794 }, { "epoch": 4.03, "learning_rate": 1.212333831273671e-05, "loss": 0.2203, "step": 66796 }, { "epoch": 4.03, "learning_rate": 1.2121829402622487e-05, "loss": 0.2467, "step": 66798 }, { "epoch": 4.03, "learning_rate": 1.2120320492508262e-05, "loss": 0.3116, "step": 66800 }, { "epoch": 4.03, "learning_rate": 1.2118811582394039e-05, "loss": 0.2679, "step": 66802 }, { "epoch": 4.03, "learning_rate": 1.2117302672279812e-05, "loss": 0.2934, "step": 66804 }, { "epoch": 4.03, "learning_rate": 1.2115793762165588e-05, "loss": 0.2931, "step": 66806 }, { "epoch": 4.03, "learning_rate": 1.2114284852051363e-05, "loss": 0.1483, "step": 66808 }, { "epoch": 4.03, "learning_rate": 1.211277594193714e-05, "loss": 0.2189, "step": 66810 }, { "epoch": 4.03, "learning_rate": 1.2111267031822914e-05, "loss": 0.1596, "step": 66812 }, { "epoch": 4.03, "learning_rate": 1.2109758121708691e-05, "loss": 0.1525, "step": 66814 }, { "epoch": 4.03, "learning_rate": 1.2108249211594466e-05, "loss": 0.3465, "step": 66816 }, { "epoch": 4.03, "learning_rate": 1.2106740301480242e-05, "loss": 0.1995, "step": 66818 }, { "epoch": 4.03, "learning_rate": 1.2105231391366017e-05, "loss": 0.3459, "step": 66820 }, { "epoch": 4.03, "learning_rate": 1.2103722481251792e-05, "loss": 0.2154, "step": 66822 }, { "epoch": 4.03, "learning_rate": 1.2102213571137567e-05, "loss": 0.2628, "step": 66824 }, { "epoch": 4.03, "learning_rate": 1.2100704661023343e-05, "loss": 0.1782, "step": 66826 }, { "epoch": 4.03, "learning_rate": 1.2099195750909118e-05, "loss": 0.2044, "step": 66828 }, { "epoch": 4.03, "learning_rate": 1.2097686840794895e-05, "loss": 0.2661, "step": 66830 }, { "epoch": 4.03, "learning_rate": 1.209617793068067e-05, "loss": 0.1989, "step": 66832 }, { "epoch": 4.03, "learning_rate": 1.2094669020566446e-05, "loss": 0.1553, "step": 66834 }, { "epoch": 4.03, "learning_rate": 1.209316011045222e-05, "loss": 0.2886, "step": 66836 }, { "epoch": 4.03, "learning_rate": 1.2091651200337996e-05, "loss": 0.2437, "step": 66838 }, { "epoch": 4.03, "learning_rate": 1.2090142290223772e-05, "loss": 0.2387, "step": 66840 }, { "epoch": 4.03, "learning_rate": 1.2088633380109547e-05, "loss": 0.3196, "step": 66842 }, { "epoch": 4.03, "learning_rate": 1.2087124469995323e-05, "loss": 0.2472, "step": 66844 }, { "epoch": 4.03, "learning_rate": 1.2085615559881098e-05, "loss": 0.3031, "step": 66846 }, { "epoch": 4.03, "learning_rate": 1.2084106649766875e-05, "loss": 0.192, "step": 66848 }, { "epoch": 4.03, "learning_rate": 1.208259773965265e-05, "loss": 0.2253, "step": 66850 }, { "epoch": 4.04, "learning_rate": 1.2081088829538426e-05, "loss": 0.2539, "step": 66852 }, { "epoch": 4.04, "learning_rate": 1.20795799194242e-05, "loss": 0.3276, "step": 66854 }, { "epoch": 4.04, "learning_rate": 1.2078071009309976e-05, "loss": 0.275, "step": 66856 }, { "epoch": 4.04, "learning_rate": 1.207656209919575e-05, "loss": 0.2727, "step": 66858 }, { "epoch": 4.04, "learning_rate": 1.2075053189081527e-05, "loss": 0.2078, "step": 66860 }, { "epoch": 4.04, "learning_rate": 1.2073544278967302e-05, "loss": 0.365, "step": 66862 }, { "epoch": 4.04, "learning_rate": 1.2072035368853078e-05, "loss": 0.2396, "step": 66864 }, { "epoch": 4.04, "learning_rate": 1.2070526458738853e-05, "loss": 0.3074, "step": 66866 }, { "epoch": 4.04, "learning_rate": 1.206901754862463e-05, "loss": 0.1198, "step": 66868 }, { "epoch": 4.04, "learning_rate": 1.2067508638510405e-05, "loss": 0.2317, "step": 66870 }, { "epoch": 4.04, "learning_rate": 1.206599972839618e-05, "loss": 0.3015, "step": 66872 }, { "epoch": 4.04, "learning_rate": 1.2064490818281956e-05, "loss": 0.266, "step": 66874 }, { "epoch": 4.04, "learning_rate": 1.206298190816773e-05, "loss": 0.2564, "step": 66876 }, { "epoch": 4.04, "learning_rate": 1.2061472998053507e-05, "loss": 0.2417, "step": 66878 }, { "epoch": 4.04, "learning_rate": 1.2059964087939282e-05, "loss": 0.1737, "step": 66880 }, { "epoch": 4.04, "learning_rate": 1.2058455177825059e-05, "loss": 0.2278, "step": 66882 }, { "epoch": 4.04, "learning_rate": 1.2056946267710833e-05, "loss": 0.175, "step": 66884 }, { "epoch": 4.04, "learning_rate": 1.2055437357596608e-05, "loss": 0.2578, "step": 66886 }, { "epoch": 4.04, "learning_rate": 1.2053928447482383e-05, "loss": 0.2099, "step": 66888 }, { "epoch": 4.04, "learning_rate": 1.205241953736816e-05, "loss": 0.2077, "step": 66890 }, { "epoch": 4.04, "learning_rate": 1.2050910627253934e-05, "loss": 0.2519, "step": 66892 }, { "epoch": 4.04, "learning_rate": 1.2049401717139711e-05, "loss": 0.3739, "step": 66894 }, { "epoch": 4.04, "learning_rate": 1.2047892807025486e-05, "loss": 0.1337, "step": 66896 }, { "epoch": 4.04, "learning_rate": 1.2046383896911262e-05, "loss": 0.1779, "step": 66898 }, { "epoch": 4.04, "learning_rate": 1.2044874986797037e-05, "loss": 0.19, "step": 66900 }, { "epoch": 4.04, "learning_rate": 1.2043366076682814e-05, "loss": 0.2555, "step": 66902 }, { "epoch": 4.04, "learning_rate": 1.2041857166568588e-05, "loss": 0.2271, "step": 66904 }, { "epoch": 4.04, "learning_rate": 1.2040348256454363e-05, "loss": 0.2606, "step": 66906 }, { "epoch": 4.04, "learning_rate": 1.203883934634014e-05, "loss": 0.2285, "step": 66908 }, { "epoch": 4.04, "learning_rate": 1.2037330436225914e-05, "loss": 0.2728, "step": 66910 }, { "epoch": 4.04, "learning_rate": 1.2035821526111691e-05, "loss": 0.323, "step": 66912 }, { "epoch": 4.04, "learning_rate": 1.2034312615997466e-05, "loss": 0.1679, "step": 66914 }, { "epoch": 4.04, "learning_rate": 1.2032803705883242e-05, "loss": 0.3657, "step": 66916 }, { "epoch": 4.04, "learning_rate": 1.2031294795769017e-05, "loss": 0.2736, "step": 66918 }, { "epoch": 4.04, "learning_rate": 1.2029785885654792e-05, "loss": 0.2947, "step": 66920 }, { "epoch": 4.04, "learning_rate": 1.2028276975540567e-05, "loss": 0.2222, "step": 66922 }, { "epoch": 4.04, "learning_rate": 1.2026768065426343e-05, "loss": 0.2536, "step": 66924 }, { "epoch": 4.04, "learning_rate": 1.2025259155312118e-05, "loss": 0.3059, "step": 66926 }, { "epoch": 4.04, "learning_rate": 1.2023750245197895e-05, "loss": 0.2825, "step": 66928 }, { "epoch": 4.04, "learning_rate": 1.202224133508367e-05, "loss": 0.381, "step": 66930 }, { "epoch": 4.04, "learning_rate": 1.2020732424969446e-05, "loss": 0.2254, "step": 66932 }, { "epoch": 4.04, "learning_rate": 1.201922351485522e-05, "loss": 0.2235, "step": 66934 }, { "epoch": 4.04, "learning_rate": 1.2017714604740996e-05, "loss": 0.1665, "step": 66936 }, { "epoch": 4.04, "learning_rate": 1.201620569462677e-05, "loss": 0.2821, "step": 66938 }, { "epoch": 4.04, "learning_rate": 1.2014696784512547e-05, "loss": 0.3189, "step": 66940 }, { "epoch": 4.04, "learning_rate": 1.2013187874398322e-05, "loss": 0.2979, "step": 66942 }, { "epoch": 4.04, "learning_rate": 1.2011678964284098e-05, "loss": 0.264, "step": 66944 }, { "epoch": 4.04, "learning_rate": 1.2010170054169873e-05, "loss": 0.2094, "step": 66946 }, { "epoch": 4.04, "learning_rate": 1.200866114405565e-05, "loss": 0.2788, "step": 66948 }, { "epoch": 4.04, "learning_rate": 1.2007152233941424e-05, "loss": 0.3105, "step": 66950 }, { "epoch": 4.04, "learning_rate": 1.20056433238272e-05, "loss": 0.2373, "step": 66952 }, { "epoch": 4.04, "learning_rate": 1.2004134413712976e-05, "loss": 0.1988, "step": 66954 }, { "epoch": 4.04, "learning_rate": 1.200262550359875e-05, "loss": 0.2018, "step": 66956 }, { "epoch": 4.04, "learning_rate": 1.2001116593484527e-05, "loss": 0.3148, "step": 66958 }, { "epoch": 4.04, "learning_rate": 1.1999607683370302e-05, "loss": 0.1135, "step": 66960 }, { "epoch": 4.04, "learning_rate": 1.1998098773256078e-05, "loss": 0.2282, "step": 66962 }, { "epoch": 4.04, "learning_rate": 1.1996589863141853e-05, "loss": 0.2064, "step": 66964 }, { "epoch": 4.04, "learning_rate": 1.199508095302763e-05, "loss": 0.163, "step": 66966 }, { "epoch": 4.04, "learning_rate": 1.1993572042913405e-05, "loss": 0.2149, "step": 66968 }, { "epoch": 4.04, "learning_rate": 1.199206313279918e-05, "loss": 0.183, "step": 66970 }, { "epoch": 4.04, "learning_rate": 1.1990554222684954e-05, "loss": 0.2238, "step": 66972 }, { "epoch": 4.04, "learning_rate": 1.198904531257073e-05, "loss": 0.3914, "step": 66974 }, { "epoch": 4.04, "learning_rate": 1.1987536402456506e-05, "loss": 0.2378, "step": 66976 }, { "epoch": 4.04, "learning_rate": 1.1986027492342282e-05, "loss": 0.2588, "step": 66978 }, { "epoch": 4.04, "learning_rate": 1.1984518582228057e-05, "loss": 0.2398, "step": 66980 }, { "epoch": 4.04, "learning_rate": 1.1983009672113833e-05, "loss": 0.2208, "step": 66982 }, { "epoch": 4.04, "learning_rate": 1.1981500761999608e-05, "loss": 0.234, "step": 66984 }, { "epoch": 4.04, "learning_rate": 1.1979991851885383e-05, "loss": 0.118, "step": 66986 }, { "epoch": 4.04, "learning_rate": 1.197848294177116e-05, "loss": 0.1479, "step": 66988 }, { "epoch": 4.04, "learning_rate": 1.1976974031656934e-05, "loss": 0.2, "step": 66990 }, { "epoch": 4.04, "learning_rate": 1.1975465121542711e-05, "loss": 0.1926, "step": 66992 }, { "epoch": 4.04, "learning_rate": 1.1973956211428486e-05, "loss": 0.3346, "step": 66994 }, { "epoch": 4.04, "learning_rate": 1.1972447301314262e-05, "loss": 0.2337, "step": 66996 }, { "epoch": 4.04, "learning_rate": 1.1970938391200037e-05, "loss": 0.2787, "step": 66998 }, { "epoch": 4.04, "learning_rate": 1.1969429481085814e-05, "loss": 0.1867, "step": 67000 }, { "epoch": 4.04, "learning_rate": 1.1967920570971587e-05, "loss": 0.1531, "step": 67002 }, { "epoch": 4.04, "learning_rate": 1.1966411660857363e-05, "loss": 0.1756, "step": 67004 }, { "epoch": 4.04, "learning_rate": 1.1964902750743138e-05, "loss": 0.2014, "step": 67006 }, { "epoch": 4.04, "learning_rate": 1.1963393840628915e-05, "loss": 0.1718, "step": 67008 }, { "epoch": 4.04, "learning_rate": 1.196188493051469e-05, "loss": 0.2699, "step": 67010 }, { "epoch": 4.04, "learning_rate": 1.1960376020400466e-05, "loss": 0.2716, "step": 67012 }, { "epoch": 4.04, "learning_rate": 1.195886711028624e-05, "loss": 0.2, "step": 67014 }, { "epoch": 4.04, "learning_rate": 1.1957358200172017e-05, "loss": 0.2951, "step": 67016 }, { "epoch": 4.05, "learning_rate": 1.1955849290057792e-05, "loss": 0.2737, "step": 67018 }, { "epoch": 4.05, "learning_rate": 1.1954340379943567e-05, "loss": 0.1553, "step": 67020 }, { "epoch": 4.05, "learning_rate": 1.1952831469829343e-05, "loss": 0.2506, "step": 67022 }, { "epoch": 4.05, "learning_rate": 1.1951322559715118e-05, "loss": 0.2138, "step": 67024 }, { "epoch": 4.05, "learning_rate": 1.1949813649600895e-05, "loss": 0.2969, "step": 67026 }, { "epoch": 4.05, "learning_rate": 1.194830473948667e-05, "loss": 0.2423, "step": 67028 }, { "epoch": 4.05, "learning_rate": 1.1946795829372446e-05, "loss": 0.208, "step": 67030 }, { "epoch": 4.05, "learning_rate": 1.194528691925822e-05, "loss": 0.2748, "step": 67032 }, { "epoch": 4.05, "learning_rate": 1.1943778009143996e-05, "loss": 0.2591, "step": 67034 }, { "epoch": 4.05, "learning_rate": 1.194226909902977e-05, "loss": 0.2384, "step": 67036 }, { "epoch": 4.05, "learning_rate": 1.1940760188915547e-05, "loss": 0.3, "step": 67038 }, { "epoch": 4.05, "learning_rate": 1.1939251278801322e-05, "loss": 0.2364, "step": 67040 }, { "epoch": 4.05, "learning_rate": 1.1937742368687098e-05, "loss": 0.2564, "step": 67042 }, { "epoch": 4.05, "learning_rate": 1.1936233458572873e-05, "loss": 0.2177, "step": 67044 }, { "epoch": 4.05, "learning_rate": 1.193472454845865e-05, "loss": 0.2414, "step": 67046 }, { "epoch": 4.05, "learning_rate": 1.1933215638344424e-05, "loss": 0.1677, "step": 67048 }, { "epoch": 4.05, "learning_rate": 1.1931706728230201e-05, "loss": 0.2679, "step": 67050 }, { "epoch": 4.05, "learning_rate": 1.1930197818115974e-05, "loss": 0.1601, "step": 67052 }, { "epoch": 4.05, "learning_rate": 1.192868890800175e-05, "loss": 0.1745, "step": 67054 }, { "epoch": 4.05, "learning_rate": 1.1927179997887525e-05, "loss": 0.1916, "step": 67056 }, { "epoch": 4.05, "learning_rate": 1.1925671087773302e-05, "loss": 0.2337, "step": 67058 }, { "epoch": 4.05, "learning_rate": 1.1924162177659077e-05, "loss": 0.1538, "step": 67060 }, { "epoch": 4.05, "learning_rate": 1.1922653267544853e-05, "loss": 0.1774, "step": 67062 }, { "epoch": 4.05, "learning_rate": 1.1921144357430628e-05, "loss": 0.2189, "step": 67064 }, { "epoch": 4.05, "learning_rate": 1.1919635447316405e-05, "loss": 0.3119, "step": 67066 }, { "epoch": 4.05, "learning_rate": 1.191812653720218e-05, "loss": 0.3273, "step": 67068 }, { "epoch": 4.05, "learning_rate": 1.1916617627087954e-05, "loss": 0.2103, "step": 67070 }, { "epoch": 4.05, "learning_rate": 1.191510871697373e-05, "loss": 0.2948, "step": 67072 }, { "epoch": 4.05, "learning_rate": 1.1913599806859506e-05, "loss": 0.235, "step": 67074 }, { "epoch": 4.05, "learning_rate": 1.1912090896745282e-05, "loss": 0.2717, "step": 67076 }, { "epoch": 4.05, "learning_rate": 1.1910581986631057e-05, "loss": 0.2048, "step": 67078 }, { "epoch": 4.05, "learning_rate": 1.1909073076516833e-05, "loss": 0.2643, "step": 67080 }, { "epoch": 4.05, "learning_rate": 1.1907564166402608e-05, "loss": 0.2146, "step": 67082 }, { "epoch": 4.05, "learning_rate": 1.1906055256288383e-05, "loss": 0.2087, "step": 67084 }, { "epoch": 4.05, "learning_rate": 1.1904546346174158e-05, "loss": 0.439, "step": 67086 }, { "epoch": 4.05, "learning_rate": 1.1903037436059934e-05, "loss": 0.267, "step": 67088 }, { "epoch": 4.05, "learning_rate": 1.190152852594571e-05, "loss": 0.2275, "step": 67090 }, { "epoch": 4.05, "learning_rate": 1.1900019615831486e-05, "loss": 0.2611, "step": 67092 }, { "epoch": 4.05, "learning_rate": 1.189851070571726e-05, "loss": 0.2083, "step": 67094 }, { "epoch": 4.05, "learning_rate": 1.1897001795603037e-05, "loss": 0.2534, "step": 67096 }, { "epoch": 4.05, "learning_rate": 1.1895492885488812e-05, "loss": 0.2102, "step": 67098 }, { "epoch": 4.05, "learning_rate": 1.1893983975374587e-05, "loss": 0.1837, "step": 67100 }, { "epoch": 4.05, "learning_rate": 1.1892475065260363e-05, "loss": 0.4866, "step": 67102 }, { "epoch": 4.05, "learning_rate": 1.1890966155146138e-05, "loss": 0.1462, "step": 67104 }, { "epoch": 4.05, "learning_rate": 1.1889457245031915e-05, "loss": 0.2725, "step": 67106 }, { "epoch": 4.05, "learning_rate": 1.188794833491769e-05, "loss": 0.2468, "step": 67108 }, { "epoch": 4.05, "learning_rate": 1.1886439424803466e-05, "loss": 0.2286, "step": 67110 }, { "epoch": 4.05, "learning_rate": 1.188493051468924e-05, "loss": 0.2441, "step": 67112 }, { "epoch": 4.05, "learning_rate": 1.1883421604575017e-05, "loss": 0.2099, "step": 67114 }, { "epoch": 4.05, "learning_rate": 1.188191269446079e-05, "loss": 0.2435, "step": 67116 }, { "epoch": 4.05, "learning_rate": 1.1880403784346567e-05, "loss": 0.2487, "step": 67118 }, { "epoch": 4.05, "learning_rate": 1.1878894874232342e-05, "loss": 0.2732, "step": 67120 }, { "epoch": 4.05, "learning_rate": 1.1877385964118118e-05, "loss": 0.185, "step": 67122 }, { "epoch": 4.05, "learning_rate": 1.1875877054003893e-05, "loss": 0.3782, "step": 67124 }, { "epoch": 4.05, "learning_rate": 1.187436814388967e-05, "loss": 0.2003, "step": 67126 }, { "epoch": 4.05, "learning_rate": 1.1872859233775444e-05, "loss": 0.2013, "step": 67128 }, { "epoch": 4.05, "learning_rate": 1.1871350323661221e-05, "loss": 0.3532, "step": 67130 }, { "epoch": 4.05, "learning_rate": 1.1869841413546996e-05, "loss": 0.1814, "step": 67132 }, { "epoch": 4.05, "learning_rate": 1.186833250343277e-05, "loss": 0.3632, "step": 67134 }, { "epoch": 4.05, "learning_rate": 1.1866823593318547e-05, "loss": 0.16, "step": 67136 }, { "epoch": 4.05, "learning_rate": 1.1865314683204322e-05, "loss": 0.2089, "step": 67138 }, { "epoch": 4.05, "learning_rate": 1.1863805773090098e-05, "loss": 0.1919, "step": 67140 }, { "epoch": 4.05, "learning_rate": 1.1862296862975873e-05, "loss": 0.2116, "step": 67142 }, { "epoch": 4.05, "learning_rate": 1.186078795286165e-05, "loss": 0.1791, "step": 67144 }, { "epoch": 4.05, "learning_rate": 1.1859279042747425e-05, "loss": 0.2795, "step": 67146 }, { "epoch": 4.05, "learning_rate": 1.1857770132633201e-05, "loss": 0.2999, "step": 67148 }, { "epoch": 4.05, "learning_rate": 1.1856261222518974e-05, "loss": 0.2465, "step": 67150 }, { "epoch": 4.05, "learning_rate": 1.185475231240475e-05, "loss": 0.2916, "step": 67152 }, { "epoch": 4.05, "learning_rate": 1.1853243402290525e-05, "loss": 0.2994, "step": 67154 }, { "epoch": 4.05, "learning_rate": 1.1851734492176302e-05, "loss": 0.2135, "step": 67156 }, { "epoch": 4.05, "learning_rate": 1.1850225582062077e-05, "loss": 0.1952, "step": 67158 }, { "epoch": 4.05, "learning_rate": 1.1848716671947853e-05, "loss": 0.1969, "step": 67160 }, { "epoch": 4.05, "learning_rate": 1.1847207761833628e-05, "loss": 0.3182, "step": 67162 }, { "epoch": 4.05, "learning_rate": 1.1845698851719405e-05, "loss": 0.2837, "step": 67164 }, { "epoch": 4.05, "learning_rate": 1.1844189941605178e-05, "loss": 0.1648, "step": 67166 }, { "epoch": 4.05, "learning_rate": 1.1842681031490954e-05, "loss": 0.2441, "step": 67168 }, { "epoch": 4.05, "learning_rate": 1.1841172121376729e-05, "loss": 0.3206, "step": 67170 }, { "epoch": 4.05, "learning_rate": 1.1839663211262506e-05, "loss": 0.2339, "step": 67172 }, { "epoch": 4.05, "learning_rate": 1.183815430114828e-05, "loss": 0.2475, "step": 67174 }, { "epoch": 4.05, "learning_rate": 1.1836645391034057e-05, "loss": 0.1975, "step": 67176 }, { "epoch": 4.05, "learning_rate": 1.1835136480919832e-05, "loss": 0.1545, "step": 67178 }, { "epoch": 4.05, "learning_rate": 1.1833627570805608e-05, "loss": 0.2007, "step": 67180 }, { "epoch": 4.05, "learning_rate": 1.1832118660691383e-05, "loss": 0.2864, "step": 67182 }, { "epoch": 4.06, "learning_rate": 1.1830609750577158e-05, "loss": 0.1659, "step": 67184 }, { "epoch": 4.06, "learning_rate": 1.1829100840462934e-05, "loss": 0.4743, "step": 67186 }, { "epoch": 4.06, "learning_rate": 1.182759193034871e-05, "loss": 0.3581, "step": 67188 }, { "epoch": 4.06, "learning_rate": 1.1826083020234486e-05, "loss": 0.1873, "step": 67190 }, { "epoch": 4.06, "learning_rate": 1.182457411012026e-05, "loss": 0.177, "step": 67192 }, { "epoch": 4.06, "learning_rate": 1.1823065200006037e-05, "loss": 0.3433, "step": 67194 }, { "epoch": 4.06, "learning_rate": 1.1821556289891812e-05, "loss": 0.2414, "step": 67196 }, { "epoch": 4.06, "learning_rate": 1.1820047379777587e-05, "loss": 0.2256, "step": 67198 }, { "epoch": 4.06, "learning_rate": 1.1818538469663362e-05, "loss": 0.1719, "step": 67200 }, { "epoch": 4.06, "learning_rate": 1.1817029559549138e-05, "loss": 0.2975, "step": 67202 }, { "epoch": 4.06, "learning_rate": 1.1815520649434913e-05, "loss": 0.2293, "step": 67204 }, { "epoch": 4.06, "learning_rate": 1.181401173932069e-05, "loss": 0.1919, "step": 67206 }, { "epoch": 4.06, "learning_rate": 1.1812502829206464e-05, "loss": 0.2859, "step": 67208 }, { "epoch": 4.06, "learning_rate": 1.181099391909224e-05, "loss": 0.1399, "step": 67210 }, { "epoch": 4.06, "learning_rate": 1.1809485008978016e-05, "loss": 0.1901, "step": 67212 }, { "epoch": 4.06, "learning_rate": 1.1807976098863792e-05, "loss": 0.2492, "step": 67214 }, { "epoch": 4.06, "learning_rate": 1.1806467188749567e-05, "loss": 0.2725, "step": 67216 }, { "epoch": 4.06, "learning_rate": 1.1804958278635342e-05, "loss": 0.2086, "step": 67218 }, { "epoch": 4.06, "learning_rate": 1.1803449368521118e-05, "loss": 0.288, "step": 67220 }, { "epoch": 4.06, "learning_rate": 1.1801940458406893e-05, "loss": 0.2646, "step": 67222 }, { "epoch": 4.06, "learning_rate": 1.180043154829267e-05, "loss": 0.358, "step": 67224 }, { "epoch": 4.06, "learning_rate": 1.1798922638178444e-05, "loss": 0.2667, "step": 67226 }, { "epoch": 4.06, "learning_rate": 1.1797413728064221e-05, "loss": 0.1886, "step": 67228 }, { "epoch": 4.06, "learning_rate": 1.1795904817949996e-05, "loss": 0.214, "step": 67230 }, { "epoch": 4.06, "learning_rate": 1.179439590783577e-05, "loss": 0.1764, "step": 67232 }, { "epoch": 4.06, "learning_rate": 1.1792886997721545e-05, "loss": 0.1978, "step": 67234 }, { "epoch": 4.06, "learning_rate": 1.1791378087607322e-05, "loss": 0.3293, "step": 67236 }, { "epoch": 4.06, "learning_rate": 1.1789869177493097e-05, "loss": 0.2193, "step": 67238 }, { "epoch": 4.06, "learning_rate": 1.1788360267378873e-05, "loss": 0.2867, "step": 67240 }, { "epoch": 4.06, "learning_rate": 1.1786851357264648e-05, "loss": 0.1977, "step": 67242 }, { "epoch": 4.06, "learning_rate": 1.1785342447150425e-05, "loss": 0.1472, "step": 67244 }, { "epoch": 4.06, "learning_rate": 1.17838335370362e-05, "loss": 0.2378, "step": 67246 }, { "epoch": 4.06, "learning_rate": 1.1782324626921974e-05, "loss": 0.166, "step": 67248 }, { "epoch": 4.06, "learning_rate": 1.178081571680775e-05, "loss": 0.2147, "step": 67250 }, { "epoch": 4.06, "learning_rate": 1.1779306806693526e-05, "loss": 0.2882, "step": 67252 }, { "epoch": 4.06, "learning_rate": 1.1777797896579302e-05, "loss": 0.2406, "step": 67254 }, { "epoch": 4.06, "learning_rate": 1.1776288986465077e-05, "loss": 0.2291, "step": 67256 }, { "epoch": 4.06, "learning_rate": 1.1774780076350853e-05, "loss": 0.2728, "step": 67258 }, { "epoch": 4.06, "learning_rate": 1.1773271166236628e-05, "loss": 0.1815, "step": 67260 }, { "epoch": 4.06, "learning_rate": 1.1771762256122405e-05, "loss": 0.2011, "step": 67262 }, { "epoch": 4.06, "learning_rate": 1.1770253346008178e-05, "loss": 0.3019, "step": 67264 }, { "epoch": 4.06, "learning_rate": 1.1768744435893954e-05, "loss": 0.2536, "step": 67266 }, { "epoch": 4.06, "learning_rate": 1.176723552577973e-05, "loss": 0.3323, "step": 67268 }, { "epoch": 4.06, "learning_rate": 1.1765726615665506e-05, "loss": 0.2071, "step": 67270 }, { "epoch": 4.06, "learning_rate": 1.176421770555128e-05, "loss": 0.1954, "step": 67272 }, { "epoch": 4.06, "learning_rate": 1.1762708795437057e-05, "loss": 0.2462, "step": 67274 }, { "epoch": 4.06, "learning_rate": 1.1761199885322832e-05, "loss": 0.1826, "step": 67276 }, { "epoch": 4.06, "learning_rate": 1.1759690975208608e-05, "loss": 0.2887, "step": 67278 }, { "epoch": 4.06, "learning_rate": 1.1758182065094383e-05, "loss": 0.1928, "step": 67280 }, { "epoch": 4.06, "learning_rate": 1.1756673154980158e-05, "loss": 0.2112, "step": 67282 }, { "epoch": 4.06, "learning_rate": 1.1755164244865933e-05, "loss": 0.1493, "step": 67284 }, { "epoch": 4.06, "learning_rate": 1.175365533475171e-05, "loss": 0.2303, "step": 67286 }, { "epoch": 4.06, "learning_rate": 1.1752146424637484e-05, "loss": 0.184, "step": 67288 }, { "epoch": 4.06, "learning_rate": 1.175063751452326e-05, "loss": 0.1852, "step": 67290 }, { "epoch": 4.06, "learning_rate": 1.1749128604409035e-05, "loss": 0.3675, "step": 67292 }, { "epoch": 4.06, "learning_rate": 1.1747619694294812e-05, "loss": 0.2598, "step": 67294 }, { "epoch": 4.06, "learning_rate": 1.1746110784180587e-05, "loss": 0.2162, "step": 67296 }, { "epoch": 4.06, "learning_rate": 1.1744601874066362e-05, "loss": 0.2096, "step": 67298 }, { "epoch": 4.06, "learning_rate": 1.1743092963952138e-05, "loss": 0.3843, "step": 67300 }, { "epoch": 4.06, "learning_rate": 1.1741584053837913e-05, "loss": 0.2572, "step": 67302 }, { "epoch": 4.06, "learning_rate": 1.174007514372369e-05, "loss": 0.2063, "step": 67304 }, { "epoch": 4.06, "learning_rate": 1.1738566233609464e-05, "loss": 0.1802, "step": 67306 }, { "epoch": 4.06, "learning_rate": 1.173705732349524e-05, "loss": 0.2285, "step": 67308 }, { "epoch": 4.06, "learning_rate": 1.1735548413381016e-05, "loss": 0.2914, "step": 67310 }, { "epoch": 4.06, "learning_rate": 1.1734039503266792e-05, "loss": 0.2474, "step": 67312 }, { "epoch": 4.06, "learning_rate": 1.1732530593152565e-05, "loss": 0.2522, "step": 67314 }, { "epoch": 4.06, "learning_rate": 1.1731021683038342e-05, "loss": 0.2439, "step": 67316 }, { "epoch": 4.06, "learning_rate": 1.1729512772924117e-05, "loss": 0.3023, "step": 67318 }, { "epoch": 4.06, "learning_rate": 1.1728003862809893e-05, "loss": 0.1755, "step": 67320 }, { "epoch": 4.06, "learning_rate": 1.1726494952695668e-05, "loss": 0.2778, "step": 67322 }, { "epoch": 4.06, "learning_rate": 1.1724986042581444e-05, "loss": 0.2318, "step": 67324 }, { "epoch": 4.06, "learning_rate": 1.172347713246722e-05, "loss": 0.1839, "step": 67326 }, { "epoch": 4.06, "learning_rate": 1.1721968222352996e-05, "loss": 0.1531, "step": 67328 }, { "epoch": 4.06, "learning_rate": 1.172045931223877e-05, "loss": 0.1973, "step": 67330 }, { "epoch": 4.06, "learning_rate": 1.1718950402124545e-05, "loss": 0.2713, "step": 67332 }, { "epoch": 4.06, "learning_rate": 1.1717441492010322e-05, "loss": 0.2842, "step": 67334 }, { "epoch": 4.06, "learning_rate": 1.1715932581896097e-05, "loss": 0.2684, "step": 67336 }, { "epoch": 4.06, "learning_rate": 1.1714423671781873e-05, "loss": 0.2568, "step": 67338 }, { "epoch": 4.06, "learning_rate": 1.1712914761667648e-05, "loss": 0.2068, "step": 67340 }, { "epoch": 4.06, "learning_rate": 1.1711405851553425e-05, "loss": 0.262, "step": 67342 }, { "epoch": 4.06, "learning_rate": 1.17098969414392e-05, "loss": 0.3395, "step": 67344 }, { "epoch": 4.06, "learning_rate": 1.1708388031324974e-05, "loss": 0.2262, "step": 67346 }, { "epoch": 4.06, "learning_rate": 1.1706879121210749e-05, "loss": 0.2341, "step": 67348 }, { "epoch": 4.07, "learning_rate": 1.1705370211096526e-05, "loss": 0.2691, "step": 67350 }, { "epoch": 4.07, "learning_rate": 1.17038613009823e-05, "loss": 0.2616, "step": 67352 }, { "epoch": 4.07, "learning_rate": 1.1702352390868077e-05, "loss": 0.245, "step": 67354 }, { "epoch": 4.07, "learning_rate": 1.1700843480753852e-05, "loss": 0.2867, "step": 67356 }, { "epoch": 4.07, "learning_rate": 1.1699334570639628e-05, "loss": 0.2254, "step": 67358 }, { "epoch": 4.07, "learning_rate": 1.1697825660525403e-05, "loss": 0.3054, "step": 67360 }, { "epoch": 4.07, "learning_rate": 1.169631675041118e-05, "loss": 0.2022, "step": 67362 }, { "epoch": 4.07, "learning_rate": 1.1694807840296953e-05, "loss": 0.219, "step": 67364 }, { "epoch": 4.07, "learning_rate": 1.169329893018273e-05, "loss": 0.2592, "step": 67366 }, { "epoch": 4.07, "learning_rate": 1.1691790020068506e-05, "loss": 0.2272, "step": 67368 }, { "epoch": 4.07, "learning_rate": 1.169028110995428e-05, "loss": 0.2046, "step": 67370 }, { "epoch": 4.07, "learning_rate": 1.1688772199840057e-05, "loss": 0.2905, "step": 67372 }, { "epoch": 4.07, "learning_rate": 1.1687263289725832e-05, "loss": 0.2442, "step": 67374 }, { "epoch": 4.07, "learning_rate": 1.1685754379611608e-05, "loss": 0.1883, "step": 67376 }, { "epoch": 4.07, "learning_rate": 1.1684245469497383e-05, "loss": 0.2195, "step": 67378 }, { "epoch": 4.07, "learning_rate": 1.1682736559383158e-05, "loss": 0.2896, "step": 67380 }, { "epoch": 4.07, "learning_rate": 1.1681227649268933e-05, "loss": 0.1821, "step": 67382 }, { "epoch": 4.07, "learning_rate": 1.167971873915471e-05, "loss": 0.1915, "step": 67384 }, { "epoch": 4.07, "learning_rate": 1.1678209829040484e-05, "loss": 0.3964, "step": 67386 }, { "epoch": 4.07, "learning_rate": 1.167670091892626e-05, "loss": 0.3266, "step": 67388 }, { "epoch": 4.07, "learning_rate": 1.1675192008812036e-05, "loss": 0.2094, "step": 67390 }, { "epoch": 4.07, "learning_rate": 1.1673683098697812e-05, "loss": 0.2725, "step": 67392 }, { "epoch": 4.07, "learning_rate": 1.1672174188583587e-05, "loss": 0.1759, "step": 67394 }, { "epoch": 4.07, "learning_rate": 1.1670665278469362e-05, "loss": 0.1975, "step": 67396 }, { "epoch": 4.07, "learning_rate": 1.1669156368355136e-05, "loss": 0.1879, "step": 67398 }, { "epoch": 4.07, "learning_rate": 1.1667647458240913e-05, "loss": 0.3703, "step": 67400 }, { "epoch": 4.07, "learning_rate": 1.1666138548126688e-05, "loss": 0.2403, "step": 67402 }, { "epoch": 4.07, "learning_rate": 1.1664629638012464e-05, "loss": 0.3055, "step": 67404 }, { "epoch": 4.07, "learning_rate": 1.1663120727898239e-05, "loss": 0.1723, "step": 67406 }, { "epoch": 4.07, "learning_rate": 1.1661611817784016e-05, "loss": 0.335, "step": 67408 }, { "epoch": 4.07, "learning_rate": 1.166010290766979e-05, "loss": 0.2174, "step": 67410 }, { "epoch": 4.07, "learning_rate": 1.1658593997555565e-05, "loss": 0.2877, "step": 67412 }, { "epoch": 4.07, "learning_rate": 1.1657085087441342e-05, "loss": 0.3601, "step": 67414 }, { "epoch": 4.07, "learning_rate": 1.1655576177327117e-05, "loss": 0.2239, "step": 67416 }, { "epoch": 4.07, "learning_rate": 1.1654067267212893e-05, "loss": 0.318, "step": 67418 }, { "epoch": 4.07, "learning_rate": 1.1652558357098668e-05, "loss": 0.208, "step": 67420 }, { "epoch": 4.07, "learning_rate": 1.1651049446984444e-05, "loss": 0.2531, "step": 67422 }, { "epoch": 4.07, "learning_rate": 1.164954053687022e-05, "loss": 0.2382, "step": 67424 }, { "epoch": 4.07, "learning_rate": 1.1648031626755996e-05, "loss": 0.4201, "step": 67426 }, { "epoch": 4.07, "learning_rate": 1.1646522716641769e-05, "loss": 0.2489, "step": 67428 }, { "epoch": 4.07, "learning_rate": 1.1645013806527545e-05, "loss": 0.2745, "step": 67430 }, { "epoch": 4.07, "learning_rate": 1.164350489641332e-05, "loss": 0.3183, "step": 67432 }, { "epoch": 4.07, "learning_rate": 1.1641995986299097e-05, "loss": 0.2139, "step": 67434 }, { "epoch": 4.07, "learning_rate": 1.1640487076184872e-05, "loss": 0.2118, "step": 67436 }, { "epoch": 4.07, "learning_rate": 1.1638978166070648e-05, "loss": 0.1531, "step": 67438 }, { "epoch": 4.07, "learning_rate": 1.1637469255956423e-05, "loss": 0.2005, "step": 67440 }, { "epoch": 4.07, "learning_rate": 1.16359603458422e-05, "loss": 0.3318, "step": 67442 }, { "epoch": 4.07, "learning_rate": 1.1634451435727974e-05, "loss": 0.368, "step": 67444 }, { "epoch": 4.07, "learning_rate": 1.1632942525613749e-05, "loss": 0.2052, "step": 67446 }, { "epoch": 4.07, "learning_rate": 1.1631433615499526e-05, "loss": 0.1692, "step": 67448 }, { "epoch": 4.07, "learning_rate": 1.16299247053853e-05, "loss": 0.2113, "step": 67450 }, { "epoch": 4.07, "learning_rate": 1.1628415795271077e-05, "loss": 0.2944, "step": 67452 }, { "epoch": 4.07, "learning_rate": 1.1626906885156852e-05, "loss": 0.2344, "step": 67454 }, { "epoch": 4.07, "learning_rate": 1.1625397975042628e-05, "loss": 0.1985, "step": 67456 }, { "epoch": 4.07, "learning_rate": 1.1623889064928403e-05, "loss": 0.1425, "step": 67458 }, { "epoch": 4.07, "learning_rate": 1.162238015481418e-05, "loss": 0.1861, "step": 67460 }, { "epoch": 4.07, "learning_rate": 1.1620871244699953e-05, "loss": 0.1448, "step": 67462 }, { "epoch": 4.07, "learning_rate": 1.161936233458573e-05, "loss": 0.25, "step": 67464 }, { "epoch": 4.07, "learning_rate": 1.1617853424471504e-05, "loss": 0.2987, "step": 67466 }, { "epoch": 4.07, "learning_rate": 1.161634451435728e-05, "loss": 0.1742, "step": 67468 }, { "epoch": 4.07, "learning_rate": 1.1614835604243055e-05, "loss": 0.1894, "step": 67470 }, { "epoch": 4.07, "learning_rate": 1.1613326694128832e-05, "loss": 0.2975, "step": 67472 }, { "epoch": 4.07, "learning_rate": 1.1611817784014607e-05, "loss": 0.3395, "step": 67474 }, { "epoch": 4.07, "learning_rate": 1.1610308873900383e-05, "loss": 0.151, "step": 67476 }, { "epoch": 4.07, "learning_rate": 1.1608799963786156e-05, "loss": 0.2605, "step": 67478 }, { "epoch": 4.07, "learning_rate": 1.1607291053671933e-05, "loss": 0.1598, "step": 67480 }, { "epoch": 4.07, "learning_rate": 1.1605782143557708e-05, "loss": 0.2441, "step": 67482 }, { "epoch": 4.07, "learning_rate": 1.1604273233443484e-05, "loss": 0.2967, "step": 67484 }, { "epoch": 4.07, "learning_rate": 1.160276432332926e-05, "loss": 0.2213, "step": 67486 }, { "epoch": 4.07, "learning_rate": 1.1601255413215036e-05, "loss": 0.2723, "step": 67488 }, { "epoch": 4.07, "learning_rate": 1.1599746503100812e-05, "loss": 0.174, "step": 67490 }, { "epoch": 4.07, "learning_rate": 1.1598237592986587e-05, "loss": 0.3174, "step": 67492 }, { "epoch": 4.07, "learning_rate": 1.1596728682872362e-05, "loss": 0.4342, "step": 67494 }, { "epoch": 4.07, "learning_rate": 1.1595219772758137e-05, "loss": 0.2755, "step": 67496 }, { "epoch": 4.07, "learning_rate": 1.1593710862643913e-05, "loss": 0.2052, "step": 67498 }, { "epoch": 4.07, "learning_rate": 1.1592201952529688e-05, "loss": 0.2095, "step": 67500 }, { "epoch": 4.07, "learning_rate": 1.1590693042415464e-05, "loss": 0.1631, "step": 67502 }, { "epoch": 4.07, "learning_rate": 1.158918413230124e-05, "loss": 0.1713, "step": 67504 }, { "epoch": 4.07, "learning_rate": 1.1587675222187016e-05, "loss": 0.4317, "step": 67506 }, { "epoch": 4.07, "learning_rate": 1.158616631207279e-05, "loss": 0.3346, "step": 67508 }, { "epoch": 4.07, "learning_rate": 1.1584657401958565e-05, "loss": 0.3392, "step": 67510 }, { "epoch": 4.07, "learning_rate": 1.158314849184434e-05, "loss": 0.1722, "step": 67512 }, { "epoch": 4.07, "learning_rate": 1.1581639581730117e-05, "loss": 0.2139, "step": 67514 }, { "epoch": 4.08, "learning_rate": 1.1580130671615891e-05, "loss": 0.2172, "step": 67516 }, { "epoch": 4.08, "learning_rate": 1.1578621761501668e-05, "loss": 0.182, "step": 67518 }, { "epoch": 4.08, "learning_rate": 1.1577112851387443e-05, "loss": 0.363, "step": 67520 }, { "epoch": 4.08, "learning_rate": 1.157560394127322e-05, "loss": 0.3105, "step": 67522 }, { "epoch": 4.08, "learning_rate": 1.1574095031158994e-05, "loss": 0.3285, "step": 67524 }, { "epoch": 4.08, "learning_rate": 1.157258612104477e-05, "loss": 0.2212, "step": 67526 }, { "epoch": 4.08, "learning_rate": 1.1571077210930545e-05, "loss": 0.2347, "step": 67528 }, { "epoch": 4.08, "learning_rate": 1.156956830081632e-05, "loss": 0.2541, "step": 67530 }, { "epoch": 4.08, "learning_rate": 1.1568059390702097e-05, "loss": 0.3354, "step": 67532 }, { "epoch": 4.08, "learning_rate": 1.1566550480587872e-05, "loss": 0.2571, "step": 67534 }, { "epoch": 4.08, "learning_rate": 1.1565041570473648e-05, "loss": 0.3172, "step": 67536 }, { "epoch": 4.08, "learning_rate": 1.1563532660359423e-05, "loss": 0.1819, "step": 67538 }, { "epoch": 4.08, "learning_rate": 1.15620237502452e-05, "loss": 0.2628, "step": 67540 }, { "epoch": 4.08, "learning_rate": 1.1560514840130974e-05, "loss": 0.2311, "step": 67542 }, { "epoch": 4.08, "learning_rate": 1.1559005930016749e-05, "loss": 0.2119, "step": 67544 }, { "epoch": 4.08, "learning_rate": 1.1557497019902524e-05, "loss": 0.3319, "step": 67546 }, { "epoch": 4.08, "learning_rate": 1.15559881097883e-05, "loss": 0.2603, "step": 67548 }, { "epoch": 4.08, "learning_rate": 1.1554479199674075e-05, "loss": 0.3154, "step": 67550 }, { "epoch": 4.08, "learning_rate": 1.1552970289559852e-05, "loss": 0.1368, "step": 67552 }, { "epoch": 4.08, "learning_rate": 1.1551461379445627e-05, "loss": 0.2478, "step": 67554 }, { "epoch": 4.08, "learning_rate": 1.1549952469331403e-05, "loss": 0.2049, "step": 67556 }, { "epoch": 4.08, "learning_rate": 1.1548443559217178e-05, "loss": 0.2106, "step": 67558 }, { "epoch": 4.08, "learning_rate": 1.1546934649102953e-05, "loss": 0.1502, "step": 67560 }, { "epoch": 4.08, "learning_rate": 1.154542573898873e-05, "loss": 0.2959, "step": 67562 }, { "epoch": 4.08, "learning_rate": 1.1543916828874504e-05, "loss": 0.238, "step": 67564 }, { "epoch": 4.08, "learning_rate": 1.154240791876028e-05, "loss": 0.3191, "step": 67566 }, { "epoch": 4.08, "learning_rate": 1.1540899008646055e-05, "loss": 0.223, "step": 67568 }, { "epoch": 4.08, "learning_rate": 1.1539390098531832e-05, "loss": 0.2925, "step": 67570 }, { "epoch": 4.08, "learning_rate": 1.1537881188417607e-05, "loss": 0.31, "step": 67572 }, { "epoch": 4.08, "learning_rate": 1.1536372278303383e-05, "loss": 0.2497, "step": 67574 }, { "epoch": 4.08, "learning_rate": 1.1534863368189156e-05, "loss": 0.288, "step": 67576 }, { "epoch": 4.08, "learning_rate": 1.1533354458074933e-05, "loss": 0.1865, "step": 67578 }, { "epoch": 4.08, "learning_rate": 1.1531845547960708e-05, "loss": 0.2005, "step": 67580 }, { "epoch": 4.08, "learning_rate": 1.1530336637846484e-05, "loss": 0.189, "step": 67582 }, { "epoch": 4.08, "learning_rate": 1.1528827727732259e-05, "loss": 0.2745, "step": 67584 }, { "epoch": 4.08, "learning_rate": 1.1527318817618036e-05, "loss": 0.2564, "step": 67586 }, { "epoch": 4.08, "learning_rate": 1.152580990750381e-05, "loss": 0.2249, "step": 67588 }, { "epoch": 4.08, "learning_rate": 1.1524300997389587e-05, "loss": 0.2226, "step": 67590 }, { "epoch": 4.08, "learning_rate": 1.1522792087275362e-05, "loss": 0.1657, "step": 67592 }, { "epoch": 4.08, "learning_rate": 1.1521283177161137e-05, "loss": 0.2016, "step": 67594 }, { "epoch": 4.08, "learning_rate": 1.1519774267046911e-05, "loss": 0.2461, "step": 67596 }, { "epoch": 4.08, "learning_rate": 1.1518265356932688e-05, "loss": 0.2207, "step": 67598 }, { "epoch": 4.08, "learning_rate": 1.1516756446818463e-05, "loss": 0.2602, "step": 67600 }, { "epoch": 4.08, "learning_rate": 1.151524753670424e-05, "loss": 0.2904, "step": 67602 }, { "epoch": 4.08, "learning_rate": 1.1513738626590016e-05, "loss": 0.378, "step": 67604 }, { "epoch": 4.08, "learning_rate": 1.151222971647579e-05, "loss": 0.2191, "step": 67606 }, { "epoch": 4.08, "learning_rate": 1.1510720806361567e-05, "loss": 0.2459, "step": 67608 }, { "epoch": 4.08, "learning_rate": 1.150921189624734e-05, "loss": 0.316, "step": 67610 }, { "epoch": 4.08, "learning_rate": 1.1507702986133117e-05, "loss": 0.1566, "step": 67612 }, { "epoch": 4.08, "learning_rate": 1.1506194076018892e-05, "loss": 0.1614, "step": 67614 }, { "epoch": 4.08, "learning_rate": 1.1504685165904668e-05, "loss": 0.2809, "step": 67616 }, { "epoch": 4.08, "learning_rate": 1.1503176255790443e-05, "loss": 0.2028, "step": 67618 }, { "epoch": 4.08, "learning_rate": 1.150166734567622e-05, "loss": 0.2298, "step": 67620 }, { "epoch": 4.08, "learning_rate": 1.1500158435561994e-05, "loss": 0.3085, "step": 67622 }, { "epoch": 4.08, "learning_rate": 1.149864952544777e-05, "loss": 0.2653, "step": 67624 }, { "epoch": 4.08, "learning_rate": 1.1497140615333544e-05, "loss": 0.2258, "step": 67626 }, { "epoch": 4.08, "learning_rate": 1.149563170521932e-05, "loss": 0.1208, "step": 67628 }, { "epoch": 4.08, "learning_rate": 1.1494122795105095e-05, "loss": 0.232, "step": 67630 }, { "epoch": 4.08, "learning_rate": 1.1492613884990872e-05, "loss": 0.2366, "step": 67632 }, { "epoch": 4.08, "learning_rate": 1.1491104974876647e-05, "loss": 0.1493, "step": 67634 }, { "epoch": 4.08, "learning_rate": 1.1489596064762423e-05, "loss": 0.2397, "step": 67636 }, { "epoch": 4.08, "learning_rate": 1.1488087154648198e-05, "loss": 0.2326, "step": 67638 }, { "epoch": 4.08, "learning_rate": 1.1486578244533974e-05, "loss": 0.1541, "step": 67640 }, { "epoch": 4.08, "learning_rate": 1.148506933441975e-05, "loss": 0.3276, "step": 67642 }, { "epoch": 4.08, "learning_rate": 1.1483560424305524e-05, "loss": 0.3161, "step": 67644 }, { "epoch": 4.08, "learning_rate": 1.14820515141913e-05, "loss": 0.1405, "step": 67646 }, { "epoch": 4.08, "learning_rate": 1.1480542604077075e-05, "loss": 0.2006, "step": 67648 }, { "epoch": 4.08, "learning_rate": 1.1479033693962852e-05, "loss": 0.2453, "step": 67650 }, { "epoch": 4.08, "learning_rate": 1.1477524783848627e-05, "loss": 0.2199, "step": 67652 }, { "epoch": 4.08, "learning_rate": 1.1476015873734403e-05, "loss": 0.1941, "step": 67654 }, { "epoch": 4.08, "learning_rate": 1.1474506963620178e-05, "loss": 0.2748, "step": 67656 }, { "epoch": 4.08, "learning_rate": 1.1472998053505953e-05, "loss": 0.145, "step": 67658 }, { "epoch": 4.08, "learning_rate": 1.1471489143391728e-05, "loss": 0.2824, "step": 67660 }, { "epoch": 4.08, "learning_rate": 1.1469980233277504e-05, "loss": 0.1352, "step": 67662 }, { "epoch": 4.08, "learning_rate": 1.1468471323163279e-05, "loss": 0.1984, "step": 67664 }, { "epoch": 4.08, "learning_rate": 1.1466962413049055e-05, "loss": 0.1483, "step": 67666 }, { "epoch": 4.08, "learning_rate": 1.146545350293483e-05, "loss": 0.324, "step": 67668 }, { "epoch": 4.08, "learning_rate": 1.1463944592820607e-05, "loss": 0.1936, "step": 67670 }, { "epoch": 4.08, "learning_rate": 1.1462435682706382e-05, "loss": 0.2228, "step": 67672 }, { "epoch": 4.08, "learning_rate": 1.1460926772592158e-05, "loss": 0.2696, "step": 67674 }, { "epoch": 4.08, "learning_rate": 1.1459417862477933e-05, "loss": 0.2623, "step": 67676 }, { "epoch": 4.08, "learning_rate": 1.1457908952363708e-05, "loss": 0.298, "step": 67678 }, { "epoch": 4.08, "learning_rate": 1.1456400042249484e-05, "loss": 0.5316, "step": 67680 }, { "epoch": 4.09, "learning_rate": 1.1454891132135259e-05, "loss": 0.1716, "step": 67682 }, { "epoch": 4.09, "learning_rate": 1.1453382222021036e-05, "loss": 0.2006, "step": 67684 }, { "epoch": 4.09, "learning_rate": 1.145187331190681e-05, "loss": 0.1707, "step": 67686 }, { "epoch": 4.09, "learning_rate": 1.1450364401792587e-05, "loss": 0.2621, "step": 67688 }, { "epoch": 4.09, "learning_rate": 1.1448855491678362e-05, "loss": 0.3139, "step": 67690 }, { "epoch": 4.09, "learning_rate": 1.1447346581564137e-05, "loss": 0.2143, "step": 67692 }, { "epoch": 4.09, "learning_rate": 1.1445837671449911e-05, "loss": 0.3163, "step": 67694 }, { "epoch": 4.09, "learning_rate": 1.1444328761335688e-05, "loss": 0.1621, "step": 67696 }, { "epoch": 4.09, "learning_rate": 1.1442819851221463e-05, "loss": 0.2992, "step": 67698 }, { "epoch": 4.09, "learning_rate": 1.144131094110724e-05, "loss": 0.3054, "step": 67700 }, { "epoch": 4.09, "learning_rate": 1.1439802030993014e-05, "loss": 0.3996, "step": 67702 }, { "epoch": 4.09, "learning_rate": 1.143829312087879e-05, "loss": 0.2181, "step": 67704 }, { "epoch": 4.09, "learning_rate": 1.1436784210764565e-05, "loss": 0.113, "step": 67706 }, { "epoch": 4.09, "learning_rate": 1.143527530065034e-05, "loss": 0.2646, "step": 67708 }, { "epoch": 4.09, "learning_rate": 1.1433766390536115e-05, "loss": 0.1978, "step": 67710 }, { "epoch": 4.09, "learning_rate": 1.1432257480421892e-05, "loss": 0.1745, "step": 67712 }, { "epoch": 4.09, "learning_rate": 1.1430748570307666e-05, "loss": 0.2225, "step": 67714 }, { "epoch": 4.09, "learning_rate": 1.1429239660193443e-05, "loss": 0.2353, "step": 67716 }, { "epoch": 4.09, "learning_rate": 1.1427730750079218e-05, "loss": 0.2526, "step": 67718 }, { "epoch": 4.09, "learning_rate": 1.1426221839964994e-05, "loss": 0.152, "step": 67720 }, { "epoch": 4.09, "learning_rate": 1.1424712929850769e-05, "loss": 0.2108, "step": 67722 }, { "epoch": 4.09, "learning_rate": 1.1423204019736544e-05, "loss": 0.2956, "step": 67724 }, { "epoch": 4.09, "learning_rate": 1.142169510962232e-05, "loss": 0.2333, "step": 67726 }, { "epoch": 4.09, "learning_rate": 1.1420186199508095e-05, "loss": 0.2617, "step": 67728 }, { "epoch": 4.09, "learning_rate": 1.1418677289393872e-05, "loss": 0.244, "step": 67730 }, { "epoch": 4.09, "learning_rate": 1.1417168379279647e-05, "loss": 0.2143, "step": 67732 }, { "epoch": 4.09, "learning_rate": 1.1415659469165423e-05, "loss": 0.2848, "step": 67734 }, { "epoch": 4.09, "learning_rate": 1.1414150559051198e-05, "loss": 0.3607, "step": 67736 }, { "epoch": 4.09, "learning_rate": 1.1412641648936974e-05, "loss": 0.2157, "step": 67738 }, { "epoch": 4.09, "learning_rate": 1.1411132738822748e-05, "loss": 0.2871, "step": 67740 }, { "epoch": 4.09, "learning_rate": 1.1409623828708524e-05, "loss": 0.1662, "step": 67742 }, { "epoch": 4.09, "learning_rate": 1.1408114918594299e-05, "loss": 0.2809, "step": 67744 }, { "epoch": 4.09, "learning_rate": 1.1406606008480075e-05, "loss": 0.2456, "step": 67746 }, { "epoch": 4.09, "learning_rate": 1.140509709836585e-05, "loss": 0.19, "step": 67748 }, { "epoch": 4.09, "learning_rate": 1.1403588188251627e-05, "loss": 0.2452, "step": 67750 }, { "epoch": 4.09, "learning_rate": 1.1402079278137402e-05, "loss": 0.2629, "step": 67752 }, { "epoch": 4.09, "learning_rate": 1.1400570368023178e-05, "loss": 0.3464, "step": 67754 }, { "epoch": 4.09, "learning_rate": 1.1399061457908953e-05, "loss": 0.2293, "step": 67756 }, { "epoch": 4.09, "learning_rate": 1.1397552547794728e-05, "loss": 0.2932, "step": 67758 }, { "epoch": 4.09, "learning_rate": 1.1396043637680504e-05, "loss": 0.3115, "step": 67760 }, { "epoch": 4.09, "learning_rate": 1.1394534727566279e-05, "loss": 0.1432, "step": 67762 }, { "epoch": 4.09, "learning_rate": 1.1393025817452056e-05, "loss": 0.1966, "step": 67764 }, { "epoch": 4.09, "learning_rate": 1.139151690733783e-05, "loss": 0.2624, "step": 67766 }, { "epoch": 4.09, "learning_rate": 1.1390007997223607e-05, "loss": 0.1398, "step": 67768 }, { "epoch": 4.09, "learning_rate": 1.1388499087109382e-05, "loss": 0.2774, "step": 67770 }, { "epoch": 4.09, "learning_rate": 1.1386990176995158e-05, "loss": 0.2353, "step": 67772 }, { "epoch": 4.09, "learning_rate": 1.1385481266880931e-05, "loss": 0.1699, "step": 67774 }, { "epoch": 4.09, "learning_rate": 1.1383972356766708e-05, "loss": 0.2703, "step": 67776 }, { "epoch": 4.09, "learning_rate": 1.1382463446652483e-05, "loss": 0.2309, "step": 67778 }, { "epoch": 4.09, "learning_rate": 1.138095453653826e-05, "loss": 0.2059, "step": 67780 }, { "epoch": 4.09, "learning_rate": 1.1379445626424034e-05, "loss": 0.2365, "step": 67782 }, { "epoch": 4.09, "learning_rate": 1.137793671630981e-05, "loss": 0.2214, "step": 67784 }, { "epoch": 4.09, "learning_rate": 1.1376427806195585e-05, "loss": 0.2213, "step": 67786 }, { "epoch": 4.09, "learning_rate": 1.1374918896081362e-05, "loss": 0.2896, "step": 67788 }, { "epoch": 4.09, "learning_rate": 1.1373409985967137e-05, "loss": 0.1271, "step": 67790 }, { "epoch": 4.09, "learning_rate": 1.1371901075852911e-05, "loss": 0.2045, "step": 67792 }, { "epoch": 4.09, "learning_rate": 1.1370392165738688e-05, "loss": 0.2127, "step": 67794 }, { "epoch": 4.09, "learning_rate": 1.1368883255624463e-05, "loss": 0.129, "step": 67796 }, { "epoch": 4.09, "learning_rate": 1.136737434551024e-05, "loss": 0.179, "step": 67798 }, { "epoch": 4.09, "learning_rate": 1.1365865435396014e-05, "loss": 0.142, "step": 67800 }, { "epoch": 4.09, "learning_rate": 1.136435652528179e-05, "loss": 0.1529, "step": 67802 }, { "epoch": 4.09, "learning_rate": 1.1362847615167565e-05, "loss": 0.3162, "step": 67804 }, { "epoch": 4.09, "learning_rate": 1.136133870505334e-05, "loss": 0.4256, "step": 67806 }, { "epoch": 4.09, "learning_rate": 1.1359829794939115e-05, "loss": 0.2277, "step": 67808 }, { "epoch": 4.09, "learning_rate": 1.1358320884824892e-05, "loss": 0.2585, "step": 67810 }, { "epoch": 4.09, "learning_rate": 1.1356811974710666e-05, "loss": 0.3149, "step": 67812 }, { "epoch": 4.09, "learning_rate": 1.1355303064596443e-05, "loss": 0.1723, "step": 67814 }, { "epoch": 4.09, "learning_rate": 1.1353794154482218e-05, "loss": 0.3774, "step": 67816 }, { "epoch": 4.09, "learning_rate": 1.1352285244367994e-05, "loss": 0.19, "step": 67818 }, { "epoch": 4.09, "learning_rate": 1.1350776334253769e-05, "loss": 0.179, "step": 67820 }, { "epoch": 4.09, "learning_rate": 1.1349267424139544e-05, "loss": 0.2187, "step": 67822 }, { "epoch": 4.09, "learning_rate": 1.1347758514025319e-05, "loss": 0.2048, "step": 67824 }, { "epoch": 4.09, "learning_rate": 1.1346249603911095e-05, "loss": 0.2692, "step": 67826 }, { "epoch": 4.09, "learning_rate": 1.134474069379687e-05, "loss": 0.3873, "step": 67828 }, { "epoch": 4.09, "learning_rate": 1.1343231783682647e-05, "loss": 0.1601, "step": 67830 }, { "epoch": 4.09, "learning_rate": 1.1341722873568421e-05, "loss": 0.242, "step": 67832 }, { "epoch": 4.09, "learning_rate": 1.1340213963454198e-05, "loss": 0.2326, "step": 67834 }, { "epoch": 4.09, "learning_rate": 1.1338705053339973e-05, "loss": 0.3026, "step": 67836 }, { "epoch": 4.09, "learning_rate": 1.133719614322575e-05, "loss": 0.3132, "step": 67838 }, { "epoch": 4.09, "learning_rate": 1.1335687233111524e-05, "loss": 0.2076, "step": 67840 }, { "epoch": 4.09, "learning_rate": 1.1334178322997299e-05, "loss": 0.279, "step": 67842 }, { "epoch": 4.09, "learning_rate": 1.1332669412883075e-05, "loss": 0.1492, "step": 67844 }, { "epoch": 4.1, "learning_rate": 1.133116050276885e-05, "loss": 0.2256, "step": 67846 }, { "epoch": 4.1, "learning_rate": 1.1329651592654627e-05, "loss": 0.1839, "step": 67848 }, { "epoch": 4.1, "learning_rate": 1.1328142682540402e-05, "loss": 0.2396, "step": 67850 }, { "epoch": 4.1, "learning_rate": 1.1326633772426178e-05, "loss": 0.3584, "step": 67852 }, { "epoch": 4.1, "learning_rate": 1.1325124862311953e-05, "loss": 0.2195, "step": 67854 }, { "epoch": 4.1, "learning_rate": 1.1323615952197728e-05, "loss": 0.23, "step": 67856 }, { "epoch": 4.1, "learning_rate": 1.1322107042083503e-05, "loss": 0.2138, "step": 67858 }, { "epoch": 4.1, "learning_rate": 1.1320598131969279e-05, "loss": 0.1897, "step": 67860 }, { "epoch": 4.1, "learning_rate": 1.1319089221855054e-05, "loss": 0.3011, "step": 67862 }, { "epoch": 4.1, "learning_rate": 1.131758031174083e-05, "loss": 0.3007, "step": 67864 }, { "epoch": 4.1, "learning_rate": 1.1316071401626605e-05, "loss": 0.4536, "step": 67866 }, { "epoch": 4.1, "learning_rate": 1.1314562491512382e-05, "loss": 0.1689, "step": 67868 }, { "epoch": 4.1, "learning_rate": 1.1313053581398157e-05, "loss": 0.1906, "step": 67870 }, { "epoch": 4.1, "learning_rate": 1.1311544671283931e-05, "loss": 0.2525, "step": 67872 }, { "epoch": 4.1, "learning_rate": 1.1310035761169708e-05, "loss": 0.3475, "step": 67874 }, { "epoch": 4.1, "learning_rate": 1.1308526851055483e-05, "loss": 0.2416, "step": 67876 }, { "epoch": 4.1, "learning_rate": 1.130701794094126e-05, "loss": 0.1929, "step": 67878 }, { "epoch": 4.1, "learning_rate": 1.1305509030827034e-05, "loss": 0.181, "step": 67880 }, { "epoch": 4.1, "learning_rate": 1.130400012071281e-05, "loss": 0.2124, "step": 67882 }, { "epoch": 4.1, "learning_rate": 1.1302491210598585e-05, "loss": 0.2492, "step": 67884 }, { "epoch": 4.1, "learning_rate": 1.1300982300484362e-05, "loss": 0.2857, "step": 67886 }, { "epoch": 4.1, "learning_rate": 1.1299473390370135e-05, "loss": 0.3431, "step": 67888 }, { "epoch": 4.1, "learning_rate": 1.1297964480255912e-05, "loss": 0.2743, "step": 67890 }, { "epoch": 4.1, "learning_rate": 1.1296455570141686e-05, "loss": 0.2526, "step": 67892 }, { "epoch": 4.1, "learning_rate": 1.1294946660027463e-05, "loss": 0.2732, "step": 67894 }, { "epoch": 4.1, "learning_rate": 1.1293437749913238e-05, "loss": 0.3234, "step": 67896 }, { "epoch": 4.1, "learning_rate": 1.1291928839799014e-05, "loss": 0.246, "step": 67898 }, { "epoch": 4.1, "learning_rate": 1.1290419929684789e-05, "loss": 0.1527, "step": 67900 }, { "epoch": 4.1, "learning_rate": 1.1288911019570566e-05, "loss": 0.2964, "step": 67902 }, { "epoch": 4.1, "learning_rate": 1.128740210945634e-05, "loss": 0.255, "step": 67904 }, { "epoch": 4.1, "learning_rate": 1.1285893199342115e-05, "loss": 0.3532, "step": 67906 }, { "epoch": 4.1, "learning_rate": 1.1284384289227892e-05, "loss": 0.2022, "step": 67908 }, { "epoch": 4.1, "learning_rate": 1.1282875379113666e-05, "loss": 0.1499, "step": 67910 }, { "epoch": 4.1, "learning_rate": 1.1281366468999443e-05, "loss": 0.2793, "step": 67912 }, { "epoch": 4.1, "learning_rate": 1.1279857558885218e-05, "loss": 0.2071, "step": 67914 }, { "epoch": 4.1, "learning_rate": 1.1278348648770994e-05, "loss": 0.2408, "step": 67916 }, { "epoch": 4.1, "learning_rate": 1.1276839738656769e-05, "loss": 0.2272, "step": 67918 }, { "epoch": 4.1, "learning_rate": 1.1275330828542546e-05, "loss": 0.2244, "step": 67920 }, { "epoch": 4.1, "learning_rate": 1.1273821918428319e-05, "loss": 0.1795, "step": 67922 }, { "epoch": 4.1, "learning_rate": 1.1272313008314095e-05, "loss": 0.2874, "step": 67924 }, { "epoch": 4.1, "learning_rate": 1.127080409819987e-05, "loss": 0.1824, "step": 67926 }, { "epoch": 4.1, "learning_rate": 1.1269295188085647e-05, "loss": 0.5273, "step": 67928 }, { "epoch": 4.1, "learning_rate": 1.1267786277971421e-05, "loss": 0.2911, "step": 67930 }, { "epoch": 4.1, "learning_rate": 1.1266277367857198e-05, "loss": 0.1964, "step": 67932 }, { "epoch": 4.1, "learning_rate": 1.1264768457742973e-05, "loss": 0.2988, "step": 67934 }, { "epoch": 4.1, "learning_rate": 1.126325954762875e-05, "loss": 0.1884, "step": 67936 }, { "epoch": 4.1, "learning_rate": 1.1261750637514522e-05, "loss": 0.3918, "step": 67938 }, { "epoch": 4.1, "learning_rate": 1.1260241727400299e-05, "loss": 0.2342, "step": 67940 }, { "epoch": 4.1, "learning_rate": 1.1258732817286074e-05, "loss": 0.2811, "step": 67942 }, { "epoch": 4.1, "learning_rate": 1.125722390717185e-05, "loss": 0.173, "step": 67944 }, { "epoch": 4.1, "learning_rate": 1.1255714997057625e-05, "loss": 0.1593, "step": 67946 }, { "epoch": 4.1, "learning_rate": 1.1254206086943402e-05, "loss": 0.238, "step": 67948 }, { "epoch": 4.1, "learning_rate": 1.1252697176829176e-05, "loss": 0.1937, "step": 67950 }, { "epoch": 4.1, "learning_rate": 1.1251188266714953e-05, "loss": 0.2184, "step": 67952 }, { "epoch": 4.1, "learning_rate": 1.1249679356600728e-05, "loss": 0.2034, "step": 67954 }, { "epoch": 4.1, "learning_rate": 1.1248170446486503e-05, "loss": 0.2837, "step": 67956 }, { "epoch": 4.1, "learning_rate": 1.1246661536372279e-05, "loss": 0.2529, "step": 67958 }, { "epoch": 4.1, "learning_rate": 1.1245152626258054e-05, "loss": 0.2947, "step": 67960 }, { "epoch": 4.1, "learning_rate": 1.124364371614383e-05, "loss": 0.2699, "step": 67962 }, { "epoch": 4.1, "learning_rate": 1.1242134806029605e-05, "loss": 0.2709, "step": 67964 }, { "epoch": 4.1, "learning_rate": 1.1240625895915382e-05, "loss": 0.1982, "step": 67966 }, { "epoch": 4.1, "learning_rate": 1.1239116985801157e-05, "loss": 0.2954, "step": 67968 }, { "epoch": 4.1, "learning_rate": 1.1237608075686931e-05, "loss": 0.2227, "step": 67970 }, { "epoch": 4.1, "learning_rate": 1.1236099165572706e-05, "loss": 0.2067, "step": 67972 }, { "epoch": 4.1, "learning_rate": 1.1234590255458483e-05, "loss": 0.2302, "step": 67974 }, { "epoch": 4.1, "learning_rate": 1.1233081345344258e-05, "loss": 0.3051, "step": 67976 }, { "epoch": 4.1, "learning_rate": 1.1231572435230034e-05, "loss": 0.3147, "step": 67978 }, { "epoch": 4.1, "learning_rate": 1.1230063525115809e-05, "loss": 0.3182, "step": 67980 }, { "epoch": 4.1, "learning_rate": 1.1228554615001585e-05, "loss": 0.2945, "step": 67982 }, { "epoch": 4.1, "learning_rate": 1.122704570488736e-05, "loss": 0.1789, "step": 67984 }, { "epoch": 4.1, "learning_rate": 1.1225536794773137e-05, "loss": 0.2577, "step": 67986 }, { "epoch": 4.1, "learning_rate": 1.1224027884658912e-05, "loss": 0.2733, "step": 67988 }, { "epoch": 4.1, "learning_rate": 1.1222518974544686e-05, "loss": 0.2751, "step": 67990 }, { "epoch": 4.1, "learning_rate": 1.1221010064430463e-05, "loss": 0.1943, "step": 67992 }, { "epoch": 4.1, "learning_rate": 1.1219501154316238e-05, "loss": 0.3024, "step": 67994 }, { "epoch": 4.1, "learning_rate": 1.1217992244202014e-05, "loss": 0.2434, "step": 67996 }, { "epoch": 4.1, "learning_rate": 1.1216483334087789e-05, "loss": 0.1863, "step": 67998 }, { "epoch": 4.1, "learning_rate": 1.1214974423973566e-05, "loss": 0.2578, "step": 68000 }, { "epoch": 4.1, "learning_rate": 1.121346551385934e-05, "loss": 0.1796, "step": 68002 }, { "epoch": 4.1, "learning_rate": 1.1211956603745115e-05, "loss": 0.2714, "step": 68004 }, { "epoch": 4.1, "learning_rate": 1.121044769363089e-05, "loss": 0.2119, "step": 68006 }, { "epoch": 4.1, "learning_rate": 1.1208938783516667e-05, "loss": 0.2243, "step": 68008 }, { "epoch": 4.1, "learning_rate": 1.1207429873402441e-05, "loss": 0.2603, "step": 68010 }, { "epoch": 4.11, "learning_rate": 1.1205920963288218e-05, "loss": 0.2794, "step": 68012 }, { "epoch": 4.11, "learning_rate": 1.1204412053173993e-05, "loss": 0.1916, "step": 68014 }, { "epoch": 4.11, "learning_rate": 1.120290314305977e-05, "loss": 0.1546, "step": 68016 }, { "epoch": 4.11, "learning_rate": 1.1201394232945544e-05, "loss": 0.2487, "step": 68018 }, { "epoch": 4.11, "learning_rate": 1.1199885322831319e-05, "loss": 0.2847, "step": 68020 }, { "epoch": 4.11, "learning_rate": 1.1198376412717095e-05, "loss": 0.2253, "step": 68022 }, { "epoch": 4.11, "learning_rate": 1.119686750260287e-05, "loss": 0.162, "step": 68024 }, { "epoch": 4.11, "learning_rate": 1.1195358592488647e-05, "loss": 0.3218, "step": 68026 }, { "epoch": 4.11, "learning_rate": 1.1193849682374421e-05, "loss": 0.2294, "step": 68028 }, { "epoch": 4.11, "learning_rate": 1.1192340772260198e-05, "loss": 0.1865, "step": 68030 }, { "epoch": 4.11, "learning_rate": 1.1190831862145973e-05, "loss": 0.2354, "step": 68032 }, { "epoch": 4.11, "learning_rate": 1.118932295203175e-05, "loss": 0.2273, "step": 68034 }, { "epoch": 4.11, "learning_rate": 1.1187814041917522e-05, "loss": 0.2103, "step": 68036 }, { "epoch": 4.11, "learning_rate": 1.1186305131803299e-05, "loss": 0.1683, "step": 68038 }, { "epoch": 4.11, "learning_rate": 1.1184796221689074e-05, "loss": 0.2375, "step": 68040 }, { "epoch": 4.11, "learning_rate": 1.118328731157485e-05, "loss": 0.2493, "step": 68042 }, { "epoch": 4.11, "learning_rate": 1.1181778401460625e-05, "loss": 0.2737, "step": 68044 }, { "epoch": 4.11, "learning_rate": 1.1180269491346402e-05, "loss": 0.2364, "step": 68046 }, { "epoch": 4.11, "learning_rate": 1.1178760581232176e-05, "loss": 0.2282, "step": 68048 }, { "epoch": 4.11, "learning_rate": 1.1177251671117953e-05, "loss": 0.3028, "step": 68050 }, { "epoch": 4.11, "learning_rate": 1.1175742761003726e-05, "loss": 0.2931, "step": 68052 }, { "epoch": 4.11, "learning_rate": 1.1174233850889503e-05, "loss": 0.3548, "step": 68054 }, { "epoch": 4.11, "learning_rate": 1.1172724940775277e-05, "loss": 0.2529, "step": 68056 }, { "epoch": 4.11, "learning_rate": 1.1171216030661054e-05, "loss": 0.3241, "step": 68058 }, { "epoch": 4.11, "learning_rate": 1.1169707120546829e-05, "loss": 0.2393, "step": 68060 }, { "epoch": 4.11, "learning_rate": 1.1168198210432605e-05, "loss": 0.1998, "step": 68062 }, { "epoch": 4.11, "learning_rate": 1.116668930031838e-05, "loss": 0.1863, "step": 68064 }, { "epoch": 4.11, "learning_rate": 1.1165180390204157e-05, "loss": 0.2181, "step": 68066 }, { "epoch": 4.11, "learning_rate": 1.1163671480089931e-05, "loss": 0.1998, "step": 68068 }, { "epoch": 4.11, "learning_rate": 1.1162162569975706e-05, "loss": 0.2599, "step": 68070 }, { "epoch": 4.11, "learning_rate": 1.1160653659861483e-05, "loss": 0.3183, "step": 68072 }, { "epoch": 4.11, "learning_rate": 1.1159144749747258e-05, "loss": 0.2939, "step": 68074 }, { "epoch": 4.11, "learning_rate": 1.1157635839633034e-05, "loss": 0.3655, "step": 68076 }, { "epoch": 4.11, "learning_rate": 1.1156126929518809e-05, "loss": 0.3309, "step": 68078 }, { "epoch": 4.11, "learning_rate": 1.1154618019404585e-05, "loss": 0.1704, "step": 68080 }, { "epoch": 4.11, "learning_rate": 1.115310910929036e-05, "loss": 0.204, "step": 68082 }, { "epoch": 4.11, "learning_rate": 1.1151600199176137e-05, "loss": 0.1804, "step": 68084 }, { "epoch": 4.11, "learning_rate": 1.115009128906191e-05, "loss": 0.2553, "step": 68086 }, { "epoch": 4.11, "learning_rate": 1.1148582378947686e-05, "loss": 0.2855, "step": 68088 }, { "epoch": 4.11, "learning_rate": 1.1147073468833461e-05, "loss": 0.3418, "step": 68090 }, { "epoch": 4.11, "learning_rate": 1.1145564558719238e-05, "loss": 0.2659, "step": 68092 }, { "epoch": 4.11, "learning_rate": 1.1144055648605013e-05, "loss": 0.2329, "step": 68094 }, { "epoch": 4.11, "learning_rate": 1.1142546738490789e-05, "loss": 0.1497, "step": 68096 }, { "epoch": 4.11, "learning_rate": 1.1141037828376564e-05, "loss": 0.2278, "step": 68098 }, { "epoch": 4.11, "learning_rate": 1.113952891826234e-05, "loss": 0.1716, "step": 68100 }, { "epoch": 4.11, "learning_rate": 1.1138020008148115e-05, "loss": 0.2069, "step": 68102 }, { "epoch": 4.11, "learning_rate": 1.113651109803389e-05, "loss": 0.2464, "step": 68104 }, { "epoch": 4.11, "learning_rate": 1.1135002187919667e-05, "loss": 0.1746, "step": 68106 }, { "epoch": 4.11, "learning_rate": 1.1133493277805441e-05, "loss": 0.1779, "step": 68108 }, { "epoch": 4.11, "learning_rate": 1.1131984367691218e-05, "loss": 0.2009, "step": 68110 }, { "epoch": 4.11, "learning_rate": 1.1130475457576993e-05, "loss": 0.21, "step": 68112 }, { "epoch": 4.11, "learning_rate": 1.112896654746277e-05, "loss": 0.3955, "step": 68114 }, { "epoch": 4.11, "learning_rate": 1.1127457637348544e-05, "loss": 0.1819, "step": 68116 }, { "epoch": 4.11, "learning_rate": 1.1125948727234319e-05, "loss": 0.2094, "step": 68118 }, { "epoch": 4.11, "learning_rate": 1.1124439817120094e-05, "loss": 0.2277, "step": 68120 }, { "epoch": 4.11, "learning_rate": 1.112293090700587e-05, "loss": 0.2412, "step": 68122 }, { "epoch": 4.11, "learning_rate": 1.1121421996891645e-05, "loss": 0.2309, "step": 68124 }, { "epoch": 4.11, "learning_rate": 1.1119913086777422e-05, "loss": 0.3253, "step": 68126 }, { "epoch": 4.11, "learning_rate": 1.1118404176663196e-05, "loss": 0.1975, "step": 68128 }, { "epoch": 4.11, "learning_rate": 1.1116895266548973e-05, "loss": 0.1556, "step": 68130 }, { "epoch": 4.11, "learning_rate": 1.1115386356434748e-05, "loss": 0.2385, "step": 68132 }, { "epoch": 4.11, "learning_rate": 1.1113877446320524e-05, "loss": 0.2381, "step": 68134 }, { "epoch": 4.11, "learning_rate": 1.1112368536206299e-05, "loss": 0.2648, "step": 68136 }, { "epoch": 4.11, "learning_rate": 1.1110859626092074e-05, "loss": 0.195, "step": 68138 }, { "epoch": 4.11, "learning_rate": 1.110935071597785e-05, "loss": 0.2648, "step": 68140 }, { "epoch": 4.11, "learning_rate": 1.1107841805863625e-05, "loss": 0.1823, "step": 68142 }, { "epoch": 4.11, "learning_rate": 1.1106332895749402e-05, "loss": 0.1662, "step": 68144 }, { "epoch": 4.11, "learning_rate": 1.1104823985635176e-05, "loss": 0.1725, "step": 68146 }, { "epoch": 4.11, "learning_rate": 1.1103315075520953e-05, "loss": 0.3129, "step": 68148 }, { "epoch": 4.11, "learning_rate": 1.1101806165406728e-05, "loss": 0.1947, "step": 68150 }, { "epoch": 4.11, "learning_rate": 1.1100297255292503e-05, "loss": 0.2049, "step": 68152 }, { "epoch": 4.11, "learning_rate": 1.1098788345178277e-05, "loss": 0.2967, "step": 68154 }, { "epoch": 4.11, "learning_rate": 1.1097279435064054e-05, "loss": 0.1885, "step": 68156 }, { "epoch": 4.11, "learning_rate": 1.1095770524949829e-05, "loss": 0.2042, "step": 68158 }, { "epoch": 4.11, "learning_rate": 1.1094261614835605e-05, "loss": 0.3245, "step": 68160 }, { "epoch": 4.11, "learning_rate": 1.109275270472138e-05, "loss": 0.3246, "step": 68162 }, { "epoch": 4.11, "learning_rate": 1.1091243794607157e-05, "loss": 0.212, "step": 68164 }, { "epoch": 4.11, "learning_rate": 1.1089734884492931e-05, "loss": 0.2263, "step": 68166 }, { "epoch": 4.11, "learning_rate": 1.1088225974378706e-05, "loss": 0.2394, "step": 68168 }, { "epoch": 4.11, "learning_rate": 1.1086717064264481e-05, "loss": 0.291, "step": 68170 }, { "epoch": 4.11, "learning_rate": 1.1085208154150258e-05, "loss": 0.3465, "step": 68172 }, { "epoch": 4.11, "learning_rate": 1.1083699244036032e-05, "loss": 0.3473, "step": 68174 }, { "epoch": 4.11, "learning_rate": 1.1082190333921809e-05, "loss": 0.2465, "step": 68176 }, { "epoch": 4.12, "learning_rate": 1.1080681423807584e-05, "loss": 0.1589, "step": 68178 }, { "epoch": 4.12, "learning_rate": 1.107917251369336e-05, "loss": 0.3362, "step": 68180 }, { "epoch": 4.12, "learning_rate": 1.1077663603579135e-05, "loss": 0.2406, "step": 68182 }, { "epoch": 4.12, "learning_rate": 1.107615469346491e-05, "loss": 0.3621, "step": 68184 }, { "epoch": 4.12, "learning_rate": 1.1074645783350686e-05, "loss": 0.2388, "step": 68186 }, { "epoch": 4.12, "learning_rate": 1.1073136873236461e-05, "loss": 0.3217, "step": 68188 }, { "epoch": 4.12, "learning_rate": 1.1071627963122238e-05, "loss": 0.3007, "step": 68190 }, { "epoch": 4.12, "learning_rate": 1.1070119053008013e-05, "loss": 0.2017, "step": 68192 }, { "epoch": 4.12, "learning_rate": 1.1068610142893789e-05, "loss": 0.1764, "step": 68194 }, { "epoch": 4.12, "learning_rate": 1.1067101232779564e-05, "loss": 0.3536, "step": 68196 }, { "epoch": 4.12, "learning_rate": 1.106559232266534e-05, "loss": 0.2681, "step": 68198 }, { "epoch": 4.12, "learning_rate": 1.1064083412551114e-05, "loss": 0.1656, "step": 68200 }, { "epoch": 4.12, "learning_rate": 1.106257450243689e-05, "loss": 0.2173, "step": 68202 }, { "epoch": 4.12, "learning_rate": 1.1061065592322665e-05, "loss": 0.2849, "step": 68204 }, { "epoch": 4.12, "learning_rate": 1.1059556682208441e-05, "loss": 0.2679, "step": 68206 }, { "epoch": 4.12, "learning_rate": 1.1058047772094216e-05, "loss": 0.2429, "step": 68208 }, { "epoch": 4.12, "learning_rate": 1.1056538861979993e-05, "loss": 0.1978, "step": 68210 }, { "epoch": 4.12, "learning_rate": 1.1055029951865768e-05, "loss": 0.2753, "step": 68212 }, { "epoch": 4.12, "learning_rate": 1.1053521041751544e-05, "loss": 0.2042, "step": 68214 }, { "epoch": 4.12, "learning_rate": 1.1052012131637319e-05, "loss": 0.1897, "step": 68216 }, { "epoch": 4.12, "learning_rate": 1.1050503221523094e-05, "loss": 0.3849, "step": 68218 }, { "epoch": 4.12, "learning_rate": 1.104899431140887e-05, "loss": 0.1792, "step": 68220 }, { "epoch": 4.12, "learning_rate": 1.1047485401294645e-05, "loss": 0.1699, "step": 68222 }, { "epoch": 4.12, "learning_rate": 1.1045976491180422e-05, "loss": 0.257, "step": 68224 }, { "epoch": 4.12, "learning_rate": 1.1044467581066196e-05, "loss": 0.2484, "step": 68226 }, { "epoch": 4.12, "learning_rate": 1.1042958670951973e-05, "loss": 0.1689, "step": 68228 }, { "epoch": 4.12, "learning_rate": 1.1041449760837748e-05, "loss": 0.1786, "step": 68230 }, { "epoch": 4.12, "learning_rate": 1.1039940850723524e-05, "loss": 0.2368, "step": 68232 }, { "epoch": 4.12, "learning_rate": 1.1038431940609297e-05, "loss": 0.2731, "step": 68234 }, { "epoch": 4.12, "learning_rate": 1.1036923030495074e-05, "loss": 0.1724, "step": 68236 }, { "epoch": 4.12, "learning_rate": 1.1035414120380849e-05, "loss": 0.2259, "step": 68238 }, { "epoch": 4.12, "learning_rate": 1.1033905210266625e-05, "loss": 0.2377, "step": 68240 }, { "epoch": 4.12, "learning_rate": 1.10323963001524e-05, "loss": 0.1962, "step": 68242 }, { "epoch": 4.12, "learning_rate": 1.1030887390038177e-05, "loss": 0.2072, "step": 68244 }, { "epoch": 4.12, "learning_rate": 1.1029378479923951e-05, "loss": 0.254, "step": 68246 }, { "epoch": 4.12, "learning_rate": 1.1027869569809728e-05, "loss": 0.2159, "step": 68248 }, { "epoch": 4.12, "learning_rate": 1.1026360659695503e-05, "loss": 0.3056, "step": 68250 }, { "epoch": 4.12, "learning_rate": 1.1024851749581278e-05, "loss": 0.2087, "step": 68252 }, { "epoch": 4.12, "learning_rate": 1.1023342839467054e-05, "loss": 0.1802, "step": 68254 }, { "epoch": 4.12, "learning_rate": 1.1021833929352829e-05, "loss": 0.2069, "step": 68256 }, { "epoch": 4.12, "learning_rate": 1.1020325019238605e-05, "loss": 0.3635, "step": 68258 }, { "epoch": 4.12, "learning_rate": 1.101881610912438e-05, "loss": 0.2381, "step": 68260 }, { "epoch": 4.12, "learning_rate": 1.1017307199010157e-05, "loss": 0.181, "step": 68262 }, { "epoch": 4.12, "learning_rate": 1.1015798288895932e-05, "loss": 0.1825, "step": 68264 }, { "epoch": 4.12, "learning_rate": 1.1014289378781706e-05, "loss": 0.2055, "step": 68266 }, { "epoch": 4.12, "learning_rate": 1.1012780468667481e-05, "loss": 0.2872, "step": 68268 }, { "epoch": 4.12, "learning_rate": 1.1011271558553258e-05, "loss": 0.2292, "step": 68270 }, { "epoch": 4.12, "learning_rate": 1.1009762648439032e-05, "loss": 0.2944, "step": 68272 }, { "epoch": 4.12, "learning_rate": 1.1008253738324809e-05, "loss": 0.1656, "step": 68274 }, { "epoch": 4.12, "learning_rate": 1.1006744828210584e-05, "loss": 0.1744, "step": 68276 }, { "epoch": 4.12, "learning_rate": 1.100523591809636e-05, "loss": 0.2424, "step": 68278 }, { "epoch": 4.12, "learning_rate": 1.1003727007982135e-05, "loss": 0.1821, "step": 68280 }, { "epoch": 4.12, "learning_rate": 1.100221809786791e-05, "loss": 0.2351, "step": 68282 }, { "epoch": 4.12, "learning_rate": 1.1000709187753685e-05, "loss": 0.1841, "step": 68284 }, { "epoch": 4.12, "learning_rate": 1.0999200277639461e-05, "loss": 0.1198, "step": 68286 }, { "epoch": 4.12, "learning_rate": 1.0997691367525236e-05, "loss": 0.3893, "step": 68288 }, { "epoch": 4.12, "learning_rate": 1.0996182457411013e-05, "loss": 0.2466, "step": 68290 }, { "epoch": 4.12, "learning_rate": 1.0994673547296787e-05, "loss": 0.2021, "step": 68292 }, { "epoch": 4.12, "learning_rate": 1.0993164637182564e-05, "loss": 0.2156, "step": 68294 }, { "epoch": 4.12, "learning_rate": 1.0991655727068339e-05, "loss": 0.2213, "step": 68296 }, { "epoch": 4.12, "learning_rate": 1.0990146816954115e-05, "loss": 0.1849, "step": 68298 }, { "epoch": 4.12, "learning_rate": 1.098863790683989e-05, "loss": 0.2063, "step": 68300 }, { "epoch": 4.12, "learning_rate": 1.0987128996725665e-05, "loss": 0.2097, "step": 68302 }, { "epoch": 4.12, "learning_rate": 1.0985620086611441e-05, "loss": 0.1897, "step": 68304 }, { "epoch": 4.12, "learning_rate": 1.0984111176497216e-05, "loss": 0.1925, "step": 68306 }, { "epoch": 4.12, "learning_rate": 1.0982602266382993e-05, "loss": 0.3066, "step": 68308 }, { "epoch": 4.12, "learning_rate": 1.0981093356268768e-05, "loss": 0.2088, "step": 68310 }, { "epoch": 4.12, "learning_rate": 1.0979584446154544e-05, "loss": 0.2144, "step": 68312 }, { "epoch": 4.12, "learning_rate": 1.0978075536040319e-05, "loss": 0.224, "step": 68314 }, { "epoch": 4.12, "learning_rate": 1.0976566625926094e-05, "loss": 0.1929, "step": 68316 }, { "epoch": 4.12, "learning_rate": 1.0975057715811869e-05, "loss": 0.3202, "step": 68318 }, { "epoch": 4.12, "learning_rate": 1.0973548805697645e-05, "loss": 0.4292, "step": 68320 }, { "epoch": 4.12, "learning_rate": 1.097203989558342e-05, "loss": 0.2735, "step": 68322 }, { "epoch": 4.12, "learning_rate": 1.0970530985469196e-05, "loss": 0.2829, "step": 68324 }, { "epoch": 4.12, "learning_rate": 1.0969022075354971e-05, "loss": 0.1756, "step": 68326 }, { "epoch": 4.12, "learning_rate": 1.0967513165240748e-05, "loss": 0.2828, "step": 68328 }, { "epoch": 4.12, "learning_rate": 1.0966004255126523e-05, "loss": 0.1741, "step": 68330 }, { "epoch": 4.12, "learning_rate": 1.0964495345012297e-05, "loss": 0.2748, "step": 68332 }, { "epoch": 4.12, "learning_rate": 1.0962986434898074e-05, "loss": 0.2757, "step": 68334 }, { "epoch": 4.12, "learning_rate": 1.0961477524783849e-05, "loss": 0.2739, "step": 68336 }, { "epoch": 4.12, "learning_rate": 1.0959968614669625e-05, "loss": 0.2729, "step": 68338 }, { "epoch": 4.12, "learning_rate": 1.09584597045554e-05, "loss": 0.2757, "step": 68340 }, { "epoch": 4.12, "learning_rate": 1.0956950794441177e-05, "loss": 0.1759, "step": 68342 }, { "epoch": 4.13, "learning_rate": 1.0955441884326951e-05, "loss": 0.3054, "step": 68344 }, { "epoch": 4.13, "learning_rate": 1.0953932974212728e-05, "loss": 0.2421, "step": 68346 }, { "epoch": 4.13, "learning_rate": 1.0952424064098501e-05, "loss": 0.1377, "step": 68348 }, { "epoch": 4.13, "learning_rate": 1.0950915153984278e-05, "loss": 0.2332, "step": 68350 }, { "epoch": 4.13, "learning_rate": 1.0949406243870052e-05, "loss": 0.1817, "step": 68352 }, { "epoch": 4.13, "learning_rate": 1.0947897333755829e-05, "loss": 0.1645, "step": 68354 }, { "epoch": 4.13, "learning_rate": 1.0946388423641604e-05, "loss": 0.2416, "step": 68356 }, { "epoch": 4.13, "learning_rate": 1.094487951352738e-05, "loss": 0.1869, "step": 68358 }, { "epoch": 4.13, "learning_rate": 1.0943370603413155e-05, "loss": 0.3414, "step": 68360 }, { "epoch": 4.13, "learning_rate": 1.0941861693298932e-05, "loss": 0.3144, "step": 68362 }, { "epoch": 4.13, "learning_rate": 1.0940352783184706e-05, "loss": 0.2794, "step": 68364 }, { "epoch": 4.13, "learning_rate": 1.0938843873070481e-05, "loss": 0.1729, "step": 68366 }, { "epoch": 4.13, "learning_rate": 1.0937334962956258e-05, "loss": 0.3242, "step": 68368 }, { "epoch": 4.13, "learning_rate": 1.0935826052842033e-05, "loss": 0.2383, "step": 68370 }, { "epoch": 4.13, "learning_rate": 1.0934317142727809e-05, "loss": 0.2464, "step": 68372 }, { "epoch": 4.13, "learning_rate": 1.0932808232613584e-05, "loss": 0.2764, "step": 68374 }, { "epoch": 4.13, "learning_rate": 1.093129932249936e-05, "loss": 0.1675, "step": 68376 }, { "epoch": 4.13, "learning_rate": 1.0929790412385135e-05, "loss": 0.1936, "step": 68378 }, { "epoch": 4.13, "learning_rate": 1.0928281502270912e-05, "loss": 0.1375, "step": 68380 }, { "epoch": 4.13, "learning_rate": 1.0926772592156685e-05, "loss": 0.2536, "step": 68382 }, { "epoch": 4.13, "learning_rate": 1.0925263682042461e-05, "loss": 0.3169, "step": 68384 }, { "epoch": 4.13, "learning_rate": 1.0923754771928236e-05, "loss": 0.3127, "step": 68386 }, { "epoch": 4.13, "learning_rate": 1.0922245861814013e-05, "loss": 0.2572, "step": 68388 }, { "epoch": 4.13, "learning_rate": 1.0920736951699787e-05, "loss": 0.2156, "step": 68390 }, { "epoch": 4.13, "learning_rate": 1.0919228041585564e-05, "loss": 0.2694, "step": 68392 }, { "epoch": 4.13, "learning_rate": 1.0917719131471339e-05, "loss": 0.1701, "step": 68394 }, { "epoch": 4.13, "learning_rate": 1.0916210221357115e-05, "loss": 0.2654, "step": 68396 }, { "epoch": 4.13, "learning_rate": 1.0914701311242888e-05, "loss": 0.2726, "step": 68398 }, { "epoch": 4.13, "learning_rate": 1.0913192401128665e-05, "loss": 0.2048, "step": 68400 }, { "epoch": 4.13, "learning_rate": 1.091168349101444e-05, "loss": 0.2508, "step": 68402 }, { "epoch": 4.13, "learning_rate": 1.0910174580900216e-05, "loss": 0.2609, "step": 68404 }, { "epoch": 4.13, "learning_rate": 1.0908665670785991e-05, "loss": 0.3851, "step": 68406 }, { "epoch": 4.13, "learning_rate": 1.0907156760671768e-05, "loss": 0.2591, "step": 68408 }, { "epoch": 4.13, "learning_rate": 1.0905647850557542e-05, "loss": 0.2365, "step": 68410 }, { "epoch": 4.13, "learning_rate": 1.0904138940443319e-05, "loss": 0.1899, "step": 68412 }, { "epoch": 4.13, "learning_rate": 1.0902630030329094e-05, "loss": 0.1767, "step": 68414 }, { "epoch": 4.13, "learning_rate": 1.0901121120214869e-05, "loss": 0.2339, "step": 68416 }, { "epoch": 4.13, "learning_rate": 1.0899612210100645e-05, "loss": 0.1322, "step": 68418 }, { "epoch": 4.13, "learning_rate": 1.089810329998642e-05, "loss": 0.2766, "step": 68420 }, { "epoch": 4.13, "learning_rate": 1.0896594389872196e-05, "loss": 0.2556, "step": 68422 }, { "epoch": 4.13, "learning_rate": 1.0895085479757971e-05, "loss": 0.2639, "step": 68424 }, { "epoch": 4.13, "learning_rate": 1.0893576569643748e-05, "loss": 0.1992, "step": 68426 }, { "epoch": 4.13, "learning_rate": 1.0892067659529523e-05, "loss": 0.3758, "step": 68428 }, { "epoch": 4.13, "learning_rate": 1.0890558749415297e-05, "loss": 0.2125, "step": 68430 }, { "epoch": 4.13, "learning_rate": 1.0889049839301072e-05, "loss": 0.2507, "step": 68432 }, { "epoch": 4.13, "learning_rate": 1.0887540929186849e-05, "loss": 0.2459, "step": 68434 }, { "epoch": 4.13, "learning_rate": 1.0886032019072624e-05, "loss": 0.1679, "step": 68436 }, { "epoch": 4.13, "learning_rate": 1.08845231089584e-05, "loss": 0.1837, "step": 68438 }, { "epoch": 4.13, "learning_rate": 1.0883014198844175e-05, "loss": 0.1926, "step": 68440 }, { "epoch": 4.13, "learning_rate": 1.0881505288729951e-05, "loss": 0.291, "step": 68442 }, { "epoch": 4.13, "learning_rate": 1.0879996378615726e-05, "loss": 0.2428, "step": 68444 }, { "epoch": 4.13, "learning_rate": 1.0878487468501503e-05, "loss": 0.1839, "step": 68446 }, { "epoch": 4.13, "learning_rate": 1.0876978558387278e-05, "loss": 0.2479, "step": 68448 }, { "epoch": 4.13, "learning_rate": 1.0875469648273052e-05, "loss": 0.2716, "step": 68450 }, { "epoch": 4.13, "learning_rate": 1.0873960738158829e-05, "loss": 0.2605, "step": 68452 }, { "epoch": 4.13, "learning_rate": 1.0872451828044604e-05, "loss": 0.2461, "step": 68454 }, { "epoch": 4.13, "learning_rate": 1.087094291793038e-05, "loss": 0.2416, "step": 68456 }, { "epoch": 4.13, "learning_rate": 1.0869434007816155e-05, "loss": 0.2744, "step": 68458 }, { "epoch": 4.13, "learning_rate": 1.0867925097701932e-05, "loss": 0.2209, "step": 68460 }, { "epoch": 4.13, "learning_rate": 1.0866416187587706e-05, "loss": 0.1842, "step": 68462 }, { "epoch": 4.13, "learning_rate": 1.0864907277473481e-05, "loss": 0.2967, "step": 68464 }, { "epoch": 4.13, "learning_rate": 1.0863398367359256e-05, "loss": 0.2529, "step": 68466 }, { "epoch": 4.13, "learning_rate": 1.0861889457245033e-05, "loss": 0.2528, "step": 68468 }, { "epoch": 4.13, "learning_rate": 1.0860380547130807e-05, "loss": 0.2306, "step": 68470 }, { "epoch": 4.13, "learning_rate": 1.0858871637016584e-05, "loss": 0.3495, "step": 68472 }, { "epoch": 4.13, "learning_rate": 1.0857362726902359e-05, "loss": 0.2903, "step": 68474 }, { "epoch": 4.13, "learning_rate": 1.0855853816788135e-05, "loss": 0.2525, "step": 68476 }, { "epoch": 4.13, "learning_rate": 1.085434490667391e-05, "loss": 0.2965, "step": 68478 }, { "epoch": 4.13, "learning_rate": 1.0852835996559685e-05, "loss": 0.2241, "step": 68480 }, { "epoch": 4.13, "learning_rate": 1.0851327086445461e-05, "loss": 0.2728, "step": 68482 }, { "epoch": 4.13, "learning_rate": 1.0849818176331236e-05, "loss": 0.2376, "step": 68484 }, { "epoch": 4.13, "learning_rate": 1.0848309266217013e-05, "loss": 0.2153, "step": 68486 }, { "epoch": 4.13, "learning_rate": 1.0846800356102788e-05, "loss": 0.1577, "step": 68488 }, { "epoch": 4.13, "learning_rate": 1.0845291445988564e-05, "loss": 0.4014, "step": 68490 }, { "epoch": 4.13, "learning_rate": 1.0843782535874339e-05, "loss": 0.2075, "step": 68492 }, { "epoch": 4.13, "learning_rate": 1.0842273625760115e-05, "loss": 0.2775, "step": 68494 }, { "epoch": 4.13, "learning_rate": 1.0840764715645889e-05, "loss": 0.3279, "step": 68496 }, { "epoch": 4.13, "learning_rate": 1.0839255805531665e-05, "loss": 0.2263, "step": 68498 }, { "epoch": 4.13, "learning_rate": 1.083774689541744e-05, "loss": 0.2261, "step": 68500 }, { "epoch": 4.13, "learning_rate": 1.0836237985303216e-05, "loss": 0.2646, "step": 68502 }, { "epoch": 4.13, "learning_rate": 1.0834729075188991e-05, "loss": 0.2107, "step": 68504 }, { "epoch": 4.13, "learning_rate": 1.0833220165074768e-05, "loss": 0.1927, "step": 68506 }, { "epoch": 4.13, "learning_rate": 1.0831711254960543e-05, "loss": 0.2412, "step": 68508 }, { "epoch": 4.14, "learning_rate": 1.0830202344846319e-05, "loss": 0.2184, "step": 68510 }, { "epoch": 4.14, "learning_rate": 1.0828693434732092e-05, "loss": 0.1987, "step": 68512 }, { "epoch": 4.14, "learning_rate": 1.0827184524617869e-05, "loss": 0.2039, "step": 68514 }, { "epoch": 4.14, "learning_rate": 1.0825675614503643e-05, "loss": 0.1404, "step": 68516 }, { "epoch": 4.14, "learning_rate": 1.082416670438942e-05, "loss": 0.201, "step": 68518 }, { "epoch": 4.14, "learning_rate": 1.0822657794275195e-05, "loss": 0.2363, "step": 68520 }, { "epoch": 4.14, "learning_rate": 1.0821148884160971e-05, "loss": 0.2844, "step": 68522 }, { "epoch": 4.14, "learning_rate": 1.0819639974046746e-05, "loss": 0.2026, "step": 68524 }, { "epoch": 4.14, "learning_rate": 1.0818131063932523e-05, "loss": 0.2314, "step": 68526 }, { "epoch": 4.14, "learning_rate": 1.0816622153818297e-05, "loss": 0.1651, "step": 68528 }, { "epoch": 4.14, "learning_rate": 1.0815113243704072e-05, "loss": 0.2981, "step": 68530 }, { "epoch": 4.14, "learning_rate": 1.0813604333589849e-05, "loss": 0.3103, "step": 68532 }, { "epoch": 4.14, "learning_rate": 1.0812095423475624e-05, "loss": 0.2834, "step": 68534 }, { "epoch": 4.14, "learning_rate": 1.08105865133614e-05, "loss": 0.3544, "step": 68536 }, { "epoch": 4.14, "learning_rate": 1.0809077603247175e-05, "loss": 0.1608, "step": 68538 }, { "epoch": 4.14, "learning_rate": 1.0807568693132951e-05, "loss": 0.2362, "step": 68540 }, { "epoch": 4.14, "learning_rate": 1.0806059783018726e-05, "loss": 0.2176, "step": 68542 }, { "epoch": 4.14, "learning_rate": 1.0804550872904503e-05, "loss": 0.3192, "step": 68544 }, { "epoch": 4.14, "learning_rate": 1.0803041962790276e-05, "loss": 0.3063, "step": 68546 }, { "epoch": 4.14, "learning_rate": 1.0801533052676052e-05, "loss": 0.2326, "step": 68548 }, { "epoch": 4.14, "learning_rate": 1.0800024142561827e-05, "loss": 0.2131, "step": 68550 }, { "epoch": 4.14, "learning_rate": 1.0798515232447604e-05, "loss": 0.4313, "step": 68552 }, { "epoch": 4.14, "learning_rate": 1.0797006322333379e-05, "loss": 0.1754, "step": 68554 }, { "epoch": 4.14, "learning_rate": 1.0795497412219155e-05, "loss": 0.2738, "step": 68556 }, { "epoch": 4.14, "learning_rate": 1.079398850210493e-05, "loss": 0.253, "step": 68558 }, { "epoch": 4.14, "learning_rate": 1.0792479591990706e-05, "loss": 0.2096, "step": 68560 }, { "epoch": 4.14, "learning_rate": 1.0790970681876481e-05, "loss": 0.2553, "step": 68562 }, { "epoch": 4.14, "learning_rate": 1.0789461771762256e-05, "loss": 0.1947, "step": 68564 }, { "epoch": 4.14, "learning_rate": 1.0787952861648033e-05, "loss": 0.2047, "step": 68566 }, { "epoch": 4.14, "learning_rate": 1.0786443951533807e-05, "loss": 0.1367, "step": 68568 }, { "epoch": 4.14, "learning_rate": 1.0784935041419584e-05, "loss": 0.2031, "step": 68570 }, { "epoch": 4.14, "learning_rate": 1.0783426131305359e-05, "loss": 0.1172, "step": 68572 }, { "epoch": 4.14, "learning_rate": 1.0781917221191135e-05, "loss": 0.1937, "step": 68574 }, { "epoch": 4.14, "learning_rate": 1.078040831107691e-05, "loss": 0.2339, "step": 68576 }, { "epoch": 4.14, "learning_rate": 1.0778899400962685e-05, "loss": 0.2998, "step": 68578 }, { "epoch": 4.14, "learning_rate": 1.077739049084846e-05, "loss": 0.2331, "step": 68580 }, { "epoch": 4.14, "learning_rate": 1.0775881580734236e-05, "loss": 0.2338, "step": 68582 }, { "epoch": 4.14, "learning_rate": 1.0774372670620011e-05, "loss": 0.2582, "step": 68584 }, { "epoch": 4.14, "learning_rate": 1.0772863760505788e-05, "loss": 0.2679, "step": 68586 }, { "epoch": 4.14, "learning_rate": 1.0771354850391562e-05, "loss": 0.2237, "step": 68588 }, { "epoch": 4.14, "learning_rate": 1.0769845940277339e-05, "loss": 0.2813, "step": 68590 }, { "epoch": 4.14, "learning_rate": 1.0768337030163114e-05, "loss": 0.2927, "step": 68592 }, { "epoch": 4.14, "learning_rate": 1.0766828120048889e-05, "loss": 0.2273, "step": 68594 }, { "epoch": 4.14, "learning_rate": 1.0765319209934663e-05, "loss": 0.178, "step": 68596 }, { "epoch": 4.14, "learning_rate": 1.076381029982044e-05, "loss": 0.2336, "step": 68598 }, { "epoch": 4.14, "learning_rate": 1.0762301389706215e-05, "loss": 0.2428, "step": 68600 }, { "epoch": 4.14, "learning_rate": 1.0760792479591991e-05, "loss": 0.3539, "step": 68602 }, { "epoch": 4.14, "learning_rate": 1.0759283569477768e-05, "loss": 0.1968, "step": 68604 }, { "epoch": 4.14, "learning_rate": 1.0757774659363543e-05, "loss": 0.2515, "step": 68606 }, { "epoch": 4.14, "learning_rate": 1.0756265749249319e-05, "loss": 0.2966, "step": 68608 }, { "epoch": 4.14, "learning_rate": 1.0754756839135094e-05, "loss": 0.1927, "step": 68610 }, { "epoch": 4.14, "learning_rate": 1.0753247929020869e-05, "loss": 0.1999, "step": 68612 }, { "epoch": 4.14, "learning_rate": 1.0751739018906644e-05, "loss": 0.2723, "step": 68614 }, { "epoch": 4.14, "learning_rate": 1.075023010879242e-05, "loss": 0.3072, "step": 68616 }, { "epoch": 4.14, "learning_rate": 1.0748721198678195e-05, "loss": 0.2765, "step": 68618 }, { "epoch": 4.14, "learning_rate": 1.0747212288563971e-05, "loss": 0.321, "step": 68620 }, { "epoch": 4.14, "learning_rate": 1.0745703378449746e-05, "loss": 0.2908, "step": 68622 }, { "epoch": 4.14, "learning_rate": 1.0744194468335523e-05, "loss": 0.2702, "step": 68624 }, { "epoch": 4.14, "learning_rate": 1.0742685558221298e-05, "loss": 0.1928, "step": 68626 }, { "epoch": 4.14, "learning_rate": 1.0741176648107072e-05, "loss": 0.1948, "step": 68628 }, { "epoch": 4.14, "learning_rate": 1.0739667737992847e-05, "loss": 0.1727, "step": 68630 }, { "epoch": 4.14, "learning_rate": 1.0738158827878624e-05, "loss": 0.4031, "step": 68632 }, { "epoch": 4.14, "learning_rate": 1.0736649917764398e-05, "loss": 0.2297, "step": 68634 }, { "epoch": 4.14, "learning_rate": 1.0735141007650175e-05, "loss": 0.2409, "step": 68636 }, { "epoch": 4.14, "learning_rate": 1.073363209753595e-05, "loss": 0.2733, "step": 68638 }, { "epoch": 4.14, "learning_rate": 1.0732123187421726e-05, "loss": 0.3051, "step": 68640 }, { "epoch": 4.14, "learning_rate": 1.0730614277307501e-05, "loss": 0.2994, "step": 68642 }, { "epoch": 4.14, "learning_rate": 1.0729105367193276e-05, "loss": 0.231, "step": 68644 }, { "epoch": 4.14, "learning_rate": 1.0727596457079052e-05, "loss": 0.2142, "step": 68646 }, { "epoch": 4.14, "learning_rate": 1.0726087546964827e-05, "loss": 0.18, "step": 68648 }, { "epoch": 4.14, "learning_rate": 1.0724578636850604e-05, "loss": 0.2141, "step": 68650 }, { "epoch": 4.14, "learning_rate": 1.0723069726736379e-05, "loss": 0.1943, "step": 68652 }, { "epoch": 4.14, "learning_rate": 1.0721560816622155e-05, "loss": 0.2715, "step": 68654 }, { "epoch": 4.14, "learning_rate": 1.072005190650793e-05, "loss": 0.2098, "step": 68656 }, { "epoch": 4.14, "learning_rate": 1.0718542996393706e-05, "loss": 0.2066, "step": 68658 }, { "epoch": 4.14, "learning_rate": 1.071703408627948e-05, "loss": 0.1657, "step": 68660 }, { "epoch": 4.14, "learning_rate": 1.0715525176165256e-05, "loss": 0.2307, "step": 68662 }, { "epoch": 4.14, "learning_rate": 1.0714016266051031e-05, "loss": 0.2867, "step": 68664 }, { "epoch": 4.14, "learning_rate": 1.0712507355936807e-05, "loss": 0.3386, "step": 68666 }, { "epoch": 4.14, "learning_rate": 1.0710998445822582e-05, "loss": 0.2042, "step": 68668 }, { "epoch": 4.14, "learning_rate": 1.0709489535708359e-05, "loss": 0.2016, "step": 68670 }, { "epoch": 4.14, "learning_rate": 1.0707980625594134e-05, "loss": 0.2142, "step": 68672 }, { "epoch": 4.14, "learning_rate": 1.070647171547991e-05, "loss": 0.19, "step": 68674 }, { "epoch": 4.15, "learning_rate": 1.0704962805365685e-05, "loss": 0.3127, "step": 68676 }, { "epoch": 4.15, "learning_rate": 1.070345389525146e-05, "loss": 0.1412, "step": 68678 }, { "epoch": 4.15, "learning_rate": 1.0701944985137236e-05, "loss": 0.2943, "step": 68680 }, { "epoch": 4.15, "learning_rate": 1.0700436075023011e-05, "loss": 0.1833, "step": 68682 }, { "epoch": 4.15, "learning_rate": 1.0698927164908788e-05, "loss": 0.3744, "step": 68684 }, { "epoch": 4.15, "learning_rate": 1.0697418254794562e-05, "loss": 0.22, "step": 68686 }, { "epoch": 4.15, "learning_rate": 1.0695909344680339e-05, "loss": 0.1568, "step": 68688 }, { "epoch": 4.15, "learning_rate": 1.0694400434566114e-05, "loss": 0.2392, "step": 68690 }, { "epoch": 4.15, "learning_rate": 1.069289152445189e-05, "loss": 0.2334, "step": 68692 }, { "epoch": 4.15, "learning_rate": 1.0691382614337663e-05, "loss": 0.1975, "step": 68694 }, { "epoch": 4.15, "learning_rate": 1.068987370422344e-05, "loss": 0.3333, "step": 68696 }, { "epoch": 4.15, "learning_rate": 1.0688364794109215e-05, "loss": 0.2173, "step": 68698 }, { "epoch": 4.15, "learning_rate": 1.0686855883994991e-05, "loss": 0.2928, "step": 68700 }, { "epoch": 4.15, "learning_rate": 1.0685346973880766e-05, "loss": 0.2183, "step": 68702 }, { "epoch": 4.15, "learning_rate": 1.0683838063766543e-05, "loss": 0.2682, "step": 68704 }, { "epoch": 4.15, "learning_rate": 1.0682329153652317e-05, "loss": 0.2213, "step": 68706 }, { "epoch": 4.15, "learning_rate": 1.0680820243538094e-05, "loss": 0.2247, "step": 68708 }, { "epoch": 4.15, "learning_rate": 1.0679311333423867e-05, "loss": 0.1633, "step": 68710 }, { "epoch": 4.15, "learning_rate": 1.0677802423309644e-05, "loss": 0.2064, "step": 68712 }, { "epoch": 4.15, "learning_rate": 1.0676293513195418e-05, "loss": 0.2334, "step": 68714 }, { "epoch": 4.15, "learning_rate": 1.0674784603081195e-05, "loss": 0.167, "step": 68716 }, { "epoch": 4.15, "learning_rate": 1.067327569296697e-05, "loss": 0.2853, "step": 68718 }, { "epoch": 4.15, "learning_rate": 1.0671766782852746e-05, "loss": 0.2263, "step": 68720 }, { "epoch": 4.15, "learning_rate": 1.0670257872738523e-05, "loss": 0.2653, "step": 68722 }, { "epoch": 4.15, "learning_rate": 1.0668748962624298e-05, "loss": 0.3341, "step": 68724 }, { "epoch": 4.15, "learning_rate": 1.0667240052510072e-05, "loss": 0.2592, "step": 68726 }, { "epoch": 4.15, "learning_rate": 1.0665731142395847e-05, "loss": 0.2177, "step": 68728 }, { "epoch": 4.15, "learning_rate": 1.0664222232281624e-05, "loss": 0.2008, "step": 68730 }, { "epoch": 4.15, "learning_rate": 1.0662713322167399e-05, "loss": 0.303, "step": 68732 }, { "epoch": 4.15, "learning_rate": 1.0661204412053175e-05, "loss": 0.1894, "step": 68734 }, { "epoch": 4.15, "learning_rate": 1.065969550193895e-05, "loss": 0.158, "step": 68736 }, { "epoch": 4.15, "learning_rate": 1.0658186591824726e-05, "loss": 0.1509, "step": 68738 }, { "epoch": 4.15, "learning_rate": 1.0656677681710501e-05, "loss": 0.2305, "step": 68740 }, { "epoch": 4.15, "learning_rate": 1.0655168771596276e-05, "loss": 0.4389, "step": 68742 }, { "epoch": 4.15, "learning_rate": 1.065365986148205e-05, "loss": 0.2193, "step": 68744 }, { "epoch": 4.15, "learning_rate": 1.0652150951367827e-05, "loss": 0.2022, "step": 68746 }, { "epoch": 4.15, "learning_rate": 1.0650642041253602e-05, "loss": 0.3525, "step": 68748 }, { "epoch": 4.15, "learning_rate": 1.0649133131139379e-05, "loss": 0.2241, "step": 68750 }, { "epoch": 4.15, "learning_rate": 1.0647624221025153e-05, "loss": 0.2164, "step": 68752 }, { "epoch": 4.15, "learning_rate": 1.064611531091093e-05, "loss": 0.3027, "step": 68754 }, { "epoch": 4.15, "learning_rate": 1.0644606400796705e-05, "loss": 0.2517, "step": 68756 }, { "epoch": 4.15, "learning_rate": 1.0643097490682481e-05, "loss": 0.183, "step": 68758 }, { "epoch": 4.15, "learning_rate": 1.0641588580568256e-05, "loss": 0.2458, "step": 68760 }, { "epoch": 4.15, "learning_rate": 1.0640079670454031e-05, "loss": 0.2568, "step": 68762 }, { "epoch": 4.15, "learning_rate": 1.0638570760339807e-05, "loss": 0.2158, "step": 68764 }, { "epoch": 4.15, "learning_rate": 1.0637061850225582e-05, "loss": 0.2161, "step": 68766 }, { "epoch": 4.15, "learning_rate": 1.0635552940111359e-05, "loss": 0.4136, "step": 68768 }, { "epoch": 4.15, "learning_rate": 1.0634044029997134e-05, "loss": 0.3183, "step": 68770 }, { "epoch": 4.15, "learning_rate": 1.063253511988291e-05, "loss": 0.2414, "step": 68772 }, { "epoch": 4.15, "learning_rate": 1.0631026209768685e-05, "loss": 0.2911, "step": 68774 }, { "epoch": 4.15, "learning_rate": 1.062951729965446e-05, "loss": 0.1727, "step": 68776 }, { "epoch": 4.15, "learning_rate": 1.0628008389540235e-05, "loss": 0.2404, "step": 68778 }, { "epoch": 4.15, "learning_rate": 1.0626499479426011e-05, "loss": 0.2166, "step": 68780 }, { "epoch": 4.15, "learning_rate": 1.0624990569311786e-05, "loss": 0.2715, "step": 68782 }, { "epoch": 4.15, "learning_rate": 1.0623481659197562e-05, "loss": 0.1868, "step": 68784 }, { "epoch": 4.15, "learning_rate": 1.0621972749083337e-05, "loss": 0.2236, "step": 68786 }, { "epoch": 4.15, "learning_rate": 1.0620463838969114e-05, "loss": 0.1988, "step": 68788 }, { "epoch": 4.15, "learning_rate": 1.0618954928854889e-05, "loss": 0.1806, "step": 68790 }, { "epoch": 4.15, "learning_rate": 1.0617446018740663e-05, "loss": 0.1537, "step": 68792 }, { "epoch": 4.15, "learning_rate": 1.061593710862644e-05, "loss": 0.1537, "step": 68794 }, { "epoch": 4.15, "learning_rate": 1.0614428198512215e-05, "loss": 0.3467, "step": 68796 }, { "epoch": 4.15, "learning_rate": 1.0612919288397991e-05, "loss": 0.2268, "step": 68798 }, { "epoch": 4.15, "learning_rate": 1.0611410378283766e-05, "loss": 0.2762, "step": 68800 }, { "epoch": 4.15, "learning_rate": 1.0609901468169543e-05, "loss": 0.1922, "step": 68802 }, { "epoch": 4.15, "learning_rate": 1.0608392558055317e-05, "loss": 0.2611, "step": 68804 }, { "epoch": 4.15, "learning_rate": 1.0606883647941094e-05, "loss": 0.2515, "step": 68806 }, { "epoch": 4.15, "learning_rate": 1.0605374737826867e-05, "loss": 0.1269, "step": 68808 }, { "epoch": 4.15, "learning_rate": 1.0603865827712644e-05, "loss": 0.2215, "step": 68810 }, { "epoch": 4.15, "learning_rate": 1.0602356917598418e-05, "loss": 0.3502, "step": 68812 }, { "epoch": 4.15, "learning_rate": 1.0600848007484195e-05, "loss": 0.1771, "step": 68814 }, { "epoch": 4.15, "learning_rate": 1.059933909736997e-05, "loss": 0.2337, "step": 68816 }, { "epoch": 4.15, "learning_rate": 1.0597830187255746e-05, "loss": 0.212, "step": 68818 }, { "epoch": 4.15, "learning_rate": 1.0596321277141521e-05, "loss": 0.2002, "step": 68820 }, { "epoch": 4.15, "learning_rate": 1.0594812367027298e-05, "loss": 0.2813, "step": 68822 }, { "epoch": 4.15, "learning_rate": 1.059330345691307e-05, "loss": 0.1759, "step": 68824 }, { "epoch": 4.15, "learning_rate": 1.0591794546798847e-05, "loss": 0.2341, "step": 68826 }, { "epoch": 4.15, "learning_rate": 1.0590285636684622e-05, "loss": 0.1369, "step": 68828 }, { "epoch": 4.15, "learning_rate": 1.0588776726570399e-05, "loss": 0.1881, "step": 68830 }, { "epoch": 4.15, "learning_rate": 1.0587267816456173e-05, "loss": 0.2271, "step": 68832 }, { "epoch": 4.15, "learning_rate": 1.058575890634195e-05, "loss": 0.2069, "step": 68834 }, { "epoch": 4.15, "learning_rate": 1.0584249996227725e-05, "loss": 0.1993, "step": 68836 }, { "epoch": 4.15, "learning_rate": 1.0582741086113501e-05, "loss": 0.2475, "step": 68838 }, { "epoch": 4.15, "learning_rate": 1.0581232175999278e-05, "loss": 0.2916, "step": 68840 }, { "epoch": 4.16, "learning_rate": 1.0579723265885051e-05, "loss": 0.2809, "step": 68842 }, { "epoch": 4.16, "learning_rate": 1.0578214355770827e-05, "loss": 0.2199, "step": 68844 }, { "epoch": 4.16, "learning_rate": 1.0576705445656602e-05, "loss": 0.3733, "step": 68846 }, { "epoch": 4.16, "learning_rate": 1.0575196535542379e-05, "loss": 0.3369, "step": 68848 }, { "epoch": 4.16, "learning_rate": 1.0573687625428154e-05, "loss": 0.2014, "step": 68850 }, { "epoch": 4.16, "learning_rate": 1.057217871531393e-05, "loss": 0.2259, "step": 68852 }, { "epoch": 4.16, "learning_rate": 1.0570669805199705e-05, "loss": 0.2152, "step": 68854 }, { "epoch": 4.16, "learning_rate": 1.0569160895085481e-05, "loss": 0.2353, "step": 68856 }, { "epoch": 4.16, "learning_rate": 1.0567651984971255e-05, "loss": 0.1773, "step": 68858 }, { "epoch": 4.16, "learning_rate": 1.0566143074857031e-05, "loss": 0.3378, "step": 68860 }, { "epoch": 4.16, "learning_rate": 1.0564634164742806e-05, "loss": 0.2262, "step": 68862 }, { "epoch": 4.16, "learning_rate": 1.0563125254628582e-05, "loss": 0.1414, "step": 68864 }, { "epoch": 4.16, "learning_rate": 1.0561616344514357e-05, "loss": 0.1733, "step": 68866 }, { "epoch": 4.16, "learning_rate": 1.0560107434400134e-05, "loss": 0.235, "step": 68868 }, { "epoch": 4.16, "learning_rate": 1.0558598524285909e-05, "loss": 0.2022, "step": 68870 }, { "epoch": 4.16, "learning_rate": 1.0557089614171685e-05, "loss": 0.1696, "step": 68872 }, { "epoch": 4.16, "learning_rate": 1.055558070405746e-05, "loss": 0.2792, "step": 68874 }, { "epoch": 4.16, "learning_rate": 1.0554071793943235e-05, "loss": 0.2093, "step": 68876 }, { "epoch": 4.16, "learning_rate": 1.0552562883829011e-05, "loss": 0.2675, "step": 68878 }, { "epoch": 4.16, "learning_rate": 1.0551053973714786e-05, "loss": 0.1854, "step": 68880 }, { "epoch": 4.16, "learning_rate": 1.0549545063600563e-05, "loss": 0.1973, "step": 68882 }, { "epoch": 4.16, "learning_rate": 1.0548036153486337e-05, "loss": 0.2992, "step": 68884 }, { "epoch": 4.16, "learning_rate": 1.0546527243372114e-05, "loss": 0.2148, "step": 68886 }, { "epoch": 4.16, "learning_rate": 1.0545018333257889e-05, "loss": 0.2581, "step": 68888 }, { "epoch": 4.16, "learning_rate": 1.0543509423143663e-05, "loss": 0.264, "step": 68890 }, { "epoch": 4.16, "learning_rate": 1.0542000513029438e-05, "loss": 0.2138, "step": 68892 }, { "epoch": 4.16, "learning_rate": 1.0540491602915215e-05, "loss": 0.2952, "step": 68894 }, { "epoch": 4.16, "learning_rate": 1.053898269280099e-05, "loss": 0.2679, "step": 68896 }, { "epoch": 4.16, "learning_rate": 1.0537473782686766e-05, "loss": 0.2654, "step": 68898 }, { "epoch": 4.16, "learning_rate": 1.0535964872572541e-05, "loss": 0.2147, "step": 68900 }, { "epoch": 4.16, "learning_rate": 1.0534455962458317e-05, "loss": 0.1598, "step": 68902 }, { "epoch": 4.16, "learning_rate": 1.0532947052344092e-05, "loss": 0.2854, "step": 68904 }, { "epoch": 4.16, "learning_rate": 1.0531438142229867e-05, "loss": 0.1433, "step": 68906 }, { "epoch": 4.16, "learning_rate": 1.0529929232115644e-05, "loss": 0.266, "step": 68908 }, { "epoch": 4.16, "learning_rate": 1.0528420322001418e-05, "loss": 0.163, "step": 68910 }, { "epoch": 4.16, "learning_rate": 1.0526911411887195e-05, "loss": 0.2916, "step": 68912 }, { "epoch": 4.16, "learning_rate": 1.052540250177297e-05, "loss": 0.3312, "step": 68914 }, { "epoch": 4.16, "learning_rate": 1.0523893591658746e-05, "loss": 0.1921, "step": 68916 }, { "epoch": 4.16, "learning_rate": 1.0522384681544521e-05, "loss": 0.1933, "step": 68918 }, { "epoch": 4.16, "learning_rate": 1.0520875771430298e-05, "loss": 0.3465, "step": 68920 }, { "epoch": 4.16, "learning_rate": 1.0519366861316072e-05, "loss": 0.2402, "step": 68922 }, { "epoch": 4.16, "learning_rate": 1.0517857951201847e-05, "loss": 0.3544, "step": 68924 }, { "epoch": 4.16, "learning_rate": 1.0516349041087622e-05, "loss": 0.1911, "step": 68926 }, { "epoch": 4.16, "learning_rate": 1.0514840130973399e-05, "loss": 0.1635, "step": 68928 }, { "epoch": 4.16, "learning_rate": 1.0513331220859173e-05, "loss": 0.2151, "step": 68930 }, { "epoch": 4.16, "learning_rate": 1.051182231074495e-05, "loss": 0.1792, "step": 68932 }, { "epoch": 4.16, "learning_rate": 1.0510313400630725e-05, "loss": 0.1617, "step": 68934 }, { "epoch": 4.16, "learning_rate": 1.0508804490516501e-05, "loss": 0.2804, "step": 68936 }, { "epoch": 4.16, "learning_rate": 1.0507295580402276e-05, "loss": 0.2044, "step": 68938 }, { "epoch": 4.16, "learning_rate": 1.0505786670288051e-05, "loss": 0.164, "step": 68940 }, { "epoch": 4.16, "learning_rate": 1.0504277760173826e-05, "loss": 0.1789, "step": 68942 }, { "epoch": 4.16, "learning_rate": 1.0502768850059602e-05, "loss": 0.2103, "step": 68944 }, { "epoch": 4.16, "learning_rate": 1.0501259939945377e-05, "loss": 0.2579, "step": 68946 }, { "epoch": 4.16, "learning_rate": 1.0499751029831154e-05, "loss": 0.2702, "step": 68948 }, { "epoch": 4.16, "learning_rate": 1.0498242119716928e-05, "loss": 0.279, "step": 68950 }, { "epoch": 4.16, "learning_rate": 1.0496733209602705e-05, "loss": 0.3012, "step": 68952 }, { "epoch": 4.16, "learning_rate": 1.049522429948848e-05, "loss": 0.2885, "step": 68954 }, { "epoch": 4.16, "learning_rate": 1.0493715389374255e-05, "loss": 0.3922, "step": 68956 }, { "epoch": 4.16, "learning_rate": 1.0492206479260031e-05, "loss": 0.3137, "step": 68958 }, { "epoch": 4.16, "learning_rate": 1.0490697569145806e-05, "loss": 0.1865, "step": 68960 }, { "epoch": 4.16, "learning_rate": 1.0489188659031582e-05, "loss": 0.2162, "step": 68962 }, { "epoch": 4.16, "learning_rate": 1.0487679748917357e-05, "loss": 0.2368, "step": 68964 }, { "epoch": 4.16, "learning_rate": 1.0486170838803134e-05, "loss": 0.183, "step": 68966 }, { "epoch": 4.16, "learning_rate": 1.0484661928688909e-05, "loss": 0.1863, "step": 68968 }, { "epoch": 4.16, "learning_rate": 1.0483153018574685e-05, "loss": 0.2105, "step": 68970 }, { "epoch": 4.16, "learning_rate": 1.0481644108460458e-05, "loss": 0.2962, "step": 68972 }, { "epoch": 4.16, "learning_rate": 1.0480135198346235e-05, "loss": 0.2629, "step": 68974 }, { "epoch": 4.16, "learning_rate": 1.047862628823201e-05, "loss": 0.1888, "step": 68976 }, { "epoch": 4.16, "learning_rate": 1.0477117378117786e-05, "loss": 0.1945, "step": 68978 }, { "epoch": 4.16, "learning_rate": 1.0475608468003561e-05, "loss": 0.1623, "step": 68980 }, { "epoch": 4.16, "learning_rate": 1.0474099557889337e-05, "loss": 0.2594, "step": 68982 }, { "epoch": 4.16, "learning_rate": 1.0472590647775112e-05, "loss": 0.1898, "step": 68984 }, { "epoch": 4.16, "learning_rate": 1.0471081737660889e-05, "loss": 0.2221, "step": 68986 }, { "epoch": 4.16, "learning_rate": 1.0469572827546664e-05, "loss": 0.3006, "step": 68988 }, { "epoch": 4.16, "learning_rate": 1.0468063917432438e-05, "loss": 0.2235, "step": 68990 }, { "epoch": 4.16, "learning_rate": 1.0466555007318215e-05, "loss": 0.1533, "step": 68992 }, { "epoch": 4.16, "learning_rate": 1.046504609720399e-05, "loss": 0.2194, "step": 68994 }, { "epoch": 4.16, "learning_rate": 1.0463537187089766e-05, "loss": 0.1933, "step": 68996 }, { "epoch": 4.16, "learning_rate": 1.0462028276975541e-05, "loss": 0.2497, "step": 68998 }, { "epoch": 4.16, "learning_rate": 1.0460519366861318e-05, "loss": 0.1986, "step": 69000 }, { "epoch": 4.16, "learning_rate": 1.0459010456747092e-05, "loss": 0.1847, "step": 69002 }, { "epoch": 4.16, "learning_rate": 1.0457501546632869e-05, "loss": 0.3942, "step": 69004 }, { "epoch": 4.17, "learning_rate": 1.0455992636518642e-05, "loss": 0.2435, "step": 69006 }, { "epoch": 4.17, "learning_rate": 1.0454483726404418e-05, "loss": 0.2226, "step": 69008 }, { "epoch": 4.17, "learning_rate": 1.0452974816290193e-05, "loss": 0.1913, "step": 69010 }, { "epoch": 4.17, "learning_rate": 1.045146590617597e-05, "loss": 0.2373, "step": 69012 }, { "epoch": 4.17, "learning_rate": 1.0449956996061745e-05, "loss": 0.203, "step": 69014 }, { "epoch": 4.17, "learning_rate": 1.0448448085947521e-05, "loss": 0.2206, "step": 69016 }, { "epoch": 4.17, "learning_rate": 1.0446939175833296e-05, "loss": 0.1879, "step": 69018 }, { "epoch": 4.17, "learning_rate": 1.0445430265719072e-05, "loss": 0.2684, "step": 69020 }, { "epoch": 4.17, "learning_rate": 1.0443921355604847e-05, "loss": 0.2234, "step": 69022 }, { "epoch": 4.17, "learning_rate": 1.0442412445490622e-05, "loss": 0.2904, "step": 69024 }, { "epoch": 4.17, "learning_rate": 1.0440903535376399e-05, "loss": 0.217, "step": 69026 }, { "epoch": 4.17, "learning_rate": 1.0439394625262173e-05, "loss": 0.2643, "step": 69028 }, { "epoch": 4.17, "learning_rate": 1.043788571514795e-05, "loss": 0.2062, "step": 69030 }, { "epoch": 4.17, "learning_rate": 1.0436376805033725e-05, "loss": 0.1127, "step": 69032 }, { "epoch": 4.17, "learning_rate": 1.0434867894919501e-05, "loss": 0.3406, "step": 69034 }, { "epoch": 4.17, "learning_rate": 1.0433358984805276e-05, "loss": 0.2377, "step": 69036 }, { "epoch": 4.17, "learning_rate": 1.0431850074691051e-05, "loss": 0.2636, "step": 69038 }, { "epoch": 4.17, "learning_rate": 1.0430341164576826e-05, "loss": 0.2075, "step": 69040 }, { "epoch": 4.17, "learning_rate": 1.0428832254462602e-05, "loss": 0.2509, "step": 69042 }, { "epoch": 4.17, "learning_rate": 1.0427323344348377e-05, "loss": 0.181, "step": 69044 }, { "epoch": 4.17, "learning_rate": 1.0425814434234154e-05, "loss": 0.2417, "step": 69046 }, { "epoch": 4.17, "learning_rate": 1.0424305524119928e-05, "loss": 0.2603, "step": 69048 }, { "epoch": 4.17, "learning_rate": 1.0422796614005705e-05, "loss": 0.2715, "step": 69050 }, { "epoch": 4.17, "learning_rate": 1.042128770389148e-05, "loss": 0.1908, "step": 69052 }, { "epoch": 4.17, "learning_rate": 1.0419778793777255e-05, "loss": 0.196, "step": 69054 }, { "epoch": 4.17, "learning_rate": 1.041826988366303e-05, "loss": 0.2119, "step": 69056 }, { "epoch": 4.17, "learning_rate": 1.0416760973548806e-05, "loss": 0.2416, "step": 69058 }, { "epoch": 4.17, "learning_rate": 1.041525206343458e-05, "loss": 0.2014, "step": 69060 }, { "epoch": 4.17, "learning_rate": 1.0413743153320357e-05, "loss": 0.3022, "step": 69062 }, { "epoch": 4.17, "learning_rate": 1.0412234243206132e-05, "loss": 0.206, "step": 69064 }, { "epoch": 4.17, "learning_rate": 1.0410725333091909e-05, "loss": 0.2773, "step": 69066 }, { "epoch": 4.17, "learning_rate": 1.0409216422977683e-05, "loss": 0.268, "step": 69068 }, { "epoch": 4.17, "learning_rate": 1.040770751286346e-05, "loss": 0.2648, "step": 69070 }, { "epoch": 4.17, "learning_rate": 1.0406198602749235e-05, "loss": 0.1807, "step": 69072 }, { "epoch": 4.17, "learning_rate": 1.040468969263501e-05, "loss": 0.1635, "step": 69074 }, { "epoch": 4.17, "learning_rate": 1.0403180782520786e-05, "loss": 0.3544, "step": 69076 }, { "epoch": 4.17, "learning_rate": 1.0401671872406561e-05, "loss": 0.2267, "step": 69078 }, { "epoch": 4.17, "learning_rate": 1.0400162962292337e-05, "loss": 0.1749, "step": 69080 }, { "epoch": 4.17, "learning_rate": 1.0398654052178112e-05, "loss": 0.2362, "step": 69082 }, { "epoch": 4.17, "learning_rate": 1.0397145142063889e-05, "loss": 0.2374, "step": 69084 }, { "epoch": 4.17, "learning_rate": 1.0395636231949664e-05, "loss": 0.2848, "step": 69086 }, { "epoch": 4.17, "learning_rate": 1.0394127321835438e-05, "loss": 0.3159, "step": 69088 }, { "epoch": 4.17, "learning_rate": 1.0392618411721213e-05, "loss": 0.252, "step": 69090 }, { "epoch": 4.17, "learning_rate": 1.039110950160699e-05, "loss": 0.1264, "step": 69092 }, { "epoch": 4.17, "learning_rate": 1.0389600591492765e-05, "loss": 0.2172, "step": 69094 }, { "epoch": 4.17, "learning_rate": 1.0388091681378541e-05, "loss": 0.174, "step": 69096 }, { "epoch": 4.17, "learning_rate": 1.0386582771264316e-05, "loss": 0.2574, "step": 69098 }, { "epoch": 4.17, "learning_rate": 1.0385073861150092e-05, "loss": 0.2376, "step": 69100 }, { "epoch": 4.17, "learning_rate": 1.0383564951035867e-05, "loss": 0.2452, "step": 69102 }, { "epoch": 4.17, "learning_rate": 1.0382056040921642e-05, "loss": 0.272, "step": 69104 }, { "epoch": 4.17, "learning_rate": 1.0380547130807419e-05, "loss": 0.1752, "step": 69106 }, { "epoch": 4.17, "learning_rate": 1.0379038220693193e-05, "loss": 0.1434, "step": 69108 }, { "epoch": 4.17, "learning_rate": 1.037752931057897e-05, "loss": 0.1646, "step": 69110 }, { "epoch": 4.17, "learning_rate": 1.0376020400464745e-05, "loss": 0.3221, "step": 69112 }, { "epoch": 4.17, "learning_rate": 1.0374511490350521e-05, "loss": 0.2077, "step": 69114 }, { "epoch": 4.17, "learning_rate": 1.0373002580236296e-05, "loss": 0.1566, "step": 69116 }, { "epoch": 4.17, "learning_rate": 1.0371493670122073e-05, "loss": 0.191, "step": 69118 }, { "epoch": 4.17, "learning_rate": 1.0369984760007846e-05, "loss": 0.2254, "step": 69120 }, { "epoch": 4.17, "learning_rate": 1.0368475849893622e-05, "loss": 0.229, "step": 69122 }, { "epoch": 4.17, "learning_rate": 1.0366966939779397e-05, "loss": 0.3509, "step": 69124 }, { "epoch": 4.17, "learning_rate": 1.0365458029665174e-05, "loss": 0.2537, "step": 69126 }, { "epoch": 4.17, "learning_rate": 1.0363949119550948e-05, "loss": 0.2546, "step": 69128 }, { "epoch": 4.17, "learning_rate": 1.0362440209436725e-05, "loss": 0.246, "step": 69130 }, { "epoch": 4.17, "learning_rate": 1.03609312993225e-05, "loss": 0.1513, "step": 69132 }, { "epoch": 4.17, "learning_rate": 1.0359422389208276e-05, "loss": 0.268, "step": 69134 }, { "epoch": 4.17, "learning_rate": 1.0357913479094051e-05, "loss": 0.1741, "step": 69136 }, { "epoch": 4.17, "learning_rate": 1.0356404568979826e-05, "loss": 0.2065, "step": 69138 }, { "epoch": 4.17, "learning_rate": 1.0354895658865602e-05, "loss": 0.1664, "step": 69140 }, { "epoch": 4.17, "learning_rate": 1.0353386748751377e-05, "loss": 0.1387, "step": 69142 }, { "epoch": 4.17, "learning_rate": 1.0351877838637154e-05, "loss": 0.1882, "step": 69144 }, { "epoch": 4.17, "learning_rate": 1.0350368928522928e-05, "loss": 0.1736, "step": 69146 }, { "epoch": 4.17, "learning_rate": 1.0348860018408705e-05, "loss": 0.1375, "step": 69148 }, { "epoch": 4.17, "learning_rate": 1.034735110829448e-05, "loss": 0.2683, "step": 69150 }, { "epoch": 4.17, "learning_rate": 1.0345842198180256e-05, "loss": 0.2402, "step": 69152 }, { "epoch": 4.17, "learning_rate": 1.034433328806603e-05, "loss": 0.1554, "step": 69154 }, { "epoch": 4.17, "learning_rate": 1.0343578833008918e-05, "loss": 0.2296, "step": 69156 }, { "epoch": 4.17, "learning_rate": 1.0342069922894694e-05, "loss": 0.1771, "step": 69158 }, { "epoch": 4.17, "learning_rate": 1.0340561012780469e-05, "loss": 0.2987, "step": 69160 }, { "epoch": 4.17, "learning_rate": 1.0339052102666244e-05, "loss": 0.3642, "step": 69162 }, { "epoch": 4.17, "learning_rate": 1.033754319255202e-05, "loss": 0.2193, "step": 69164 }, { "epoch": 4.17, "learning_rate": 1.0336034282437795e-05, "loss": 0.1959, "step": 69166 }, { "epoch": 4.17, "learning_rate": 1.0334525372323572e-05, "loss": 0.238, "step": 69168 }, { "epoch": 4.17, "learning_rate": 1.0333016462209347e-05, "loss": 0.3289, "step": 69170 }, { "epoch": 4.18, "learning_rate": 1.0331507552095123e-05, "loss": 0.2466, "step": 69172 }, { "epoch": 4.18, "learning_rate": 1.0329998641980898e-05, "loss": 0.2122, "step": 69174 }, { "epoch": 4.18, "learning_rate": 1.0328489731866674e-05, "loss": 0.296, "step": 69176 }, { "epoch": 4.18, "learning_rate": 1.0326980821752447e-05, "loss": 0.2425, "step": 69178 }, { "epoch": 4.18, "learning_rate": 1.0325471911638224e-05, "loss": 0.298, "step": 69180 }, { "epoch": 4.18, "learning_rate": 1.0323963001523999e-05, "loss": 0.269, "step": 69182 }, { "epoch": 4.18, "learning_rate": 1.0322454091409775e-05, "loss": 0.3487, "step": 69184 }, { "epoch": 4.18, "learning_rate": 1.032094518129555e-05, "loss": 0.2489, "step": 69186 }, { "epoch": 4.18, "learning_rate": 1.0319436271181327e-05, "loss": 0.2575, "step": 69188 }, { "epoch": 4.18, "learning_rate": 1.0317927361067101e-05, "loss": 0.17, "step": 69190 }, { "epoch": 4.18, "learning_rate": 1.0316418450952878e-05, "loss": 0.3422, "step": 69192 }, { "epoch": 4.18, "learning_rate": 1.0314909540838653e-05, "loss": 0.2954, "step": 69194 }, { "epoch": 4.18, "learning_rate": 1.0313400630724428e-05, "loss": 0.232, "step": 69196 }, { "epoch": 4.18, "learning_rate": 1.0311891720610204e-05, "loss": 0.2104, "step": 69198 }, { "epoch": 4.18, "learning_rate": 1.0310382810495979e-05, "loss": 0.3317, "step": 69200 }, { "epoch": 4.18, "learning_rate": 1.0308873900381755e-05, "loss": 0.1108, "step": 69202 }, { "epoch": 4.18, "learning_rate": 1.030736499026753e-05, "loss": 0.3107, "step": 69204 }, { "epoch": 4.18, "learning_rate": 1.0305856080153307e-05, "loss": 0.2719, "step": 69206 }, { "epoch": 4.18, "learning_rate": 1.0304347170039082e-05, "loss": 0.2925, "step": 69208 }, { "epoch": 4.18, "learning_rate": 1.0302838259924858e-05, "loss": 0.2343, "step": 69210 }, { "epoch": 4.18, "learning_rate": 1.0301329349810631e-05, "loss": 0.1248, "step": 69212 }, { "epoch": 4.18, "learning_rate": 1.0299820439696408e-05, "loss": 0.1085, "step": 69214 }, { "epoch": 4.18, "learning_rate": 1.0298311529582183e-05, "loss": 0.188, "step": 69216 }, { "epoch": 4.18, "learning_rate": 1.0296802619467959e-05, "loss": 0.2501, "step": 69218 }, { "epoch": 4.18, "learning_rate": 1.0295293709353734e-05, "loss": 0.2885, "step": 69220 }, { "epoch": 4.18, "learning_rate": 1.029378479923951e-05, "loss": 0.2074, "step": 69222 }, { "epoch": 4.18, "learning_rate": 1.0292275889125285e-05, "loss": 0.2048, "step": 69224 }, { "epoch": 4.18, "learning_rate": 1.0290766979011062e-05, "loss": 0.3472, "step": 69226 }, { "epoch": 4.18, "learning_rate": 1.0289258068896835e-05, "loss": 0.2557, "step": 69228 }, { "epoch": 4.18, "learning_rate": 1.0287749158782611e-05, "loss": 0.304, "step": 69230 }, { "epoch": 4.18, "learning_rate": 1.0286240248668386e-05, "loss": 0.1826, "step": 69232 }, { "epoch": 4.18, "learning_rate": 1.0284731338554163e-05, "loss": 0.2853, "step": 69234 }, { "epoch": 4.18, "learning_rate": 1.0283222428439938e-05, "loss": 0.2792, "step": 69236 }, { "epoch": 4.18, "learning_rate": 1.0281713518325714e-05, "loss": 0.1723, "step": 69238 }, { "epoch": 4.18, "learning_rate": 1.0280204608211489e-05, "loss": 0.2735, "step": 69240 }, { "epoch": 4.18, "learning_rate": 1.0278695698097265e-05, "loss": 0.2138, "step": 69242 }, { "epoch": 4.18, "learning_rate": 1.027718678798304e-05, "loss": 0.386, "step": 69244 }, { "epoch": 4.18, "learning_rate": 1.0275677877868815e-05, "loss": 0.1862, "step": 69246 }, { "epoch": 4.18, "learning_rate": 1.0274168967754592e-05, "loss": 0.283, "step": 69248 }, { "epoch": 4.18, "learning_rate": 1.0272660057640366e-05, "loss": 0.1427, "step": 69250 }, { "epoch": 4.18, "learning_rate": 1.0271151147526143e-05, "loss": 0.2364, "step": 69252 }, { "epoch": 4.18, "learning_rate": 1.0269642237411918e-05, "loss": 0.1967, "step": 69254 }, { "epoch": 4.18, "learning_rate": 1.0268133327297694e-05, "loss": 0.2237, "step": 69256 }, { "epoch": 4.18, "learning_rate": 1.0266624417183469e-05, "loss": 0.3044, "step": 69258 }, { "epoch": 4.18, "learning_rate": 1.0265115507069244e-05, "loss": 0.314, "step": 69260 }, { "epoch": 4.18, "learning_rate": 1.0263606596955019e-05, "loss": 0.263, "step": 69262 }, { "epoch": 4.18, "learning_rate": 1.0262097686840795e-05, "loss": 0.175, "step": 69264 }, { "epoch": 4.18, "learning_rate": 1.026058877672657e-05, "loss": 0.3427, "step": 69266 }, { "epoch": 4.18, "learning_rate": 1.0259079866612347e-05, "loss": 0.1956, "step": 69268 }, { "epoch": 4.18, "learning_rate": 1.0257570956498121e-05, "loss": 0.2885, "step": 69270 }, { "epoch": 4.18, "learning_rate": 1.0256062046383898e-05, "loss": 0.2291, "step": 69272 }, { "epoch": 4.18, "learning_rate": 1.0254553136269673e-05, "loss": 0.2683, "step": 69274 }, { "epoch": 4.18, "learning_rate": 1.025304422615545e-05, "loss": 0.2585, "step": 69276 }, { "epoch": 4.18, "learning_rate": 1.0251535316041224e-05, "loss": 0.2609, "step": 69278 }, { "epoch": 4.18, "learning_rate": 1.0250026405926999e-05, "loss": 0.224, "step": 69280 }, { "epoch": 4.18, "learning_rate": 1.0248517495812775e-05, "loss": 0.2164, "step": 69282 }, { "epoch": 4.18, "learning_rate": 1.024700858569855e-05, "loss": 0.2143, "step": 69284 }, { "epoch": 4.18, "learning_rate": 1.0245499675584327e-05, "loss": 0.2337, "step": 69286 }, { "epoch": 4.18, "learning_rate": 1.0243990765470102e-05, "loss": 0.2241, "step": 69288 }, { "epoch": 4.18, "learning_rate": 1.0242481855355878e-05, "loss": 0.2394, "step": 69290 }, { "epoch": 4.18, "learning_rate": 1.0240972945241653e-05, "loss": 0.2233, "step": 69292 }, { "epoch": 4.18, "learning_rate": 1.0239464035127428e-05, "loss": 0.2514, "step": 69294 }, { "epoch": 4.18, "learning_rate": 1.0237955125013203e-05, "loss": 0.2492, "step": 69296 }, { "epoch": 4.18, "learning_rate": 1.0236446214898979e-05, "loss": 0.1889, "step": 69298 }, { "epoch": 4.18, "learning_rate": 1.0234937304784754e-05, "loss": 0.3127, "step": 69300 }, { "epoch": 4.18, "learning_rate": 1.023342839467053e-05, "loss": 0.1547, "step": 69302 }, { "epoch": 4.18, "learning_rate": 1.0231919484556305e-05, "loss": 0.3183, "step": 69304 }, { "epoch": 4.18, "learning_rate": 1.0230410574442082e-05, "loss": 0.197, "step": 69306 }, { "epoch": 4.18, "learning_rate": 1.0228901664327857e-05, "loss": 0.2119, "step": 69308 }, { "epoch": 4.18, "learning_rate": 1.0227392754213631e-05, "loss": 0.2174, "step": 69310 }, { "epoch": 4.18, "learning_rate": 1.0225883844099408e-05, "loss": 0.2306, "step": 69312 }, { "epoch": 4.18, "learning_rate": 1.0224374933985183e-05, "loss": 0.1751, "step": 69314 }, { "epoch": 4.18, "learning_rate": 1.022286602387096e-05, "loss": 0.1965, "step": 69316 }, { "epoch": 4.18, "learning_rate": 1.0221357113756734e-05, "loss": 0.3431, "step": 69318 }, { "epoch": 4.18, "learning_rate": 1.021984820364251e-05, "loss": 0.2296, "step": 69320 }, { "epoch": 4.18, "learning_rate": 1.0218339293528285e-05, "loss": 0.2031, "step": 69322 }, { "epoch": 4.18, "learning_rate": 1.0216830383414062e-05, "loss": 0.2644, "step": 69324 }, { "epoch": 4.18, "learning_rate": 1.0215321473299835e-05, "loss": 0.3267, "step": 69326 }, { "epoch": 4.18, "learning_rate": 1.0213812563185611e-05, "loss": 0.1479, "step": 69328 }, { "epoch": 4.18, "learning_rate": 1.0212303653071386e-05, "loss": 0.2781, "step": 69330 }, { "epoch": 4.18, "learning_rate": 1.0210794742957163e-05, "loss": 0.2248, "step": 69332 }, { "epoch": 4.18, "learning_rate": 1.0209285832842938e-05, "loss": 0.2711, "step": 69334 }, { "epoch": 4.18, "learning_rate": 1.0207776922728714e-05, "loss": 0.2314, "step": 69336 }, { "epoch": 4.19, "learning_rate": 1.0206268012614489e-05, "loss": 0.2543, "step": 69338 }, { "epoch": 4.19, "learning_rate": 1.0204759102500265e-05, "loss": 0.1306, "step": 69340 }, { "epoch": 4.19, "learning_rate": 1.020325019238604e-05, "loss": 0.1941, "step": 69342 }, { "epoch": 4.19, "learning_rate": 1.0201741282271815e-05, "loss": 0.5237, "step": 69344 }, { "epoch": 4.19, "learning_rate": 1.020023237215759e-05, "loss": 0.2929, "step": 69346 }, { "epoch": 4.19, "learning_rate": 1.0198723462043366e-05, "loss": 0.1141, "step": 69348 }, { "epoch": 4.19, "learning_rate": 1.0197214551929141e-05, "loss": 0.2473, "step": 69350 }, { "epoch": 4.19, "learning_rate": 1.0195705641814918e-05, "loss": 0.2276, "step": 69352 }, { "epoch": 4.19, "learning_rate": 1.0194196731700693e-05, "loss": 0.4351, "step": 69354 }, { "epoch": 4.19, "learning_rate": 1.0192687821586469e-05, "loss": 0.3304, "step": 69356 }, { "epoch": 4.19, "learning_rate": 1.0191178911472244e-05, "loss": 0.1862, "step": 69358 }, { "epoch": 4.19, "learning_rate": 1.0189670001358019e-05, "loss": 0.2252, "step": 69360 }, { "epoch": 4.19, "learning_rate": 1.0188161091243795e-05, "loss": 0.1885, "step": 69362 }, { "epoch": 4.19, "learning_rate": 1.018665218112957e-05, "loss": 0.3471, "step": 69364 }, { "epoch": 4.19, "learning_rate": 1.0185897726072458e-05, "loss": 0.2081, "step": 69366 }, { "epoch": 4.19, "learning_rate": 1.0184388815958233e-05, "loss": 0.1799, "step": 69368 }, { "epoch": 4.19, "learning_rate": 1.018287990584401e-05, "loss": 0.1346, "step": 69370 }, { "epoch": 4.19, "learning_rate": 1.0181370995729784e-05, "loss": 0.2483, "step": 69372 }, { "epoch": 4.19, "learning_rate": 1.0179862085615561e-05, "loss": 0.2559, "step": 69374 }, { "epoch": 4.19, "learning_rate": 1.0178353175501336e-05, "loss": 0.2031, "step": 69376 }, { "epoch": 4.19, "learning_rate": 1.0176844265387112e-05, "loss": 0.2247, "step": 69378 }, { "epoch": 4.19, "learning_rate": 1.0175335355272887e-05, "loss": 0.1865, "step": 69380 }, { "epoch": 4.19, "learning_rate": 1.0173826445158664e-05, "loss": 0.1424, "step": 69382 }, { "epoch": 4.19, "learning_rate": 1.0172317535044438e-05, "loss": 0.2414, "step": 69384 }, { "epoch": 4.19, "learning_rate": 1.0170808624930213e-05, "loss": 0.2415, "step": 69386 }, { "epoch": 4.19, "learning_rate": 1.0169299714815988e-05, "loss": 0.1922, "step": 69388 }, { "epoch": 4.19, "learning_rate": 1.0167790804701765e-05, "loss": 0.2431, "step": 69390 }, { "epoch": 4.19, "learning_rate": 1.016628189458754e-05, "loss": 0.2109, "step": 69392 }, { "epoch": 4.19, "learning_rate": 1.0164772984473316e-05, "loss": 0.2213, "step": 69394 }, { "epoch": 4.19, "learning_rate": 1.016326407435909e-05, "loss": 0.3725, "step": 69396 }, { "epoch": 4.19, "learning_rate": 1.0161755164244867e-05, "loss": 0.3282, "step": 69398 }, { "epoch": 4.19, "learning_rate": 1.0160246254130642e-05, "loss": 0.2603, "step": 69400 }, { "epoch": 4.19, "learning_rate": 1.0158737344016417e-05, "loss": 0.2422, "step": 69402 }, { "epoch": 4.19, "learning_rate": 1.0157228433902192e-05, "loss": 0.1004, "step": 69404 }, { "epoch": 4.19, "learning_rate": 1.0155719523787968e-05, "loss": 0.23, "step": 69406 }, { "epoch": 4.19, "learning_rate": 1.0154210613673743e-05, "loss": 0.292, "step": 69408 }, { "epoch": 4.19, "learning_rate": 1.015270170355952e-05, "loss": 0.2282, "step": 69410 }, { "epoch": 4.19, "learning_rate": 1.0151192793445294e-05, "loss": 0.1499, "step": 69412 }, { "epoch": 4.19, "learning_rate": 1.0149683883331071e-05, "loss": 0.2567, "step": 69414 }, { "epoch": 4.19, "learning_rate": 1.0148174973216846e-05, "loss": 0.2251, "step": 69416 }, { "epoch": 4.19, "learning_rate": 1.014666606310262e-05, "loss": 0.2787, "step": 69418 }, { "epoch": 4.19, "learning_rate": 1.0145157152988397e-05, "loss": 0.1944, "step": 69420 }, { "epoch": 4.19, "learning_rate": 1.0143648242874172e-05, "loss": 0.1955, "step": 69422 }, { "epoch": 4.19, "learning_rate": 1.0142139332759948e-05, "loss": 0.245, "step": 69424 }, { "epoch": 4.19, "learning_rate": 1.0140630422645723e-05, "loss": 0.1971, "step": 69426 }, { "epoch": 4.19, "learning_rate": 1.01391215125315e-05, "loss": 0.2051, "step": 69428 }, { "epoch": 4.19, "learning_rate": 1.0137612602417275e-05, "loss": 0.2809, "step": 69430 }, { "epoch": 4.19, "learning_rate": 1.0136103692303051e-05, "loss": 0.1881, "step": 69432 }, { "epoch": 4.19, "learning_rate": 1.0134594782188824e-05, "loss": 0.2518, "step": 69434 }, { "epoch": 4.19, "learning_rate": 1.01330858720746e-05, "loss": 0.2053, "step": 69436 }, { "epoch": 4.19, "learning_rate": 1.0131576961960376e-05, "loss": 0.1919, "step": 69438 }, { "epoch": 4.19, "learning_rate": 1.0130068051846152e-05, "loss": 0.2749, "step": 69440 }, { "epoch": 4.19, "learning_rate": 1.0128559141731927e-05, "loss": 0.2471, "step": 69442 }, { "epoch": 4.19, "learning_rate": 1.0127050231617703e-05, "loss": 0.1401, "step": 69444 }, { "epoch": 4.19, "learning_rate": 1.0125541321503478e-05, "loss": 0.2773, "step": 69446 }, { "epoch": 4.19, "learning_rate": 1.0124032411389255e-05, "loss": 0.2415, "step": 69448 }, { "epoch": 4.19, "learning_rate": 1.012252350127503e-05, "loss": 0.4395, "step": 69450 }, { "epoch": 4.19, "learning_rate": 1.0121014591160804e-05, "loss": 0.2373, "step": 69452 }, { "epoch": 4.19, "learning_rate": 1.0119505681046581e-05, "loss": 0.2305, "step": 69454 }, { "epoch": 4.19, "learning_rate": 1.0117996770932356e-05, "loss": 0.2911, "step": 69456 }, { "epoch": 4.19, "learning_rate": 1.0116487860818132e-05, "loss": 0.1918, "step": 69458 }, { "epoch": 4.19, "learning_rate": 1.0114978950703907e-05, "loss": 0.1769, "step": 69460 }, { "epoch": 4.19, "learning_rate": 1.0113470040589684e-05, "loss": 0.2243, "step": 69462 }, { "epoch": 4.19, "learning_rate": 1.0111961130475458e-05, "loss": 0.2312, "step": 69464 }, { "epoch": 4.19, "learning_rate": 1.0110452220361235e-05, "loss": 0.191, "step": 69466 }, { "epoch": 4.19, "learning_rate": 1.0108943310247008e-05, "loss": 0.2276, "step": 69468 }, { "epoch": 4.19, "learning_rate": 1.0107434400132785e-05, "loss": 0.1676, "step": 69470 }, { "epoch": 4.19, "learning_rate": 1.010592549001856e-05, "loss": 0.2259, "step": 69472 }, { "epoch": 4.19, "learning_rate": 1.0104416579904336e-05, "loss": 0.1959, "step": 69474 }, { "epoch": 4.19, "learning_rate": 1.010290766979011e-05, "loss": 0.2311, "step": 69476 }, { "epoch": 4.19, "learning_rate": 1.0101398759675887e-05, "loss": 0.196, "step": 69478 }, { "epoch": 4.19, "learning_rate": 1.0099889849561662e-05, "loss": 0.1931, "step": 69480 }, { "epoch": 4.19, "learning_rate": 1.0098380939447439e-05, "loss": 0.2856, "step": 69482 }, { "epoch": 4.19, "learning_rate": 1.0096872029333213e-05, "loss": 0.362, "step": 69484 }, { "epoch": 4.19, "learning_rate": 1.0095363119218988e-05, "loss": 0.2804, "step": 69486 }, { "epoch": 4.19, "learning_rate": 1.0093854209104765e-05, "loss": 0.1363, "step": 69488 }, { "epoch": 4.19, "learning_rate": 1.009234529899054e-05, "loss": 0.249, "step": 69490 }, { "epoch": 4.19, "learning_rate": 1.0090836388876316e-05, "loss": 0.2376, "step": 69492 }, { "epoch": 4.19, "learning_rate": 1.008932747876209e-05, "loss": 0.1456, "step": 69494 }, { "epoch": 4.19, "learning_rate": 1.0087818568647867e-05, "loss": 0.3409, "step": 69496 }, { "epoch": 4.19, "learning_rate": 1.0086309658533642e-05, "loss": 0.1965, "step": 69498 }, { "epoch": 4.19, "learning_rate": 1.0084800748419417e-05, "loss": 0.2329, "step": 69500 }, { "epoch": 4.19, "learning_rate": 1.0083291838305192e-05, "loss": 0.2587, "step": 69502 }, { "epoch": 4.2, "learning_rate": 1.0081782928190968e-05, "loss": 0.1929, "step": 69504 }, { "epoch": 4.2, "learning_rate": 1.0080274018076743e-05, "loss": 0.2518, "step": 69506 }, { "epoch": 4.2, "learning_rate": 1.007876510796252e-05, "loss": 0.1617, "step": 69508 }, { "epoch": 4.2, "learning_rate": 1.0077256197848294e-05, "loss": 0.2062, "step": 69510 }, { "epoch": 4.2, "learning_rate": 1.0075747287734071e-05, "loss": 0.2428, "step": 69512 }, { "epoch": 4.2, "learning_rate": 1.0074238377619846e-05, "loss": 0.1386, "step": 69514 }, { "epoch": 4.2, "learning_rate": 1.007272946750562e-05, "loss": 0.166, "step": 69516 }, { "epoch": 4.2, "learning_rate": 1.0071220557391395e-05, "loss": 0.134, "step": 69518 }, { "epoch": 4.2, "learning_rate": 1.0069711647277172e-05, "loss": 0.1899, "step": 69520 }, { "epoch": 4.2, "learning_rate": 1.0068202737162947e-05, "loss": 0.2201, "step": 69522 }, { "epoch": 4.2, "learning_rate": 1.0066693827048723e-05, "loss": 0.2586, "step": 69524 }, { "epoch": 4.2, "learning_rate": 1.0065184916934498e-05, "loss": 0.2081, "step": 69526 }, { "epoch": 4.2, "learning_rate": 1.0063676006820275e-05, "loss": 0.1781, "step": 69528 }, { "epoch": 4.2, "learning_rate": 1.006216709670605e-05, "loss": 0.2949, "step": 69530 }, { "epoch": 4.2, "learning_rate": 1.0060658186591824e-05, "loss": 0.208, "step": 69532 }, { "epoch": 4.2, "learning_rate": 1.00591492764776e-05, "loss": 0.3276, "step": 69534 }, { "epoch": 4.2, "learning_rate": 1.0057640366363376e-05, "loss": 0.2003, "step": 69536 }, { "epoch": 4.2, "learning_rate": 1.0056131456249152e-05, "loss": 0.2257, "step": 69538 }, { "epoch": 4.2, "learning_rate": 1.0054622546134927e-05, "loss": 0.2135, "step": 69540 }, { "epoch": 4.2, "learning_rate": 1.0053113636020703e-05, "loss": 0.1785, "step": 69542 }, { "epoch": 4.2, "learning_rate": 1.0051604725906478e-05, "loss": 0.2774, "step": 69544 }, { "epoch": 4.2, "learning_rate": 1.0050095815792255e-05, "loss": 0.2316, "step": 69546 }, { "epoch": 4.2, "learning_rate": 1.004858690567803e-05, "loss": 0.3144, "step": 69548 }, { "epoch": 4.2, "learning_rate": 1.0047077995563804e-05, "loss": 0.2233, "step": 69550 }, { "epoch": 4.2, "learning_rate": 1.004556908544958e-05, "loss": 0.1869, "step": 69552 }, { "epoch": 4.2, "learning_rate": 1.0044060175335356e-05, "loss": 0.2929, "step": 69554 }, { "epoch": 4.2, "learning_rate": 1.004255126522113e-05, "loss": 0.3529, "step": 69556 }, { "epoch": 4.2, "learning_rate": 1.0041042355106907e-05, "loss": 0.2106, "step": 69558 }, { "epoch": 4.2, "learning_rate": 1.0039533444992682e-05, "loss": 0.1857, "step": 69560 }, { "epoch": 4.2, "learning_rate": 1.0038024534878458e-05, "loss": 0.3028, "step": 69562 }, { "epoch": 4.2, "learning_rate": 1.0036515624764233e-05, "loss": 0.1477, "step": 69564 }, { "epoch": 4.2, "learning_rate": 1.0035006714650008e-05, "loss": 0.2372, "step": 69566 }, { "epoch": 4.2, "learning_rate": 1.0033497804535785e-05, "loss": 0.2381, "step": 69568 }, { "epoch": 4.2, "learning_rate": 1.003198889442156e-05, "loss": 0.3396, "step": 69570 }, { "epoch": 4.2, "learning_rate": 1.0030479984307336e-05, "loss": 0.1806, "step": 69572 }, { "epoch": 4.2, "learning_rate": 1.002897107419311e-05, "loss": 0.3887, "step": 69574 }, { "epoch": 4.2, "learning_rate": 1.0027462164078887e-05, "loss": 0.2063, "step": 69576 }, { "epoch": 4.2, "learning_rate": 1.0025953253964662e-05, "loss": 0.1849, "step": 69578 }, { "epoch": 4.2, "learning_rate": 1.0024444343850439e-05, "loss": 0.1936, "step": 69580 }, { "epoch": 4.2, "learning_rate": 1.0022935433736212e-05, "loss": 0.2522, "step": 69582 }, { "epoch": 4.2, "learning_rate": 1.0021426523621988e-05, "loss": 0.251, "step": 69584 }, { "epoch": 4.2, "learning_rate": 1.0019917613507763e-05, "loss": 0.1553, "step": 69586 }, { "epoch": 4.2, "learning_rate": 1.001840870339354e-05, "loss": 0.2625, "step": 69588 }, { "epoch": 4.2, "learning_rate": 1.0016899793279314e-05, "loss": 0.2569, "step": 69590 }, { "epoch": 4.2, "learning_rate": 1.0015390883165091e-05, "loss": 0.2317, "step": 69592 }, { "epoch": 4.2, "learning_rate": 1.0013881973050866e-05, "loss": 0.2241, "step": 69594 }, { "epoch": 4.2, "learning_rate": 1.0012373062936642e-05, "loss": 0.2783, "step": 69596 }, { "epoch": 4.2, "learning_rate": 1.0010864152822415e-05, "loss": 0.2641, "step": 69598 }, { "epoch": 4.2, "learning_rate": 1.0009355242708192e-05, "loss": 0.4118, "step": 69600 }, { "epoch": 4.2, "learning_rate": 1.0007846332593968e-05, "loss": 0.2289, "step": 69602 }, { "epoch": 4.2, "learning_rate": 1.0006337422479743e-05, "loss": 0.2114, "step": 69604 }, { "epoch": 4.2, "learning_rate": 1.000482851236552e-05, "loss": 0.2524, "step": 69606 }, { "epoch": 4.2, "learning_rate": 1.0003319602251295e-05, "loss": 0.2173, "step": 69608 }, { "epoch": 4.2, "learning_rate": 1.0001810692137071e-05, "loss": 0.2147, "step": 69610 }, { "epoch": 4.2, "learning_rate": 1.0000301782022846e-05, "loss": 0.2187, "step": 69612 }, { "epoch": 4.2, "learning_rate": 9.998792871908622e-06, "loss": 0.1914, "step": 69614 }, { "epoch": 4.2, "learning_rate": 9.997283961794395e-06, "loss": 0.2987, "step": 69616 }, { "epoch": 4.2, "learning_rate": 9.995775051680172e-06, "loss": 0.2387, "step": 69618 }, { "epoch": 4.2, "learning_rate": 9.994266141565947e-06, "loss": 0.2099, "step": 69620 }, { "epoch": 4.2, "learning_rate": 9.992757231451723e-06, "loss": 0.2571, "step": 69622 }, { "epoch": 4.2, "learning_rate": 9.991248321337498e-06, "loss": 0.2502, "step": 69624 }, { "epoch": 4.2, "learning_rate": 9.989739411223275e-06, "loss": 0.2481, "step": 69626 }, { "epoch": 4.2, "learning_rate": 9.98823050110905e-06, "loss": 0.2509, "step": 69628 }, { "epoch": 4.2, "learning_rate": 9.986721590994826e-06, "loss": 0.2453, "step": 69630 }, { "epoch": 4.2, "learning_rate": 9.985212680880599e-06, "loss": 0.1254, "step": 69632 }, { "epoch": 4.2, "learning_rate": 9.983703770766376e-06, "loss": 0.2416, "step": 69634 }, { "epoch": 4.2, "learning_rate": 9.98219486065215e-06, "loss": 0.2153, "step": 69636 }, { "epoch": 4.2, "learning_rate": 9.980685950537927e-06, "loss": 0.2704, "step": 69638 }, { "epoch": 4.2, "learning_rate": 9.979177040423702e-06, "loss": 0.1685, "step": 69640 }, { "epoch": 4.2, "learning_rate": 9.977668130309478e-06, "loss": 0.207, "step": 69642 }, { "epoch": 4.2, "learning_rate": 9.976159220195253e-06, "loss": 0.363, "step": 69644 }, { "epoch": 4.2, "learning_rate": 9.97465031008103e-06, "loss": 0.1584, "step": 69646 }, { "epoch": 4.2, "learning_rate": 9.973141399966804e-06, "loss": 0.1515, "step": 69648 }, { "epoch": 4.2, "learning_rate": 9.97163248985258e-06, "loss": 0.2334, "step": 69650 }, { "epoch": 4.2, "learning_rate": 9.970123579738356e-06, "loss": 0.2175, "step": 69652 }, { "epoch": 4.2, "learning_rate": 9.96861466962413e-06, "loss": 0.2896, "step": 69654 }, { "epoch": 4.2, "learning_rate": 9.967105759509907e-06, "loss": 0.2554, "step": 69656 }, { "epoch": 4.2, "learning_rate": 9.965596849395682e-06, "loss": 0.2686, "step": 69658 }, { "epoch": 4.2, "learning_rate": 9.964087939281458e-06, "loss": 0.2302, "step": 69660 }, { "epoch": 4.2, "learning_rate": 9.962579029167233e-06, "loss": 0.2207, "step": 69662 }, { "epoch": 4.2, "learning_rate": 9.961070119053008e-06, "loss": 0.2633, "step": 69664 }, { "epoch": 4.2, "learning_rate": 9.959561208938783e-06, "loss": 0.3444, "step": 69666 }, { "epoch": 4.2, "learning_rate": 9.95805229882456e-06, "loss": 0.2368, "step": 69668 }, { "epoch": 4.21, "learning_rate": 9.956543388710334e-06, "loss": 0.2598, "step": 69670 }, { "epoch": 4.21, "learning_rate": 9.95503447859611e-06, "loss": 0.2832, "step": 69672 }, { "epoch": 4.21, "learning_rate": 9.953525568481886e-06, "loss": 0.2174, "step": 69674 }, { "epoch": 4.21, "learning_rate": 9.952016658367662e-06, "loss": 0.1798, "step": 69676 }, { "epoch": 4.21, "learning_rate": 9.950507748253437e-06, "loss": 0.2857, "step": 69678 }, { "epoch": 4.21, "learning_rate": 9.948998838139212e-06, "loss": 0.269, "step": 69680 }, { "epoch": 4.21, "learning_rate": 9.947489928024988e-06, "loss": 0.3197, "step": 69682 }, { "epoch": 4.21, "learning_rate": 9.945981017910763e-06, "loss": 0.2552, "step": 69684 }, { "epoch": 4.21, "learning_rate": 9.94447210779654e-06, "loss": 0.2512, "step": 69686 }, { "epoch": 4.21, "learning_rate": 9.942963197682314e-06, "loss": 0.3822, "step": 69688 }, { "epoch": 4.21, "learning_rate": 9.941454287568091e-06, "loss": 0.1862, "step": 69690 }, { "epoch": 4.21, "learning_rate": 9.939945377453866e-06, "loss": 0.3149, "step": 69692 }, { "epoch": 4.21, "learning_rate": 9.938436467339642e-06, "loss": 0.2895, "step": 69694 }, { "epoch": 4.21, "learning_rate": 9.936927557225417e-06, "loss": 0.2079, "step": 69696 }, { "epoch": 4.21, "learning_rate": 9.935418647111192e-06, "loss": 0.1787, "step": 69698 }, { "epoch": 4.21, "learning_rate": 9.933909736996967e-06, "loss": 0.1782, "step": 69700 }, { "epoch": 4.21, "learning_rate": 9.932400826882743e-06, "loss": 0.2537, "step": 69702 }, { "epoch": 4.21, "learning_rate": 9.930891916768518e-06, "loss": 0.194, "step": 69704 }, { "epoch": 4.21, "learning_rate": 9.929383006654295e-06, "loss": 0.1642, "step": 69706 }, { "epoch": 4.21, "learning_rate": 9.92787409654007e-06, "loss": 0.2029, "step": 69708 }, { "epoch": 4.21, "learning_rate": 9.926365186425846e-06, "loss": 0.2203, "step": 69710 }, { "epoch": 4.21, "learning_rate": 9.92485627631162e-06, "loss": 0.2446, "step": 69712 }, { "epoch": 4.21, "learning_rate": 9.923347366197396e-06, "loss": 0.245, "step": 69714 }, { "epoch": 4.21, "learning_rate": 9.92183845608317e-06, "loss": 0.2978, "step": 69716 }, { "epoch": 4.21, "learning_rate": 9.920329545968947e-06, "loss": 0.1867, "step": 69718 }, { "epoch": 4.21, "learning_rate": 9.918820635854723e-06, "loss": 0.3423, "step": 69720 }, { "epoch": 4.21, "learning_rate": 9.917311725740498e-06, "loss": 0.1988, "step": 69722 }, { "epoch": 4.21, "learning_rate": 9.915802815626275e-06, "loss": 0.2404, "step": 69724 }, { "epoch": 4.21, "learning_rate": 9.91429390551205e-06, "loss": 0.107, "step": 69726 }, { "epoch": 4.21, "learning_rate": 9.912784995397826e-06, "loss": 0.2663, "step": 69728 }, { "epoch": 4.21, "learning_rate": 9.9112760852836e-06, "loss": 0.1498, "step": 69730 }, { "epoch": 4.21, "learning_rate": 9.909767175169376e-06, "loss": 0.2373, "step": 69732 }, { "epoch": 4.21, "learning_rate": 9.90825826505515e-06, "loss": 0.2461, "step": 69734 }, { "epoch": 4.21, "learning_rate": 9.906749354940927e-06, "loss": 0.3581, "step": 69736 }, { "epoch": 4.21, "learning_rate": 9.905240444826702e-06, "loss": 0.2575, "step": 69738 }, { "epoch": 4.21, "learning_rate": 9.903731534712478e-06, "loss": 0.252, "step": 69740 }, { "epoch": 4.21, "learning_rate": 9.902222624598253e-06, "loss": 0.3216, "step": 69742 }, { "epoch": 4.21, "learning_rate": 9.90071371448403e-06, "loss": 0.2685, "step": 69744 }, { "epoch": 4.21, "learning_rate": 9.899204804369803e-06, "loss": 0.2105, "step": 69746 }, { "epoch": 4.21, "learning_rate": 9.89769589425558e-06, "loss": 0.2368, "step": 69748 }, { "epoch": 4.21, "learning_rate": 9.896186984141354e-06, "loss": 0.2882, "step": 69750 }, { "epoch": 4.21, "learning_rate": 9.89467807402713e-06, "loss": 0.2638, "step": 69752 }, { "epoch": 4.21, "learning_rate": 9.893169163912905e-06, "loss": 0.1527, "step": 69754 }, { "epoch": 4.21, "learning_rate": 9.891660253798682e-06, "loss": 0.1695, "step": 69756 }, { "epoch": 4.21, "learning_rate": 9.890151343684457e-06, "loss": 0.1786, "step": 69758 }, { "epoch": 4.21, "learning_rate": 9.888642433570233e-06, "loss": 0.1915, "step": 69760 }, { "epoch": 4.21, "learning_rate": 9.887133523456008e-06, "loss": 0.395, "step": 69762 }, { "epoch": 4.21, "learning_rate": 9.885624613341783e-06, "loss": 0.3528, "step": 69764 }, { "epoch": 4.21, "learning_rate": 9.88411570322756e-06, "loss": 0.1501, "step": 69766 }, { "epoch": 4.21, "learning_rate": 9.882606793113334e-06, "loss": 0.2701, "step": 69768 }, { "epoch": 4.21, "learning_rate": 9.88109788299911e-06, "loss": 0.3002, "step": 69770 }, { "epoch": 4.21, "learning_rate": 9.879588972884886e-06, "loss": 0.2358, "step": 69772 }, { "epoch": 4.21, "learning_rate": 9.878080062770662e-06, "loss": 0.3546, "step": 69774 }, { "epoch": 4.21, "learning_rate": 9.876571152656437e-06, "loss": 0.1088, "step": 69776 }, { "epoch": 4.21, "learning_rate": 9.875062242542213e-06, "loss": 0.279, "step": 69778 }, { "epoch": 4.21, "learning_rate": 9.873553332427987e-06, "loss": 0.2471, "step": 69780 }, { "epoch": 4.21, "learning_rate": 9.872044422313763e-06, "loss": 0.2584, "step": 69782 }, { "epoch": 4.21, "learning_rate": 9.870535512199538e-06, "loss": 0.1908, "step": 69784 }, { "epoch": 4.21, "learning_rate": 9.869026602085314e-06, "loss": 0.183, "step": 69786 }, { "epoch": 4.21, "learning_rate": 9.86751769197109e-06, "loss": 0.225, "step": 69788 }, { "epoch": 4.21, "learning_rate": 9.866008781856866e-06, "loss": 0.1871, "step": 69790 }, { "epoch": 4.21, "learning_rate": 9.86449987174264e-06, "loss": 0.2604, "step": 69792 }, { "epoch": 4.21, "learning_rate": 9.862990961628417e-06, "loss": 0.2468, "step": 69794 }, { "epoch": 4.21, "learning_rate": 9.861482051514192e-06, "loss": 0.2381, "step": 69796 }, { "epoch": 4.21, "learning_rate": 9.859973141399967e-06, "loss": 0.1619, "step": 69798 }, { "epoch": 4.21, "learning_rate": 9.858464231285743e-06, "loss": 0.2816, "step": 69800 }, { "epoch": 4.21, "learning_rate": 9.856955321171518e-06, "loss": 0.1689, "step": 69802 }, { "epoch": 4.21, "learning_rate": 9.855446411057295e-06, "loss": 0.2357, "step": 69804 }, { "epoch": 4.21, "learning_rate": 9.85393750094307e-06, "loss": 0.3367, "step": 69806 }, { "epoch": 4.21, "learning_rate": 9.852428590828846e-06, "loss": 0.1434, "step": 69808 }, { "epoch": 4.21, "learning_rate": 9.85091968071462e-06, "loss": 0.2467, "step": 69810 }, { "epoch": 4.21, "learning_rate": 9.849410770600396e-06, "loss": 0.2286, "step": 69812 }, { "epoch": 4.21, "learning_rate": 9.84790186048617e-06, "loss": 0.2529, "step": 69814 }, { "epoch": 4.21, "learning_rate": 9.846392950371947e-06, "loss": 0.2143, "step": 69816 }, { "epoch": 4.21, "learning_rate": 9.844884040257722e-06, "loss": 0.2056, "step": 69818 }, { "epoch": 4.21, "learning_rate": 9.843375130143498e-06, "loss": 0.3559, "step": 69820 }, { "epoch": 4.21, "learning_rate": 9.841866220029273e-06, "loss": 0.1678, "step": 69822 }, { "epoch": 4.21, "learning_rate": 9.84035730991505e-06, "loss": 0.205, "step": 69824 }, { "epoch": 4.21, "learning_rate": 9.838848399800824e-06, "loss": 0.2728, "step": 69826 }, { "epoch": 4.21, "learning_rate": 9.8373394896866e-06, "loss": 0.4148, "step": 69828 }, { "epoch": 4.21, "learning_rate": 9.835830579572374e-06, "loss": 0.2152, "step": 69830 }, { "epoch": 4.21, "learning_rate": 9.83432166945815e-06, "loss": 0.2165, "step": 69832 }, { "epoch": 4.21, "learning_rate": 9.832812759343925e-06, "loss": 0.1635, "step": 69834 }, { "epoch": 4.22, "learning_rate": 9.831303849229702e-06, "loss": 0.3187, "step": 69836 }, { "epoch": 4.22, "learning_rate": 9.829794939115478e-06, "loss": 0.3043, "step": 69838 }, { "epoch": 4.22, "learning_rate": 9.828286029001253e-06, "loss": 0.3007, "step": 69840 }, { "epoch": 4.22, "learning_rate": 9.82677711888703e-06, "loss": 0.4103, "step": 69842 }, { "epoch": 4.22, "learning_rate": 9.825268208772803e-06, "loss": 0.1531, "step": 69844 }, { "epoch": 4.22, "learning_rate": 9.82375929865858e-06, "loss": 0.3252, "step": 69846 }, { "epoch": 4.22, "learning_rate": 9.822250388544354e-06, "loss": 0.1509, "step": 69848 }, { "epoch": 4.22, "learning_rate": 9.82074147843013e-06, "loss": 0.2291, "step": 69850 }, { "epoch": 4.22, "learning_rate": 9.819232568315906e-06, "loss": 0.1909, "step": 69852 }, { "epoch": 4.22, "learning_rate": 9.817723658201682e-06, "loss": 0.302, "step": 69854 }, { "epoch": 4.22, "learning_rate": 9.816214748087457e-06, "loss": 0.2507, "step": 69856 }, { "epoch": 4.22, "learning_rate": 9.814705837973233e-06, "loss": 0.2469, "step": 69858 }, { "epoch": 4.22, "learning_rate": 9.813196927859008e-06, "loss": 0.2778, "step": 69860 }, { "epoch": 4.22, "learning_rate": 9.811688017744783e-06, "loss": 0.3221, "step": 69862 }, { "epoch": 4.22, "learning_rate": 9.810179107630558e-06, "loss": 0.4142, "step": 69864 }, { "epoch": 4.22, "learning_rate": 9.808670197516334e-06, "loss": 0.2633, "step": 69866 }, { "epoch": 4.22, "learning_rate": 9.80716128740211e-06, "loss": 0.2153, "step": 69868 }, { "epoch": 4.22, "learning_rate": 9.805652377287886e-06, "loss": 0.2182, "step": 69870 }, { "epoch": 4.22, "learning_rate": 9.80414346717366e-06, "loss": 0.1688, "step": 69872 }, { "epoch": 4.22, "learning_rate": 9.802634557059437e-06, "loss": 0.2498, "step": 69874 }, { "epoch": 4.22, "learning_rate": 9.801125646945212e-06, "loss": 0.2055, "step": 69876 }, { "epoch": 4.22, "learning_rate": 9.799616736830987e-06, "loss": 0.2478, "step": 69878 }, { "epoch": 4.22, "learning_rate": 9.798107826716763e-06, "loss": 0.267, "step": 69880 }, { "epoch": 4.22, "learning_rate": 9.796598916602538e-06, "loss": 0.3841, "step": 69882 }, { "epoch": 4.22, "learning_rate": 9.795090006488314e-06, "loss": 0.3957, "step": 69884 }, { "epoch": 4.22, "learning_rate": 9.79358109637409e-06, "loss": 0.3087, "step": 69886 }, { "epoch": 4.22, "learning_rate": 9.792072186259866e-06, "loss": 0.2314, "step": 69888 }, { "epoch": 4.22, "learning_rate": 9.79056327614564e-06, "loss": 0.1464, "step": 69890 }, { "epoch": 4.22, "learning_rate": 9.789054366031417e-06, "loss": 0.2655, "step": 69892 }, { "epoch": 4.22, "learning_rate": 9.78754545591719e-06, "loss": 0.2233, "step": 69894 }, { "epoch": 4.22, "learning_rate": 9.786036545802967e-06, "loss": 0.2945, "step": 69896 }, { "epoch": 4.22, "learning_rate": 9.784527635688742e-06, "loss": 0.1887, "step": 69898 }, { "epoch": 4.22, "learning_rate": 9.783018725574518e-06, "loss": 0.1932, "step": 69900 }, { "epoch": 4.22, "learning_rate": 9.781509815460293e-06, "loss": 0.2918, "step": 69902 }, { "epoch": 4.22, "learning_rate": 9.78000090534607e-06, "loss": 0.2284, "step": 69904 }, { "epoch": 4.22, "learning_rate": 9.778491995231844e-06, "loss": 0.1451, "step": 69906 }, { "epoch": 4.22, "learning_rate": 9.77698308511762e-06, "loss": 0.3642, "step": 69908 }, { "epoch": 4.22, "learning_rate": 9.775474175003396e-06, "loss": 0.2883, "step": 69910 }, { "epoch": 4.22, "learning_rate": 9.77396526488917e-06, "loss": 0.2161, "step": 69912 }, { "epoch": 4.22, "learning_rate": 9.772456354774947e-06, "loss": 0.3084, "step": 69914 }, { "epoch": 4.22, "learning_rate": 9.770947444660722e-06, "loss": 0.2486, "step": 69916 }, { "epoch": 4.22, "learning_rate": 9.769438534546498e-06, "loss": 0.2045, "step": 69918 }, { "epoch": 4.22, "learning_rate": 9.767929624432273e-06, "loss": 0.1443, "step": 69920 }, { "epoch": 4.22, "learning_rate": 9.76642071431805e-06, "loss": 0.1741, "step": 69922 }, { "epoch": 4.22, "learning_rate": 9.764911804203824e-06, "loss": 0.4002, "step": 69924 }, { "epoch": 4.22, "learning_rate": 9.763402894089601e-06, "loss": 0.2759, "step": 69926 }, { "epoch": 4.22, "learning_rate": 9.761893983975374e-06, "loss": 0.2777, "step": 69928 }, { "epoch": 4.22, "learning_rate": 9.76038507386115e-06, "loss": 0.392, "step": 69930 }, { "epoch": 4.22, "learning_rate": 9.758876163746925e-06, "loss": 0.1718, "step": 69932 }, { "epoch": 4.22, "learning_rate": 9.757367253632702e-06, "loss": 0.2128, "step": 69934 }, { "epoch": 4.22, "learning_rate": 9.755858343518477e-06, "loss": 0.2285, "step": 69936 }, { "epoch": 4.22, "learning_rate": 9.754349433404253e-06, "loss": 0.2077, "step": 69938 }, { "epoch": 4.22, "learning_rate": 9.752840523290028e-06, "loss": 0.2503, "step": 69940 }, { "epoch": 4.22, "learning_rate": 9.751331613175805e-06, "loss": 0.2415, "step": 69942 }, { "epoch": 4.22, "learning_rate": 9.749822703061578e-06, "loss": 0.1604, "step": 69944 }, { "epoch": 4.22, "learning_rate": 9.748313792947354e-06, "loss": 0.1824, "step": 69946 }, { "epoch": 4.22, "learning_rate": 9.746804882833129e-06, "loss": 0.2396, "step": 69948 }, { "epoch": 4.22, "learning_rate": 9.745295972718906e-06, "loss": 0.3217, "step": 69950 }, { "epoch": 4.22, "learning_rate": 9.74378706260468e-06, "loss": 0.2588, "step": 69952 }, { "epoch": 4.22, "learning_rate": 9.742278152490457e-06, "loss": 0.2103, "step": 69954 }, { "epoch": 4.22, "learning_rate": 9.740769242376233e-06, "loss": 0.3139, "step": 69956 }, { "epoch": 4.22, "learning_rate": 9.739260332262008e-06, "loss": 0.196, "step": 69958 }, { "epoch": 4.22, "learning_rate": 9.737751422147783e-06, "loss": 0.2212, "step": 69960 }, { "epoch": 4.22, "learning_rate": 9.736242512033558e-06, "loss": 0.2124, "step": 69962 }, { "epoch": 4.22, "learning_rate": 9.734733601919334e-06, "loss": 0.1644, "step": 69964 }, { "epoch": 4.22, "learning_rate": 9.73322469180511e-06, "loss": 0.3182, "step": 69966 }, { "epoch": 4.22, "learning_rate": 9.731715781690886e-06, "loss": 0.2165, "step": 69968 }, { "epoch": 4.22, "learning_rate": 9.73020687157666e-06, "loss": 0.3533, "step": 69970 }, { "epoch": 4.22, "learning_rate": 9.728697961462437e-06, "loss": 0.2598, "step": 69972 }, { "epoch": 4.22, "learning_rate": 9.727189051348212e-06, "loss": 0.1539, "step": 69974 }, { "epoch": 4.22, "learning_rate": 9.725680141233987e-06, "loss": 0.2192, "step": 69976 }, { "epoch": 4.22, "learning_rate": 9.724171231119761e-06, "loss": 0.2037, "step": 69978 }, { "epoch": 4.22, "learning_rate": 9.722662321005538e-06, "loss": 0.2158, "step": 69980 }, { "epoch": 4.22, "learning_rate": 9.721153410891313e-06, "loss": 0.1759, "step": 69982 }, { "epoch": 4.22, "learning_rate": 9.71964450077709e-06, "loss": 0.2534, "step": 69984 }, { "epoch": 4.22, "learning_rate": 9.718135590662864e-06, "loss": 0.2389, "step": 69986 }, { "epoch": 4.22, "learning_rate": 9.71662668054864e-06, "loss": 0.1636, "step": 69988 }, { "epoch": 4.22, "learning_rate": 9.715117770434415e-06, "loss": 0.3557, "step": 69990 }, { "epoch": 4.22, "learning_rate": 9.71360886032019e-06, "loss": 0.255, "step": 69992 }, { "epoch": 4.22, "learning_rate": 9.712099950205967e-06, "loss": 0.1578, "step": 69994 }, { "epoch": 4.22, "learning_rate": 9.710591040091742e-06, "loss": 0.2235, "step": 69996 }, { "epoch": 4.22, "learning_rate": 9.709082129977518e-06, "loss": 0.2192, "step": 69998 }, { "epoch": 4.23, "learning_rate": 9.707573219863293e-06, "loss": 0.2248, "step": 70000 }, { "epoch": 4.23, "learning_rate": 9.70606430974907e-06, "loss": 0.2508, "step": 70002 }, { "epoch": 4.23, "learning_rate": 9.704555399634844e-06, "loss": 0.1949, "step": 70004 }, { "epoch": 4.23, "learning_rate": 9.70304648952062e-06, "loss": 0.2834, "step": 70006 }, { "epoch": 4.23, "learning_rate": 9.701537579406396e-06, "loss": 0.108, "step": 70008 }, { "epoch": 4.23, "learning_rate": 9.70002866929217e-06, "loss": 0.2126, "step": 70010 }, { "epoch": 4.23, "learning_rate": 9.698519759177945e-06, "loss": 0.2301, "step": 70012 }, { "epoch": 4.23, "learning_rate": 9.697010849063722e-06, "loss": 0.2252, "step": 70014 }, { "epoch": 4.23, "learning_rate": 9.695501938949497e-06, "loss": 0.2848, "step": 70016 }, { "epoch": 4.23, "learning_rate": 9.693993028835273e-06, "loss": 0.1494, "step": 70018 }, { "epoch": 4.23, "learning_rate": 9.692484118721048e-06, "loss": 0.2107, "step": 70020 }, { "epoch": 4.23, "learning_rate": 9.690975208606824e-06, "loss": 0.168, "step": 70022 }, { "epoch": 4.23, "learning_rate": 9.6894662984926e-06, "loss": 0.2399, "step": 70024 }, { "epoch": 4.23, "learning_rate": 9.687957388378374e-06, "loss": 0.3148, "step": 70026 }, { "epoch": 4.23, "learning_rate": 9.68644847826415e-06, "loss": 0.262, "step": 70028 }, { "epoch": 4.23, "learning_rate": 9.684939568149925e-06, "loss": 0.1869, "step": 70030 }, { "epoch": 4.23, "learning_rate": 9.683430658035702e-06, "loss": 0.236, "step": 70032 }, { "epoch": 4.23, "learning_rate": 9.681921747921477e-06, "loss": 0.2355, "step": 70034 }, { "epoch": 4.23, "learning_rate": 9.680412837807253e-06, "loss": 0.3138, "step": 70036 }, { "epoch": 4.23, "learning_rate": 9.678903927693028e-06, "loss": 0.1713, "step": 70038 }, { "epoch": 4.23, "learning_rate": 9.677395017578805e-06, "loss": 0.2974, "step": 70040 }, { "epoch": 4.23, "learning_rate": 9.675886107464578e-06, "loss": 0.2109, "step": 70042 }, { "epoch": 4.23, "learning_rate": 9.674377197350354e-06, "loss": 0.3095, "step": 70044 }, { "epoch": 4.23, "learning_rate": 9.672868287236129e-06, "loss": 0.3208, "step": 70046 }, { "epoch": 4.23, "learning_rate": 9.671359377121906e-06, "loss": 0.2615, "step": 70048 }, { "epoch": 4.23, "learning_rate": 9.66985046700768e-06, "loss": 0.252, "step": 70050 }, { "epoch": 4.23, "learning_rate": 9.668341556893457e-06, "loss": 0.2366, "step": 70052 }, { "epoch": 4.23, "learning_rate": 9.666832646779232e-06, "loss": 0.2101, "step": 70054 }, { "epoch": 4.23, "learning_rate": 9.665323736665008e-06, "loss": 0.3049, "step": 70056 }, { "epoch": 4.23, "learning_rate": 9.663814826550781e-06, "loss": 0.2254, "step": 70058 }, { "epoch": 4.23, "learning_rate": 9.662305916436558e-06, "loss": 0.2677, "step": 70060 }, { "epoch": 4.23, "learning_rate": 9.660797006322333e-06, "loss": 0.2401, "step": 70062 }, { "epoch": 4.23, "learning_rate": 9.65928809620811e-06, "loss": 0.2309, "step": 70064 }, { "epoch": 4.23, "learning_rate": 9.657779186093884e-06, "loss": 0.1403, "step": 70066 }, { "epoch": 4.23, "learning_rate": 9.65627027597966e-06, "loss": 0.2753, "step": 70068 }, { "epoch": 4.23, "learning_rate": 9.654761365865435e-06, "loss": 0.1819, "step": 70070 }, { "epoch": 4.23, "learning_rate": 9.653252455751212e-06, "loss": 0.2277, "step": 70072 }, { "epoch": 4.23, "learning_rate": 9.651743545636987e-06, "loss": 0.2797, "step": 70074 }, { "epoch": 4.23, "learning_rate": 9.650234635522762e-06, "loss": 0.1905, "step": 70076 }, { "epoch": 4.23, "learning_rate": 9.648725725408538e-06, "loss": 0.3656, "step": 70078 }, { "epoch": 4.23, "learning_rate": 9.647216815294313e-06, "loss": 0.2766, "step": 70080 }, { "epoch": 4.23, "learning_rate": 9.64570790518009e-06, "loss": 0.2221, "step": 70082 }, { "epoch": 4.23, "learning_rate": 9.644198995065864e-06, "loss": 0.1965, "step": 70084 }, { "epoch": 4.23, "learning_rate": 9.64269008495164e-06, "loss": 0.1401, "step": 70086 }, { "epoch": 4.23, "learning_rate": 9.641181174837416e-06, "loss": 0.2049, "step": 70088 }, { "epoch": 4.23, "learning_rate": 9.639672264723192e-06, "loss": 0.2037, "step": 70090 }, { "epoch": 4.23, "learning_rate": 9.638163354608965e-06, "loss": 0.2057, "step": 70092 }, { "epoch": 4.23, "learning_rate": 9.636654444494742e-06, "loss": 0.1802, "step": 70094 }, { "epoch": 4.23, "learning_rate": 9.635145534380517e-06, "loss": 0.1204, "step": 70096 }, { "epoch": 4.23, "learning_rate": 9.633636624266293e-06, "loss": 0.3018, "step": 70098 }, { "epoch": 4.23, "learning_rate": 9.632127714152068e-06, "loss": 0.132, "step": 70100 }, { "epoch": 4.23, "learning_rate": 9.630618804037844e-06, "loss": 0.3936, "step": 70102 }, { "epoch": 4.23, "learning_rate": 9.62910989392362e-06, "loss": 0.2058, "step": 70104 }, { "epoch": 4.23, "learning_rate": 9.627600983809396e-06, "loss": 0.2244, "step": 70106 }, { "epoch": 4.23, "learning_rate": 9.62609207369517e-06, "loss": 0.2228, "step": 70108 }, { "epoch": 4.23, "learning_rate": 9.624583163580945e-06, "loss": 0.1842, "step": 70110 }, { "epoch": 4.23, "learning_rate": 9.623074253466722e-06, "loss": 0.291, "step": 70112 }, { "epoch": 4.23, "learning_rate": 9.621565343352497e-06, "loss": 0.3131, "step": 70114 }, { "epoch": 4.23, "learning_rate": 9.620056433238273e-06, "loss": 0.2639, "step": 70116 }, { "epoch": 4.23, "learning_rate": 9.618547523124048e-06, "loss": 0.2519, "step": 70118 }, { "epoch": 4.23, "learning_rate": 9.617038613009825e-06, "loss": 0.3195, "step": 70120 }, { "epoch": 4.23, "learning_rate": 9.6155297028956e-06, "loss": 0.277, "step": 70122 }, { "epoch": 4.23, "learning_rate": 9.614020792781374e-06, "loss": 0.2076, "step": 70124 }, { "epoch": 4.23, "learning_rate": 9.612511882667149e-06, "loss": 0.2995, "step": 70126 }, { "epoch": 4.23, "learning_rate": 9.611002972552925e-06, "loss": 0.2477, "step": 70128 }, { "epoch": 4.23, "learning_rate": 9.6094940624387e-06, "loss": 0.1553, "step": 70130 }, { "epoch": 4.23, "learning_rate": 9.607985152324477e-06, "loss": 0.297, "step": 70132 }, { "epoch": 4.23, "learning_rate": 9.606476242210252e-06, "loss": 0.2808, "step": 70134 }, { "epoch": 4.23, "learning_rate": 9.604967332096028e-06, "loss": 0.2543, "step": 70136 }, { "epoch": 4.23, "learning_rate": 9.603458421981803e-06, "loss": 0.2485, "step": 70138 }, { "epoch": 4.23, "learning_rate": 9.601949511867578e-06, "loss": 0.4393, "step": 70140 }, { "epoch": 4.23, "learning_rate": 9.600440601753354e-06, "loss": 0.3459, "step": 70142 }, { "epoch": 4.23, "learning_rate": 9.598931691639129e-06, "loss": 0.2157, "step": 70144 }, { "epoch": 4.23, "learning_rate": 9.597422781524906e-06, "loss": 0.1833, "step": 70146 }, { "epoch": 4.23, "learning_rate": 9.59591387141068e-06, "loss": 0.218, "step": 70148 }, { "epoch": 4.23, "learning_rate": 9.594404961296457e-06, "loss": 0.233, "step": 70150 }, { "epoch": 4.23, "learning_rate": 9.592896051182232e-06, "loss": 0.1879, "step": 70152 }, { "epoch": 4.23, "learning_rate": 9.591387141068008e-06, "loss": 0.2791, "step": 70154 }, { "epoch": 4.23, "learning_rate": 9.589878230953781e-06, "loss": 0.3255, "step": 70156 }, { "epoch": 4.23, "learning_rate": 9.588369320839558e-06, "loss": 0.1706, "step": 70158 }, { "epoch": 4.23, "learning_rate": 9.586860410725333e-06, "loss": 0.1603, "step": 70160 }, { "epoch": 4.23, "learning_rate": 9.58535150061111e-06, "loss": 0.2394, "step": 70162 }, { "epoch": 4.23, "learning_rate": 9.583842590496884e-06, "loss": 0.3231, "step": 70164 }, { "epoch": 4.24, "learning_rate": 9.58233368038266e-06, "loss": 0.1804, "step": 70166 }, { "epoch": 4.24, "learning_rate": 9.580824770268435e-06, "loss": 0.2373, "step": 70168 }, { "epoch": 4.24, "learning_rate": 9.579315860154212e-06, "loss": 0.2325, "step": 70170 }, { "epoch": 4.24, "learning_rate": 9.577806950039987e-06, "loss": 0.2001, "step": 70172 }, { "epoch": 4.24, "learning_rate": 9.576298039925762e-06, "loss": 0.1552, "step": 70174 }, { "epoch": 4.24, "learning_rate": 9.574789129811536e-06, "loss": 0.2052, "step": 70176 }, { "epoch": 4.24, "learning_rate": 9.573280219697313e-06, "loss": 0.1847, "step": 70178 }, { "epoch": 4.24, "learning_rate": 9.571771309583088e-06, "loss": 0.2223, "step": 70180 }, { "epoch": 4.24, "learning_rate": 9.570262399468864e-06, "loss": 0.4182, "step": 70182 }, { "epoch": 4.24, "learning_rate": 9.568753489354639e-06, "loss": 0.2413, "step": 70184 }, { "epoch": 4.24, "learning_rate": 9.567244579240416e-06, "loss": 0.1957, "step": 70186 }, { "epoch": 4.24, "learning_rate": 9.56573566912619e-06, "loss": 0.2417, "step": 70188 }, { "epoch": 4.24, "learning_rate": 9.564226759011965e-06, "loss": 0.2146, "step": 70190 }, { "epoch": 4.24, "learning_rate": 9.562717848897742e-06, "loss": 0.1723, "step": 70192 }, { "epoch": 4.24, "learning_rate": 9.561208938783517e-06, "loss": 0.173, "step": 70194 }, { "epoch": 4.24, "learning_rate": 9.559700028669293e-06, "loss": 0.3077, "step": 70196 }, { "epoch": 4.24, "learning_rate": 9.558191118555068e-06, "loss": 0.2486, "step": 70198 }, { "epoch": 4.24, "learning_rate": 9.556682208440844e-06, "loss": 0.297, "step": 70200 }, { "epoch": 4.24, "learning_rate": 9.55517329832662e-06, "loss": 0.2407, "step": 70202 }, { "epoch": 4.24, "learning_rate": 9.553664388212396e-06, "loss": 0.2124, "step": 70204 }, { "epoch": 4.24, "learning_rate": 9.552155478098169e-06, "loss": 0.3162, "step": 70206 }, { "epoch": 4.24, "learning_rate": 9.550646567983945e-06, "loss": 0.1676, "step": 70208 }, { "epoch": 4.24, "learning_rate": 9.54913765786972e-06, "loss": 0.1345, "step": 70210 }, { "epoch": 4.24, "learning_rate": 9.547628747755497e-06, "loss": 0.2883, "step": 70212 }, { "epoch": 4.24, "learning_rate": 9.546119837641272e-06, "loss": 0.2396, "step": 70214 }, { "epoch": 4.24, "learning_rate": 9.544610927527048e-06, "loss": 0.3279, "step": 70216 }, { "epoch": 4.24, "learning_rate": 9.543102017412823e-06, "loss": 0.2133, "step": 70218 }, { "epoch": 4.24, "learning_rate": 9.5415931072986e-06, "loss": 0.2277, "step": 70220 }, { "epoch": 4.24, "learning_rate": 9.540084197184374e-06, "loss": 0.1624, "step": 70222 }, { "epoch": 4.24, "learning_rate": 9.538575287070149e-06, "loss": 0.2129, "step": 70224 }, { "epoch": 4.24, "learning_rate": 9.537066376955926e-06, "loss": 0.3018, "step": 70226 }, { "epoch": 4.24, "learning_rate": 9.5355574668417e-06, "loss": 0.2586, "step": 70228 }, { "epoch": 4.24, "learning_rate": 9.534048556727477e-06, "loss": 0.3219, "step": 70230 }, { "epoch": 4.24, "learning_rate": 9.532539646613252e-06, "loss": 0.2761, "step": 70232 }, { "epoch": 4.24, "learning_rate": 9.531030736499028e-06, "loss": 0.2244, "step": 70234 }, { "epoch": 4.24, "learning_rate": 9.529521826384803e-06, "loss": 0.2065, "step": 70236 }, { "epoch": 4.24, "learning_rate": 9.52801291627058e-06, "loss": 0.3072, "step": 70238 }, { "epoch": 4.24, "learning_rate": 9.526504006156353e-06, "loss": 0.2457, "step": 70240 }, { "epoch": 4.24, "learning_rate": 9.52499509604213e-06, "loss": 0.1292, "step": 70242 }, { "epoch": 4.24, "learning_rate": 9.523486185927904e-06, "loss": 0.1332, "step": 70244 }, { "epoch": 4.24, "learning_rate": 9.52197727581368e-06, "loss": 0.1952, "step": 70246 }, { "epoch": 4.24, "learning_rate": 9.520468365699455e-06, "loss": 0.1632, "step": 70248 }, { "epoch": 4.24, "learning_rate": 9.518959455585232e-06, "loss": 0.1697, "step": 70250 }, { "epoch": 4.24, "learning_rate": 9.517450545471007e-06, "loss": 0.1959, "step": 70252 }, { "epoch": 4.24, "learning_rate": 9.515941635356783e-06, "loss": 0.1704, "step": 70254 }, { "epoch": 4.24, "learning_rate": 9.514432725242558e-06, "loss": 0.1667, "step": 70256 }, { "epoch": 4.24, "learning_rate": 9.512923815128333e-06, "loss": 0.146, "step": 70258 }, { "epoch": 4.24, "learning_rate": 9.51141490501411e-06, "loss": 0.1554, "step": 70260 }, { "epoch": 4.24, "learning_rate": 9.509905994899884e-06, "loss": 0.3212, "step": 70262 }, { "epoch": 4.24, "learning_rate": 9.50839708478566e-06, "loss": 0.2218, "step": 70264 }, { "epoch": 4.24, "learning_rate": 9.506888174671435e-06, "loss": 0.2368, "step": 70266 }, { "epoch": 4.24, "learning_rate": 9.505379264557212e-06, "loss": 0.2833, "step": 70268 }, { "epoch": 4.24, "learning_rate": 9.503870354442987e-06, "loss": 0.3369, "step": 70270 }, { "epoch": 4.24, "learning_rate": 9.502361444328762e-06, "loss": 0.2375, "step": 70272 }, { "epoch": 4.24, "learning_rate": 9.500852534214536e-06, "loss": 0.1641, "step": 70274 }, { "epoch": 4.24, "learning_rate": 9.499343624100313e-06, "loss": 0.2963, "step": 70276 }, { "epoch": 4.24, "learning_rate": 9.497834713986088e-06, "loss": 0.2064, "step": 70278 }, { "epoch": 4.24, "learning_rate": 9.496325803871864e-06, "loss": 0.2774, "step": 70280 }, { "epoch": 4.24, "learning_rate": 9.494816893757639e-06, "loss": 0.247, "step": 70282 }, { "epoch": 4.24, "learning_rate": 9.493307983643416e-06, "loss": 0.261, "step": 70284 }, { "epoch": 4.24, "learning_rate": 9.49179907352919e-06, "loss": 0.2629, "step": 70286 }, { "epoch": 4.24, "learning_rate": 9.490290163414965e-06, "loss": 0.2504, "step": 70288 }, { "epoch": 4.24, "learning_rate": 9.48878125330074e-06, "loss": 0.2221, "step": 70290 }, { "epoch": 4.24, "learning_rate": 9.487272343186517e-06, "loss": 0.3495, "step": 70292 }, { "epoch": 4.24, "learning_rate": 9.485763433072291e-06, "loss": 0.1537, "step": 70294 }, { "epoch": 4.24, "learning_rate": 9.484254522958068e-06, "loss": 0.172, "step": 70296 }, { "epoch": 4.24, "learning_rate": 9.482745612843843e-06, "loss": 0.283, "step": 70298 }, { "epoch": 4.24, "learning_rate": 9.48123670272962e-06, "loss": 0.2062, "step": 70300 }, { "epoch": 4.24, "learning_rate": 9.479727792615394e-06, "loss": 0.2221, "step": 70302 }, { "epoch": 4.24, "learning_rate": 9.478218882501169e-06, "loss": 0.2343, "step": 70304 }, { "epoch": 4.24, "learning_rate": 9.476709972386945e-06, "loss": 0.1914, "step": 70306 }, { "epoch": 4.24, "learning_rate": 9.47520106227272e-06, "loss": 0.2211, "step": 70308 }, { "epoch": 4.24, "learning_rate": 9.473692152158497e-06, "loss": 0.2587, "step": 70310 }, { "epoch": 4.24, "learning_rate": 9.472937697101385e-06, "loss": 0.3661, "step": 70312 }, { "epoch": 4.24, "learning_rate": 9.47142878698716e-06, "loss": 0.2324, "step": 70314 }, { "epoch": 4.24, "learning_rate": 9.469919876872935e-06, "loss": 0.1448, "step": 70316 }, { "epoch": 4.24, "learning_rate": 9.468410966758711e-06, "loss": 0.1982, "step": 70318 }, { "epoch": 4.24, "learning_rate": 9.466902056644486e-06, "loss": 0.2728, "step": 70320 }, { "epoch": 4.24, "learning_rate": 9.465393146530262e-06, "loss": 0.2573, "step": 70322 }, { "epoch": 4.24, "learning_rate": 9.463884236416037e-06, "loss": 0.3748, "step": 70324 }, { "epoch": 4.24, "learning_rate": 9.462375326301814e-06, "loss": 0.1865, "step": 70326 }, { "epoch": 4.24, "learning_rate": 9.460866416187589e-06, "loss": 0.1361, "step": 70328 }, { "epoch": 4.24, "learning_rate": 9.459357506073363e-06, "loss": 0.1356, "step": 70330 }, { "epoch": 4.25, "learning_rate": 9.457848595959138e-06, "loss": 0.256, "step": 70332 }, { "epoch": 4.25, "learning_rate": 9.456339685844915e-06, "loss": 0.2535, "step": 70334 }, { "epoch": 4.25, "learning_rate": 9.45483077573069e-06, "loss": 0.1994, "step": 70336 }, { "epoch": 4.25, "learning_rate": 9.453321865616466e-06, "loss": 0.236, "step": 70338 }, { "epoch": 4.25, "learning_rate": 9.451812955502241e-06, "loss": 0.2465, "step": 70340 }, { "epoch": 4.25, "learning_rate": 9.450304045388017e-06, "loss": 0.1931, "step": 70342 }, { "epoch": 4.25, "learning_rate": 9.448795135273792e-06, "loss": 0.192, "step": 70344 }, { "epoch": 4.25, "learning_rate": 9.447286225159567e-06, "loss": 0.1792, "step": 70346 }, { "epoch": 4.25, "learning_rate": 9.445777315045342e-06, "loss": 0.2355, "step": 70348 }, { "epoch": 4.25, "learning_rate": 9.444268404931118e-06, "loss": 0.1192, "step": 70350 }, { "epoch": 4.25, "learning_rate": 9.442759494816893e-06, "loss": 0.2643, "step": 70352 }, { "epoch": 4.25, "learning_rate": 9.44125058470267e-06, "loss": 0.1607, "step": 70354 }, { "epoch": 4.25, "learning_rate": 9.439741674588445e-06, "loss": 0.2404, "step": 70356 }, { "epoch": 4.25, "learning_rate": 9.438232764474221e-06, "loss": 0.1674, "step": 70358 }, { "epoch": 4.25, "learning_rate": 9.436723854359996e-06, "loss": 0.2601, "step": 70360 }, { "epoch": 4.25, "learning_rate": 9.435214944245772e-06, "loss": 0.2046, "step": 70362 }, { "epoch": 4.25, "learning_rate": 9.433706034131547e-06, "loss": 0.2173, "step": 70364 }, { "epoch": 4.25, "learning_rate": 9.432197124017322e-06, "loss": 0.1586, "step": 70366 }, { "epoch": 4.25, "learning_rate": 9.430688213903099e-06, "loss": 0.3126, "step": 70368 }, { "epoch": 4.25, "learning_rate": 9.429179303788873e-06, "loss": 0.1588, "step": 70370 }, { "epoch": 4.25, "learning_rate": 9.42767039367465e-06, "loss": 0.2647, "step": 70372 }, { "epoch": 4.25, "learning_rate": 9.426161483560425e-06, "loss": 0.1842, "step": 70374 }, { "epoch": 4.25, "learning_rate": 9.424652573446201e-06, "loss": 0.2038, "step": 70376 }, { "epoch": 4.25, "learning_rate": 9.423143663331976e-06, "loss": 0.244, "step": 70378 }, { "epoch": 4.25, "learning_rate": 9.421634753217751e-06, "loss": 0.1883, "step": 70380 }, { "epoch": 4.25, "learning_rate": 9.420125843103526e-06, "loss": 0.2147, "step": 70382 }, { "epoch": 4.25, "learning_rate": 9.418616932989302e-06, "loss": 0.2114, "step": 70384 }, { "epoch": 4.25, "learning_rate": 9.417108022875077e-06, "loss": 0.3521, "step": 70386 }, { "epoch": 4.25, "learning_rate": 9.415599112760854e-06, "loss": 0.1741, "step": 70388 }, { "epoch": 4.25, "learning_rate": 9.414090202646628e-06, "loss": 0.172, "step": 70390 }, { "epoch": 4.25, "learning_rate": 9.412581292532405e-06, "loss": 0.557, "step": 70392 }, { "epoch": 4.25, "learning_rate": 9.41107238241818e-06, "loss": 0.2822, "step": 70394 }, { "epoch": 4.25, "learning_rate": 9.409563472303955e-06, "loss": 0.2381, "step": 70396 }, { "epoch": 4.25, "learning_rate": 9.408054562189731e-06, "loss": 0.2245, "step": 70398 }, { "epoch": 4.25, "learning_rate": 9.406545652075506e-06, "loss": 0.1613, "step": 70400 }, { "epoch": 4.25, "learning_rate": 9.405036741961282e-06, "loss": 0.2361, "step": 70402 }, { "epoch": 4.25, "learning_rate": 9.403527831847057e-06, "loss": 0.2831, "step": 70404 }, { "epoch": 4.25, "learning_rate": 9.402018921732834e-06, "loss": 0.2929, "step": 70406 }, { "epoch": 4.25, "learning_rate": 9.400510011618609e-06, "loss": 0.2576, "step": 70408 }, { "epoch": 4.25, "learning_rate": 9.399001101504385e-06, "loss": 0.3541, "step": 70410 }, { "epoch": 4.25, "learning_rate": 9.397492191390158e-06, "loss": 0.276, "step": 70412 }, { "epoch": 4.25, "learning_rate": 9.395983281275935e-06, "loss": 0.317, "step": 70414 }, { "epoch": 4.25, "learning_rate": 9.39447437116171e-06, "loss": 0.2287, "step": 70416 }, { "epoch": 4.25, "learning_rate": 9.392965461047486e-06, "loss": 0.2834, "step": 70418 }, { "epoch": 4.25, "learning_rate": 9.39145655093326e-06, "loss": 0.3107, "step": 70420 }, { "epoch": 4.25, "learning_rate": 9.389947640819037e-06, "loss": 0.2282, "step": 70422 }, { "epoch": 4.25, "learning_rate": 9.388438730704812e-06, "loss": 0.354, "step": 70424 }, { "epoch": 4.25, "learning_rate": 9.386929820590589e-06, "loss": 0.1955, "step": 70426 }, { "epoch": 4.25, "learning_rate": 9.385420910476363e-06, "loss": 0.2847, "step": 70428 }, { "epoch": 4.25, "learning_rate": 9.383912000362138e-06, "loss": 0.1962, "step": 70430 }, { "epoch": 4.25, "learning_rate": 9.382403090247915e-06, "loss": 0.3615, "step": 70432 }, { "epoch": 4.25, "learning_rate": 9.38089418013369e-06, "loss": 0.3519, "step": 70434 }, { "epoch": 4.25, "learning_rate": 9.379385270019466e-06, "loss": 0.2342, "step": 70436 }, { "epoch": 4.25, "learning_rate": 9.377876359905241e-06, "loss": 0.2785, "step": 70438 }, { "epoch": 4.25, "learning_rate": 9.376367449791017e-06, "loss": 0.2371, "step": 70440 }, { "epoch": 4.25, "learning_rate": 9.374858539676792e-06, "loss": 0.1582, "step": 70442 }, { "epoch": 4.25, "learning_rate": 9.373349629562569e-06, "loss": 0.2403, "step": 70444 }, { "epoch": 4.25, "learning_rate": 9.371840719448342e-06, "loss": 0.1783, "step": 70446 }, { "epoch": 4.25, "learning_rate": 9.370331809334118e-06, "loss": 0.2002, "step": 70448 }, { "epoch": 4.25, "learning_rate": 9.368822899219893e-06, "loss": 0.1617, "step": 70450 }, { "epoch": 4.25, "learning_rate": 9.36731398910567e-06, "loss": 0.1726, "step": 70452 }, { "epoch": 4.25, "learning_rate": 9.365805078991445e-06, "loss": 0.4509, "step": 70454 }, { "epoch": 4.25, "learning_rate": 9.364296168877221e-06, "loss": 0.2098, "step": 70456 }, { "epoch": 4.25, "learning_rate": 9.362787258762996e-06, "loss": 0.2406, "step": 70458 }, { "epoch": 4.25, "learning_rate": 9.361278348648772e-06, "loss": 0.2197, "step": 70460 }, { "epoch": 4.25, "learning_rate": 9.359769438534546e-06, "loss": 0.2668, "step": 70462 }, { "epoch": 4.25, "learning_rate": 9.358260528420322e-06, "loss": 0.2643, "step": 70464 }, { "epoch": 4.25, "learning_rate": 9.356751618306097e-06, "loss": 0.2408, "step": 70466 }, { "epoch": 4.25, "learning_rate": 9.355242708191873e-06, "loss": 0.207, "step": 70468 }, { "epoch": 4.25, "learning_rate": 9.353733798077648e-06, "loss": 0.3512, "step": 70470 }, { "epoch": 4.25, "learning_rate": 9.352224887963425e-06, "loss": 0.1638, "step": 70472 }, { "epoch": 4.25, "learning_rate": 9.3507159778492e-06, "loss": 0.3992, "step": 70474 }, { "epoch": 4.25, "learning_rate": 9.349207067734976e-06, "loss": 0.2074, "step": 70476 }, { "epoch": 4.25, "learning_rate": 9.347698157620751e-06, "loss": 0.2784, "step": 70478 }, { "epoch": 4.25, "learning_rate": 9.346189247506526e-06, "loss": 0.2625, "step": 70480 }, { "epoch": 4.25, "learning_rate": 9.344680337392302e-06, "loss": 0.3144, "step": 70482 }, { "epoch": 4.25, "learning_rate": 9.343171427278077e-06, "loss": 0.2136, "step": 70484 }, { "epoch": 4.25, "learning_rate": 9.341662517163854e-06, "loss": 0.2633, "step": 70486 }, { "epoch": 4.25, "learning_rate": 9.340153607049628e-06, "loss": 0.2058, "step": 70488 }, { "epoch": 4.25, "learning_rate": 9.338644696935405e-06, "loss": 0.2367, "step": 70490 }, { "epoch": 4.25, "learning_rate": 9.33713578682118e-06, "loss": 0.2069, "step": 70492 }, { "epoch": 4.25, "learning_rate": 9.335626876706955e-06, "loss": 0.1827, "step": 70494 }, { "epoch": 4.25, "learning_rate": 9.33411796659273e-06, "loss": 0.2324, "step": 70496 }, { "epoch": 4.26, "learning_rate": 9.332609056478506e-06, "loss": 0.3059, "step": 70498 }, { "epoch": 4.26, "learning_rate": 9.33110014636428e-06, "loss": 0.2225, "step": 70500 }, { "epoch": 4.26, "learning_rate": 9.329591236250057e-06, "loss": 0.2296, "step": 70502 }, { "epoch": 4.26, "learning_rate": 9.328082326135832e-06, "loss": 0.2013, "step": 70504 }, { "epoch": 4.26, "learning_rate": 9.326573416021609e-06, "loss": 0.1738, "step": 70506 }, { "epoch": 4.26, "learning_rate": 9.325064505907383e-06, "loss": 0.1884, "step": 70508 }, { "epoch": 4.26, "learning_rate": 9.32355559579316e-06, "loss": 0.2502, "step": 70510 }, { "epoch": 4.26, "learning_rate": 9.322046685678935e-06, "loss": 0.2231, "step": 70512 }, { "epoch": 4.26, "learning_rate": 9.32053777556471e-06, "loss": 0.3169, "step": 70514 }, { "epoch": 4.26, "learning_rate": 9.319028865450486e-06, "loss": 0.1927, "step": 70516 }, { "epoch": 4.26, "learning_rate": 9.317519955336261e-06, "loss": 0.1917, "step": 70518 }, { "epoch": 4.26, "learning_rate": 9.316011045222037e-06, "loss": 0.242, "step": 70520 }, { "epoch": 4.26, "learning_rate": 9.314502135107812e-06, "loss": 0.2619, "step": 70522 }, { "epoch": 4.26, "learning_rate": 9.312993224993589e-06, "loss": 0.2385, "step": 70524 }, { "epoch": 4.26, "learning_rate": 9.311484314879364e-06, "loss": 0.2951, "step": 70526 }, { "epoch": 4.26, "learning_rate": 9.309975404765138e-06, "loss": 0.1504, "step": 70528 }, { "epoch": 4.26, "learning_rate": 9.308466494650913e-06, "loss": 0.2384, "step": 70530 }, { "epoch": 4.26, "learning_rate": 9.30695758453669e-06, "loss": 0.2073, "step": 70532 }, { "epoch": 4.26, "learning_rate": 9.305448674422465e-06, "loss": 0.1958, "step": 70534 }, { "epoch": 4.26, "learning_rate": 9.303939764308241e-06, "loss": 0.2478, "step": 70536 }, { "epoch": 4.26, "learning_rate": 9.302430854194016e-06, "loss": 0.3072, "step": 70538 }, { "epoch": 4.26, "learning_rate": 9.300921944079792e-06, "loss": 0.2825, "step": 70540 }, { "epoch": 4.26, "learning_rate": 9.299413033965567e-06, "loss": 0.1919, "step": 70542 }, { "epoch": 4.26, "learning_rate": 9.297904123851342e-06, "loss": 0.3097, "step": 70544 }, { "epoch": 4.26, "learning_rate": 9.296395213737119e-06, "loss": 0.1547, "step": 70546 }, { "epoch": 4.26, "learning_rate": 9.294886303622893e-06, "loss": 0.1304, "step": 70548 }, { "epoch": 4.26, "learning_rate": 9.29337739350867e-06, "loss": 0.2011, "step": 70550 }, { "epoch": 4.26, "learning_rate": 9.291868483394445e-06, "loss": 0.2427, "step": 70552 }, { "epoch": 4.26, "learning_rate": 9.290359573280221e-06, "loss": 0.2812, "step": 70554 }, { "epoch": 4.26, "learning_rate": 9.288850663165996e-06, "loss": 0.1908, "step": 70556 }, { "epoch": 4.26, "learning_rate": 9.287341753051773e-06, "loss": 0.3516, "step": 70558 }, { "epoch": 4.26, "learning_rate": 9.285832842937546e-06, "loss": 0.2689, "step": 70560 }, { "epoch": 4.26, "learning_rate": 9.284323932823322e-06, "loss": 0.2487, "step": 70562 }, { "epoch": 4.26, "learning_rate": 9.282815022709097e-06, "loss": 0.2963, "step": 70564 }, { "epoch": 4.26, "learning_rate": 9.281306112594873e-06, "loss": 0.1949, "step": 70566 }, { "epoch": 4.26, "learning_rate": 9.279797202480648e-06, "loss": 0.21, "step": 70568 }, { "epoch": 4.26, "learning_rate": 9.278288292366425e-06, "loss": 0.2222, "step": 70570 }, { "epoch": 4.26, "learning_rate": 9.2767793822522e-06, "loss": 0.1651, "step": 70572 }, { "epoch": 4.26, "learning_rate": 9.275270472137976e-06, "loss": 0.2823, "step": 70574 }, { "epoch": 4.26, "learning_rate": 9.27376156202375e-06, "loss": 0.3673, "step": 70576 }, { "epoch": 4.26, "learning_rate": 9.272252651909526e-06, "loss": 0.4236, "step": 70578 }, { "epoch": 4.26, "learning_rate": 9.2707437417953e-06, "loss": 0.2513, "step": 70580 }, { "epoch": 4.26, "learning_rate": 9.269234831681077e-06, "loss": 0.1956, "step": 70582 }, { "epoch": 4.26, "learning_rate": 9.267725921566852e-06, "loss": 0.2561, "step": 70584 }, { "epoch": 4.26, "learning_rate": 9.266217011452628e-06, "loss": 0.2435, "step": 70586 }, { "epoch": 4.26, "learning_rate": 9.264708101338403e-06, "loss": 0.3683, "step": 70588 }, { "epoch": 4.26, "learning_rate": 9.26319919122418e-06, "loss": 0.2778, "step": 70590 }, { "epoch": 4.26, "learning_rate": 9.261690281109955e-06, "loss": 0.1991, "step": 70592 }, { "epoch": 4.26, "learning_rate": 9.26018137099573e-06, "loss": 0.1472, "step": 70594 }, { "epoch": 4.26, "learning_rate": 9.258672460881506e-06, "loss": 0.3046, "step": 70596 }, { "epoch": 4.26, "learning_rate": 9.25716355076728e-06, "loss": 0.2782, "step": 70598 }, { "epoch": 4.26, "learning_rate": 9.255654640653057e-06, "loss": 0.1967, "step": 70600 }, { "epoch": 4.26, "learning_rate": 9.254145730538832e-06, "loss": 0.2578, "step": 70602 }, { "epoch": 4.26, "learning_rate": 9.252636820424609e-06, "loss": 0.2716, "step": 70604 }, { "epoch": 4.26, "learning_rate": 9.251127910310383e-06, "loss": 0.351, "step": 70606 }, { "epoch": 4.26, "learning_rate": 9.24961900019616e-06, "loss": 0.1881, "step": 70608 }, { "epoch": 4.26, "learning_rate": 9.248110090081933e-06, "loss": 0.2292, "step": 70610 }, { "epoch": 4.26, "learning_rate": 9.24660117996771e-06, "loss": 0.2689, "step": 70612 }, { "epoch": 4.26, "learning_rate": 9.245092269853484e-06, "loss": 0.1863, "step": 70614 }, { "epoch": 4.26, "learning_rate": 9.243583359739261e-06, "loss": 0.2632, "step": 70616 }, { "epoch": 4.26, "learning_rate": 9.242074449625036e-06, "loss": 0.1545, "step": 70618 }, { "epoch": 4.26, "learning_rate": 9.240565539510812e-06, "loss": 0.2603, "step": 70620 }, { "epoch": 4.26, "learning_rate": 9.239056629396587e-06, "loss": 0.2742, "step": 70622 }, { "epoch": 4.26, "learning_rate": 9.237547719282364e-06, "loss": 0.232, "step": 70624 }, { "epoch": 4.26, "learning_rate": 9.236038809168138e-06, "loss": 0.2572, "step": 70626 }, { "epoch": 4.26, "learning_rate": 9.234529899053913e-06, "loss": 0.2007, "step": 70628 }, { "epoch": 4.26, "learning_rate": 9.23302098893969e-06, "loss": 0.2661, "step": 70630 }, { "epoch": 4.26, "learning_rate": 9.231512078825465e-06, "loss": 0.1995, "step": 70632 }, { "epoch": 4.26, "learning_rate": 9.230003168711241e-06, "loss": 0.2627, "step": 70634 }, { "epoch": 4.26, "learning_rate": 9.228494258597016e-06, "loss": 0.144, "step": 70636 }, { "epoch": 4.26, "learning_rate": 9.226985348482792e-06, "loss": 0.2711, "step": 70638 }, { "epoch": 4.26, "learning_rate": 9.225476438368567e-06, "loss": 0.2114, "step": 70640 }, { "epoch": 4.26, "learning_rate": 9.223967528254342e-06, "loss": 0.3332, "step": 70642 }, { "epoch": 4.26, "learning_rate": 9.222458618140117e-06, "loss": 0.1063, "step": 70644 }, { "epoch": 4.26, "learning_rate": 9.220949708025893e-06, "loss": 0.2513, "step": 70646 }, { "epoch": 4.26, "learning_rate": 9.219440797911668e-06, "loss": 0.1608, "step": 70648 }, { "epoch": 4.26, "learning_rate": 9.217931887797445e-06, "loss": 0.193, "step": 70650 }, { "epoch": 4.26, "learning_rate": 9.21642297768322e-06, "loss": 0.14, "step": 70652 }, { "epoch": 4.26, "learning_rate": 9.214914067568996e-06, "loss": 0.3455, "step": 70654 }, { "epoch": 4.26, "learning_rate": 9.213405157454771e-06, "loss": 0.1212, "step": 70656 }, { "epoch": 4.26, "learning_rate": 9.211896247340546e-06, "loss": 0.1754, "step": 70658 }, { "epoch": 4.26, "learning_rate": 9.210387337226322e-06, "loss": 0.307, "step": 70660 }, { "epoch": 4.26, "learning_rate": 9.208878427112097e-06, "loss": 0.2544, "step": 70662 }, { "epoch": 4.27, "learning_rate": 9.207369516997874e-06, "loss": 0.2291, "step": 70664 }, { "epoch": 4.27, "learning_rate": 9.205860606883648e-06, "loss": 0.2114, "step": 70666 }, { "epoch": 4.27, "learning_rate": 9.204351696769425e-06, "loss": 0.1786, "step": 70668 }, { "epoch": 4.27, "learning_rate": 9.2028427866552e-06, "loss": 0.176, "step": 70670 }, { "epoch": 4.27, "learning_rate": 9.201333876540976e-06, "loss": 0.1712, "step": 70672 }, { "epoch": 4.27, "learning_rate": 9.199824966426751e-06, "loss": 0.166, "step": 70674 }, { "epoch": 4.27, "learning_rate": 9.198316056312526e-06, "loss": 0.3146, "step": 70676 }, { "epoch": 4.27, "learning_rate": 9.1968071461983e-06, "loss": 0.2633, "step": 70678 }, { "epoch": 4.27, "learning_rate": 9.195298236084077e-06, "loss": 0.1887, "step": 70680 }, { "epoch": 4.27, "learning_rate": 9.193789325969852e-06, "loss": 0.385, "step": 70682 }, { "epoch": 4.27, "learning_rate": 9.192280415855628e-06, "loss": 0.199, "step": 70684 }, { "epoch": 4.27, "learning_rate": 9.190771505741403e-06, "loss": 0.2694, "step": 70686 }, { "epoch": 4.27, "learning_rate": 9.18926259562718e-06, "loss": 0.2543, "step": 70688 }, { "epoch": 4.27, "learning_rate": 9.187753685512955e-06, "loss": 0.2884, "step": 70690 }, { "epoch": 4.27, "learning_rate": 9.18624477539873e-06, "loss": 0.2357, "step": 70692 }, { "epoch": 4.27, "learning_rate": 9.184735865284504e-06, "loss": 0.1895, "step": 70694 }, { "epoch": 4.27, "learning_rate": 9.18322695517028e-06, "loss": 0.2538, "step": 70696 }, { "epoch": 4.27, "learning_rate": 9.181718045056056e-06, "loss": 0.1888, "step": 70698 }, { "epoch": 4.27, "learning_rate": 9.180209134941832e-06, "loss": 0.166, "step": 70700 }, { "epoch": 4.27, "learning_rate": 9.178700224827607e-06, "loss": 0.2688, "step": 70702 }, { "epoch": 4.27, "learning_rate": 9.177191314713383e-06, "loss": 0.2778, "step": 70704 }, { "epoch": 4.27, "learning_rate": 9.175682404599158e-06, "loss": 0.1842, "step": 70706 }, { "epoch": 4.27, "learning_rate": 9.174173494484933e-06, "loss": 0.1737, "step": 70708 }, { "epoch": 4.27, "learning_rate": 9.17266458437071e-06, "loss": 0.2597, "step": 70710 }, { "epoch": 4.27, "learning_rate": 9.171155674256484e-06, "loss": 0.4097, "step": 70712 }, { "epoch": 4.27, "learning_rate": 9.169646764142261e-06, "loss": 0.2025, "step": 70714 }, { "epoch": 4.27, "learning_rate": 9.168137854028036e-06, "loss": 0.2378, "step": 70716 }, { "epoch": 4.27, "learning_rate": 9.166628943913812e-06, "loss": 0.1796, "step": 70718 }, { "epoch": 4.27, "learning_rate": 9.165120033799587e-06, "loss": 0.233, "step": 70720 }, { "epoch": 4.27, "learning_rate": 9.163611123685364e-06, "loss": 0.2941, "step": 70722 }, { "epoch": 4.27, "learning_rate": 9.162102213571137e-06, "loss": 0.2488, "step": 70724 }, { "epoch": 4.27, "learning_rate": 9.160593303456913e-06, "loss": 0.2374, "step": 70726 }, { "epoch": 4.27, "learning_rate": 9.159084393342688e-06, "loss": 0.2242, "step": 70728 }, { "epoch": 4.27, "learning_rate": 9.157575483228465e-06, "loss": 0.3077, "step": 70730 }, { "epoch": 4.27, "learning_rate": 9.15606657311424e-06, "loss": 0.2708, "step": 70732 }, { "epoch": 4.27, "learning_rate": 9.154557663000016e-06, "loss": 0.2951, "step": 70734 }, { "epoch": 4.27, "learning_rate": 9.15304875288579e-06, "loss": 0.1799, "step": 70736 }, { "epoch": 4.27, "learning_rate": 9.151539842771567e-06, "loss": 0.1924, "step": 70738 }, { "epoch": 4.27, "learning_rate": 9.150030932657342e-06, "loss": 0.2158, "step": 70740 }, { "epoch": 4.27, "learning_rate": 9.148522022543117e-06, "loss": 0.2037, "step": 70742 }, { "epoch": 4.27, "learning_rate": 9.147013112428893e-06, "loss": 0.235, "step": 70744 }, { "epoch": 4.27, "learning_rate": 9.145504202314668e-06, "loss": 0.1215, "step": 70746 }, { "epoch": 4.27, "learning_rate": 9.143995292200445e-06, "loss": 0.1966, "step": 70748 }, { "epoch": 4.27, "learning_rate": 9.14248638208622e-06, "loss": 0.1626, "step": 70750 }, { "epoch": 4.27, "learning_rate": 9.140977471971996e-06, "loss": 0.2379, "step": 70752 }, { "epoch": 4.27, "learning_rate": 9.139468561857771e-06, "loss": 0.1796, "step": 70754 }, { "epoch": 4.27, "learning_rate": 9.137959651743547e-06, "loss": 0.2399, "step": 70756 }, { "epoch": 4.27, "learning_rate": 9.13645074162932e-06, "loss": 0.2638, "step": 70758 }, { "epoch": 4.27, "learning_rate": 9.134941831515097e-06, "loss": 0.2679, "step": 70760 }, { "epoch": 4.27, "learning_rate": 9.133432921400872e-06, "loss": 0.2124, "step": 70762 }, { "epoch": 4.27, "learning_rate": 9.131924011286648e-06, "loss": 0.2453, "step": 70764 }, { "epoch": 4.27, "learning_rate": 9.130415101172423e-06, "loss": 0.2863, "step": 70766 }, { "epoch": 4.27, "learning_rate": 9.1289061910582e-06, "loss": 0.2174, "step": 70768 }, { "epoch": 4.27, "learning_rate": 9.127397280943975e-06, "loss": 0.284, "step": 70770 }, { "epoch": 4.27, "learning_rate": 9.125888370829751e-06, "loss": 0.2202, "step": 70772 }, { "epoch": 4.27, "learning_rate": 9.124379460715526e-06, "loss": 0.2046, "step": 70774 }, { "epoch": 4.27, "learning_rate": 9.1228705506013e-06, "loss": 0.2382, "step": 70776 }, { "epoch": 4.27, "learning_rate": 9.121361640487077e-06, "loss": 0.1359, "step": 70778 }, { "epoch": 4.27, "learning_rate": 9.119852730372852e-06, "loss": 0.2143, "step": 70780 }, { "epoch": 4.27, "learning_rate": 9.118343820258629e-06, "loss": 0.2422, "step": 70782 }, { "epoch": 4.27, "learning_rate": 9.116834910144403e-06, "loss": 0.2808, "step": 70784 }, { "epoch": 4.27, "learning_rate": 9.11532600003018e-06, "loss": 0.3154, "step": 70786 }, { "epoch": 4.27, "learning_rate": 9.113817089915955e-06, "loss": 0.2703, "step": 70788 }, { "epoch": 4.27, "learning_rate": 9.11230817980173e-06, "loss": 0.229, "step": 70790 }, { "epoch": 4.27, "learning_rate": 9.110799269687504e-06, "loss": 0.1632, "step": 70792 }, { "epoch": 4.27, "learning_rate": 9.10929035957328e-06, "loss": 0.2719, "step": 70794 }, { "epoch": 4.27, "learning_rate": 9.107781449459056e-06, "loss": 0.241, "step": 70796 }, { "epoch": 4.27, "learning_rate": 9.106272539344832e-06, "loss": 0.2242, "step": 70798 }, { "epoch": 4.27, "learning_rate": 9.104763629230607e-06, "loss": 0.3184, "step": 70800 }, { "epoch": 4.27, "learning_rate": 9.103254719116384e-06, "loss": 0.2463, "step": 70802 }, { "epoch": 4.27, "learning_rate": 9.101745809002158e-06, "loss": 0.2137, "step": 70804 }, { "epoch": 4.27, "learning_rate": 9.100236898887933e-06, "loss": 0.2331, "step": 70806 }, { "epoch": 4.27, "learning_rate": 9.098727988773708e-06, "loss": 0.2155, "step": 70808 }, { "epoch": 4.27, "learning_rate": 9.097219078659484e-06, "loss": 0.256, "step": 70810 }, { "epoch": 4.27, "learning_rate": 9.09571016854526e-06, "loss": 0.1823, "step": 70812 }, { "epoch": 4.27, "learning_rate": 9.094201258431036e-06, "loss": 0.2343, "step": 70814 }, { "epoch": 4.27, "learning_rate": 9.09269234831681e-06, "loss": 0.2368, "step": 70816 }, { "epoch": 4.27, "learning_rate": 9.091183438202587e-06, "loss": 0.1594, "step": 70818 }, { "epoch": 4.27, "learning_rate": 9.089674528088362e-06, "loss": 0.2518, "step": 70820 }, { "epoch": 4.27, "learning_rate": 9.088165617974138e-06, "loss": 0.2477, "step": 70822 }, { "epoch": 4.27, "learning_rate": 9.086656707859913e-06, "loss": 0.3023, "step": 70824 }, { "epoch": 4.27, "learning_rate": 9.085147797745688e-06, "loss": 0.18, "step": 70826 }, { "epoch": 4.27, "learning_rate": 9.083638887631465e-06, "loss": 0.1671, "step": 70828 }, { "epoch": 4.28, "learning_rate": 9.08212997751724e-06, "loss": 0.2988, "step": 70830 }, { "epoch": 4.28, "learning_rate": 9.080621067403016e-06, "loss": 0.2565, "step": 70832 }, { "epoch": 4.28, "learning_rate": 9.07911215728879e-06, "loss": 0.2232, "step": 70834 }, { "epoch": 4.28, "learning_rate": 9.077603247174567e-06, "loss": 0.3474, "step": 70836 }, { "epoch": 4.28, "learning_rate": 9.076094337060342e-06, "loss": 0.1897, "step": 70838 }, { "epoch": 4.28, "learning_rate": 9.074585426946117e-06, "loss": 0.2116, "step": 70840 }, { "epoch": 4.28, "learning_rate": 9.073076516831892e-06, "loss": 0.167, "step": 70842 }, { "epoch": 4.28, "learning_rate": 9.071567606717668e-06, "loss": 0.116, "step": 70844 }, { "epoch": 4.28, "learning_rate": 9.070058696603443e-06, "loss": 0.1947, "step": 70846 }, { "epoch": 4.28, "learning_rate": 9.06854978648922e-06, "loss": 0.1706, "step": 70848 }, { "epoch": 4.28, "learning_rate": 9.067040876374994e-06, "loss": 0.2736, "step": 70850 }, { "epoch": 4.28, "learning_rate": 9.065531966260771e-06, "loss": 0.1913, "step": 70852 }, { "epoch": 4.28, "learning_rate": 9.064023056146546e-06, "loss": 0.2504, "step": 70854 }, { "epoch": 4.28, "learning_rate": 9.06251414603232e-06, "loss": 0.2454, "step": 70856 }, { "epoch": 4.28, "learning_rate": 9.061005235918097e-06, "loss": 0.2794, "step": 70858 }, { "epoch": 4.28, "learning_rate": 9.059496325803872e-06, "loss": 0.4673, "step": 70860 }, { "epoch": 4.28, "learning_rate": 9.057987415689648e-06, "loss": 0.1565, "step": 70862 }, { "epoch": 4.28, "learning_rate": 9.056478505575423e-06, "loss": 0.1605, "step": 70864 }, { "epoch": 4.28, "learning_rate": 9.0549695954612e-06, "loss": 0.1895, "step": 70866 }, { "epoch": 4.28, "learning_rate": 9.053460685346975e-06, "loss": 0.244, "step": 70868 }, { "epoch": 4.28, "learning_rate": 9.051951775232751e-06, "loss": 0.1907, "step": 70870 }, { "epoch": 4.28, "learning_rate": 9.050442865118524e-06, "loss": 0.1931, "step": 70872 }, { "epoch": 4.28, "learning_rate": 9.0489339550043e-06, "loss": 0.3594, "step": 70874 }, { "epoch": 4.28, "learning_rate": 9.047425044890076e-06, "loss": 0.345, "step": 70876 }, { "epoch": 4.28, "learning_rate": 9.045916134775852e-06, "loss": 0.2439, "step": 70878 }, { "epoch": 4.28, "learning_rate": 9.044407224661627e-06, "loss": 0.2407, "step": 70880 }, { "epoch": 4.28, "learning_rate": 9.042898314547403e-06, "loss": 0.2772, "step": 70882 }, { "epoch": 4.28, "learning_rate": 9.041389404433178e-06, "loss": 0.2246, "step": 70884 }, { "epoch": 4.28, "learning_rate": 9.039880494318955e-06, "loss": 0.2259, "step": 70886 }, { "epoch": 4.28, "learning_rate": 9.038371584204728e-06, "loss": 0.2681, "step": 70888 }, { "epoch": 4.28, "learning_rate": 9.036862674090504e-06, "loss": 0.2107, "step": 70890 }, { "epoch": 4.28, "learning_rate": 9.035353763976281e-06, "loss": 0.1922, "step": 70892 }, { "epoch": 4.28, "learning_rate": 9.033844853862056e-06, "loss": 0.253, "step": 70894 }, { "epoch": 4.28, "learning_rate": 9.032335943747832e-06, "loss": 0.3146, "step": 70896 }, { "epoch": 4.28, "learning_rate": 9.030827033633607e-06, "loss": 0.2072, "step": 70898 }, { "epoch": 4.28, "learning_rate": 9.029318123519384e-06, "loss": 0.3769, "step": 70900 }, { "epoch": 4.28, "learning_rate": 9.027809213405158e-06, "loss": 0.2497, "step": 70902 }, { "epoch": 4.28, "learning_rate": 9.026300303290935e-06, "loss": 0.1746, "step": 70904 }, { "epoch": 4.28, "learning_rate": 9.024791393176708e-06, "loss": 0.2115, "step": 70906 }, { "epoch": 4.28, "learning_rate": 9.023282483062485e-06, "loss": 0.3721, "step": 70908 }, { "epoch": 4.28, "learning_rate": 9.02177357294826e-06, "loss": 0.3004, "step": 70910 }, { "epoch": 4.28, "learning_rate": 9.020264662834036e-06, "loss": 0.389, "step": 70912 }, { "epoch": 4.28, "learning_rate": 9.01875575271981e-06, "loss": 0.203, "step": 70914 }, { "epoch": 4.28, "learning_rate": 9.017246842605587e-06, "loss": 0.1884, "step": 70916 }, { "epoch": 4.28, "learning_rate": 9.015737932491362e-06, "loss": 0.2409, "step": 70918 }, { "epoch": 4.28, "learning_rate": 9.014229022377139e-06, "loss": 0.1841, "step": 70920 }, { "epoch": 4.28, "learning_rate": 9.012720112262912e-06, "loss": 0.3062, "step": 70922 }, { "epoch": 4.28, "learning_rate": 9.011211202148688e-06, "loss": 0.275, "step": 70924 }, { "epoch": 4.28, "learning_rate": 9.009702292034463e-06, "loss": 0.3681, "step": 70926 }, { "epoch": 4.28, "learning_rate": 9.00819338192024e-06, "loss": 0.2469, "step": 70928 }, { "epoch": 4.28, "learning_rate": 9.006684471806014e-06, "loss": 0.2677, "step": 70930 }, { "epoch": 4.28, "learning_rate": 9.00517556169179e-06, "loss": 0.2867, "step": 70932 }, { "epoch": 4.28, "learning_rate": 9.003666651577566e-06, "loss": 0.2199, "step": 70934 }, { "epoch": 4.28, "learning_rate": 9.002157741463342e-06, "loss": 0.3952, "step": 70936 }, { "epoch": 4.28, "learning_rate": 9.000648831349117e-06, "loss": 0.2443, "step": 70938 }, { "epoch": 4.28, "learning_rate": 8.999139921234892e-06, "loss": 0.2154, "step": 70940 }, { "epoch": 4.28, "learning_rate": 8.997631011120668e-06, "loss": 0.1671, "step": 70942 }, { "epoch": 4.28, "learning_rate": 8.996122101006443e-06, "loss": 0.234, "step": 70944 }, { "epoch": 4.28, "learning_rate": 8.99461319089222e-06, "loss": 0.2539, "step": 70946 }, { "epoch": 4.28, "learning_rate": 8.993104280777994e-06, "loss": 0.2112, "step": 70948 }, { "epoch": 4.28, "learning_rate": 8.991595370663771e-06, "loss": 0.1882, "step": 70950 }, { "epoch": 4.28, "learning_rate": 8.990086460549546e-06, "loss": 0.2326, "step": 70952 }, { "epoch": 4.28, "learning_rate": 8.98857755043532e-06, "loss": 0.1638, "step": 70954 }, { "epoch": 4.28, "learning_rate": 8.987068640321095e-06, "loss": 0.346, "step": 70956 }, { "epoch": 4.28, "learning_rate": 8.985559730206872e-06, "loss": 0.3349, "step": 70958 }, { "epoch": 4.28, "learning_rate": 8.984050820092647e-06, "loss": 0.1143, "step": 70960 }, { "epoch": 4.28, "learning_rate": 8.982541909978423e-06, "loss": 0.1702, "step": 70962 }, { "epoch": 4.28, "learning_rate": 8.981032999864198e-06, "loss": 0.3343, "step": 70964 }, { "epoch": 4.28, "learning_rate": 8.979524089749975e-06, "loss": 0.2162, "step": 70966 }, { "epoch": 4.28, "learning_rate": 8.97801517963575e-06, "loss": 0.2095, "step": 70968 }, { "epoch": 4.28, "learning_rate": 8.976506269521524e-06, "loss": 0.4497, "step": 70970 }, { "epoch": 4.28, "learning_rate": 8.9749973594073e-06, "loss": 0.1981, "step": 70972 }, { "epoch": 4.28, "learning_rate": 8.973488449293076e-06, "loss": 0.2518, "step": 70974 }, { "epoch": 4.28, "learning_rate": 8.971979539178852e-06, "loss": 0.2557, "step": 70976 }, { "epoch": 4.28, "learning_rate": 8.970470629064627e-06, "loss": 0.2979, "step": 70978 }, { "epoch": 4.28, "learning_rate": 8.968961718950403e-06, "loss": 0.1608, "step": 70980 }, { "epoch": 4.28, "learning_rate": 8.967452808836178e-06, "loss": 0.2059, "step": 70982 }, { "epoch": 4.28, "learning_rate": 8.965943898721955e-06, "loss": 0.2178, "step": 70984 }, { "epoch": 4.28, "learning_rate": 8.96443498860773e-06, "loss": 0.2767, "step": 70986 }, { "epoch": 4.28, "learning_rate": 8.962926078493504e-06, "loss": 0.1978, "step": 70988 }, { "epoch": 4.28, "learning_rate": 8.96141716837928e-06, "loss": 0.2894, "step": 70990 }, { "epoch": 4.28, "learning_rate": 8.959908258265056e-06, "loss": 0.2182, "step": 70992 }, { "epoch": 4.29, "learning_rate": 8.95839934815083e-06, "loss": 0.2644, "step": 70994 }, { "epoch": 4.29, "learning_rate": 8.956890438036607e-06, "loss": 0.2054, "step": 70996 }, { "epoch": 4.29, "learning_rate": 8.955381527922382e-06, "loss": 0.289, "step": 70998 }, { "epoch": 4.29, "learning_rate": 8.953872617808158e-06, "loss": 0.1806, "step": 71000 }, { "epoch": 4.29, "learning_rate": 8.952363707693933e-06, "loss": 0.2536, "step": 71002 }, { "epoch": 4.29, "learning_rate": 8.950854797579708e-06, "loss": 0.2629, "step": 71004 }, { "epoch": 4.29, "learning_rate": 8.949345887465483e-06, "loss": 0.1971, "step": 71006 }, { "epoch": 4.29, "learning_rate": 8.94783697735126e-06, "loss": 0.2611, "step": 71008 }, { "epoch": 4.29, "learning_rate": 8.946328067237034e-06, "loss": 0.3807, "step": 71010 }, { "epoch": 4.29, "learning_rate": 8.94481915712281e-06, "loss": 0.1823, "step": 71012 }, { "epoch": 4.29, "learning_rate": 8.943310247008587e-06, "loss": 0.2398, "step": 71014 }, { "epoch": 4.29, "learning_rate": 8.941801336894362e-06, "loss": 0.2425, "step": 71016 }, { "epoch": 4.29, "learning_rate": 8.940292426780139e-06, "loss": 0.2383, "step": 71018 }, { "epoch": 4.29, "learning_rate": 8.938783516665912e-06, "loss": 0.1909, "step": 71020 }, { "epoch": 4.29, "learning_rate": 8.937274606551688e-06, "loss": 0.1891, "step": 71022 }, { "epoch": 4.29, "learning_rate": 8.935765696437463e-06, "loss": 0.1436, "step": 71024 }, { "epoch": 4.29, "learning_rate": 8.93425678632324e-06, "loss": 0.3364, "step": 71026 }, { "epoch": 4.29, "learning_rate": 8.932747876209014e-06, "loss": 0.2631, "step": 71028 }, { "epoch": 4.29, "learning_rate": 8.931238966094791e-06, "loss": 0.3618, "step": 71030 }, { "epoch": 4.29, "learning_rate": 8.929730055980566e-06, "loss": 0.2163, "step": 71032 }, { "epoch": 4.29, "learning_rate": 8.928221145866342e-06, "loss": 0.2243, "step": 71034 }, { "epoch": 4.29, "learning_rate": 8.926712235752115e-06, "loss": 0.1585, "step": 71036 }, { "epoch": 4.29, "learning_rate": 8.925203325637892e-06, "loss": 0.3382, "step": 71038 }, { "epoch": 4.29, "learning_rate": 8.923694415523667e-06, "loss": 0.1981, "step": 71040 }, { "epoch": 4.29, "learning_rate": 8.922185505409443e-06, "loss": 0.1598, "step": 71042 }, { "epoch": 4.29, "learning_rate": 8.920676595295218e-06, "loss": 0.1946, "step": 71044 }, { "epoch": 4.29, "learning_rate": 8.919167685180994e-06, "loss": 0.1299, "step": 71046 }, { "epoch": 4.29, "learning_rate": 8.91765877506677e-06, "loss": 0.1208, "step": 71048 }, { "epoch": 4.29, "learning_rate": 8.916149864952546e-06, "loss": 0.2341, "step": 71050 }, { "epoch": 4.29, "learning_rate": 8.91464095483832e-06, "loss": 0.2632, "step": 71052 }, { "epoch": 4.29, "learning_rate": 8.913132044724095e-06, "loss": 0.1787, "step": 71054 }, { "epoch": 4.29, "learning_rate": 8.911623134609872e-06, "loss": 0.2229, "step": 71056 }, { "epoch": 4.29, "learning_rate": 8.910114224495647e-06, "loss": 0.2049, "step": 71058 }, { "epoch": 4.29, "learning_rate": 8.908605314381423e-06, "loss": 0.2005, "step": 71060 }, { "epoch": 4.29, "learning_rate": 8.907096404267198e-06, "loss": 0.2635, "step": 71062 }, { "epoch": 4.29, "learning_rate": 8.905587494152975e-06, "loss": 0.1878, "step": 71064 }, { "epoch": 4.29, "learning_rate": 8.90407858403875e-06, "loss": 0.1878, "step": 71066 }, { "epoch": 4.29, "learning_rate": 8.902569673924526e-06, "loss": 0.2522, "step": 71068 }, { "epoch": 4.29, "learning_rate": 8.901060763810299e-06, "loss": 0.2275, "step": 71070 }, { "epoch": 4.29, "learning_rate": 8.899551853696076e-06, "loss": 0.2953, "step": 71072 }, { "epoch": 4.29, "learning_rate": 8.89804294358185e-06, "loss": 0.2425, "step": 71074 }, { "epoch": 4.29, "learning_rate": 8.896534033467627e-06, "loss": 0.2164, "step": 71076 }, { "epoch": 4.29, "learning_rate": 8.895025123353402e-06, "loss": 0.2356, "step": 71078 }, { "epoch": 4.29, "learning_rate": 8.893516213239178e-06, "loss": 0.1258, "step": 71080 }, { "epoch": 4.29, "learning_rate": 8.892007303124953e-06, "loss": 0.2193, "step": 71082 }, { "epoch": 4.29, "learning_rate": 8.89049839301073e-06, "loss": 0.2252, "step": 71084 }, { "epoch": 4.29, "learning_rate": 8.888989482896504e-06, "loss": 0.2294, "step": 71086 }, { "epoch": 4.29, "learning_rate": 8.88748057278228e-06, "loss": 0.3913, "step": 71088 }, { "epoch": 4.29, "learning_rate": 8.885971662668056e-06, "loss": 0.3027, "step": 71090 }, { "epoch": 4.29, "learning_rate": 8.88446275255383e-06, "loss": 0.1814, "step": 71092 }, { "epoch": 4.29, "learning_rate": 8.882953842439607e-06, "loss": 0.2104, "step": 71094 }, { "epoch": 4.29, "learning_rate": 8.881444932325382e-06, "loss": 0.1866, "step": 71096 }, { "epoch": 4.29, "learning_rate": 8.879936022211158e-06, "loss": 0.2766, "step": 71098 }, { "epoch": 4.29, "learning_rate": 8.878427112096933e-06, "loss": 0.2455, "step": 71100 }, { "epoch": 4.29, "learning_rate": 8.876918201982708e-06, "loss": 0.184, "step": 71102 }, { "epoch": 4.29, "learning_rate": 8.875409291868483e-06, "loss": 0.2604, "step": 71104 }, { "epoch": 4.29, "learning_rate": 8.87390038175426e-06, "loss": 0.3065, "step": 71106 }, { "epoch": 4.29, "learning_rate": 8.872391471640034e-06, "loss": 0.1926, "step": 71108 }, { "epoch": 4.29, "learning_rate": 8.87088256152581e-06, "loss": 0.3061, "step": 71110 }, { "epoch": 4.29, "learning_rate": 8.869373651411586e-06, "loss": 0.2519, "step": 71112 }, { "epoch": 4.29, "learning_rate": 8.867864741297362e-06, "loss": 0.151, "step": 71114 }, { "epoch": 4.29, "learning_rate": 8.866355831183137e-06, "loss": 0.2457, "step": 71116 }, { "epoch": 4.29, "learning_rate": 8.864846921068912e-06, "loss": 0.279, "step": 71118 }, { "epoch": 4.29, "learning_rate": 8.863338010954687e-06, "loss": 0.2077, "step": 71120 }, { "epoch": 4.29, "learning_rate": 8.861829100840463e-06, "loss": 0.1417, "step": 71122 }, { "epoch": 4.29, "learning_rate": 8.860320190726238e-06, "loss": 0.156, "step": 71124 }, { "epoch": 4.29, "learning_rate": 8.858811280612014e-06, "loss": 0.2123, "step": 71126 }, { "epoch": 4.29, "learning_rate": 8.85730237049779e-06, "loss": 0.3076, "step": 71128 }, { "epoch": 4.29, "learning_rate": 8.855793460383566e-06, "loss": 0.4017, "step": 71130 }, { "epoch": 4.29, "learning_rate": 8.854284550269342e-06, "loss": 0.1498, "step": 71132 }, { "epoch": 4.29, "learning_rate": 8.852775640155117e-06, "loss": 0.3066, "step": 71134 }, { "epoch": 4.29, "learning_rate": 8.851266730040892e-06, "loss": 0.2783, "step": 71136 }, { "epoch": 4.29, "learning_rate": 8.849757819926667e-06, "loss": 0.1899, "step": 71138 }, { "epoch": 4.29, "learning_rate": 8.848248909812443e-06, "loss": 0.2022, "step": 71140 }, { "epoch": 4.29, "learning_rate": 8.846739999698218e-06, "loss": 0.2248, "step": 71142 }, { "epoch": 4.29, "learning_rate": 8.845231089583995e-06, "loss": 0.2205, "step": 71144 }, { "epoch": 4.29, "learning_rate": 8.84372217946977e-06, "loss": 0.2303, "step": 71146 }, { "epoch": 4.29, "learning_rate": 8.842213269355546e-06, "loss": 0.2253, "step": 71148 }, { "epoch": 4.29, "learning_rate": 8.84070435924132e-06, "loss": 0.2606, "step": 71150 }, { "epoch": 4.29, "learning_rate": 8.839195449127096e-06, "loss": 0.1574, "step": 71152 }, { "epoch": 4.29, "learning_rate": 8.83768653901287e-06, "loss": 0.2422, "step": 71154 }, { "epoch": 4.29, "learning_rate": 8.836177628898647e-06, "loss": 0.3145, "step": 71156 }, { "epoch": 4.29, "learning_rate": 8.834668718784422e-06, "loss": 0.3237, "step": 71158 }, { "epoch": 4.3, "learning_rate": 8.833159808670198e-06, "loss": 0.1888, "step": 71160 }, { "epoch": 4.3, "learning_rate": 8.831650898555973e-06, "loss": 0.2293, "step": 71162 }, { "epoch": 4.3, "learning_rate": 8.83014198844175e-06, "loss": 0.1149, "step": 71164 }, { "epoch": 4.3, "learning_rate": 8.828633078327524e-06, "loss": 0.2504, "step": 71166 }, { "epoch": 4.3, "learning_rate": 8.827124168213299e-06, "loss": 0.2635, "step": 71168 }, { "epoch": 4.3, "learning_rate": 8.825615258099076e-06, "loss": 0.2577, "step": 71170 }, { "epoch": 4.3, "learning_rate": 8.82410634798485e-06, "loss": 0.3947, "step": 71172 }, { "epoch": 4.3, "learning_rate": 8.822597437870627e-06, "loss": 0.2246, "step": 71174 }, { "epoch": 4.3, "learning_rate": 8.821088527756402e-06, "loss": 0.1307, "step": 71176 }, { "epoch": 4.3, "learning_rate": 8.819579617642178e-06, "loss": 0.1898, "step": 71178 }, { "epoch": 4.3, "learning_rate": 8.818070707527953e-06, "loss": 0.1787, "step": 71180 }, { "epoch": 4.3, "learning_rate": 8.81656179741373e-06, "loss": 0.2348, "step": 71182 }, { "epoch": 4.3, "learning_rate": 8.815052887299503e-06, "loss": 0.1386, "step": 71184 }, { "epoch": 4.3, "learning_rate": 8.81354397718528e-06, "loss": 0.228, "step": 71186 }, { "epoch": 4.3, "learning_rate": 8.812035067071054e-06, "loss": 0.1571, "step": 71188 }, { "epoch": 4.3, "learning_rate": 8.81052615695683e-06, "loss": 0.2201, "step": 71190 }, { "epoch": 4.3, "learning_rate": 8.809017246842605e-06, "loss": 0.3543, "step": 71192 }, { "epoch": 4.3, "learning_rate": 8.807508336728382e-06, "loss": 0.2943, "step": 71194 }, { "epoch": 4.3, "learning_rate": 8.805999426614157e-06, "loss": 0.1847, "step": 71196 }, { "epoch": 4.3, "learning_rate": 8.804490516499933e-06, "loss": 0.2649, "step": 71198 }, { "epoch": 4.3, "learning_rate": 8.802981606385708e-06, "loss": 0.2251, "step": 71200 }, { "epoch": 4.3, "learning_rate": 8.801472696271483e-06, "loss": 0.295, "step": 71202 }, { "epoch": 4.3, "learning_rate": 8.79996378615726e-06, "loss": 0.1738, "step": 71204 }, { "epoch": 4.3, "learning_rate": 8.798454876043034e-06, "loss": 0.3136, "step": 71206 }, { "epoch": 4.3, "learning_rate": 8.79694596592881e-06, "loss": 0.2862, "step": 71208 }, { "epoch": 4.3, "learning_rate": 8.795437055814586e-06, "loss": 0.2394, "step": 71210 }, { "epoch": 4.3, "learning_rate": 8.793928145700362e-06, "loss": 0.2626, "step": 71212 }, { "epoch": 4.3, "learning_rate": 8.792419235586137e-06, "loss": 0.3319, "step": 71214 }, { "epoch": 4.3, "learning_rate": 8.790910325471913e-06, "loss": 0.2285, "step": 71216 }, { "epoch": 4.3, "learning_rate": 8.789401415357687e-06, "loss": 0.2116, "step": 71218 }, { "epoch": 4.3, "learning_rate": 8.787892505243463e-06, "loss": 0.2711, "step": 71220 }, { "epoch": 4.3, "learning_rate": 8.786383595129238e-06, "loss": 0.3161, "step": 71222 }, { "epoch": 4.3, "learning_rate": 8.784874685015014e-06, "loss": 0.2372, "step": 71224 }, { "epoch": 4.3, "learning_rate": 8.78336577490079e-06, "loss": 0.4521, "step": 71226 }, { "epoch": 4.3, "learning_rate": 8.781856864786566e-06, "loss": 0.2803, "step": 71228 }, { "epoch": 4.3, "learning_rate": 8.78034795467234e-06, "loss": 0.2654, "step": 71230 }, { "epoch": 4.3, "learning_rate": 8.778839044558117e-06, "loss": 0.2442, "step": 71232 }, { "epoch": 4.3, "learning_rate": 8.77733013444389e-06, "loss": 0.1728, "step": 71234 }, { "epoch": 4.3, "learning_rate": 8.775821224329667e-06, "loss": 0.3112, "step": 71236 }, { "epoch": 4.3, "learning_rate": 8.774312314215442e-06, "loss": 0.2449, "step": 71238 }, { "epoch": 4.3, "learning_rate": 8.772803404101218e-06, "loss": 0.2301, "step": 71240 }, { "epoch": 4.3, "learning_rate": 8.771294493986993e-06, "loss": 0.2318, "step": 71242 }, { "epoch": 4.3, "learning_rate": 8.76978558387277e-06, "loss": 0.2048, "step": 71244 }, { "epoch": 4.3, "learning_rate": 8.768276673758544e-06, "loss": 0.2207, "step": 71246 }, { "epoch": 4.3, "learning_rate": 8.76676776364432e-06, "loss": 0.2289, "step": 71248 }, { "epoch": 4.3, "learning_rate": 8.765258853530096e-06, "loss": 0.2132, "step": 71250 }, { "epoch": 4.3, "learning_rate": 8.76374994341587e-06, "loss": 0.143, "step": 71252 }, { "epoch": 4.3, "learning_rate": 8.762241033301647e-06, "loss": 0.3529, "step": 71254 }, { "epoch": 4.3, "learning_rate": 8.760732123187422e-06, "loss": 0.2971, "step": 71256 }, { "epoch": 4.3, "learning_rate": 8.759223213073198e-06, "loss": 0.2359, "step": 71258 }, { "epoch": 4.3, "learning_rate": 8.757714302958973e-06, "loss": 0.2397, "step": 71260 }, { "epoch": 4.3, "learning_rate": 8.75620539284475e-06, "loss": 0.1987, "step": 71262 }, { "epoch": 4.3, "learning_rate": 8.754696482730524e-06, "loss": 0.2071, "step": 71264 }, { "epoch": 4.3, "learning_rate": 8.7531875726163e-06, "loss": 0.2278, "step": 71266 }, { "epoch": 4.3, "learning_rate": 8.751678662502074e-06, "loss": 0.2525, "step": 71268 }, { "epoch": 4.3, "learning_rate": 8.75016975238785e-06, "loss": 0.164, "step": 71270 }, { "epoch": 4.3, "learning_rate": 8.748660842273625e-06, "loss": 0.1542, "step": 71272 }, { "epoch": 4.3, "learning_rate": 8.747151932159402e-06, "loss": 0.1637, "step": 71274 }, { "epoch": 4.3, "learning_rate": 8.745643022045177e-06, "loss": 0.2538, "step": 71276 }, { "epoch": 4.3, "learning_rate": 8.744134111930953e-06, "loss": 0.1636, "step": 71278 }, { "epoch": 4.3, "learning_rate": 8.742625201816728e-06, "loss": 0.2929, "step": 71280 }, { "epoch": 4.3, "learning_rate": 8.741116291702503e-06, "loss": 0.1823, "step": 71282 }, { "epoch": 4.3, "learning_rate": 8.73960738158828e-06, "loss": 0.1517, "step": 71284 }, { "epoch": 4.3, "learning_rate": 8.738098471474054e-06, "loss": 0.2394, "step": 71286 }, { "epoch": 4.3, "learning_rate": 8.73658956135983e-06, "loss": 0.2428, "step": 71288 }, { "epoch": 4.3, "learning_rate": 8.735080651245605e-06, "loss": 0.2529, "step": 71290 }, { "epoch": 4.3, "learning_rate": 8.733571741131382e-06, "loss": 0.1687, "step": 71292 }, { "epoch": 4.3, "learning_rate": 8.732062831017157e-06, "loss": 0.1891, "step": 71294 }, { "epoch": 4.3, "learning_rate": 8.730553920902933e-06, "loss": 0.2085, "step": 71296 }, { "epoch": 4.3, "learning_rate": 8.729045010788708e-06, "loss": 0.2727, "step": 71298 }, { "epoch": 4.3, "learning_rate": 8.727536100674483e-06, "loss": 0.234, "step": 71300 }, { "epoch": 4.3, "learning_rate": 8.726027190560258e-06, "loss": 0.1945, "step": 71302 }, { "epoch": 4.3, "learning_rate": 8.724518280446034e-06, "loss": 0.1887, "step": 71304 }, { "epoch": 4.3, "learning_rate": 8.723009370331809e-06, "loss": 0.2146, "step": 71306 }, { "epoch": 4.3, "learning_rate": 8.721500460217586e-06, "loss": 0.234, "step": 71308 }, { "epoch": 4.3, "learning_rate": 8.71999155010336e-06, "loss": 0.2199, "step": 71310 }, { "epoch": 4.3, "learning_rate": 8.718482639989137e-06, "loss": 0.2134, "step": 71312 }, { "epoch": 4.3, "learning_rate": 8.716973729874912e-06, "loss": 0.3072, "step": 71314 }, { "epoch": 4.3, "learning_rate": 8.715464819760687e-06, "loss": 0.1791, "step": 71316 }, { "epoch": 4.3, "learning_rate": 8.713955909646463e-06, "loss": 0.2011, "step": 71318 }, { "epoch": 4.3, "learning_rate": 8.712446999532238e-06, "loss": 0.3317, "step": 71320 }, { "epoch": 4.3, "learning_rate": 8.710938089418014e-06, "loss": 0.1797, "step": 71322 }, { "epoch": 4.3, "learning_rate": 8.70942917930379e-06, "loss": 0.1759, "step": 71324 }, { "epoch": 4.31, "learning_rate": 8.707920269189566e-06, "loss": 0.2066, "step": 71326 }, { "epoch": 4.31, "learning_rate": 8.70641135907534e-06, "loss": 0.323, "step": 71328 }, { "epoch": 4.31, "learning_rate": 8.704902448961117e-06, "loss": 0.2486, "step": 71330 }, { "epoch": 4.31, "learning_rate": 8.70339353884689e-06, "loss": 0.1496, "step": 71332 }, { "epoch": 4.31, "learning_rate": 8.701884628732667e-06, "loss": 0.2239, "step": 71334 }, { "epoch": 4.31, "learning_rate": 8.700375718618442e-06, "loss": 0.2816, "step": 71336 }, { "epoch": 4.31, "learning_rate": 8.698866808504218e-06, "loss": 0.1891, "step": 71338 }, { "epoch": 4.31, "learning_rate": 8.697357898389993e-06, "loss": 0.1786, "step": 71340 }, { "epoch": 4.31, "learning_rate": 8.69584898827577e-06, "loss": 0.2895, "step": 71342 }, { "epoch": 4.31, "learning_rate": 8.694340078161544e-06, "loss": 0.289, "step": 71344 }, { "epoch": 4.31, "learning_rate": 8.69283116804732e-06, "loss": 0.2346, "step": 71346 }, { "epoch": 4.31, "learning_rate": 8.691322257933094e-06, "loss": 0.2231, "step": 71348 }, { "epoch": 4.31, "learning_rate": 8.68981334781887e-06, "loss": 0.3893, "step": 71350 }, { "epoch": 4.31, "learning_rate": 8.688304437704645e-06, "loss": 0.2759, "step": 71352 }, { "epoch": 4.31, "learning_rate": 8.686795527590422e-06, "loss": 0.1197, "step": 71354 }, { "epoch": 4.31, "learning_rate": 8.685286617476197e-06, "loss": 0.1451, "step": 71356 }, { "epoch": 4.31, "learning_rate": 8.683777707361973e-06, "loss": 0.2876, "step": 71358 }, { "epoch": 4.31, "learning_rate": 8.682268797247748e-06, "loss": 0.239, "step": 71360 }, { "epoch": 4.31, "learning_rate": 8.680759887133524e-06, "loss": 0.3375, "step": 71362 }, { "epoch": 4.31, "learning_rate": 8.6792509770193e-06, "loss": 0.2067, "step": 71364 }, { "epoch": 4.31, "learning_rate": 8.677742066905074e-06, "loss": 0.1904, "step": 71366 }, { "epoch": 4.31, "learning_rate": 8.67623315679085e-06, "loss": 0.2094, "step": 71368 }, { "epoch": 4.31, "learning_rate": 8.674724246676625e-06, "loss": 0.264, "step": 71370 }, { "epoch": 4.31, "learning_rate": 8.673215336562402e-06, "loss": 0.2091, "step": 71372 }, { "epoch": 4.31, "learning_rate": 8.671706426448177e-06, "loss": 0.288, "step": 71374 }, { "epoch": 4.31, "learning_rate": 8.670197516333953e-06, "loss": 0.2367, "step": 71376 }, { "epoch": 4.31, "learning_rate": 8.668688606219728e-06, "loss": 0.2271, "step": 71378 }, { "epoch": 4.31, "learning_rate": 8.667179696105505e-06, "loss": 0.2473, "step": 71380 }, { "epoch": 4.31, "learning_rate": 8.665670785991278e-06, "loss": 0.2379, "step": 71382 }, { "epoch": 4.31, "learning_rate": 8.664161875877054e-06, "loss": 0.1474, "step": 71384 }, { "epoch": 4.31, "learning_rate": 8.662652965762829e-06, "loss": 0.1954, "step": 71386 }, { "epoch": 4.31, "learning_rate": 8.661144055648606e-06, "loss": 0.3072, "step": 71388 }, { "epoch": 4.31, "learning_rate": 8.65963514553438e-06, "loss": 0.2849, "step": 71390 }, { "epoch": 4.31, "learning_rate": 8.658126235420157e-06, "loss": 0.2176, "step": 71392 }, { "epoch": 4.31, "learning_rate": 8.656617325305932e-06, "loss": 0.2941, "step": 71394 }, { "epoch": 4.31, "learning_rate": 8.655108415191708e-06, "loss": 0.2678, "step": 71396 }, { "epoch": 4.31, "learning_rate": 8.653599505077483e-06, "loss": 0.1678, "step": 71398 }, { "epoch": 4.31, "learning_rate": 8.652090594963258e-06, "loss": 0.1636, "step": 71400 }, { "epoch": 4.31, "learning_rate": 8.650581684849034e-06, "loss": 0.2206, "step": 71402 }, { "epoch": 4.31, "learning_rate": 8.64907277473481e-06, "loss": 0.3391, "step": 71404 }, { "epoch": 4.31, "learning_rate": 8.647563864620586e-06, "loss": 0.2732, "step": 71406 }, { "epoch": 4.31, "learning_rate": 8.64605495450636e-06, "loss": 0.2224, "step": 71408 }, { "epoch": 4.31, "learning_rate": 8.644546044392137e-06, "loss": 0.316, "step": 71410 }, { "epoch": 4.31, "learning_rate": 8.643037134277912e-06, "loss": 0.2501, "step": 71412 }, { "epoch": 4.31, "learning_rate": 8.641528224163687e-06, "loss": 0.3026, "step": 71414 }, { "epoch": 4.31, "learning_rate": 8.640019314049461e-06, "loss": 0.2192, "step": 71416 }, { "epoch": 4.31, "learning_rate": 8.638510403935238e-06, "loss": 0.2572, "step": 71418 }, { "epoch": 4.31, "learning_rate": 8.637001493821013e-06, "loss": 0.3964, "step": 71420 }, { "epoch": 4.31, "learning_rate": 8.63549258370679e-06, "loss": 0.3292, "step": 71422 }, { "epoch": 4.31, "learning_rate": 8.633983673592564e-06, "loss": 0.2653, "step": 71424 }, { "epoch": 4.31, "learning_rate": 8.63247476347834e-06, "loss": 0.2432, "step": 71426 }, { "epoch": 4.31, "learning_rate": 8.630965853364115e-06, "loss": 0.2203, "step": 71428 }, { "epoch": 4.31, "learning_rate": 8.62945694324989e-06, "loss": 0.2056, "step": 71430 }, { "epoch": 4.31, "learning_rate": 8.627948033135667e-06, "loss": 0.2928, "step": 71432 }, { "epoch": 4.31, "learning_rate": 8.626439123021442e-06, "loss": 0.2012, "step": 71434 }, { "epoch": 4.31, "learning_rate": 8.624930212907218e-06, "loss": 0.2231, "step": 71436 }, { "epoch": 4.31, "learning_rate": 8.623421302792993e-06, "loss": 0.1942, "step": 71438 }, { "epoch": 4.31, "learning_rate": 8.62191239267877e-06, "loss": 0.2701, "step": 71440 }, { "epoch": 4.31, "learning_rate": 8.620403482564544e-06, "loss": 0.182, "step": 71442 }, { "epoch": 4.31, "learning_rate": 8.61889457245032e-06, "loss": 0.2536, "step": 71444 }, { "epoch": 4.31, "learning_rate": 8.617385662336096e-06, "loss": 0.1809, "step": 71446 }, { "epoch": 4.31, "learning_rate": 8.61587675222187e-06, "loss": 0.1774, "step": 71448 }, { "epoch": 4.31, "learning_rate": 8.614367842107645e-06, "loss": 0.2606, "step": 71450 }, { "epoch": 4.31, "learning_rate": 8.612858931993422e-06, "loss": 0.2228, "step": 71452 }, { "epoch": 4.31, "learning_rate": 8.611350021879197e-06, "loss": 0.2611, "step": 71454 }, { "epoch": 4.31, "learning_rate": 8.609841111764973e-06, "loss": 0.2069, "step": 71456 }, { "epoch": 4.31, "learning_rate": 8.608332201650748e-06, "loss": 0.144, "step": 71458 }, { "epoch": 4.31, "learning_rate": 8.606823291536524e-06, "loss": 0.2325, "step": 71460 }, { "epoch": 4.31, "learning_rate": 8.6053143814223e-06, "loss": 0.2584, "step": 71462 }, { "epoch": 4.31, "learning_rate": 8.603805471308074e-06, "loss": 0.1612, "step": 71464 }, { "epoch": 4.31, "learning_rate": 8.602296561193849e-06, "loss": 0.152, "step": 71466 }, { "epoch": 4.31, "learning_rate": 8.600787651079625e-06, "loss": 0.4688, "step": 71468 }, { "epoch": 4.31, "learning_rate": 8.5992787409654e-06, "loss": 0.2009, "step": 71470 }, { "epoch": 4.31, "learning_rate": 8.597769830851177e-06, "loss": 0.2559, "step": 71472 }, { "epoch": 4.31, "learning_rate": 8.596260920736952e-06, "loss": 0.2273, "step": 71474 }, { "epoch": 4.31, "learning_rate": 8.594752010622728e-06, "loss": 0.3083, "step": 71476 }, { "epoch": 4.31, "learning_rate": 8.593243100508503e-06, "loss": 0.2056, "step": 71478 }, { "epoch": 4.31, "learning_rate": 8.591734190394278e-06, "loss": 0.2607, "step": 71480 }, { "epoch": 4.31, "learning_rate": 8.590225280280054e-06, "loss": 0.1321, "step": 71482 }, { "epoch": 4.31, "learning_rate": 8.588716370165829e-06, "loss": 0.2697, "step": 71484 }, { "epoch": 4.31, "learning_rate": 8.587207460051606e-06, "loss": 0.2335, "step": 71486 }, { "epoch": 4.31, "learning_rate": 8.58569854993738e-06, "loss": 0.3279, "step": 71488 }, { "epoch": 4.31, "learning_rate": 8.584189639823157e-06, "loss": 0.1907, "step": 71490 }, { "epoch": 4.32, "learning_rate": 8.582680729708932e-06, "loss": 0.1913, "step": 71492 }, { "epoch": 4.32, "learning_rate": 8.581171819594708e-06, "loss": 0.2365, "step": 71494 }, { "epoch": 4.32, "learning_rate": 8.579662909480481e-06, "loss": 0.242, "step": 71496 }, { "epoch": 4.32, "learning_rate": 8.578153999366258e-06, "loss": 0.2104, "step": 71498 }, { "epoch": 4.32, "learning_rate": 8.576645089252033e-06, "loss": 0.1811, "step": 71500 }, { "epoch": 4.32, "learning_rate": 8.57513617913781e-06, "loss": 0.2983, "step": 71502 }, { "epoch": 4.32, "learning_rate": 8.573627269023584e-06, "loss": 0.2459, "step": 71504 }, { "epoch": 4.32, "learning_rate": 8.57211835890936e-06, "loss": 0.2282, "step": 71506 }, { "epoch": 4.32, "learning_rate": 8.570609448795135e-06, "loss": 0.2839, "step": 71508 }, { "epoch": 4.32, "learning_rate": 8.569100538680912e-06, "loss": 0.2038, "step": 71510 }, { "epoch": 4.32, "learning_rate": 8.567591628566687e-06, "loss": 0.2353, "step": 71512 }, { "epoch": 4.32, "learning_rate": 8.566082718452462e-06, "loss": 0.2641, "step": 71514 }, { "epoch": 4.32, "learning_rate": 8.564573808338238e-06, "loss": 0.1298, "step": 71516 }, { "epoch": 4.32, "learning_rate": 8.563064898224013e-06, "loss": 0.307, "step": 71518 }, { "epoch": 4.32, "learning_rate": 8.56155598810979e-06, "loss": 0.155, "step": 71520 }, { "epoch": 4.32, "learning_rate": 8.560047077995564e-06, "loss": 0.4086, "step": 71522 }, { "epoch": 4.32, "learning_rate": 8.55853816788134e-06, "loss": 0.2477, "step": 71524 }, { "epoch": 4.32, "learning_rate": 8.557029257767116e-06, "loss": 0.2573, "step": 71526 }, { "epoch": 4.32, "learning_rate": 8.555520347652892e-06, "loss": 0.1888, "step": 71528 }, { "epoch": 4.32, "learning_rate": 8.554011437538665e-06, "loss": 0.3106, "step": 71530 }, { "epoch": 4.32, "learning_rate": 8.552502527424442e-06, "loss": 0.3122, "step": 71532 }, { "epoch": 4.32, "learning_rate": 8.550993617310216e-06, "loss": 0.1411, "step": 71534 }, { "epoch": 4.32, "learning_rate": 8.549484707195993e-06, "loss": 0.2069, "step": 71536 }, { "epoch": 4.32, "learning_rate": 8.547975797081768e-06, "loss": 0.3014, "step": 71538 }, { "epoch": 4.32, "learning_rate": 8.546466886967544e-06, "loss": 0.3132, "step": 71540 }, { "epoch": 4.32, "learning_rate": 8.544957976853319e-06, "loss": 0.184, "step": 71542 }, { "epoch": 4.32, "learning_rate": 8.543449066739096e-06, "loss": 0.2553, "step": 71544 }, { "epoch": 4.32, "learning_rate": 8.54194015662487e-06, "loss": 0.186, "step": 71546 }, { "epoch": 4.32, "learning_rate": 8.540431246510645e-06, "loss": 0.325, "step": 71548 }, { "epoch": 4.32, "learning_rate": 8.538922336396422e-06, "loss": 0.2449, "step": 71550 }, { "epoch": 4.32, "learning_rate": 8.537413426282197e-06, "loss": 0.1625, "step": 71552 }, { "epoch": 4.32, "learning_rate": 8.535904516167973e-06, "loss": 0.1999, "step": 71554 }, { "epoch": 4.32, "learning_rate": 8.534395606053748e-06, "loss": 0.29, "step": 71556 }, { "epoch": 4.32, "learning_rate": 8.532886695939524e-06, "loss": 0.2795, "step": 71558 }, { "epoch": 4.32, "learning_rate": 8.5313777858253e-06, "loss": 0.313, "step": 71560 }, { "epoch": 4.32, "learning_rate": 8.529868875711074e-06, "loss": 0.1855, "step": 71562 }, { "epoch": 4.32, "learning_rate": 8.528359965596849e-06, "loss": 0.2003, "step": 71564 }, { "epoch": 4.32, "learning_rate": 8.526851055482625e-06, "loss": 0.2393, "step": 71566 }, { "epoch": 4.32, "learning_rate": 8.5253421453684e-06, "loss": 0.1855, "step": 71568 }, { "epoch": 4.32, "learning_rate": 8.523833235254177e-06, "loss": 0.2346, "step": 71570 }, { "epoch": 4.32, "learning_rate": 8.522324325139952e-06, "loss": 0.2907, "step": 71572 }, { "epoch": 4.32, "learning_rate": 8.520815415025728e-06, "loss": 0.2744, "step": 71574 }, { "epoch": 4.32, "learning_rate": 8.519306504911503e-06, "loss": 0.1404, "step": 71576 }, { "epoch": 4.32, "learning_rate": 8.517797594797278e-06, "loss": 0.3177, "step": 71578 }, { "epoch": 4.32, "learning_rate": 8.516288684683053e-06, "loss": 0.2355, "step": 71580 }, { "epoch": 4.32, "learning_rate": 8.514779774568829e-06, "loss": 0.2074, "step": 71582 }, { "epoch": 4.32, "learning_rate": 8.513270864454604e-06, "loss": 0.2237, "step": 71584 }, { "epoch": 4.32, "learning_rate": 8.51176195434038e-06, "loss": 0.1644, "step": 71586 }, { "epoch": 4.32, "learning_rate": 8.510253044226155e-06, "loss": 0.1704, "step": 71588 }, { "epoch": 4.32, "learning_rate": 8.508744134111932e-06, "loss": 0.1998, "step": 71590 }, { "epoch": 4.32, "learning_rate": 8.507235223997707e-06, "loss": 0.2258, "step": 71592 }, { "epoch": 4.32, "learning_rate": 8.505726313883481e-06, "loss": 0.2109, "step": 71594 }, { "epoch": 4.32, "learning_rate": 8.504217403769258e-06, "loss": 0.2545, "step": 71596 }, { "epoch": 4.32, "learning_rate": 8.502708493655033e-06, "loss": 0.1917, "step": 71598 }, { "epoch": 4.32, "learning_rate": 8.50119958354081e-06, "loss": 0.1753, "step": 71600 }, { "epoch": 4.32, "learning_rate": 8.499690673426584e-06, "loss": 0.2118, "step": 71602 }, { "epoch": 4.32, "learning_rate": 8.49818176331236e-06, "loss": 0.2349, "step": 71604 }, { "epoch": 4.32, "learning_rate": 8.496672853198135e-06, "loss": 0.2936, "step": 71606 }, { "epoch": 4.32, "learning_rate": 8.495163943083912e-06, "loss": 0.1917, "step": 71608 }, { "epoch": 4.32, "learning_rate": 8.493655032969687e-06, "loss": 0.2371, "step": 71610 }, { "epoch": 4.32, "learning_rate": 8.492146122855462e-06, "loss": 0.1896, "step": 71612 }, { "epoch": 4.32, "learning_rate": 8.490637212741236e-06, "loss": 0.1246, "step": 71614 }, { "epoch": 4.32, "learning_rate": 8.489128302627013e-06, "loss": 0.2441, "step": 71616 }, { "epoch": 4.32, "learning_rate": 8.487619392512788e-06, "loss": 0.2359, "step": 71618 }, { "epoch": 4.32, "learning_rate": 8.486110482398564e-06, "loss": 0.3031, "step": 71620 }, { "epoch": 4.32, "learning_rate": 8.484601572284339e-06, "loss": 0.2984, "step": 71622 }, { "epoch": 4.32, "learning_rate": 8.483092662170116e-06, "loss": 0.1992, "step": 71624 }, { "epoch": 4.32, "learning_rate": 8.48158375205589e-06, "loss": 0.2112, "step": 71626 }, { "epoch": 4.32, "learning_rate": 8.480074841941665e-06, "loss": 0.2891, "step": 71628 }, { "epoch": 4.32, "learning_rate": 8.478565931827442e-06, "loss": 0.2206, "step": 71630 }, { "epoch": 4.32, "learning_rate": 8.477057021713217e-06, "loss": 0.197, "step": 71632 }, { "epoch": 4.32, "learning_rate": 8.475548111598993e-06, "loss": 0.1647, "step": 71634 }, { "epoch": 4.32, "learning_rate": 8.474039201484768e-06, "loss": 0.1667, "step": 71636 }, { "epoch": 4.32, "learning_rate": 8.472530291370544e-06, "loss": 0.1469, "step": 71638 }, { "epoch": 4.32, "learning_rate": 8.47102138125632e-06, "loss": 0.2437, "step": 71640 }, { "epoch": 4.32, "learning_rate": 8.469512471142096e-06, "loss": 0.2367, "step": 71642 }, { "epoch": 4.32, "learning_rate": 8.468003561027869e-06, "loss": 0.2606, "step": 71644 }, { "epoch": 4.32, "learning_rate": 8.466494650913645e-06, "loss": 0.2142, "step": 71646 }, { "epoch": 4.32, "learning_rate": 8.46498574079942e-06, "loss": 0.1659, "step": 71648 }, { "epoch": 4.32, "learning_rate": 8.463476830685197e-06, "loss": 0.2246, "step": 71650 }, { "epoch": 4.32, "learning_rate": 8.461967920570971e-06, "loss": 0.1985, "step": 71652 }, { "epoch": 4.32, "learning_rate": 8.460459010456748e-06, "loss": 0.2339, "step": 71654 }, { "epoch": 4.32, "learning_rate": 8.458950100342523e-06, "loss": 0.182, "step": 71656 }, { "epoch": 4.33, "learning_rate": 8.4574411902283e-06, "loss": 0.2237, "step": 71658 }, { "epoch": 4.33, "learning_rate": 8.455932280114074e-06, "loss": 0.15, "step": 71660 }, { "epoch": 4.33, "learning_rate": 8.454423369999849e-06, "loss": 0.2728, "step": 71662 }, { "epoch": 4.33, "learning_rate": 8.452914459885625e-06, "loss": 0.3013, "step": 71664 }, { "epoch": 4.33, "learning_rate": 8.4514055497714e-06, "loss": 0.2891, "step": 71666 }, { "epoch": 4.33, "learning_rate": 8.449896639657177e-06, "loss": 0.195, "step": 71668 }, { "epoch": 4.33, "learning_rate": 8.448387729542952e-06, "loss": 0.1652, "step": 71670 }, { "epoch": 4.33, "learning_rate": 8.446878819428728e-06, "loss": 0.3873, "step": 71672 }, { "epoch": 4.33, "learning_rate": 8.445369909314503e-06, "loss": 0.1426, "step": 71674 }, { "epoch": 4.33, "learning_rate": 8.44386099920028e-06, "loss": 0.2562, "step": 71676 }, { "epoch": 4.33, "learning_rate": 8.442352089086053e-06, "loss": 0.3144, "step": 71678 }, { "epoch": 4.33, "learning_rate": 8.440843178971829e-06, "loss": 0.1791, "step": 71680 }, { "epoch": 4.33, "learning_rate": 8.439334268857604e-06, "loss": 0.1792, "step": 71682 }, { "epoch": 4.33, "learning_rate": 8.43782535874338e-06, "loss": 0.1921, "step": 71684 }, { "epoch": 4.33, "learning_rate": 8.436316448629155e-06, "loss": 0.3831, "step": 71686 }, { "epoch": 4.33, "learning_rate": 8.434807538514932e-06, "loss": 0.2139, "step": 71688 }, { "epoch": 4.33, "learning_rate": 8.433298628400707e-06, "loss": 0.1963, "step": 71690 }, { "epoch": 4.33, "learning_rate": 8.431789718286483e-06, "loss": 0.2602, "step": 71692 }, { "epoch": 4.33, "learning_rate": 8.430280808172256e-06, "loss": 0.2077, "step": 71694 }, { "epoch": 4.33, "learning_rate": 8.428771898058033e-06, "loss": 0.3588, "step": 71696 }, { "epoch": 4.33, "learning_rate": 8.427262987943808e-06, "loss": 0.2871, "step": 71698 }, { "epoch": 4.33, "learning_rate": 8.425754077829584e-06, "loss": 0.2677, "step": 71700 }, { "epoch": 4.33, "learning_rate": 8.424245167715359e-06, "loss": 0.1774, "step": 71702 }, { "epoch": 4.33, "learning_rate": 8.422736257601135e-06, "loss": 0.3212, "step": 71704 }, { "epoch": 4.33, "learning_rate": 8.42122734748691e-06, "loss": 0.2016, "step": 71706 }, { "epoch": 4.33, "learning_rate": 8.419718437372687e-06, "loss": 0.2102, "step": 71708 }, { "epoch": 4.33, "learning_rate": 8.418209527258462e-06, "loss": 0.1811, "step": 71710 }, { "epoch": 4.33, "learning_rate": 8.416700617144236e-06, "loss": 0.1813, "step": 71712 }, { "epoch": 4.33, "learning_rate": 8.415191707030013e-06, "loss": 0.1501, "step": 71714 }, { "epoch": 4.33, "learning_rate": 8.413682796915788e-06, "loss": 0.1866, "step": 71716 }, { "epoch": 4.33, "learning_rate": 8.412173886801564e-06, "loss": 0.2769, "step": 71718 }, { "epoch": 4.33, "learning_rate": 8.410664976687339e-06, "loss": 0.1848, "step": 71720 }, { "epoch": 4.33, "learning_rate": 8.409156066573116e-06, "loss": 0.2659, "step": 71722 }, { "epoch": 4.33, "learning_rate": 8.40764715645889e-06, "loss": 0.1866, "step": 71724 }, { "epoch": 4.33, "learning_rate": 8.406138246344665e-06, "loss": 0.1435, "step": 71726 }, { "epoch": 4.33, "learning_rate": 8.40462933623044e-06, "loss": 0.2102, "step": 71728 }, { "epoch": 4.33, "learning_rate": 8.403120426116217e-06, "loss": 0.1366, "step": 71730 }, { "epoch": 4.33, "learning_rate": 8.401611516001991e-06, "loss": 0.2257, "step": 71732 }, { "epoch": 4.33, "learning_rate": 8.400102605887768e-06, "loss": 0.2572, "step": 71734 }, { "epoch": 4.33, "learning_rate": 8.398593695773543e-06, "loss": 0.1732, "step": 71736 }, { "epoch": 4.33, "learning_rate": 8.39708478565932e-06, "loss": 0.2432, "step": 71738 }, { "epoch": 4.33, "learning_rate": 8.395575875545094e-06, "loss": 0.3295, "step": 71740 }, { "epoch": 4.33, "learning_rate": 8.394066965430869e-06, "loss": 0.1951, "step": 71742 }, { "epoch": 4.33, "learning_rate": 8.392558055316645e-06, "loss": 0.188, "step": 71744 }, { "epoch": 4.33, "learning_rate": 8.39104914520242e-06, "loss": 0.4029, "step": 71746 }, { "epoch": 4.33, "learning_rate": 8.389540235088197e-06, "loss": 0.2031, "step": 71748 }, { "epoch": 4.33, "learning_rate": 8.388031324973972e-06, "loss": 0.2133, "step": 71750 }, { "epoch": 4.33, "learning_rate": 8.386522414859748e-06, "loss": 0.2198, "step": 71752 }, { "epoch": 4.33, "learning_rate": 8.385013504745523e-06, "loss": 0.3006, "step": 71754 }, { "epoch": 4.33, "learning_rate": 8.3835045946313e-06, "loss": 0.304, "step": 71756 }, { "epoch": 4.33, "learning_rate": 8.381995684517074e-06, "loss": 0.1939, "step": 71758 }, { "epoch": 4.33, "learning_rate": 8.380486774402849e-06, "loss": 0.2375, "step": 71760 }, { "epoch": 4.33, "learning_rate": 8.378977864288624e-06, "loss": 0.2017, "step": 71762 }, { "epoch": 4.33, "learning_rate": 8.3774689541744e-06, "loss": 0.1577, "step": 71764 }, { "epoch": 4.33, "learning_rate": 8.375960044060175e-06, "loss": 0.1589, "step": 71766 }, { "epoch": 4.33, "learning_rate": 8.374451133945952e-06, "loss": 0.1855, "step": 71768 }, { "epoch": 4.33, "learning_rate": 8.372942223831727e-06, "loss": 0.3165, "step": 71770 }, { "epoch": 4.33, "learning_rate": 8.371433313717503e-06, "loss": 0.2223, "step": 71772 }, { "epoch": 4.33, "learning_rate": 8.369924403603278e-06, "loss": 0.4287, "step": 71774 }, { "epoch": 4.33, "learning_rate": 8.368415493489053e-06, "loss": 0.175, "step": 71776 }, { "epoch": 4.33, "learning_rate": 8.36690658337483e-06, "loss": 0.2809, "step": 71778 }, { "epoch": 4.33, "learning_rate": 8.365397673260604e-06, "loss": 0.1864, "step": 71780 }, { "epoch": 4.33, "learning_rate": 8.36388876314638e-06, "loss": 0.2236, "step": 71782 }, { "epoch": 4.33, "learning_rate": 8.362379853032155e-06, "loss": 0.1377, "step": 71784 }, { "epoch": 4.33, "learning_rate": 8.360870942917932e-06, "loss": 0.2421, "step": 71786 }, { "epoch": 4.33, "learning_rate": 8.359362032803707e-06, "loss": 0.2852, "step": 71788 }, { "epoch": 4.33, "learning_rate": 8.357853122689483e-06, "loss": 0.2408, "step": 71790 }, { "epoch": 4.33, "learning_rate": 8.356344212575256e-06, "loss": 0.2173, "step": 71792 }, { "epoch": 4.33, "learning_rate": 8.354835302461033e-06, "loss": 0.3513, "step": 71794 }, { "epoch": 4.33, "learning_rate": 8.353326392346808e-06, "loss": 0.2741, "step": 71796 }, { "epoch": 4.33, "learning_rate": 8.351817482232584e-06, "loss": 0.2888, "step": 71798 }, { "epoch": 4.33, "learning_rate": 8.350308572118359e-06, "loss": 0.1852, "step": 71800 }, { "epoch": 4.33, "learning_rate": 8.348799662004135e-06, "loss": 0.252, "step": 71802 }, { "epoch": 4.33, "learning_rate": 8.34729075188991e-06, "loss": 0.2364, "step": 71804 }, { "epoch": 4.33, "learning_rate": 8.345781841775687e-06, "loss": 0.314, "step": 71806 }, { "epoch": 4.33, "learning_rate": 8.34427293166146e-06, "loss": 0.188, "step": 71808 }, { "epoch": 4.33, "learning_rate": 8.342764021547236e-06, "loss": 0.2943, "step": 71810 }, { "epoch": 4.33, "learning_rate": 8.341255111433011e-06, "loss": 0.2492, "step": 71812 }, { "epoch": 4.33, "learning_rate": 8.339746201318788e-06, "loss": 0.3498, "step": 71814 }, { "epoch": 4.33, "learning_rate": 8.338237291204563e-06, "loss": 0.2704, "step": 71816 }, { "epoch": 4.33, "learning_rate": 8.336728381090339e-06, "loss": 0.2066, "step": 71818 }, { "epoch": 4.33, "learning_rate": 8.335219470976114e-06, "loss": 0.3464, "step": 71820 }, { "epoch": 4.33, "learning_rate": 8.33371056086189e-06, "loss": 0.2655, "step": 71822 }, { "epoch": 4.34, "learning_rate": 8.332201650747665e-06, "loss": 0.2223, "step": 71824 }, { "epoch": 4.34, "learning_rate": 8.33069274063344e-06, "loss": 0.253, "step": 71826 }, { "epoch": 4.34, "learning_rate": 8.329183830519217e-06, "loss": 0.2322, "step": 71828 }, { "epoch": 4.34, "learning_rate": 8.327674920404991e-06, "loss": 0.3723, "step": 71830 }, { "epoch": 4.34, "learning_rate": 8.326166010290768e-06, "loss": 0.1634, "step": 71832 }, { "epoch": 4.34, "learning_rate": 8.324657100176543e-06, "loss": 0.2326, "step": 71834 }, { "epoch": 4.34, "learning_rate": 8.32314819006232e-06, "loss": 0.268, "step": 71836 }, { "epoch": 4.34, "learning_rate": 8.321639279948094e-06, "loss": 0.2352, "step": 71838 }, { "epoch": 4.34, "learning_rate": 8.32013036983387e-06, "loss": 0.249, "step": 71840 }, { "epoch": 4.34, "learning_rate": 8.318621459719644e-06, "loss": 0.2092, "step": 71842 }, { "epoch": 4.34, "learning_rate": 8.31711254960542e-06, "loss": 0.2243, "step": 71844 }, { "epoch": 4.34, "learning_rate": 8.315603639491195e-06, "loss": 0.3341, "step": 71846 }, { "epoch": 4.34, "learning_rate": 8.314094729376972e-06, "loss": 0.2293, "step": 71848 }, { "epoch": 4.34, "learning_rate": 8.312585819262746e-06, "loss": 0.2551, "step": 71850 }, { "epoch": 4.34, "learning_rate": 8.311076909148523e-06, "loss": 0.2408, "step": 71852 }, { "epoch": 4.34, "learning_rate": 8.309567999034298e-06, "loss": 0.3249, "step": 71854 }, { "epoch": 4.34, "learning_rate": 8.308059088920074e-06, "loss": 0.1112, "step": 71856 }, { "epoch": 4.34, "learning_rate": 8.306550178805849e-06, "loss": 0.4221, "step": 71858 }, { "epoch": 4.34, "learning_rate": 8.305041268691624e-06, "loss": 0.3128, "step": 71860 }, { "epoch": 4.34, "learning_rate": 8.3035323585774e-06, "loss": 0.2218, "step": 71862 }, { "epoch": 4.34, "learning_rate": 8.302023448463175e-06, "loss": 0.2978, "step": 71864 }, { "epoch": 4.34, "learning_rate": 8.300514538348952e-06, "loss": 0.1616, "step": 71866 }, { "epoch": 4.34, "learning_rate": 8.299005628234727e-06, "loss": 0.2152, "step": 71868 }, { "epoch": 4.34, "learning_rate": 8.297496718120503e-06, "loss": 0.1083, "step": 71870 }, { "epoch": 4.34, "learning_rate": 8.295987808006278e-06, "loss": 0.2123, "step": 71872 }, { "epoch": 4.34, "learning_rate": 8.294478897892053e-06, "loss": 0.26, "step": 71874 }, { "epoch": 4.34, "learning_rate": 8.292969987777828e-06, "loss": 0.3055, "step": 71876 }, { "epoch": 4.34, "learning_rate": 8.291461077663604e-06, "loss": 0.2269, "step": 71878 }, { "epoch": 4.34, "learning_rate": 8.289952167549379e-06, "loss": 0.1863, "step": 71880 }, { "epoch": 4.34, "learning_rate": 8.288443257435155e-06, "loss": 0.2137, "step": 71882 }, { "epoch": 4.34, "learning_rate": 8.28693434732093e-06, "loss": 0.2344, "step": 71884 }, { "epoch": 4.34, "learning_rate": 8.285425437206707e-06, "loss": 0.2144, "step": 71886 }, { "epoch": 4.34, "learning_rate": 8.283916527092482e-06, "loss": 0.3117, "step": 71888 }, { "epoch": 4.34, "learning_rate": 8.282407616978256e-06, "loss": 0.2518, "step": 71890 }, { "epoch": 4.34, "learning_rate": 8.280898706864033e-06, "loss": 0.2051, "step": 71892 }, { "epoch": 4.34, "learning_rate": 8.279389796749808e-06, "loss": 0.2183, "step": 71894 }, { "epoch": 4.34, "learning_rate": 8.277880886635584e-06, "loss": 0.2755, "step": 71896 }, { "epoch": 4.34, "learning_rate": 8.276371976521359e-06, "loss": 0.1916, "step": 71898 }, { "epoch": 4.34, "learning_rate": 8.274863066407136e-06, "loss": 0.2606, "step": 71900 }, { "epoch": 4.34, "learning_rate": 8.27335415629291e-06, "loss": 0.3736, "step": 71902 }, { "epoch": 4.34, "learning_rate": 8.271845246178687e-06, "loss": 0.1538, "step": 71904 }, { "epoch": 4.34, "learning_rate": 8.27033633606446e-06, "loss": 0.2353, "step": 71906 }, { "epoch": 4.34, "learning_rate": 8.268827425950236e-06, "loss": 0.2162, "step": 71908 }, { "epoch": 4.34, "learning_rate": 8.267318515836011e-06, "loss": 0.1805, "step": 71910 }, { "epoch": 4.34, "learning_rate": 8.265809605721788e-06, "loss": 0.1685, "step": 71912 }, { "epoch": 4.34, "learning_rate": 8.264300695607563e-06, "loss": 0.2089, "step": 71914 }, { "epoch": 4.34, "learning_rate": 8.26279178549334e-06, "loss": 0.3732, "step": 71916 }, { "epoch": 4.34, "learning_rate": 8.261282875379114e-06, "loss": 0.1951, "step": 71918 }, { "epoch": 4.34, "learning_rate": 8.25977396526489e-06, "loss": 0.2232, "step": 71920 }, { "epoch": 4.34, "learning_rate": 8.258265055150665e-06, "loss": 0.2796, "step": 71922 }, { "epoch": 4.34, "learning_rate": 8.25675614503644e-06, "loss": 0.2426, "step": 71924 }, { "epoch": 4.34, "learning_rate": 8.255247234922215e-06, "loss": 0.3273, "step": 71926 }, { "epoch": 4.34, "learning_rate": 8.253738324807991e-06, "loss": 0.2518, "step": 71928 }, { "epoch": 4.34, "learning_rate": 8.252229414693766e-06, "loss": 0.2706, "step": 71930 }, { "epoch": 4.34, "learning_rate": 8.250720504579543e-06, "loss": 0.2996, "step": 71932 }, { "epoch": 4.34, "learning_rate": 8.249211594465318e-06, "loss": 0.2612, "step": 71934 }, { "epoch": 4.34, "learning_rate": 8.247702684351094e-06, "loss": 0.2594, "step": 71936 }, { "epoch": 4.34, "learning_rate": 8.246193774236869e-06, "loss": 0.2896, "step": 71938 }, { "epoch": 4.34, "learning_rate": 8.244684864122644e-06, "loss": 0.2838, "step": 71940 }, { "epoch": 4.34, "learning_rate": 8.24317595400842e-06, "loss": 0.3444, "step": 71942 }, { "epoch": 4.34, "learning_rate": 8.241667043894195e-06, "loss": 0.2746, "step": 71944 }, { "epoch": 4.34, "learning_rate": 8.240158133779972e-06, "loss": 0.1687, "step": 71946 }, { "epoch": 4.34, "learning_rate": 8.238649223665746e-06, "loss": 0.195, "step": 71948 }, { "epoch": 4.34, "learning_rate": 8.237140313551523e-06, "loss": 0.2034, "step": 71950 }, { "epoch": 4.34, "learning_rate": 8.235631403437298e-06, "loss": 0.1997, "step": 71952 }, { "epoch": 4.34, "learning_rate": 8.234122493323074e-06, "loss": 0.2171, "step": 71954 }, { "epoch": 4.34, "learning_rate": 8.232613583208847e-06, "loss": 0.1879, "step": 71956 }, { "epoch": 4.34, "learning_rate": 8.231104673094624e-06, "loss": 0.2549, "step": 71958 }, { "epoch": 4.34, "learning_rate": 8.229595762980399e-06, "loss": 0.314, "step": 71960 }, { "epoch": 4.34, "learning_rate": 8.228086852866175e-06, "loss": 0.303, "step": 71962 }, { "epoch": 4.34, "learning_rate": 8.22657794275195e-06, "loss": 0.2666, "step": 71964 }, { "epoch": 4.34, "learning_rate": 8.225069032637727e-06, "loss": 0.2645, "step": 71966 }, { "epoch": 4.34, "learning_rate": 8.223560122523501e-06, "loss": 0.2744, "step": 71968 }, { "epoch": 4.34, "learning_rate": 8.222051212409278e-06, "loss": 0.1913, "step": 71970 }, { "epoch": 4.34, "learning_rate": 8.220542302295053e-06, "loss": 0.2328, "step": 71972 }, { "epoch": 4.34, "learning_rate": 8.219033392180828e-06, "loss": 0.211, "step": 71974 }, { "epoch": 4.34, "learning_rate": 8.217524482066604e-06, "loss": 0.2526, "step": 71976 }, { "epoch": 4.34, "learning_rate": 8.216015571952379e-06, "loss": 0.2456, "step": 71978 }, { "epoch": 4.34, "learning_rate": 8.214506661838155e-06, "loss": 0.3613, "step": 71980 }, { "epoch": 4.34, "learning_rate": 8.21299775172393e-06, "loss": 0.1445, "step": 71982 }, { "epoch": 4.34, "learning_rate": 8.211488841609707e-06, "loss": 0.3569, "step": 71984 }, { "epoch": 4.34, "learning_rate": 8.209979931495482e-06, "loss": 0.2071, "step": 71986 }, { "epoch": 4.35, "learning_rate": 8.208471021381258e-06, "loss": 0.3537, "step": 71988 }, { "epoch": 4.35, "learning_rate": 8.206962111267031e-06, "loss": 0.4322, "step": 71990 }, { "epoch": 4.35, "learning_rate": 8.205453201152808e-06, "loss": 0.3372, "step": 71992 }, { "epoch": 4.35, "learning_rate": 8.203944291038583e-06, "loss": 0.1434, "step": 71994 }, { "epoch": 4.35, "learning_rate": 8.202435380924359e-06, "loss": 0.3231, "step": 71996 }, { "epoch": 4.35, "learning_rate": 8.200926470810134e-06, "loss": 0.1689, "step": 71998 }, { "epoch": 4.35, "learning_rate": 8.19941756069591e-06, "loss": 0.1616, "step": 72000 }, { "epoch": 4.35, "learning_rate": 8.197908650581685e-06, "loss": 0.1939, "step": 72002 }, { "epoch": 4.35, "learning_rate": 8.196399740467462e-06, "loss": 0.2227, "step": 72004 }, { "epoch": 4.35, "learning_rate": 8.194890830353235e-06, "loss": 0.2831, "step": 72006 }, { "epoch": 4.35, "learning_rate": 8.193381920239011e-06, "loss": 0.1438, "step": 72008 }, { "epoch": 4.35, "learning_rate": 8.191873010124788e-06, "loss": 0.3066, "step": 72010 }, { "epoch": 4.35, "learning_rate": 8.190364100010563e-06, "loss": 0.2917, "step": 72012 }, { "epoch": 4.35, "learning_rate": 8.18885518989634e-06, "loss": 0.1355, "step": 72014 }, { "epoch": 4.35, "learning_rate": 8.187346279782114e-06, "loss": 0.2944, "step": 72016 }, { "epoch": 4.35, "learning_rate": 8.18583736966789e-06, "loss": 0.2674, "step": 72018 }, { "epoch": 4.35, "learning_rate": 8.184328459553665e-06, "loss": 0.1881, "step": 72020 }, { "epoch": 4.35, "learning_rate": 8.18281954943944e-06, "loss": 0.1858, "step": 72022 }, { "epoch": 4.35, "learning_rate": 8.181310639325215e-06, "loss": 0.3184, "step": 72024 }, { "epoch": 4.35, "learning_rate": 8.179801729210992e-06, "loss": 0.3222, "step": 72026 }, { "epoch": 4.35, "learning_rate": 8.178292819096766e-06, "loss": 0.2319, "step": 72028 }, { "epoch": 4.35, "learning_rate": 8.176783908982543e-06, "loss": 0.1479, "step": 72030 }, { "epoch": 4.35, "learning_rate": 8.175274998868318e-06, "loss": 0.2195, "step": 72032 }, { "epoch": 4.35, "learning_rate": 8.173766088754094e-06, "loss": 0.2301, "step": 72034 }, { "epoch": 4.35, "learning_rate": 8.172257178639869e-06, "loss": 0.2961, "step": 72036 }, { "epoch": 4.35, "learning_rate": 8.170748268525644e-06, "loss": 0.2049, "step": 72038 }, { "epoch": 4.35, "learning_rate": 8.169239358411419e-06, "loss": 0.207, "step": 72040 }, { "epoch": 4.35, "learning_rate": 8.167730448297195e-06, "loss": 0.3216, "step": 72042 }, { "epoch": 4.35, "learning_rate": 8.16622153818297e-06, "loss": 0.2372, "step": 72044 }, { "epoch": 4.35, "learning_rate": 8.164712628068746e-06, "loss": 0.2138, "step": 72046 }, { "epoch": 4.35, "learning_rate": 8.163203717954521e-06, "loss": 0.2392, "step": 72048 }, { "epoch": 4.35, "learning_rate": 8.161694807840298e-06, "loss": 0.2821, "step": 72050 }, { "epoch": 4.35, "learning_rate": 8.160185897726073e-06, "loss": 0.3416, "step": 72052 }, { "epoch": 4.35, "learning_rate": 8.158676987611847e-06, "loss": 0.1736, "step": 72054 }, { "epoch": 4.35, "learning_rate": 8.157168077497624e-06, "loss": 0.1952, "step": 72056 }, { "epoch": 4.35, "learning_rate": 8.155659167383399e-06, "loss": 0.2365, "step": 72058 }, { "epoch": 4.35, "learning_rate": 8.154150257269175e-06, "loss": 0.1959, "step": 72060 }, { "epoch": 4.35, "learning_rate": 8.15264134715495e-06, "loss": 0.1359, "step": 72062 }, { "epoch": 4.35, "learning_rate": 8.151132437040727e-06, "loss": 0.2029, "step": 72064 }, { "epoch": 4.35, "learning_rate": 8.149623526926501e-06, "loss": 0.1925, "step": 72066 }, { "epoch": 4.35, "learning_rate": 8.148114616812278e-06, "loss": 0.1086, "step": 72068 }, { "epoch": 4.35, "learning_rate": 8.146605706698053e-06, "loss": 0.2735, "step": 72070 }, { "epoch": 4.35, "learning_rate": 8.145096796583828e-06, "loss": 0.2554, "step": 72072 }, { "epoch": 4.35, "learning_rate": 8.143587886469602e-06, "loss": 0.1589, "step": 72074 }, { "epoch": 4.35, "learning_rate": 8.142078976355379e-06, "loss": 0.2017, "step": 72076 }, { "epoch": 4.35, "learning_rate": 8.140570066241154e-06, "loss": 0.3178, "step": 72078 }, { "epoch": 4.35, "learning_rate": 8.13906115612693e-06, "loss": 0.3403, "step": 72080 }, { "epoch": 4.35, "learning_rate": 8.137552246012705e-06, "loss": 0.174, "step": 72082 }, { "epoch": 4.35, "learning_rate": 8.136043335898482e-06, "loss": 0.2439, "step": 72084 }, { "epoch": 4.35, "learning_rate": 8.134534425784256e-06, "loss": 0.262, "step": 72086 }, { "epoch": 4.35, "learning_rate": 8.133025515670031e-06, "loss": 0.2457, "step": 72088 }, { "epoch": 4.35, "learning_rate": 8.131516605555808e-06, "loss": 0.2864, "step": 72090 }, { "epoch": 4.35, "learning_rate": 8.130007695441583e-06, "loss": 0.2915, "step": 72092 }, { "epoch": 4.35, "learning_rate": 8.128498785327359e-06, "loss": 0.3021, "step": 72094 }, { "epoch": 4.35, "learning_rate": 8.126989875213134e-06, "loss": 0.251, "step": 72096 }, { "epoch": 4.35, "learning_rate": 8.12548096509891e-06, "loss": 0.2466, "step": 72098 }, { "epoch": 4.35, "learning_rate": 8.123972054984685e-06, "loss": 0.2517, "step": 72100 }, { "epoch": 4.35, "learning_rate": 8.122463144870462e-06, "loss": 0.208, "step": 72102 }, { "epoch": 4.35, "learning_rate": 8.120954234756235e-06, "loss": 0.1712, "step": 72104 }, { "epoch": 4.35, "learning_rate": 8.119445324642011e-06, "loss": 0.3597, "step": 72106 }, { "epoch": 4.35, "learning_rate": 8.117936414527786e-06, "loss": 0.2309, "step": 72108 }, { "epoch": 4.35, "learning_rate": 8.116427504413563e-06, "loss": 0.3549, "step": 72110 }, { "epoch": 4.35, "learning_rate": 8.114918594299338e-06, "loss": 0.224, "step": 72112 }, { "epoch": 4.35, "learning_rate": 8.113409684185114e-06, "loss": 0.1693, "step": 72114 }, { "epoch": 4.35, "learning_rate": 8.111900774070889e-06, "loss": 0.2047, "step": 72116 }, { "epoch": 4.35, "learning_rate": 8.110391863956665e-06, "loss": 0.1831, "step": 72118 }, { "epoch": 4.35, "learning_rate": 8.108882953842439e-06, "loss": 0.2603, "step": 72120 }, { "epoch": 4.35, "learning_rate": 8.107374043728215e-06, "loss": 0.1967, "step": 72122 }, { "epoch": 4.35, "learning_rate": 8.10586513361399e-06, "loss": 0.277, "step": 72124 }, { "epoch": 4.35, "learning_rate": 8.104356223499766e-06, "loss": 0.3828, "step": 72126 }, { "epoch": 4.35, "learning_rate": 8.102847313385543e-06, "loss": 0.2676, "step": 72128 }, { "epoch": 4.35, "learning_rate": 8.101338403271318e-06, "loss": 0.2084, "step": 72130 }, { "epoch": 4.35, "learning_rate": 8.099829493157094e-06, "loss": 0.2321, "step": 72132 }, { "epoch": 4.35, "learning_rate": 8.098320583042869e-06, "loss": 0.2501, "step": 72134 }, { "epoch": 4.35, "learning_rate": 8.096811672928644e-06, "loss": 0.2674, "step": 72136 }, { "epoch": 4.35, "learning_rate": 8.095302762814419e-06, "loss": 0.3419, "step": 72138 }, { "epoch": 4.35, "learning_rate": 8.093793852700195e-06, "loss": 0.2833, "step": 72140 }, { "epoch": 4.35, "learning_rate": 8.09228494258597e-06, "loss": 0.3008, "step": 72142 }, { "epoch": 4.35, "learning_rate": 8.090776032471747e-06, "loss": 0.2817, "step": 72144 }, { "epoch": 4.35, "learning_rate": 8.089267122357521e-06, "loss": 0.1701, "step": 72146 }, { "epoch": 4.35, "learning_rate": 8.087758212243298e-06, "loss": 0.3417, "step": 72148 }, { "epoch": 4.35, "learning_rate": 8.086249302129073e-06, "loss": 0.1784, "step": 72150 }, { "epoch": 4.35, "learning_rate": 8.08474039201485e-06, "loss": 0.1544, "step": 72152 }, { "epoch": 4.36, "learning_rate": 8.083231481900622e-06, "loss": 0.2256, "step": 72154 }, { "epoch": 4.36, "learning_rate": 8.081722571786399e-06, "loss": 0.2598, "step": 72156 }, { "epoch": 4.36, "learning_rate": 8.080213661672174e-06, "loss": 0.2503, "step": 72158 }, { "epoch": 4.36, "learning_rate": 8.07870475155795e-06, "loss": 0.2315, "step": 72160 }, { "epoch": 4.36, "learning_rate": 8.077195841443725e-06, "loss": 0.2426, "step": 72162 }, { "epoch": 4.36, "learning_rate": 8.075686931329501e-06, "loss": 0.3146, "step": 72164 }, { "epoch": 4.36, "learning_rate": 8.074178021215276e-06, "loss": 0.1999, "step": 72166 }, { "epoch": 4.36, "learning_rate": 8.072669111101053e-06, "loss": 0.168, "step": 72168 }, { "epoch": 4.36, "learning_rate": 8.071160200986828e-06, "loss": 0.2569, "step": 72170 }, { "epoch": 4.36, "learning_rate": 8.069651290872602e-06, "loss": 0.197, "step": 72172 }, { "epoch": 4.36, "learning_rate": 8.068142380758379e-06, "loss": 0.1963, "step": 72174 }, { "epoch": 4.36, "learning_rate": 8.066633470644154e-06, "loss": 0.1779, "step": 72176 }, { "epoch": 4.36, "learning_rate": 8.06512456052993e-06, "loss": 0.1663, "step": 72178 }, { "epoch": 4.36, "learning_rate": 8.063615650415705e-06, "loss": 0.1625, "step": 72180 }, { "epoch": 4.36, "learning_rate": 8.062106740301482e-06, "loss": 0.2476, "step": 72182 }, { "epoch": 4.36, "learning_rate": 8.060597830187256e-06, "loss": 0.2034, "step": 72184 }, { "epoch": 4.36, "learning_rate": 8.059088920073031e-06, "loss": 0.2119, "step": 72186 }, { "epoch": 4.36, "learning_rate": 8.057580009958806e-06, "loss": 0.2041, "step": 72188 }, { "epoch": 4.36, "learning_rate": 8.056071099844583e-06, "loss": 0.2263, "step": 72190 }, { "epoch": 4.36, "learning_rate": 8.054562189730357e-06, "loss": 0.1794, "step": 72192 }, { "epoch": 4.36, "learning_rate": 8.053053279616134e-06, "loss": 0.1838, "step": 72194 }, { "epoch": 4.36, "learning_rate": 8.051544369501909e-06, "loss": 0.2101, "step": 72196 }, { "epoch": 4.36, "learning_rate": 8.050035459387685e-06, "loss": 0.1987, "step": 72198 }, { "epoch": 4.36, "learning_rate": 8.04852654927346e-06, "loss": 0.1766, "step": 72200 }, { "epoch": 4.36, "learning_rate": 8.047017639159235e-06, "loss": 0.3156, "step": 72202 }, { "epoch": 4.36, "learning_rate": 8.045508729045011e-06, "loss": 0.2629, "step": 72204 }, { "epoch": 4.36, "learning_rate": 8.043999818930786e-06, "loss": 0.3857, "step": 72206 }, { "epoch": 4.36, "learning_rate": 8.042490908816563e-06, "loss": 0.2956, "step": 72208 }, { "epoch": 4.36, "learning_rate": 8.040981998702338e-06, "loss": 0.264, "step": 72210 }, { "epoch": 4.36, "learning_rate": 8.039473088588114e-06, "loss": 0.2417, "step": 72212 }, { "epoch": 4.36, "learning_rate": 8.037964178473889e-06, "loss": 0.193, "step": 72214 }, { "epoch": 4.36, "learning_rate": 8.036455268359665e-06, "loss": 0.1449, "step": 72216 }, { "epoch": 4.36, "learning_rate": 8.03494635824544e-06, "loss": 0.2099, "step": 72218 }, { "epoch": 4.36, "learning_rate": 8.033437448131215e-06, "loss": 0.3222, "step": 72220 }, { "epoch": 4.36, "learning_rate": 8.03192853801699e-06, "loss": 0.3557, "step": 72222 }, { "epoch": 4.36, "learning_rate": 8.030419627902766e-06, "loss": 0.3014, "step": 72224 }, { "epoch": 4.36, "learning_rate": 8.028910717788541e-06, "loss": 0.1739, "step": 72226 }, { "epoch": 4.36, "learning_rate": 8.027401807674318e-06, "loss": 0.1688, "step": 72228 }, { "epoch": 4.36, "learning_rate": 8.025892897560093e-06, "loss": 0.1911, "step": 72230 }, { "epoch": 4.36, "learning_rate": 8.024383987445869e-06, "loss": 0.1705, "step": 72232 }, { "epoch": 4.36, "learning_rate": 8.022875077331644e-06, "loss": 0.2575, "step": 72234 }, { "epoch": 4.36, "learning_rate": 8.021366167217419e-06, "loss": 0.2458, "step": 72236 }, { "epoch": 4.36, "learning_rate": 8.019857257103194e-06, "loss": 0.3532, "step": 72238 }, { "epoch": 4.36, "learning_rate": 8.01834834698897e-06, "loss": 0.2013, "step": 72240 }, { "epoch": 4.36, "learning_rate": 8.016839436874745e-06, "loss": 0.381, "step": 72242 }, { "epoch": 4.36, "learning_rate": 8.015330526760521e-06, "loss": 0.202, "step": 72244 }, { "epoch": 4.36, "learning_rate": 8.013821616646298e-06, "loss": 0.3178, "step": 72246 }, { "epoch": 4.36, "learning_rate": 8.012312706532073e-06, "loss": 0.2466, "step": 72248 }, { "epoch": 4.36, "learning_rate": 8.01080379641785e-06, "loss": 0.1204, "step": 72250 }, { "epoch": 4.36, "learning_rate": 8.009294886303622e-06, "loss": 0.2144, "step": 72252 }, { "epoch": 4.36, "learning_rate": 8.007785976189399e-06, "loss": 0.3316, "step": 72254 }, { "epoch": 4.36, "learning_rate": 8.006277066075174e-06, "loss": 0.161, "step": 72256 }, { "epoch": 4.36, "learning_rate": 8.00476815596095e-06, "loss": 0.1354, "step": 72258 }, { "epoch": 4.36, "learning_rate": 8.003259245846725e-06, "loss": 0.2265, "step": 72260 }, { "epoch": 4.36, "learning_rate": 8.001750335732502e-06, "loss": 0.2163, "step": 72262 }, { "epoch": 4.36, "learning_rate": 8.000241425618276e-06, "loss": 0.1643, "step": 72264 }, { "epoch": 4.36, "learning_rate": 7.998732515504053e-06, "loss": 0.2665, "step": 72266 }, { "epoch": 4.36, "learning_rate": 7.997223605389826e-06, "loss": 0.255, "step": 72268 }, { "epoch": 4.36, "learning_rate": 7.995714695275602e-06, "loss": 0.3148, "step": 72270 }, { "epoch": 4.36, "learning_rate": 7.994205785161377e-06, "loss": 0.2442, "step": 72272 }, { "epoch": 4.36, "learning_rate": 7.992696875047154e-06, "loss": 0.1683, "step": 72274 }, { "epoch": 4.36, "learning_rate": 7.991187964932929e-06, "loss": 0.1946, "step": 72276 }, { "epoch": 4.36, "learning_rate": 7.989679054818705e-06, "loss": 0.2235, "step": 72278 }, { "epoch": 4.36, "learning_rate": 7.98817014470448e-06, "loss": 0.1508, "step": 72280 }, { "epoch": 4.36, "learning_rate": 7.986661234590256e-06, "loss": 0.2009, "step": 72282 }, { "epoch": 4.36, "learning_rate": 7.985152324476031e-06, "loss": 0.1774, "step": 72284 }, { "epoch": 4.36, "learning_rate": 7.983643414361806e-06, "loss": 0.2351, "step": 72286 }, { "epoch": 4.36, "learning_rate": 7.982134504247583e-06, "loss": 0.2624, "step": 72288 }, { "epoch": 4.36, "learning_rate": 7.980625594133357e-06, "loss": 0.2009, "step": 72290 }, { "epoch": 4.36, "learning_rate": 7.979116684019134e-06, "loss": 0.247, "step": 72292 }, { "epoch": 4.36, "learning_rate": 7.977607773904909e-06, "loss": 0.1354, "step": 72294 }, { "epoch": 4.36, "learning_rate": 7.976098863790685e-06, "loss": 0.249, "step": 72296 }, { "epoch": 4.36, "learning_rate": 7.97458995367646e-06, "loss": 0.2354, "step": 72298 }, { "epoch": 4.36, "learning_rate": 7.973081043562237e-06, "loss": 0.1409, "step": 72300 }, { "epoch": 4.36, "learning_rate": 7.97157213344801e-06, "loss": 0.2944, "step": 72302 }, { "epoch": 4.36, "learning_rate": 7.970063223333786e-06, "loss": 0.1985, "step": 72304 }, { "epoch": 4.36, "learning_rate": 7.968554313219561e-06, "loss": 0.2848, "step": 72306 }, { "epoch": 4.36, "learning_rate": 7.967045403105338e-06, "loss": 0.2741, "step": 72308 }, { "epoch": 4.36, "learning_rate": 7.965536492991112e-06, "loss": 0.1826, "step": 72310 }, { "epoch": 4.36, "learning_rate": 7.964027582876889e-06, "loss": 0.2198, "step": 72312 }, { "epoch": 4.36, "learning_rate": 7.962518672762664e-06, "loss": 0.2194, "step": 72314 }, { "epoch": 4.36, "learning_rate": 7.96100976264844e-06, "loss": 0.1978, "step": 72316 }, { "epoch": 4.36, "learning_rate": 7.959500852534215e-06, "loss": 0.2229, "step": 72318 }, { "epoch": 4.37, "learning_rate": 7.95799194241999e-06, "loss": 0.2305, "step": 72320 }, { "epoch": 4.37, "learning_rate": 7.956483032305766e-06, "loss": 0.2396, "step": 72322 }, { "epoch": 4.37, "learning_rate": 7.954974122191541e-06, "loss": 0.2375, "step": 72324 }, { "epoch": 4.37, "learning_rate": 7.953465212077318e-06, "loss": 0.1993, "step": 72326 }, { "epoch": 4.37, "learning_rate": 7.951956301963093e-06, "loss": 0.2031, "step": 72328 }, { "epoch": 4.37, "learning_rate": 7.950447391848869e-06, "loss": 0.3198, "step": 72330 }, { "epoch": 4.37, "learning_rate": 7.948938481734644e-06, "loss": 0.2905, "step": 72332 }, { "epoch": 4.37, "learning_rate": 7.947429571620419e-06, "loss": 0.2156, "step": 72334 }, { "epoch": 4.37, "learning_rate": 7.945920661506194e-06, "loss": 0.1992, "step": 72336 }, { "epoch": 4.37, "learning_rate": 7.94441175139197e-06, "loss": 0.3399, "step": 72338 }, { "epoch": 4.37, "learning_rate": 7.942902841277745e-06, "loss": 0.4569, "step": 72340 }, { "epoch": 4.37, "learning_rate": 7.941393931163521e-06, "loss": 0.3005, "step": 72342 }, { "epoch": 4.37, "learning_rate": 7.939885021049296e-06, "loss": 0.1809, "step": 72344 }, { "epoch": 4.37, "learning_rate": 7.938376110935073e-06, "loss": 0.2499, "step": 72346 }, { "epoch": 4.37, "learning_rate": 7.936867200820848e-06, "loss": 0.1946, "step": 72348 }, { "epoch": 4.37, "learning_rate": 7.935358290706622e-06, "loss": 0.2828, "step": 72350 }, { "epoch": 4.37, "learning_rate": 7.933849380592397e-06, "loss": 0.2617, "step": 72352 }, { "epoch": 4.37, "learning_rate": 7.932340470478174e-06, "loss": 0.2753, "step": 72354 }, { "epoch": 4.37, "learning_rate": 7.930831560363949e-06, "loss": 0.2327, "step": 72356 }, { "epoch": 4.37, "learning_rate": 7.929322650249725e-06, "loss": 0.1736, "step": 72358 }, { "epoch": 4.37, "learning_rate": 7.9278137401355e-06, "loss": 0.2225, "step": 72360 }, { "epoch": 4.37, "learning_rate": 7.926304830021276e-06, "loss": 0.1577, "step": 72362 }, { "epoch": 4.37, "learning_rate": 7.924795919907053e-06, "loss": 0.1995, "step": 72364 }, { "epoch": 4.37, "learning_rate": 7.923287009792826e-06, "loss": 0.2667, "step": 72366 }, { "epoch": 4.37, "learning_rate": 7.921778099678603e-06, "loss": 0.252, "step": 72368 }, { "epoch": 4.37, "learning_rate": 7.920269189564377e-06, "loss": 0.1354, "step": 72370 }, { "epoch": 4.37, "learning_rate": 7.918760279450154e-06, "loss": 0.2048, "step": 72372 }, { "epoch": 4.37, "learning_rate": 7.917251369335929e-06, "loss": 0.2613, "step": 72374 }, { "epoch": 4.37, "learning_rate": 7.915742459221705e-06, "loss": 0.1844, "step": 72376 }, { "epoch": 4.37, "learning_rate": 7.91423354910748e-06, "loss": 0.4884, "step": 72378 }, { "epoch": 4.37, "learning_rate": 7.912724638993257e-06, "loss": 0.2461, "step": 72380 }, { "epoch": 4.37, "learning_rate": 7.911215728879031e-06, "loss": 0.2905, "step": 72382 }, { "epoch": 4.37, "learning_rate": 7.909706818764806e-06, "loss": 0.2245, "step": 72384 }, { "epoch": 4.37, "learning_rate": 7.908197908650581e-06, "loss": 0.2013, "step": 72386 }, { "epoch": 4.37, "learning_rate": 7.906688998536358e-06, "loss": 0.2519, "step": 72388 }, { "epoch": 4.37, "learning_rate": 7.905180088422132e-06, "loss": 0.2972, "step": 72390 }, { "epoch": 4.37, "learning_rate": 7.903671178307909e-06, "loss": 0.2193, "step": 72392 }, { "epoch": 4.37, "learning_rate": 7.902162268193684e-06, "loss": 0.1956, "step": 72394 }, { "epoch": 4.37, "learning_rate": 7.90065335807946e-06, "loss": 0.1727, "step": 72396 }, { "epoch": 4.37, "learning_rate": 7.899144447965235e-06, "loss": 0.201, "step": 72398 }, { "epoch": 4.37, "learning_rate": 7.89763553785101e-06, "loss": 0.3069, "step": 72400 }, { "epoch": 4.37, "learning_rate": 7.896126627736786e-06, "loss": 0.1992, "step": 72402 }, { "epoch": 4.37, "learning_rate": 7.894617717622561e-06, "loss": 0.1696, "step": 72404 }, { "epoch": 4.37, "learning_rate": 7.893108807508338e-06, "loss": 0.2923, "step": 72406 }, { "epoch": 4.37, "learning_rate": 7.891599897394112e-06, "loss": 0.3438, "step": 72408 }, { "epoch": 4.37, "learning_rate": 7.890090987279889e-06, "loss": 0.1285, "step": 72410 }, { "epoch": 4.37, "learning_rate": 7.888582077165664e-06, "loss": 0.2085, "step": 72412 }, { "epoch": 4.37, "learning_rate": 7.88707316705144e-06, "loss": 0.2532, "step": 72414 }, { "epoch": 4.37, "learning_rate": 7.885564256937213e-06, "loss": 0.2181, "step": 72416 }, { "epoch": 4.37, "learning_rate": 7.88405534682299e-06, "loss": 0.2005, "step": 72418 }, { "epoch": 4.37, "learning_rate": 7.882546436708765e-06, "loss": 0.2367, "step": 72420 }, { "epoch": 4.37, "learning_rate": 7.881037526594541e-06, "loss": 0.2946, "step": 72422 }, { "epoch": 4.37, "learning_rate": 7.879528616480316e-06, "loss": 0.2049, "step": 72424 }, { "epoch": 4.37, "learning_rate": 7.878019706366093e-06, "loss": 0.2261, "step": 72426 }, { "epoch": 4.37, "learning_rate": 7.876510796251867e-06, "loss": 0.221, "step": 72428 }, { "epoch": 4.37, "learning_rate": 7.875001886137644e-06, "loss": 0.2373, "step": 72430 }, { "epoch": 4.37, "learning_rate": 7.873492976023419e-06, "loss": 0.1878, "step": 72432 }, { "epoch": 4.37, "learning_rate": 7.871984065909194e-06, "loss": 0.1863, "step": 72434 }, { "epoch": 4.37, "learning_rate": 7.87047515579497e-06, "loss": 0.2444, "step": 72436 }, { "epoch": 4.37, "learning_rate": 7.868966245680745e-06, "loss": 0.1485, "step": 72438 }, { "epoch": 4.37, "learning_rate": 7.867457335566521e-06, "loss": 0.2574, "step": 72440 }, { "epoch": 4.37, "learning_rate": 7.865948425452296e-06, "loss": 0.3831, "step": 72442 }, { "epoch": 4.37, "learning_rate": 7.864439515338073e-06, "loss": 0.2791, "step": 72444 }, { "epoch": 4.37, "learning_rate": 7.862930605223848e-06, "loss": 0.1887, "step": 72446 }, { "epoch": 4.37, "learning_rate": 7.861421695109622e-06, "loss": 0.2853, "step": 72448 }, { "epoch": 4.37, "learning_rate": 7.859912784995397e-06, "loss": 0.2139, "step": 72450 }, { "epoch": 4.37, "learning_rate": 7.858403874881174e-06, "loss": 0.2318, "step": 72452 }, { "epoch": 4.37, "learning_rate": 7.856894964766949e-06, "loss": 0.2271, "step": 72454 }, { "epoch": 4.37, "learning_rate": 7.855386054652725e-06, "loss": 0.2158, "step": 72456 }, { "epoch": 4.37, "learning_rate": 7.8538771445385e-06, "loss": 0.2766, "step": 72458 }, { "epoch": 4.37, "learning_rate": 7.852368234424276e-06, "loss": 0.1884, "step": 72460 }, { "epoch": 4.37, "learning_rate": 7.850859324310051e-06, "loss": 0.3117, "step": 72462 }, { "epoch": 4.37, "learning_rate": 7.849350414195828e-06, "loss": 0.154, "step": 72464 }, { "epoch": 4.37, "learning_rate": 7.847841504081601e-06, "loss": 0.276, "step": 72466 }, { "epoch": 4.37, "learning_rate": 7.846332593967377e-06, "loss": 0.2774, "step": 72468 }, { "epoch": 4.37, "learning_rate": 7.844823683853152e-06, "loss": 0.1904, "step": 72470 }, { "epoch": 4.37, "learning_rate": 7.843314773738929e-06, "loss": 0.1763, "step": 72472 }, { "epoch": 4.37, "learning_rate": 7.841805863624704e-06, "loss": 0.2302, "step": 72474 }, { "epoch": 4.37, "learning_rate": 7.84029695351048e-06, "loss": 0.2001, "step": 72476 }, { "epoch": 4.37, "learning_rate": 7.838788043396255e-06, "loss": 0.2437, "step": 72478 }, { "epoch": 4.37, "learning_rate": 7.837279133282031e-06, "loss": 0.1415, "step": 72480 }, { "epoch": 4.37, "learning_rate": 7.835770223167806e-06, "loss": 0.2099, "step": 72482 }, { "epoch": 4.37, "learning_rate": 7.834261313053581e-06, "loss": 0.2323, "step": 72484 }, { "epoch": 4.38, "learning_rate": 7.832752402939358e-06, "loss": 0.2977, "step": 72486 }, { "epoch": 4.38, "learning_rate": 7.831243492825132e-06, "loss": 0.264, "step": 72488 }, { "epoch": 4.38, "learning_rate": 7.829734582710909e-06, "loss": 0.1689, "step": 72490 }, { "epoch": 4.38, "learning_rate": 7.828225672596684e-06, "loss": 0.2259, "step": 72492 }, { "epoch": 4.38, "learning_rate": 7.82671676248246e-06, "loss": 0.1937, "step": 72494 }, { "epoch": 4.38, "learning_rate": 7.825207852368235e-06, "loss": 0.2692, "step": 72496 }, { "epoch": 4.38, "learning_rate": 7.82369894225401e-06, "loss": 0.2011, "step": 72498 }, { "epoch": 4.38, "learning_rate": 7.822190032139785e-06, "loss": 0.1667, "step": 72500 }, { "epoch": 4.38, "learning_rate": 7.820681122025561e-06, "loss": 0.3106, "step": 72502 }, { "epoch": 4.38, "learning_rate": 7.819172211911336e-06, "loss": 0.1904, "step": 72504 }, { "epoch": 4.38, "learning_rate": 7.817663301797113e-06, "loss": 0.2046, "step": 72506 }, { "epoch": 4.38, "learning_rate": 7.816154391682887e-06, "loss": 0.2602, "step": 72508 }, { "epoch": 4.38, "learning_rate": 7.814645481568664e-06, "loss": 0.1957, "step": 72510 }, { "epoch": 4.38, "learning_rate": 7.813136571454439e-06, "loss": 0.2163, "step": 72512 }, { "epoch": 4.38, "learning_rate": 7.811627661340213e-06, "loss": 0.2249, "step": 72514 }, { "epoch": 4.38, "learning_rate": 7.81011875122599e-06, "loss": 0.2503, "step": 72516 }, { "epoch": 4.38, "learning_rate": 7.808609841111765e-06, "loss": 0.1945, "step": 72518 }, { "epoch": 4.38, "learning_rate": 7.807100930997541e-06, "loss": 0.2826, "step": 72520 }, { "epoch": 4.38, "learning_rate": 7.805592020883316e-06, "loss": 0.1963, "step": 72522 }, { "epoch": 4.38, "learning_rate": 7.804083110769093e-06, "loss": 0.155, "step": 72524 }, { "epoch": 4.38, "learning_rate": 7.802574200654867e-06, "loss": 0.1696, "step": 72526 }, { "epoch": 4.38, "learning_rate": 7.801065290540644e-06, "loss": 0.3233, "step": 72528 }, { "epoch": 4.38, "learning_rate": 7.799556380426419e-06, "loss": 0.3766, "step": 72530 }, { "epoch": 4.38, "learning_rate": 7.798047470312194e-06, "loss": 0.2536, "step": 72532 }, { "epoch": 4.38, "learning_rate": 7.796538560197968e-06, "loss": 0.2637, "step": 72534 }, { "epoch": 4.38, "learning_rate": 7.795029650083745e-06, "loss": 0.2792, "step": 72536 }, { "epoch": 4.38, "learning_rate": 7.79352073996952e-06, "loss": 0.2003, "step": 72538 }, { "epoch": 4.38, "learning_rate": 7.792011829855296e-06, "loss": 0.4169, "step": 72540 }, { "epoch": 4.38, "learning_rate": 7.790502919741071e-06, "loss": 0.2194, "step": 72542 }, { "epoch": 4.38, "learning_rate": 7.788994009626848e-06, "loss": 0.2025, "step": 72544 }, { "epoch": 4.38, "learning_rate": 7.787485099512622e-06, "loss": 0.317, "step": 72546 }, { "epoch": 4.38, "learning_rate": 7.785976189398397e-06, "loss": 0.2224, "step": 72548 }, { "epoch": 4.38, "learning_rate": 7.784467279284174e-06, "loss": 0.2846, "step": 72550 }, { "epoch": 4.38, "learning_rate": 7.782958369169949e-06, "loss": 0.207, "step": 72552 }, { "epoch": 4.38, "learning_rate": 7.781449459055725e-06, "loss": 0.217, "step": 72554 }, { "epoch": 4.38, "learning_rate": 7.7799405489415e-06, "loss": 0.2561, "step": 72556 }, { "epoch": 4.38, "learning_rate": 7.778431638827276e-06, "loss": 0.3084, "step": 72558 }, { "epoch": 4.38, "learning_rate": 7.776922728713051e-06, "loss": 0.2735, "step": 72560 }, { "epoch": 4.38, "learning_rate": 7.775413818598828e-06, "loss": 0.2343, "step": 72562 }, { "epoch": 4.38, "learning_rate": 7.773904908484601e-06, "loss": 0.171, "step": 72564 }, { "epoch": 4.38, "learning_rate": 7.772395998370377e-06, "loss": 0.2744, "step": 72566 }, { "epoch": 4.38, "learning_rate": 7.770887088256152e-06, "loss": 0.1605, "step": 72568 }, { "epoch": 4.38, "learning_rate": 7.769378178141929e-06, "loss": 0.3058, "step": 72570 }, { "epoch": 4.38, "learning_rate": 7.767869268027704e-06, "loss": 0.1354, "step": 72572 }, { "epoch": 4.38, "learning_rate": 7.76636035791348e-06, "loss": 0.2501, "step": 72574 }, { "epoch": 4.38, "learning_rate": 7.764851447799255e-06, "loss": 0.2023, "step": 72576 }, { "epoch": 4.38, "learning_rate": 7.763342537685031e-06, "loss": 0.1785, "step": 72578 }, { "epoch": 4.38, "learning_rate": 7.761833627570805e-06, "loss": 0.1979, "step": 72580 }, { "epoch": 4.38, "learning_rate": 7.760324717456581e-06, "loss": 0.2622, "step": 72582 }, { "epoch": 4.38, "learning_rate": 7.758815807342356e-06, "loss": 0.2413, "step": 72584 }, { "epoch": 4.38, "learning_rate": 7.757306897228132e-06, "loss": 0.2154, "step": 72586 }, { "epoch": 4.38, "learning_rate": 7.755797987113907e-06, "loss": 0.2364, "step": 72588 }, { "epoch": 4.38, "learning_rate": 7.754289076999684e-06, "loss": 0.227, "step": 72590 }, { "epoch": 4.38, "learning_rate": 7.752780166885459e-06, "loss": 0.1642, "step": 72592 }, { "epoch": 4.38, "learning_rate": 7.751271256771235e-06, "loss": 0.1926, "step": 72594 }, { "epoch": 4.38, "learning_rate": 7.74976234665701e-06, "loss": 0.3251, "step": 72596 }, { "epoch": 4.38, "learning_rate": 7.748253436542785e-06, "loss": 0.1617, "step": 72598 }, { "epoch": 4.38, "learning_rate": 7.746744526428561e-06, "loss": 0.2338, "step": 72600 }, { "epoch": 4.38, "learning_rate": 7.745235616314336e-06, "loss": 0.2639, "step": 72602 }, { "epoch": 4.38, "learning_rate": 7.743726706200113e-06, "loss": 0.1443, "step": 72604 }, { "epoch": 4.38, "learning_rate": 7.742217796085887e-06, "loss": 0.1847, "step": 72606 }, { "epoch": 4.38, "learning_rate": 7.740708885971664e-06, "loss": 0.1447, "step": 72608 }, { "epoch": 4.38, "learning_rate": 7.739199975857439e-06, "loss": 0.2039, "step": 72610 }, { "epoch": 4.38, "learning_rate": 7.737691065743215e-06, "loss": 0.2315, "step": 72612 }, { "epoch": 4.38, "learning_rate": 7.736182155628988e-06, "loss": 0.2347, "step": 72614 }, { "epoch": 4.38, "learning_rate": 7.734673245514765e-06, "loss": 0.2406, "step": 72616 }, { "epoch": 4.38, "learning_rate": 7.73316433540054e-06, "loss": 0.2397, "step": 72618 }, { "epoch": 4.38, "learning_rate": 7.731655425286316e-06, "loss": 0.2808, "step": 72620 }, { "epoch": 4.38, "learning_rate": 7.730146515172091e-06, "loss": 0.2768, "step": 72622 }, { "epoch": 4.38, "learning_rate": 7.728637605057868e-06, "loss": 0.3269, "step": 72624 }, { "epoch": 4.38, "learning_rate": 7.727128694943642e-06, "loss": 0.2908, "step": 72626 }, { "epoch": 4.38, "learning_rate": 7.725619784829419e-06, "loss": 0.2462, "step": 72628 }, { "epoch": 4.38, "learning_rate": 7.724110874715194e-06, "loss": 0.1915, "step": 72630 }, { "epoch": 4.38, "learning_rate": 7.722601964600969e-06, "loss": 0.3615, "step": 72632 }, { "epoch": 4.38, "learning_rate": 7.721093054486745e-06, "loss": 0.41, "step": 72634 }, { "epoch": 4.38, "learning_rate": 7.71958414437252e-06, "loss": 0.2668, "step": 72636 }, { "epoch": 4.38, "learning_rate": 7.718075234258296e-06, "loss": 0.1691, "step": 72638 }, { "epoch": 4.38, "learning_rate": 7.716566324144071e-06, "loss": 0.2325, "step": 72640 }, { "epoch": 4.38, "learning_rate": 7.715057414029848e-06, "loss": 0.1949, "step": 72642 }, { "epoch": 4.38, "learning_rate": 7.713548503915623e-06, "loss": 0.2039, "step": 72644 }, { "epoch": 4.38, "learning_rate": 7.712039593801397e-06, "loss": 0.2255, "step": 72646 }, { "epoch": 4.38, "learning_rate": 7.710530683687172e-06, "loss": 0.2155, "step": 72648 }, { "epoch": 4.38, "learning_rate": 7.709021773572949e-06, "loss": 0.2042, "step": 72650 }, { "epoch": 4.39, "learning_rate": 7.707512863458723e-06, "loss": 0.2658, "step": 72652 }, { "epoch": 4.39, "learning_rate": 7.7060039533445e-06, "loss": 0.3666, "step": 72654 }, { "epoch": 4.39, "learning_rate": 7.704495043230275e-06, "loss": 0.2531, "step": 72656 }, { "epoch": 4.39, "learning_rate": 7.702986133116051e-06, "loss": 0.1453, "step": 72658 }, { "epoch": 4.39, "learning_rate": 7.701477223001826e-06, "loss": 0.2383, "step": 72660 }, { "epoch": 4.39, "learning_rate": 7.699968312887601e-06, "loss": 0.2532, "step": 72662 }, { "epoch": 4.39, "learning_rate": 7.698459402773377e-06, "loss": 0.1825, "step": 72664 }, { "epoch": 4.39, "learning_rate": 7.696950492659152e-06, "loss": 0.3147, "step": 72666 }, { "epoch": 4.39, "learning_rate": 7.695441582544929e-06, "loss": 0.2718, "step": 72668 }, { "epoch": 4.39, "learning_rate": 7.693932672430704e-06, "loss": 0.2686, "step": 72670 }, { "epoch": 4.39, "learning_rate": 7.69242376231648e-06, "loss": 0.1997, "step": 72672 }, { "epoch": 4.39, "learning_rate": 7.690914852202255e-06, "loss": 0.1864, "step": 72674 }, { "epoch": 4.39, "learning_rate": 7.689405942088031e-06, "loss": 0.1996, "step": 72676 }, { "epoch": 4.39, "learning_rate": 7.687897031973805e-06, "loss": 0.2151, "step": 72678 }, { "epoch": 4.39, "learning_rate": 7.686388121859581e-06, "loss": 0.1906, "step": 72680 }, { "epoch": 4.39, "learning_rate": 7.684879211745356e-06, "loss": 0.1644, "step": 72682 }, { "epoch": 4.39, "learning_rate": 7.683370301631132e-06, "loss": 0.1907, "step": 72684 }, { "epoch": 4.39, "learning_rate": 7.681861391516907e-06, "loss": 0.227, "step": 72686 }, { "epoch": 4.39, "learning_rate": 7.680352481402684e-06, "loss": 0.1768, "step": 72688 }, { "epoch": 4.39, "learning_rate": 7.678843571288459e-06, "loss": 0.2388, "step": 72690 }, { "epoch": 4.39, "learning_rate": 7.677334661174235e-06, "loss": 0.1865, "step": 72692 }, { "epoch": 4.39, "learning_rate": 7.67582575106001e-06, "loss": 0.226, "step": 72694 }, { "epoch": 4.39, "learning_rate": 7.674316840945785e-06, "loss": 0.1879, "step": 72696 }, { "epoch": 4.39, "learning_rate": 7.67280793083156e-06, "loss": 0.3037, "step": 72698 }, { "epoch": 4.39, "learning_rate": 7.671299020717336e-06, "loss": 0.2907, "step": 72700 }, { "epoch": 4.39, "learning_rate": 7.669790110603111e-06, "loss": 0.2016, "step": 72702 }, { "epoch": 4.39, "learning_rate": 7.668281200488887e-06, "loss": 0.2253, "step": 72704 }, { "epoch": 4.39, "learning_rate": 7.666772290374662e-06, "loss": 0.2059, "step": 72706 }, { "epoch": 4.39, "learning_rate": 7.665263380260439e-06, "loss": 0.2612, "step": 72708 }, { "epoch": 4.39, "learning_rate": 7.663754470146214e-06, "loss": 0.3155, "step": 72710 }, { "epoch": 4.39, "learning_rate": 7.662245560031988e-06, "loss": 0.2603, "step": 72712 }, { "epoch": 4.39, "learning_rate": 7.660736649917765e-06, "loss": 0.1739, "step": 72714 }, { "epoch": 4.39, "learning_rate": 7.65922773980354e-06, "loss": 0.201, "step": 72716 }, { "epoch": 4.39, "learning_rate": 7.657718829689316e-06, "loss": 0.3087, "step": 72718 }, { "epoch": 4.39, "learning_rate": 7.656209919575091e-06, "loss": 0.2191, "step": 72720 }, { "epoch": 4.39, "learning_rate": 7.654701009460868e-06, "loss": 0.2656, "step": 72722 }, { "epoch": 4.39, "learning_rate": 7.653192099346642e-06, "loss": 0.2876, "step": 72724 }, { "epoch": 4.39, "learning_rate": 7.651683189232419e-06, "loss": 0.159, "step": 72726 }, { "epoch": 4.39, "learning_rate": 7.650174279118192e-06, "loss": 0.1614, "step": 72728 }, { "epoch": 4.39, "learning_rate": 7.648665369003969e-06, "loss": 0.2339, "step": 72730 }, { "epoch": 4.39, "learning_rate": 7.647156458889743e-06, "loss": 0.1761, "step": 72732 }, { "epoch": 4.39, "learning_rate": 7.64564754877552e-06, "loss": 0.4161, "step": 72734 }, { "epoch": 4.39, "learning_rate": 7.644138638661295e-06, "loss": 0.2141, "step": 72736 }, { "epoch": 4.39, "learning_rate": 7.642629728547071e-06, "loss": 0.2595, "step": 72738 }, { "epoch": 4.39, "learning_rate": 7.641120818432846e-06, "loss": 0.2655, "step": 72740 }, { "epoch": 4.39, "learning_rate": 7.639611908318623e-06, "loss": 0.1807, "step": 72742 }, { "epoch": 4.39, "learning_rate": 7.638102998204397e-06, "loss": 0.2702, "step": 72744 }, { "epoch": 4.39, "learning_rate": 7.636594088090172e-06, "loss": 0.1695, "step": 72746 }, { "epoch": 4.39, "learning_rate": 7.635085177975949e-06, "loss": 0.1869, "step": 72748 }, { "epoch": 4.39, "learning_rate": 7.633576267861724e-06, "loss": 0.2201, "step": 72750 }, { "epoch": 4.39, "learning_rate": 7.6320673577475e-06, "loss": 0.2801, "step": 72752 }, { "epoch": 4.39, "learning_rate": 7.630558447633275e-06, "loss": 0.3404, "step": 72754 }, { "epoch": 4.39, "learning_rate": 7.6290495375190505e-06, "loss": 0.1587, "step": 72756 }, { "epoch": 4.39, "learning_rate": 7.627540627404826e-06, "loss": 0.2605, "step": 72758 }, { "epoch": 4.39, "learning_rate": 7.626031717290601e-06, "loss": 0.2418, "step": 72760 }, { "epoch": 4.39, "learning_rate": 7.624522807176377e-06, "loss": 0.3115, "step": 72762 }, { "epoch": 4.39, "learning_rate": 7.623013897062152e-06, "loss": 0.2048, "step": 72764 }, { "epoch": 4.39, "learning_rate": 7.621504986947928e-06, "loss": 0.1802, "step": 72766 }, { "epoch": 4.39, "learning_rate": 7.619996076833704e-06, "loss": 0.27, "step": 72768 }, { "epoch": 4.39, "learning_rate": 7.618487166719479e-06, "loss": 0.3832, "step": 72770 }, { "epoch": 4.39, "learning_rate": 7.616978256605255e-06, "loss": 0.2483, "step": 72772 }, { "epoch": 4.39, "learning_rate": 7.615469346491031e-06, "loss": 0.2009, "step": 72774 }, { "epoch": 4.39, "learning_rate": 7.613960436376806e-06, "loss": 0.2516, "step": 72776 }, { "epoch": 4.39, "learning_rate": 7.61245152626258e-06, "loss": 0.1983, "step": 72778 }, { "epoch": 4.39, "learning_rate": 7.610942616148356e-06, "loss": 0.183, "step": 72780 }, { "epoch": 4.39, "learning_rate": 7.609433706034132e-06, "loss": 0.2007, "step": 72782 }, { "epoch": 4.39, "learning_rate": 7.607924795919907e-06, "loss": 0.2422, "step": 72784 }, { "epoch": 4.39, "learning_rate": 7.606415885805683e-06, "loss": 0.2403, "step": 72786 }, { "epoch": 4.39, "learning_rate": 7.604906975691459e-06, "loss": 0.1699, "step": 72788 }, { "epoch": 4.39, "learning_rate": 7.603398065577234e-06, "loss": 0.2876, "step": 72790 }, { "epoch": 4.39, "learning_rate": 7.60188915546301e-06, "loss": 0.2174, "step": 72792 }, { "epoch": 4.39, "learning_rate": 7.600380245348784e-06, "loss": 0.1771, "step": 72794 }, { "epoch": 4.39, "learning_rate": 7.59887133523456e-06, "loss": 0.2331, "step": 72796 }, { "epoch": 4.39, "learning_rate": 7.597362425120335e-06, "loss": 0.2728, "step": 72798 }, { "epoch": 4.39, "learning_rate": 7.595853515006111e-06, "loss": 0.3081, "step": 72800 }, { "epoch": 4.39, "learning_rate": 7.5943446048918875e-06, "loss": 0.3037, "step": 72802 }, { "epoch": 4.39, "learning_rate": 7.592835694777663e-06, "loss": 0.2712, "step": 72804 }, { "epoch": 4.39, "learning_rate": 7.591326784663439e-06, "loss": 0.2451, "step": 72806 }, { "epoch": 4.39, "learning_rate": 7.5898178745492145e-06, "loss": 0.1919, "step": 72808 }, { "epoch": 4.39, "learning_rate": 7.5883089644349884e-06, "loss": 0.2275, "step": 72810 }, { "epoch": 4.39, "learning_rate": 7.586800054320764e-06, "loss": 0.3315, "step": 72812 }, { "epoch": 4.39, "learning_rate": 7.58529114420654e-06, "loss": 0.2874, "step": 72814 }, { "epoch": 4.39, "learning_rate": 7.5837822340923154e-06, "loss": 0.2204, "step": 72816 }, { "epoch": 4.4, "learning_rate": 7.582273323978091e-06, "loss": 0.3231, "step": 72818 }, { "epoch": 4.4, "learning_rate": 7.580764413863867e-06, "loss": 0.1769, "step": 72820 }, { "epoch": 4.4, "learning_rate": 7.5792555037496424e-06, "loss": 0.2382, "step": 72822 }, { "epoch": 4.4, "learning_rate": 7.577746593635418e-06, "loss": 0.2292, "step": 72824 }, { "epoch": 4.4, "learning_rate": 7.576237683521192e-06, "loss": 0.2014, "step": 72826 }, { "epoch": 4.4, "learning_rate": 7.574728773406968e-06, "loss": 0.2456, "step": 72828 }, { "epoch": 4.4, "learning_rate": 7.573219863292743e-06, "loss": 0.2788, "step": 72830 }, { "epoch": 4.4, "learning_rate": 7.571710953178519e-06, "loss": 0.1613, "step": 72832 }, { "epoch": 4.4, "learning_rate": 7.570202043064295e-06, "loss": 0.2172, "step": 72834 }, { "epoch": 4.4, "learning_rate": 7.56869313295007e-06, "loss": 0.2401, "step": 72836 }, { "epoch": 4.4, "learning_rate": 7.567184222835846e-06, "loss": 0.148, "step": 72838 }, { "epoch": 4.4, "learning_rate": 7.565675312721622e-06, "loss": 0.1522, "step": 72840 }, { "epoch": 4.4, "learning_rate": 7.564166402607397e-06, "loss": 0.307, "step": 72842 }, { "epoch": 4.4, "learning_rate": 7.562657492493172e-06, "loss": 0.2016, "step": 72844 }, { "epoch": 4.4, "learning_rate": 7.561148582378948e-06, "loss": 0.2133, "step": 72846 }, { "epoch": 4.4, "learning_rate": 7.5596396722647236e-06, "loss": 0.1552, "step": 72848 }, { "epoch": 4.4, "learning_rate": 7.558130762150499e-06, "loss": 0.2454, "step": 72850 }, { "epoch": 4.4, "learning_rate": 7.556621852036275e-06, "loss": 0.2725, "step": 72852 }, { "epoch": 4.4, "learning_rate": 7.5551129419220506e-06, "loss": 0.1837, "step": 72854 }, { "epoch": 4.4, "learning_rate": 7.553604031807826e-06, "loss": 0.1619, "step": 72856 }, { "epoch": 4.4, "learning_rate": 7.552095121693602e-06, "loss": 0.1796, "step": 72858 }, { "epoch": 4.4, "learning_rate": 7.550586211579376e-06, "loss": 0.2848, "step": 72860 }, { "epoch": 4.4, "learning_rate": 7.5490773014651515e-06, "loss": 0.2051, "step": 72862 }, { "epoch": 4.4, "learning_rate": 7.547568391350927e-06, "loss": 0.31, "step": 72864 }, { "epoch": 4.4, "learning_rate": 7.546059481236703e-06, "loss": 0.257, "step": 72866 }, { "epoch": 4.4, "learning_rate": 7.5445505711224785e-06, "loss": 0.2219, "step": 72868 }, { "epoch": 4.4, "learning_rate": 7.543041661008254e-06, "loss": 0.2315, "step": 72870 }, { "epoch": 4.4, "learning_rate": 7.54153275089403e-06, "loss": 0.1469, "step": 72872 }, { "epoch": 4.4, "learning_rate": 7.5400238407798055e-06, "loss": 0.2724, "step": 72874 }, { "epoch": 4.4, "learning_rate": 7.53851493066558e-06, "loss": 0.2081, "step": 72876 }, { "epoch": 4.4, "learning_rate": 7.537006020551356e-06, "loss": 0.1502, "step": 72878 }, { "epoch": 4.4, "learning_rate": 7.535497110437132e-06, "loss": 0.2834, "step": 72880 }, { "epoch": 4.4, "learning_rate": 7.533988200322907e-06, "loss": 0.2241, "step": 72882 }, { "epoch": 4.4, "learning_rate": 7.532479290208683e-06, "loss": 0.2464, "step": 72884 }, { "epoch": 4.4, "learning_rate": 7.530970380094459e-06, "loss": 0.2734, "step": 72886 }, { "epoch": 4.4, "learning_rate": 7.529461469980234e-06, "loss": 0.2632, "step": 72888 }, { "epoch": 4.4, "learning_rate": 7.52795255986601e-06, "loss": 0.1559, "step": 72890 }, { "epoch": 4.4, "learning_rate": 7.526443649751784e-06, "loss": 0.2106, "step": 72892 }, { "epoch": 4.4, "learning_rate": 7.52493473963756e-06, "loss": 0.2588, "step": 72894 }, { "epoch": 4.4, "learning_rate": 7.523425829523335e-06, "loss": 0.209, "step": 72896 }, { "epoch": 4.4, "learning_rate": 7.521916919409111e-06, "loss": 0.3438, "step": 72898 }, { "epoch": 4.4, "learning_rate": 7.520408009294887e-06, "loss": 0.3257, "step": 72900 }, { "epoch": 4.4, "learning_rate": 7.518899099180662e-06, "loss": 0.1451, "step": 72902 }, { "epoch": 4.4, "learning_rate": 7.517390189066438e-06, "loss": 0.2129, "step": 72904 }, { "epoch": 4.4, "learning_rate": 7.515881278952214e-06, "loss": 0.1768, "step": 72906 }, { "epoch": 4.4, "learning_rate": 7.514372368837988e-06, "loss": 0.1994, "step": 72908 }, { "epoch": 4.4, "learning_rate": 7.512863458723763e-06, "loss": 0.3131, "step": 72910 }, { "epoch": 4.4, "learning_rate": 7.511354548609539e-06, "loss": 0.205, "step": 72912 }, { "epoch": 4.4, "learning_rate": 7.509845638495315e-06, "loss": 0.2011, "step": 72914 }, { "epoch": 4.4, "learning_rate": 7.50833672838109e-06, "loss": 0.3239, "step": 72916 }, { "epoch": 4.4, "learning_rate": 7.506827818266866e-06, "loss": 0.2658, "step": 72918 }, { "epoch": 4.4, "learning_rate": 7.5053189081526425e-06, "loss": 0.1856, "step": 72920 }, { "epoch": 4.4, "learning_rate": 7.503809998038418e-06, "loss": 0.1189, "step": 72922 }, { "epoch": 4.4, "learning_rate": 7.502301087924194e-06, "loss": 0.2119, "step": 72924 }, { "epoch": 4.4, "learning_rate": 7.500792177809968e-06, "loss": 0.2962, "step": 72926 }, { "epoch": 4.4, "learning_rate": 7.4992832676957434e-06, "loss": 0.3306, "step": 72928 }, { "epoch": 4.4, "learning_rate": 7.497774357581519e-06, "loss": 0.1693, "step": 72930 }, { "epoch": 4.4, "learning_rate": 7.496265447467295e-06, "loss": 0.1811, "step": 72932 }, { "epoch": 4.4, "learning_rate": 7.4947565373530704e-06, "loss": 0.2264, "step": 72934 }, { "epoch": 4.4, "learning_rate": 7.493247627238846e-06, "loss": 0.2567, "step": 72936 }, { "epoch": 4.4, "learning_rate": 7.491738717124622e-06, "loss": 0.238, "step": 72938 }, { "epoch": 4.4, "learning_rate": 7.4902298070103974e-06, "loss": 0.1606, "step": 72940 }, { "epoch": 4.4, "learning_rate": 7.488720896896171e-06, "loss": 0.2812, "step": 72942 }, { "epoch": 4.4, "learning_rate": 7.487211986781947e-06, "loss": 0.3127, "step": 72944 }, { "epoch": 4.4, "learning_rate": 7.485703076667723e-06, "loss": 0.1886, "step": 72946 }, { "epoch": 4.4, "learning_rate": 7.484194166553498e-06, "loss": 0.2861, "step": 72948 }, { "epoch": 4.4, "learning_rate": 7.482685256439274e-06, "loss": 0.3299, "step": 72950 }, { "epoch": 4.4, "learning_rate": 7.48117634632505e-06, "loss": 0.205, "step": 72952 }, { "epoch": 4.4, "learning_rate": 7.479667436210825e-06, "loss": 0.1579, "step": 72954 }, { "epoch": 4.4, "learning_rate": 7.478158526096601e-06, "loss": 0.215, "step": 72956 }, { "epoch": 4.4, "learning_rate": 7.476649615982376e-06, "loss": 0.3833, "step": 72958 }, { "epoch": 4.4, "learning_rate": 7.4751407058681516e-06, "loss": 0.1629, "step": 72960 }, { "epoch": 4.4, "learning_rate": 7.473631795753927e-06, "loss": 0.1593, "step": 72962 }, { "epoch": 4.4, "learning_rate": 7.472122885639703e-06, "loss": 0.1933, "step": 72964 }, { "epoch": 4.4, "learning_rate": 7.4706139755254786e-06, "loss": 0.1826, "step": 72966 }, { "epoch": 4.4, "learning_rate": 7.469105065411254e-06, "loss": 0.2272, "step": 72968 }, { "epoch": 4.4, "learning_rate": 7.46759615529703e-06, "loss": 0.2965, "step": 72970 }, { "epoch": 4.4, "learning_rate": 7.4660872451828056e-06, "loss": 0.3097, "step": 72972 }, { "epoch": 4.4, "learning_rate": 7.4645783350685795e-06, "loss": 0.187, "step": 72974 }, { "epoch": 4.4, "learning_rate": 7.463069424954355e-06, "loss": 0.2125, "step": 72976 }, { "epoch": 4.4, "learning_rate": 7.461560514840131e-06, "loss": 0.2444, "step": 72978 }, { "epoch": 4.4, "learning_rate": 7.4600516047259065e-06, "loss": 0.2304, "step": 72980 }, { "epoch": 4.4, "learning_rate": 7.458542694611682e-06, "loss": 0.2803, "step": 72982 }, { "epoch": 4.41, "learning_rate": 7.457033784497458e-06, "loss": 0.1737, "step": 72984 }, { "epoch": 4.41, "learning_rate": 7.4555248743832335e-06, "loss": 0.1795, "step": 72986 }, { "epoch": 4.41, "learning_rate": 7.454015964269009e-06, "loss": 0.1555, "step": 72988 }, { "epoch": 4.41, "learning_rate": 7.452507054154784e-06, "loss": 0.1346, "step": 72990 }, { "epoch": 4.41, "learning_rate": 7.45099814404056e-06, "loss": 0.3176, "step": 72992 }, { "epoch": 4.41, "learning_rate": 7.449489233926335e-06, "loss": 0.2384, "step": 72994 }, { "epoch": 4.41, "learning_rate": 7.447980323812111e-06, "loss": 0.1871, "step": 72996 }, { "epoch": 4.41, "learning_rate": 7.446471413697887e-06, "loss": 0.1418, "step": 72998 }, { "epoch": 4.41, "learning_rate": 7.444962503583662e-06, "loss": 0.2894, "step": 73000 }, { "epoch": 4.41, "learning_rate": 7.443453593469438e-06, "loss": 0.3392, "step": 73002 }, { "epoch": 4.41, "learning_rate": 7.441944683355214e-06, "loss": 0.2223, "step": 73004 }, { "epoch": 4.41, "learning_rate": 7.440435773240989e-06, "loss": 0.1255, "step": 73006 }, { "epoch": 4.41, "learning_rate": 7.438926863126763e-06, "loss": 0.1644, "step": 73008 }, { "epoch": 4.41, "learning_rate": 7.437417953012539e-06, "loss": 0.2545, "step": 73010 }, { "epoch": 4.41, "learning_rate": 7.435909042898315e-06, "loss": 0.2694, "step": 73012 }, { "epoch": 4.41, "learning_rate": 7.43440013278409e-06, "loss": 0.1928, "step": 73014 }, { "epoch": 4.41, "learning_rate": 7.432891222669866e-06, "loss": 0.2962, "step": 73016 }, { "epoch": 4.41, "learning_rate": 7.431382312555642e-06, "loss": 0.274, "step": 73018 }, { "epoch": 4.41, "learning_rate": 7.429873402441417e-06, "loss": 0.1659, "step": 73020 }, { "epoch": 4.41, "learning_rate": 7.428364492327193e-06, "loss": 0.1838, "step": 73022 }, { "epoch": 4.41, "learning_rate": 7.426855582212967e-06, "loss": 0.1993, "step": 73024 }, { "epoch": 4.41, "learning_rate": 7.425346672098743e-06, "loss": 0.2142, "step": 73026 }, { "epoch": 4.41, "learning_rate": 7.423837761984518e-06, "loss": 0.2205, "step": 73028 }, { "epoch": 4.41, "learning_rate": 7.422328851870294e-06, "loss": 0.1794, "step": 73030 }, { "epoch": 4.41, "learning_rate": 7.42081994175607e-06, "loss": 0.1945, "step": 73032 }, { "epoch": 4.41, "learning_rate": 7.419311031641845e-06, "loss": 0.1473, "step": 73034 }, { "epoch": 4.41, "learning_rate": 7.417802121527621e-06, "loss": 0.2049, "step": 73036 }, { "epoch": 4.41, "learning_rate": 7.4162932114133975e-06, "loss": 0.1903, "step": 73038 }, { "epoch": 4.41, "learning_rate": 7.4147843012991715e-06, "loss": 0.2537, "step": 73040 }, { "epoch": 4.41, "learning_rate": 7.413275391184947e-06, "loss": 0.2565, "step": 73042 }, { "epoch": 4.41, "learning_rate": 7.411766481070723e-06, "loss": 0.3122, "step": 73044 }, { "epoch": 4.41, "learning_rate": 7.4102575709564985e-06, "loss": 0.2314, "step": 73046 }, { "epoch": 4.41, "learning_rate": 7.408748660842274e-06, "loss": 0.2308, "step": 73048 }, { "epoch": 4.41, "learning_rate": 7.40723975072805e-06, "loss": 0.2716, "step": 73050 }, { "epoch": 4.41, "learning_rate": 7.4057308406138255e-06, "loss": 0.43, "step": 73052 }, { "epoch": 4.41, "learning_rate": 7.404221930499601e-06, "loss": 0.1426, "step": 73054 }, { "epoch": 4.41, "learning_rate": 7.402713020385375e-06, "loss": 0.2726, "step": 73056 }, { "epoch": 4.41, "learning_rate": 7.401204110271151e-06, "loss": 0.1778, "step": 73058 }, { "epoch": 4.41, "learning_rate": 7.3996952001569264e-06, "loss": 0.1999, "step": 73060 }, { "epoch": 4.41, "learning_rate": 7.398186290042702e-06, "loss": 0.2171, "step": 73062 }, { "epoch": 4.41, "learning_rate": 7.396677379928478e-06, "loss": 0.1793, "step": 73064 }, { "epoch": 4.41, "learning_rate": 7.3951684698142534e-06, "loss": 0.2031, "step": 73066 }, { "epoch": 4.41, "learning_rate": 7.393659559700029e-06, "loss": 0.2251, "step": 73068 }, { "epoch": 4.41, "learning_rate": 7.392150649585805e-06, "loss": 0.2785, "step": 73070 }, { "epoch": 4.41, "learning_rate": 7.39064173947158e-06, "loss": 0.2001, "step": 73072 }, { "epoch": 4.41, "learning_rate": 7.389132829357355e-06, "loss": 0.1963, "step": 73074 }, { "epoch": 4.41, "learning_rate": 7.387623919243131e-06, "loss": 0.2305, "step": 73076 }, { "epoch": 4.41, "learning_rate": 7.386115009128907e-06, "loss": 0.2437, "step": 73078 }, { "epoch": 4.41, "learning_rate": 7.384606099014682e-06, "loss": 0.1648, "step": 73080 }, { "epoch": 4.41, "learning_rate": 7.383097188900458e-06, "loss": 0.1648, "step": 73082 }, { "epoch": 4.41, "learning_rate": 7.3815882787862336e-06, "loss": 0.314, "step": 73084 }, { "epoch": 4.41, "learning_rate": 7.380079368672009e-06, "loss": 0.1943, "step": 73086 }, { "epoch": 4.41, "learning_rate": 7.378570458557785e-06, "loss": 0.1355, "step": 73088 }, { "epoch": 4.41, "learning_rate": 7.377061548443559e-06, "loss": 0.2212, "step": 73090 }, { "epoch": 4.41, "learning_rate": 7.3755526383293345e-06, "loss": 0.2447, "step": 73092 }, { "epoch": 4.41, "learning_rate": 7.37404372821511e-06, "loss": 0.2785, "step": 73094 }, { "epoch": 4.41, "learning_rate": 7.372534818100886e-06, "loss": 0.1987, "step": 73096 }, { "epoch": 4.41, "learning_rate": 7.3710259079866615e-06, "loss": 0.2055, "step": 73098 }, { "epoch": 4.41, "learning_rate": 7.369516997872437e-06, "loss": 0.3548, "step": 73100 }, { "epoch": 4.41, "learning_rate": 7.368008087758213e-06, "loss": 0.3167, "step": 73102 }, { "epoch": 4.41, "learning_rate": 7.3664991776439885e-06, "loss": 0.1584, "step": 73104 }, { "epoch": 4.41, "learning_rate": 7.364990267529763e-06, "loss": 0.2664, "step": 73106 }, { "epoch": 4.41, "learning_rate": 7.363481357415539e-06, "loss": 0.267, "step": 73108 }, { "epoch": 4.41, "learning_rate": 7.361972447301315e-06, "loss": 0.3402, "step": 73110 }, { "epoch": 4.41, "learning_rate": 7.36046353718709e-06, "loss": 0.3307, "step": 73112 }, { "epoch": 4.41, "learning_rate": 7.358954627072866e-06, "loss": 0.2087, "step": 73114 }, { "epoch": 4.41, "learning_rate": 7.357445716958642e-06, "loss": 0.2239, "step": 73116 }, { "epoch": 4.41, "learning_rate": 7.355936806844417e-06, "loss": 0.223, "step": 73118 }, { "epoch": 4.41, "learning_rate": 7.354427896730193e-06, "loss": 0.2414, "step": 73120 }, { "epoch": 4.41, "learning_rate": 7.352918986615967e-06, "loss": 0.1909, "step": 73122 }, { "epoch": 4.41, "learning_rate": 7.351410076501743e-06, "loss": 0.287, "step": 73124 }, { "epoch": 4.41, "learning_rate": 7.349901166387518e-06, "loss": 0.195, "step": 73126 }, { "epoch": 4.41, "learning_rate": 7.348392256273294e-06, "loss": 0.1649, "step": 73128 }, { "epoch": 4.41, "learning_rate": 7.34688334615907e-06, "loss": 0.3783, "step": 73130 }, { "epoch": 4.41, "learning_rate": 7.345374436044845e-06, "loss": 0.2173, "step": 73132 }, { "epoch": 4.41, "learning_rate": 7.343865525930621e-06, "loss": 0.3264, "step": 73134 }, { "epoch": 4.41, "learning_rate": 7.342356615816397e-06, "loss": 0.2712, "step": 73136 }, { "epoch": 4.41, "learning_rate": 7.340847705702171e-06, "loss": 0.1852, "step": 73138 }, { "epoch": 4.41, "learning_rate": 7.339338795587946e-06, "loss": 0.1992, "step": 73140 }, { "epoch": 4.41, "learning_rate": 7.337829885473722e-06, "loss": 0.2428, "step": 73142 }, { "epoch": 4.41, "learning_rate": 7.336320975359498e-06, "loss": 0.2954, "step": 73144 }, { "epoch": 4.41, "learning_rate": 7.334812065245273e-06, "loss": 0.2926, "step": 73146 }, { "epoch": 4.42, "learning_rate": 7.333303155131049e-06, "loss": 0.2845, "step": 73148 }, { "epoch": 4.42, "learning_rate": 7.331794245016825e-06, "loss": 0.3138, "step": 73150 }, { "epoch": 4.42, "learning_rate": 7.3302853349026e-06, "loss": 0.1824, "step": 73152 }, { "epoch": 4.42, "learning_rate": 7.328776424788376e-06, "loss": 0.1779, "step": 73154 }, { "epoch": 4.42, "learning_rate": 7.327267514674151e-06, "loss": 0.1829, "step": 73156 }, { "epoch": 4.42, "learning_rate": 7.3257586045599265e-06, "loss": 0.08, "step": 73158 }, { "epoch": 4.42, "learning_rate": 7.324249694445702e-06, "loss": 0.2111, "step": 73160 }, { "epoch": 4.42, "learning_rate": 7.322740784331478e-06, "loss": 0.2942, "step": 73162 }, { "epoch": 4.42, "learning_rate": 7.3212318742172535e-06, "loss": 0.1521, "step": 73164 }, { "epoch": 4.42, "learning_rate": 7.319722964103029e-06, "loss": 0.2412, "step": 73166 }, { "epoch": 4.42, "learning_rate": 7.318214053988805e-06, "loss": 0.2495, "step": 73168 }, { "epoch": 4.42, "learning_rate": 7.3167051438745805e-06, "loss": 0.2164, "step": 73170 }, { "epoch": 4.42, "learning_rate": 7.3151962337603544e-06, "loss": 0.2274, "step": 73172 }, { "epoch": 4.42, "learning_rate": 7.31368732364613e-06, "loss": 0.1946, "step": 73174 }, { "epoch": 4.42, "learning_rate": 7.312178413531906e-06, "loss": 0.2436, "step": 73176 }, { "epoch": 4.42, "learning_rate": 7.3106695034176814e-06, "loss": 0.2138, "step": 73178 }, { "epoch": 4.42, "learning_rate": 7.309160593303457e-06, "loss": 0.204, "step": 73180 }, { "epoch": 4.42, "learning_rate": 7.307651683189233e-06, "loss": 0.24, "step": 73182 }, { "epoch": 4.42, "learning_rate": 7.3061427730750084e-06, "loss": 0.3278, "step": 73184 }, { "epoch": 4.42, "learning_rate": 7.304633862960784e-06, "loss": 0.2022, "step": 73186 }, { "epoch": 4.42, "learning_rate": 7.303124952846559e-06, "loss": 0.2885, "step": 73188 }, { "epoch": 4.42, "learning_rate": 7.301616042732335e-06, "loss": 0.1355, "step": 73190 }, { "epoch": 4.42, "learning_rate": 7.30010713261811e-06, "loss": 0.2127, "step": 73192 }, { "epoch": 4.42, "learning_rate": 7.298598222503886e-06, "loss": 0.2787, "step": 73194 }, { "epoch": 4.42, "learning_rate": 7.297089312389662e-06, "loss": 0.284, "step": 73196 }, { "epoch": 4.42, "learning_rate": 7.295580402275437e-06, "loss": 0.2057, "step": 73198 }, { "epoch": 4.42, "learning_rate": 7.294071492161213e-06, "loss": 0.2688, "step": 73200 }, { "epoch": 4.42, "learning_rate": 7.292562582046989e-06, "loss": 0.3058, "step": 73202 }, { "epoch": 4.42, "learning_rate": 7.2910536719327626e-06, "loss": 0.2883, "step": 73204 }, { "epoch": 4.42, "learning_rate": 7.289544761818538e-06, "loss": 0.2419, "step": 73206 }, { "epoch": 4.42, "learning_rate": 7.288035851704314e-06, "loss": 0.1478, "step": 73208 }, { "epoch": 4.42, "learning_rate": 7.2865269415900896e-06, "loss": 0.3015, "step": 73210 }, { "epoch": 4.42, "learning_rate": 7.285018031475865e-06, "loss": 0.2439, "step": 73212 }, { "epoch": 4.42, "learning_rate": 7.283509121361641e-06, "loss": 0.1471, "step": 73214 }, { "epoch": 4.42, "learning_rate": 7.2820002112474166e-06, "loss": 0.2085, "step": 73216 }, { "epoch": 4.42, "learning_rate": 7.280491301133192e-06, "loss": 0.2459, "step": 73218 }, { "epoch": 4.42, "learning_rate": 7.278982391018967e-06, "loss": 0.2124, "step": 73220 }, { "epoch": 4.42, "learning_rate": 7.277473480904743e-06, "loss": 0.2621, "step": 73222 }, { "epoch": 4.42, "learning_rate": 7.275964570790518e-06, "loss": 0.1931, "step": 73224 }, { "epoch": 4.42, "learning_rate": 7.274455660676294e-06, "loss": 0.1896, "step": 73226 }, { "epoch": 4.42, "learning_rate": 7.27294675056207e-06, "loss": 0.1929, "step": 73228 }, { "epoch": 4.42, "learning_rate": 7.271437840447845e-06, "loss": 0.2516, "step": 73230 }, { "epoch": 4.42, "learning_rate": 7.269928930333621e-06, "loss": 0.1737, "step": 73232 }, { "epoch": 4.42, "learning_rate": 7.268420020219397e-06, "loss": 0.276, "step": 73234 }, { "epoch": 4.42, "learning_rate": 7.266911110105172e-06, "loss": 0.2618, "step": 73236 }, { "epoch": 4.42, "learning_rate": 7.265402199990946e-06, "loss": 0.249, "step": 73238 }, { "epoch": 4.42, "learning_rate": 7.263893289876722e-06, "loss": 0.1766, "step": 73240 }, { "epoch": 4.42, "learning_rate": 7.262384379762498e-06, "loss": 0.2162, "step": 73242 }, { "epoch": 4.42, "learning_rate": 7.260875469648273e-06, "loss": 0.2957, "step": 73244 }, { "epoch": 4.42, "learning_rate": 7.259366559534049e-06, "loss": 0.1888, "step": 73246 }, { "epoch": 4.42, "learning_rate": 7.257857649419825e-06, "loss": 0.1815, "step": 73248 }, { "epoch": 4.42, "learning_rate": 7.2563487393056e-06, "loss": 0.2596, "step": 73250 }, { "epoch": 4.42, "learning_rate": 7.254839829191376e-06, "loss": 0.2074, "step": 73252 }, { "epoch": 4.42, "learning_rate": 7.25333091907715e-06, "loss": 0.3009, "step": 73254 }, { "epoch": 4.42, "learning_rate": 7.251822008962926e-06, "loss": 0.1571, "step": 73256 }, { "epoch": 4.42, "learning_rate": 7.250313098848701e-06, "loss": 0.204, "step": 73258 }, { "epoch": 4.42, "learning_rate": 7.248804188734477e-06, "loss": 0.1778, "step": 73260 }, { "epoch": 4.42, "learning_rate": 7.247295278620253e-06, "loss": 0.2618, "step": 73262 }, { "epoch": 4.42, "learning_rate": 7.245786368506028e-06, "loss": 0.2664, "step": 73264 }, { "epoch": 4.42, "learning_rate": 7.244277458391804e-06, "loss": 0.2591, "step": 73266 }, { "epoch": 4.42, "learning_rate": 7.24276854827758e-06, "loss": 0.311, "step": 73268 }, { "epoch": 4.42, "learning_rate": 7.2412596381633545e-06, "loss": 0.302, "step": 73270 }, { "epoch": 4.42, "learning_rate": 7.23975072804913e-06, "loss": 0.2185, "step": 73272 }, { "epoch": 4.42, "learning_rate": 7.238241817934906e-06, "loss": 0.1997, "step": 73274 }, { "epoch": 4.42, "learning_rate": 7.2367329078206815e-06, "loss": 0.1526, "step": 73276 }, { "epoch": 4.42, "learning_rate": 7.235223997706457e-06, "loss": 0.4416, "step": 73278 }, { "epoch": 4.42, "learning_rate": 7.233715087592233e-06, "loss": 0.2186, "step": 73280 }, { "epoch": 4.42, "learning_rate": 7.2322061774780085e-06, "loss": 0.2544, "step": 73282 }, { "epoch": 4.42, "learning_rate": 7.230697267363784e-06, "loss": 0.2229, "step": 73284 }, { "epoch": 4.42, "learning_rate": 7.229188357249558e-06, "loss": 0.3028, "step": 73286 }, { "epoch": 4.42, "learning_rate": 7.227679447135334e-06, "loss": 0.2377, "step": 73288 }, { "epoch": 4.42, "learning_rate": 7.2261705370211094e-06, "loss": 0.2078, "step": 73290 }, { "epoch": 4.42, "learning_rate": 7.224661626906885e-06, "loss": 0.2868, "step": 73292 }, { "epoch": 4.42, "learning_rate": 7.223152716792661e-06, "loss": 0.3043, "step": 73294 }, { "epoch": 4.42, "learning_rate": 7.2216438066784364e-06, "loss": 0.1558, "step": 73296 }, { "epoch": 4.42, "learning_rate": 7.220134896564212e-06, "loss": 0.2472, "step": 73298 }, { "epoch": 4.42, "learning_rate": 7.218625986449988e-06, "loss": 0.2175, "step": 73300 }, { "epoch": 4.42, "learning_rate": 7.217117076335763e-06, "loss": 0.2967, "step": 73302 }, { "epoch": 4.42, "learning_rate": 7.215608166221538e-06, "loss": 0.2845, "step": 73304 }, { "epoch": 4.42, "learning_rate": 7.214099256107314e-06, "loss": 0.3043, "step": 73306 }, { "epoch": 4.42, "learning_rate": 7.21259034599309e-06, "loss": 0.1507, "step": 73308 }, { "epoch": 4.42, "learning_rate": 7.211081435878865e-06, "loss": 0.2384, "step": 73310 }, { "epoch": 4.42, "learning_rate": 7.209572525764641e-06, "loss": 0.2465, "step": 73312 }, { "epoch": 4.43, "learning_rate": 7.208063615650417e-06, "loss": 0.1715, "step": 73314 }, { "epoch": 4.43, "learning_rate": 7.206554705536192e-06, "loss": 0.2054, "step": 73316 }, { "epoch": 4.43, "learning_rate": 7.205045795421968e-06, "loss": 0.1757, "step": 73318 }, { "epoch": 4.43, "learning_rate": 7.203536885307742e-06, "loss": 0.1592, "step": 73320 }, { "epoch": 4.43, "learning_rate": 7.2020279751935176e-06, "loss": 0.197, "step": 73322 }, { "epoch": 4.43, "learning_rate": 7.200519065079293e-06, "loss": 0.3525, "step": 73324 }, { "epoch": 4.43, "learning_rate": 7.199010154965069e-06, "loss": 0.2326, "step": 73326 }, { "epoch": 4.43, "learning_rate": 7.1975012448508446e-06, "loss": 0.2757, "step": 73328 }, { "epoch": 4.43, "learning_rate": 7.19599233473662e-06, "loss": 0.2158, "step": 73330 }, { "epoch": 4.43, "learning_rate": 7.194483424622396e-06, "loss": 0.2963, "step": 73332 }, { "epoch": 4.43, "learning_rate": 7.1929745145081716e-06, "loss": 0.1849, "step": 73334 }, { "epoch": 4.43, "learning_rate": 7.191465604393946e-06, "loss": 0.3959, "step": 73336 }, { "epoch": 4.43, "learning_rate": 7.189956694279722e-06, "loss": 0.205, "step": 73338 }, { "epoch": 4.43, "learning_rate": 7.188447784165498e-06, "loss": 0.1392, "step": 73340 }, { "epoch": 4.43, "learning_rate": 7.186938874051273e-06, "loss": 0.1685, "step": 73342 }, { "epoch": 4.43, "learning_rate": 7.185429963937049e-06, "loss": 0.2558, "step": 73344 }, { "epoch": 4.43, "learning_rate": 7.183921053822825e-06, "loss": 0.1234, "step": 73346 }, { "epoch": 4.43, "learning_rate": 7.1824121437086e-06, "loss": 0.2111, "step": 73348 }, { "epoch": 4.43, "learning_rate": 7.180903233594376e-06, "loss": 0.1835, "step": 73350 }, { "epoch": 4.43, "learning_rate": 7.17939432348015e-06, "loss": 0.1776, "step": 73352 }, { "epoch": 4.43, "learning_rate": 7.177885413365926e-06, "loss": 0.2646, "step": 73354 }, { "epoch": 4.43, "learning_rate": 7.176376503251701e-06, "loss": 0.2409, "step": 73356 }, { "epoch": 4.43, "learning_rate": 7.174867593137477e-06, "loss": 0.2302, "step": 73358 }, { "epoch": 4.43, "learning_rate": 7.173358683023253e-06, "loss": 0.3331, "step": 73360 }, { "epoch": 4.43, "learning_rate": 7.171849772909028e-06, "loss": 0.1728, "step": 73362 }, { "epoch": 4.43, "learning_rate": 7.170340862794804e-06, "loss": 0.2147, "step": 73364 }, { "epoch": 4.43, "learning_rate": 7.16883195268058e-06, "loss": 0.2362, "step": 73366 }, { "epoch": 4.43, "learning_rate": 7.167323042566354e-06, "loss": 0.215, "step": 73368 }, { "epoch": 4.43, "learning_rate": 7.165814132452129e-06, "loss": 0.159, "step": 73370 }, { "epoch": 4.43, "learning_rate": 7.164305222337905e-06, "loss": 0.4279, "step": 73372 }, { "epoch": 4.43, "learning_rate": 7.162796312223681e-06, "loss": 0.2651, "step": 73374 }, { "epoch": 4.43, "learning_rate": 7.161287402109456e-06, "loss": 0.3204, "step": 73376 }, { "epoch": 4.43, "learning_rate": 7.159778491995232e-06, "loss": 0.3418, "step": 73378 }, { "epoch": 4.43, "learning_rate": 7.158269581881008e-06, "loss": 0.2959, "step": 73380 }, { "epoch": 4.43, "learning_rate": 7.156760671766783e-06, "loss": 0.2672, "step": 73382 }, { "epoch": 4.43, "learning_rate": 7.155251761652559e-06, "loss": 0.1296, "step": 73384 }, { "epoch": 4.43, "learning_rate": 7.153742851538334e-06, "loss": 0.2256, "step": 73386 }, { "epoch": 4.43, "learning_rate": 7.1522339414241095e-06, "loss": 0.1834, "step": 73388 }, { "epoch": 4.43, "learning_rate": 7.150725031309885e-06, "loss": 0.2714, "step": 73390 }, { "epoch": 4.43, "learning_rate": 7.149216121195661e-06, "loss": 0.2013, "step": 73392 }, { "epoch": 4.43, "learning_rate": 7.1477072110814365e-06, "loss": 0.3587, "step": 73394 }, { "epoch": 4.43, "learning_rate": 7.146198300967212e-06, "loss": 0.23, "step": 73396 }, { "epoch": 4.43, "learning_rate": 7.144689390852988e-06, "loss": 0.2218, "step": 73398 }, { "epoch": 4.43, "learning_rate": 7.1431804807387635e-06, "loss": 0.2202, "step": 73400 }, { "epoch": 4.43, "learning_rate": 7.1416715706245375e-06, "loss": 0.2133, "step": 73402 }, { "epoch": 4.43, "learning_rate": 7.140162660510313e-06, "loss": 0.1614, "step": 73404 }, { "epoch": 4.43, "learning_rate": 7.138653750396089e-06, "loss": 0.1336, "step": 73406 }, { "epoch": 4.43, "learning_rate": 7.1371448402818645e-06, "loss": 0.1328, "step": 73408 }, { "epoch": 4.43, "learning_rate": 7.13563593016764e-06, "loss": 0.3729, "step": 73410 }, { "epoch": 4.43, "learning_rate": 7.134127020053416e-06, "loss": 0.2899, "step": 73412 }, { "epoch": 4.43, "learning_rate": 7.1326181099391915e-06, "loss": 0.2637, "step": 73414 }, { "epoch": 4.43, "learning_rate": 7.131109199824967e-06, "loss": 0.1474, "step": 73416 }, { "epoch": 4.43, "learning_rate": 7.129600289710742e-06, "loss": 0.1506, "step": 73418 }, { "epoch": 4.43, "learning_rate": 7.128091379596518e-06, "loss": 0.3073, "step": 73420 }, { "epoch": 4.43, "learning_rate": 7.126582469482293e-06, "loss": 0.236, "step": 73422 }, { "epoch": 4.43, "learning_rate": 7.125073559368069e-06, "loss": 0.1792, "step": 73424 }, { "epoch": 4.43, "learning_rate": 7.123564649253845e-06, "loss": 0.3317, "step": 73426 }, { "epoch": 4.43, "learning_rate": 7.12205573913962e-06, "loss": 0.1922, "step": 73428 }, { "epoch": 4.43, "learning_rate": 7.120546829025396e-06, "loss": 0.2293, "step": 73430 }, { "epoch": 4.43, "learning_rate": 7.119037918911172e-06, "loss": 0.235, "step": 73432 }, { "epoch": 4.43, "learning_rate": 7.117529008796946e-06, "loss": 0.1861, "step": 73434 }, { "epoch": 4.43, "learning_rate": 7.116020098682721e-06, "loss": 0.2999, "step": 73436 }, { "epoch": 4.43, "learning_rate": 7.114511188568497e-06, "loss": 0.1919, "step": 73438 }, { "epoch": 4.43, "learning_rate": 7.113002278454273e-06, "loss": 0.335, "step": 73440 }, { "epoch": 4.43, "learning_rate": 7.111493368340048e-06, "loss": 0.2629, "step": 73442 }, { "epoch": 4.43, "learning_rate": 7.109984458225824e-06, "loss": 0.185, "step": 73444 }, { "epoch": 4.43, "learning_rate": 7.1084755481116e-06, "loss": 0.2197, "step": 73446 }, { "epoch": 4.43, "learning_rate": 7.106966637997375e-06, "loss": 0.2906, "step": 73448 }, { "epoch": 4.43, "learning_rate": 7.10545772788315e-06, "loss": 0.2302, "step": 73450 }, { "epoch": 4.43, "learning_rate": 7.103948817768926e-06, "loss": 0.264, "step": 73452 }, { "epoch": 4.43, "learning_rate": 7.102439907654701e-06, "loss": 0.1566, "step": 73454 }, { "epoch": 4.43, "learning_rate": 7.100930997540477e-06, "loss": 0.2442, "step": 73456 }, { "epoch": 4.43, "learning_rate": 7.099422087426253e-06, "loss": 0.1623, "step": 73458 }, { "epoch": 4.43, "learning_rate": 7.097913177312028e-06, "loss": 0.1666, "step": 73460 }, { "epoch": 4.43, "learning_rate": 7.096404267197804e-06, "loss": 0.1472, "step": 73462 }, { "epoch": 4.43, "learning_rate": 7.09489535708358e-06, "loss": 0.1704, "step": 73464 }, { "epoch": 4.43, "learning_rate": 7.093386446969355e-06, "loss": 0.2544, "step": 73466 }, { "epoch": 4.43, "learning_rate": 7.091877536855129e-06, "loss": 0.2503, "step": 73468 }, { "epoch": 4.43, "learning_rate": 7.090368626740905e-06, "loss": 0.1655, "step": 73470 }, { "epoch": 4.43, "learning_rate": 7.088859716626681e-06, "loss": 0.1751, "step": 73472 }, { "epoch": 4.43, "learning_rate": 7.087350806512456e-06, "loss": 0.1547, "step": 73474 }, { "epoch": 4.43, "learning_rate": 7.085841896398232e-06, "loss": 0.2265, "step": 73476 }, { "epoch": 4.43, "learning_rate": 7.084332986284008e-06, "loss": 0.3232, "step": 73478 }, { "epoch": 4.44, "learning_rate": 7.082824076169783e-06, "loss": 0.3544, "step": 73480 }, { "epoch": 4.44, "learning_rate": 7.081315166055559e-06, "loss": 0.1863, "step": 73482 }, { "epoch": 4.44, "learning_rate": 7.079806255941333e-06, "loss": 0.2984, "step": 73484 }, { "epoch": 4.44, "learning_rate": 7.078297345827109e-06, "loss": 0.2017, "step": 73486 }, { "epoch": 4.44, "learning_rate": 7.076788435712884e-06, "loss": 0.3085, "step": 73488 }, { "epoch": 4.44, "learning_rate": 7.07527952559866e-06, "loss": 0.3389, "step": 73490 }, { "epoch": 4.44, "learning_rate": 7.073770615484436e-06, "loss": 0.1694, "step": 73492 }, { "epoch": 4.44, "learning_rate": 7.072261705370211e-06, "loss": 0.1952, "step": 73494 }, { "epoch": 4.44, "learning_rate": 7.070752795255987e-06, "loss": 0.1644, "step": 73496 }, { "epoch": 4.44, "learning_rate": 7.069243885141763e-06, "loss": 0.1724, "step": 73498 }, { "epoch": 4.44, "learning_rate": 7.0677349750275375e-06, "loss": 0.2063, "step": 73500 }, { "epoch": 4.44, "learning_rate": 7.066226064913313e-06, "loss": 0.3695, "step": 73502 }, { "epoch": 4.44, "learning_rate": 7.064717154799089e-06, "loss": 0.285, "step": 73504 }, { "epoch": 4.44, "learning_rate": 7.0632082446848645e-06, "loss": 0.1946, "step": 73506 }, { "epoch": 4.44, "learning_rate": 7.06169933457064e-06, "loss": 0.1645, "step": 73508 }, { "epoch": 4.44, "learning_rate": 7.060190424456416e-06, "loss": 0.2608, "step": 73510 }, { "epoch": 4.44, "learning_rate": 7.0586815143421915e-06, "loss": 0.1849, "step": 73512 }, { "epoch": 4.44, "learning_rate": 7.057172604227967e-06, "loss": 0.2552, "step": 73514 }, { "epoch": 4.44, "learning_rate": 7.055663694113741e-06, "loss": 0.2615, "step": 73516 }, { "epoch": 4.44, "learning_rate": 7.054154783999517e-06, "loss": 0.2131, "step": 73518 }, { "epoch": 4.44, "learning_rate": 7.0526458738852925e-06, "loss": 0.1589, "step": 73520 }, { "epoch": 4.44, "learning_rate": 7.051136963771068e-06, "loss": 0.175, "step": 73522 }, { "epoch": 4.44, "learning_rate": 7.049628053656844e-06, "loss": 0.2778, "step": 73524 }, { "epoch": 4.44, "learning_rate": 7.0481191435426195e-06, "loss": 0.17, "step": 73526 }, { "epoch": 4.44, "learning_rate": 7.046610233428395e-06, "loss": 0.1926, "step": 73528 }, { "epoch": 4.44, "learning_rate": 7.045101323314171e-06, "loss": 0.1195, "step": 73530 }, { "epoch": 4.44, "learning_rate": 7.043592413199946e-06, "loss": 0.2472, "step": 73532 }, { "epoch": 4.44, "learning_rate": 7.042083503085721e-06, "loss": 0.1662, "step": 73534 }, { "epoch": 4.44, "learning_rate": 7.040574592971497e-06, "loss": 0.2386, "step": 73536 }, { "epoch": 4.44, "learning_rate": 7.039065682857273e-06, "loss": 0.1766, "step": 73538 }, { "epoch": 4.44, "learning_rate": 7.037556772743048e-06, "loss": 0.2501, "step": 73540 }, { "epoch": 4.44, "learning_rate": 7.036047862628824e-06, "loss": 0.26, "step": 73542 }, { "epoch": 4.44, "learning_rate": 7.0345389525146e-06, "loss": 0.2046, "step": 73544 }, { "epoch": 4.44, "learning_rate": 7.033030042400375e-06, "loss": 0.4305, "step": 73546 }, { "epoch": 4.44, "learning_rate": 7.031521132286151e-06, "loss": 0.3207, "step": 73548 }, { "epoch": 4.44, "learning_rate": 7.030012222171925e-06, "loss": 0.3714, "step": 73550 }, { "epoch": 4.44, "learning_rate": 7.028503312057701e-06, "loss": 0.2523, "step": 73552 }, { "epoch": 4.44, "learning_rate": 7.026994401943476e-06, "loss": 0.342, "step": 73554 }, { "epoch": 4.44, "learning_rate": 7.025485491829252e-06, "loss": 0.2248, "step": 73556 }, { "epoch": 4.44, "learning_rate": 7.023976581715028e-06, "loss": 0.3174, "step": 73558 }, { "epoch": 4.44, "learning_rate": 7.022467671600803e-06, "loss": 0.3127, "step": 73560 }, { "epoch": 4.44, "learning_rate": 7.020958761486579e-06, "loss": 0.1344, "step": 73562 }, { "epoch": 4.44, "learning_rate": 7.019449851372355e-06, "loss": 0.2642, "step": 73564 }, { "epoch": 4.44, "learning_rate": 7.017940941258129e-06, "loss": 0.1661, "step": 73566 }, { "epoch": 4.44, "learning_rate": 7.016432031143905e-06, "loss": 0.2837, "step": 73568 }, { "epoch": 4.44, "learning_rate": 7.014923121029681e-06, "loss": 0.2601, "step": 73570 }, { "epoch": 4.44, "learning_rate": 7.013414210915456e-06, "loss": 0.1328, "step": 73572 }, { "epoch": 4.44, "learning_rate": 7.011905300801232e-06, "loss": 0.213, "step": 73574 }, { "epoch": 4.44, "learning_rate": 7.010396390687008e-06, "loss": 0.1635, "step": 73576 }, { "epoch": 4.44, "learning_rate": 7.008887480572783e-06, "loss": 0.2018, "step": 73578 }, { "epoch": 4.44, "learning_rate": 7.007378570458559e-06, "loss": 0.2071, "step": 73580 }, { "epoch": 4.44, "learning_rate": 7.005869660344333e-06, "loss": 0.2401, "step": 73582 }, { "epoch": 4.44, "learning_rate": 7.004360750230109e-06, "loss": 0.3825, "step": 73584 }, { "epoch": 4.44, "learning_rate": 7.002851840115884e-06, "loss": 0.2785, "step": 73586 }, { "epoch": 4.44, "learning_rate": 7.00134293000166e-06, "loss": 0.1922, "step": 73588 }, { "epoch": 4.44, "learning_rate": 6.999834019887436e-06, "loss": 0.261, "step": 73590 }, { "epoch": 4.44, "learning_rate": 6.998325109773211e-06, "loss": 0.2224, "step": 73592 }, { "epoch": 4.44, "learning_rate": 6.996816199658987e-06, "loss": 0.2114, "step": 73594 }, { "epoch": 4.44, "learning_rate": 6.995307289544763e-06, "loss": 0.3134, "step": 73596 }, { "epoch": 4.44, "learning_rate": 6.993798379430537e-06, "loss": 0.2004, "step": 73598 }, { "epoch": 4.44, "learning_rate": 6.992289469316312e-06, "loss": 0.2315, "step": 73600 }, { "epoch": 4.44, "learning_rate": 6.990780559202088e-06, "loss": 0.3595, "step": 73602 }, { "epoch": 4.44, "learning_rate": 6.989271649087864e-06, "loss": 0.2215, "step": 73604 }, { "epoch": 4.44, "learning_rate": 6.987762738973639e-06, "loss": 0.1144, "step": 73606 }, { "epoch": 4.44, "learning_rate": 6.986253828859415e-06, "loss": 0.34, "step": 73608 }, { "epoch": 4.44, "learning_rate": 6.984744918745191e-06, "loss": 0.2298, "step": 73610 }, { "epoch": 4.44, "learning_rate": 6.983236008630966e-06, "loss": 0.1905, "step": 73612 }, { "epoch": 4.44, "learning_rate": 6.981727098516741e-06, "loss": 0.222, "step": 73614 }, { "epoch": 4.44, "learning_rate": 6.980218188402517e-06, "loss": 0.2163, "step": 73616 }, { "epoch": 4.44, "learning_rate": 6.9787092782882925e-06, "loss": 0.217, "step": 73618 }, { "epoch": 4.44, "learning_rate": 6.977200368174068e-06, "loss": 0.2144, "step": 73620 }, { "epoch": 4.44, "learning_rate": 6.975691458059844e-06, "loss": 0.1804, "step": 73622 }, { "epoch": 4.44, "learning_rate": 6.9741825479456195e-06, "loss": 0.1765, "step": 73624 }, { "epoch": 4.44, "learning_rate": 6.972673637831395e-06, "loss": 0.2018, "step": 73626 }, { "epoch": 4.44, "learning_rate": 6.971164727717171e-06, "loss": 0.2251, "step": 73628 }, { "epoch": 4.44, "learning_rate": 6.9696558176029465e-06, "loss": 0.2176, "step": 73630 }, { "epoch": 4.44, "learning_rate": 6.9681469074887205e-06, "loss": 0.2661, "step": 73632 }, { "epoch": 4.44, "learning_rate": 6.966637997374496e-06, "loss": 0.2177, "step": 73634 }, { "epoch": 4.44, "learning_rate": 6.965129087260272e-06, "loss": 0.1981, "step": 73636 }, { "epoch": 4.44, "learning_rate": 6.9636201771460475e-06, "loss": 0.177, "step": 73638 }, { "epoch": 4.44, "learning_rate": 6.962111267031823e-06, "loss": 0.2257, "step": 73640 }, { "epoch": 4.44, "learning_rate": 6.960602356917599e-06, "loss": 0.2167, "step": 73642 }, { "epoch": 4.44, "learning_rate": 6.9590934468033745e-06, "loss": 0.3707, "step": 73644 }, { "epoch": 4.45, "learning_rate": 6.95758453668915e-06, "loss": 0.4582, "step": 73646 }, { "epoch": 4.45, "learning_rate": 6.956075626574925e-06, "loss": 0.158, "step": 73648 }, { "epoch": 4.45, "learning_rate": 6.954566716460701e-06, "loss": 0.2617, "step": 73650 }, { "epoch": 4.45, "learning_rate": 6.953057806346476e-06, "loss": 0.2857, "step": 73652 }, { "epoch": 4.45, "learning_rate": 6.951548896232252e-06, "loss": 0.2357, "step": 73654 }, { "epoch": 4.45, "learning_rate": 6.950039986118028e-06, "loss": 0.1377, "step": 73656 }, { "epoch": 4.45, "learning_rate": 6.948531076003803e-06, "loss": 0.2134, "step": 73658 }, { "epoch": 4.45, "learning_rate": 6.947022165889579e-06, "loss": 0.2601, "step": 73660 }, { "epoch": 4.45, "learning_rate": 6.945513255775355e-06, "loss": 0.3156, "step": 73662 }, { "epoch": 4.45, "learning_rate": 6.944004345661129e-06, "loss": 0.3121, "step": 73664 }, { "epoch": 4.45, "learning_rate": 6.942495435546904e-06, "loss": 0.1274, "step": 73666 }, { "epoch": 4.45, "learning_rate": 6.94098652543268e-06, "loss": 0.2191, "step": 73668 }, { "epoch": 4.45, "learning_rate": 6.939477615318456e-06, "loss": 0.282, "step": 73670 }, { "epoch": 4.45, "learning_rate": 6.937968705204231e-06, "loss": 0.2091, "step": 73672 }, { "epoch": 4.45, "learning_rate": 6.936459795090007e-06, "loss": 0.2734, "step": 73674 }, { "epoch": 4.45, "learning_rate": 6.934950884975783e-06, "loss": 0.1587, "step": 73676 }, { "epoch": 4.45, "learning_rate": 6.933441974861558e-06, "loss": 0.2968, "step": 73678 }, { "epoch": 4.45, "learning_rate": 6.931933064747333e-06, "loss": 0.2457, "step": 73680 }, { "epoch": 4.45, "learning_rate": 6.930424154633109e-06, "loss": 0.2455, "step": 73682 }, { "epoch": 4.45, "learning_rate": 6.9289152445188844e-06, "loss": 0.3382, "step": 73684 }, { "epoch": 4.45, "learning_rate": 6.92740633440466e-06, "loss": 0.2068, "step": 73686 }, { "epoch": 4.45, "learning_rate": 6.925897424290436e-06, "loss": 0.2256, "step": 73688 }, { "epoch": 4.45, "learning_rate": 6.9243885141762114e-06, "loss": 0.2529, "step": 73690 }, { "epoch": 4.45, "learning_rate": 6.922879604061987e-06, "loss": 0.156, "step": 73692 }, { "epoch": 4.45, "learning_rate": 6.921370693947763e-06, "loss": 0.1418, "step": 73694 }, { "epoch": 4.45, "learning_rate": 6.9198617838335384e-06, "loss": 0.2447, "step": 73696 }, { "epoch": 4.45, "learning_rate": 6.918352873719312e-06, "loss": 0.3268, "step": 73698 }, { "epoch": 4.45, "learning_rate": 6.916843963605088e-06, "loss": 0.1411, "step": 73700 }, { "epoch": 4.45, "learning_rate": 6.915335053490864e-06, "loss": 0.1424, "step": 73702 }, { "epoch": 4.45, "learning_rate": 6.913826143376639e-06, "loss": 0.2358, "step": 73704 }, { "epoch": 4.45, "learning_rate": 6.912317233262415e-06, "loss": 0.2485, "step": 73706 }, { "epoch": 4.45, "learning_rate": 6.910808323148191e-06, "loss": 0.2515, "step": 73708 }, { "epoch": 4.45, "learning_rate": 6.909299413033966e-06, "loss": 0.1679, "step": 73710 }, { "epoch": 4.45, "learning_rate": 6.907790502919742e-06, "loss": 0.2033, "step": 73712 }, { "epoch": 4.45, "learning_rate": 6.906281592805516e-06, "loss": 0.1965, "step": 73714 }, { "epoch": 4.45, "learning_rate": 6.904772682691292e-06, "loss": 0.2189, "step": 73716 }, { "epoch": 4.45, "learning_rate": 6.903263772577067e-06, "loss": 0.1882, "step": 73718 }, { "epoch": 4.45, "learning_rate": 6.901754862462843e-06, "loss": 0.2686, "step": 73720 }, { "epoch": 4.45, "learning_rate": 6.900245952348619e-06, "loss": 0.1544, "step": 73722 }, { "epoch": 4.45, "learning_rate": 6.898737042234394e-06, "loss": 0.2169, "step": 73724 }, { "epoch": 4.45, "learning_rate": 6.89722813212017e-06, "loss": 0.161, "step": 73726 }, { "epoch": 4.45, "learning_rate": 6.895719222005946e-06, "loss": 0.1912, "step": 73728 }, { "epoch": 4.45, "learning_rate": 6.8942103118917205e-06, "loss": 0.2528, "step": 73730 }, { "epoch": 4.45, "learning_rate": 6.892701401777496e-06, "loss": 0.348, "step": 73732 }, { "epoch": 4.45, "learning_rate": 6.891192491663272e-06, "loss": 0.1862, "step": 73734 }, { "epoch": 4.45, "learning_rate": 6.8896835815490475e-06, "loss": 0.3345, "step": 73736 }, { "epoch": 4.45, "learning_rate": 6.888174671434823e-06, "loss": 0.2949, "step": 73738 }, { "epoch": 4.45, "learning_rate": 6.886665761320599e-06, "loss": 0.2272, "step": 73740 }, { "epoch": 4.45, "learning_rate": 6.8851568512063745e-06, "loss": 0.2485, "step": 73742 }, { "epoch": 4.45, "learning_rate": 6.88364794109215e-06, "loss": 0.2622, "step": 73744 }, { "epoch": 4.45, "learning_rate": 6.882139030977924e-06, "loss": 0.1899, "step": 73746 }, { "epoch": 4.45, "learning_rate": 6.8806301208637e-06, "loss": 0.1752, "step": 73748 }, { "epoch": 4.45, "learning_rate": 6.8791212107494755e-06, "loss": 0.2334, "step": 73750 }, { "epoch": 4.45, "learning_rate": 6.877612300635251e-06, "loss": 0.2003, "step": 73752 }, { "epoch": 4.45, "learning_rate": 6.876103390521027e-06, "loss": 0.2571, "step": 73754 }, { "epoch": 4.45, "learning_rate": 6.8745944804068025e-06, "loss": 0.14, "step": 73756 }, { "epoch": 4.45, "learning_rate": 6.873085570292578e-06, "loss": 0.1939, "step": 73758 }, { "epoch": 4.45, "learning_rate": 6.871576660178354e-06, "loss": 0.2447, "step": 73760 }, { "epoch": 4.45, "learning_rate": 6.870067750064129e-06, "loss": 0.2758, "step": 73762 }, { "epoch": 4.45, "learning_rate": 6.868558839949904e-06, "loss": 0.1492, "step": 73764 }, { "epoch": 4.45, "learning_rate": 6.86704992983568e-06, "loss": 0.2099, "step": 73766 }, { "epoch": 4.45, "learning_rate": 6.865541019721456e-06, "loss": 0.2266, "step": 73768 }, { "epoch": 4.45, "learning_rate": 6.864032109607231e-06, "loss": 0.3029, "step": 73770 }, { "epoch": 4.45, "learning_rate": 6.862523199493007e-06, "loss": 0.3126, "step": 73772 }, { "epoch": 4.45, "learning_rate": 6.861014289378783e-06, "loss": 0.3091, "step": 73774 }, { "epoch": 4.45, "learning_rate": 6.859505379264558e-06, "loss": 0.3334, "step": 73776 }, { "epoch": 4.45, "learning_rate": 6.857996469150334e-06, "loss": 0.242, "step": 73778 }, { "epoch": 4.45, "learning_rate": 6.856487559036108e-06, "loss": 0.3176, "step": 73780 }, { "epoch": 4.45, "learning_rate": 6.854978648921884e-06, "loss": 0.2196, "step": 73782 }, { "epoch": 4.45, "learning_rate": 6.853469738807659e-06, "loss": 0.2734, "step": 73784 }, { "epoch": 4.45, "learning_rate": 6.851960828693435e-06, "loss": 0.3413, "step": 73786 }, { "epoch": 4.45, "learning_rate": 6.850451918579211e-06, "loss": 0.2378, "step": 73788 }, { "epoch": 4.45, "learning_rate": 6.848943008464986e-06, "loss": 0.1473, "step": 73790 }, { "epoch": 4.45, "learning_rate": 6.847434098350762e-06, "loss": 0.1558, "step": 73792 }, { "epoch": 4.45, "learning_rate": 6.845925188236538e-06, "loss": 0.2376, "step": 73794 }, { "epoch": 4.45, "learning_rate": 6.844416278122312e-06, "loss": 0.2225, "step": 73796 }, { "epoch": 4.45, "learning_rate": 6.842907368008088e-06, "loss": 0.332, "step": 73798 }, { "epoch": 4.45, "learning_rate": 6.841398457893864e-06, "loss": 0.1896, "step": 73800 }, { "epoch": 4.45, "learning_rate": 6.8398895477796394e-06, "loss": 0.1723, "step": 73802 }, { "epoch": 4.45, "learning_rate": 6.838380637665415e-06, "loss": 0.236, "step": 73804 }, { "epoch": 4.45, "learning_rate": 6.836871727551191e-06, "loss": 0.2056, "step": 73806 }, { "epoch": 4.45, "learning_rate": 6.8353628174369664e-06, "loss": 0.2185, "step": 73808 }, { "epoch": 4.45, "learning_rate": 6.833853907322742e-06, "loss": 0.1461, "step": 73810 }, { "epoch": 4.46, "learning_rate": 6.832344997208516e-06, "loss": 0.1679, "step": 73812 }, { "epoch": 4.46, "learning_rate": 6.830836087094292e-06, "loss": 0.2251, "step": 73814 }, { "epoch": 4.46, "learning_rate": 6.829327176980067e-06, "loss": 0.2737, "step": 73816 }, { "epoch": 4.46, "learning_rate": 6.827818266865843e-06, "loss": 0.2391, "step": 73818 }, { "epoch": 4.46, "learning_rate": 6.826309356751619e-06, "loss": 0.2834, "step": 73820 }, { "epoch": 4.46, "learning_rate": 6.824800446637394e-06, "loss": 0.2429, "step": 73822 }, { "epoch": 4.46, "learning_rate": 6.82329153652317e-06, "loss": 0.2338, "step": 73824 }, { "epoch": 4.46, "learning_rate": 6.821782626408946e-06, "loss": 0.2733, "step": 73826 }, { "epoch": 4.46, "learning_rate": 6.82027371629472e-06, "loss": 0.2703, "step": 73828 }, { "epoch": 4.46, "learning_rate": 6.818764806180495e-06, "loss": 0.4022, "step": 73830 }, { "epoch": 4.46, "learning_rate": 6.817255896066271e-06, "loss": 0.2872, "step": 73832 }, { "epoch": 4.46, "learning_rate": 6.815746985952047e-06, "loss": 0.2269, "step": 73834 }, { "epoch": 4.46, "learning_rate": 6.814238075837822e-06, "loss": 0.3197, "step": 73836 }, { "epoch": 4.46, "learning_rate": 6.812729165723598e-06, "loss": 0.2237, "step": 73838 }, { "epoch": 4.46, "learning_rate": 6.811220255609374e-06, "loss": 0.1847, "step": 73840 }, { "epoch": 4.46, "learning_rate": 6.809711345495149e-06, "loss": 0.2678, "step": 73842 }, { "epoch": 4.46, "learning_rate": 6.808202435380924e-06, "loss": 0.2718, "step": 73844 }, { "epoch": 4.46, "learning_rate": 6.8066935252667e-06, "loss": 0.2427, "step": 73846 }, { "epoch": 4.46, "learning_rate": 6.8051846151524755e-06, "loss": 0.2747, "step": 73848 }, { "epoch": 4.46, "learning_rate": 6.803675705038251e-06, "loss": 0.2753, "step": 73850 }, { "epoch": 4.46, "learning_rate": 6.802166794924027e-06, "loss": 0.1994, "step": 73852 }, { "epoch": 4.46, "learning_rate": 6.8006578848098025e-06, "loss": 0.1541, "step": 73854 }, { "epoch": 4.46, "learning_rate": 6.799148974695578e-06, "loss": 0.4655, "step": 73856 }, { "epoch": 4.46, "learning_rate": 6.797640064581354e-06, "loss": 0.1913, "step": 73858 }, { "epoch": 4.46, "learning_rate": 6.7961311544671295e-06, "loss": 0.279, "step": 73860 }, { "epoch": 4.46, "learning_rate": 6.7946222443529035e-06, "loss": 0.2426, "step": 73862 }, { "epoch": 4.46, "learning_rate": 6.793113334238679e-06, "loss": 0.2571, "step": 73864 }, { "epoch": 4.46, "learning_rate": 6.791604424124455e-06, "loss": 0.2369, "step": 73866 }, { "epoch": 4.46, "learning_rate": 6.7900955140102305e-06, "loss": 0.3924, "step": 73868 }, { "epoch": 4.46, "learning_rate": 6.788586603896006e-06, "loss": 0.1568, "step": 73870 }, { "epoch": 4.46, "learning_rate": 6.787077693781782e-06, "loss": 0.2276, "step": 73872 }, { "epoch": 4.46, "learning_rate": 6.7855687836675575e-06, "loss": 0.212, "step": 73874 }, { "epoch": 4.46, "learning_rate": 6.784059873553333e-06, "loss": 0.2385, "step": 73876 }, { "epoch": 4.46, "learning_rate": 6.782550963439108e-06, "loss": 0.1885, "step": 73878 }, { "epoch": 4.46, "learning_rate": 6.781042053324884e-06, "loss": 0.3276, "step": 73880 }, { "epoch": 4.46, "learning_rate": 6.779533143210659e-06, "loss": 0.2377, "step": 73882 }, { "epoch": 4.46, "learning_rate": 6.778024233096435e-06, "loss": 0.2103, "step": 73884 }, { "epoch": 4.46, "learning_rate": 6.776515322982211e-06, "loss": 0.1475, "step": 73886 }, { "epoch": 4.46, "learning_rate": 6.775006412867986e-06, "loss": 0.175, "step": 73888 }, { "epoch": 4.46, "learning_rate": 6.773497502753762e-06, "loss": 0.1624, "step": 73890 }, { "epoch": 4.46, "learning_rate": 6.771988592639538e-06, "loss": 0.2585, "step": 73892 }, { "epoch": 4.46, "learning_rate": 6.770479682525312e-06, "loss": 0.313, "step": 73894 }, { "epoch": 4.46, "learning_rate": 6.768970772411087e-06, "loss": 0.297, "step": 73896 }, { "epoch": 4.46, "learning_rate": 6.767461862296863e-06, "loss": 0.1613, "step": 73898 }, { "epoch": 4.46, "learning_rate": 6.765952952182639e-06, "loss": 0.2068, "step": 73900 }, { "epoch": 4.46, "learning_rate": 6.764444042068414e-06, "loss": 0.2069, "step": 73902 }, { "epoch": 4.46, "learning_rate": 6.76293513195419e-06, "loss": 0.1406, "step": 73904 }, { "epoch": 4.46, "learning_rate": 6.761426221839966e-06, "loss": 0.2996, "step": 73906 }, { "epoch": 4.46, "learning_rate": 6.759917311725741e-06, "loss": 0.1551, "step": 73908 }, { "epoch": 4.46, "learning_rate": 6.758408401611515e-06, "loss": 0.288, "step": 73910 }, { "epoch": 4.46, "learning_rate": 6.756899491497291e-06, "loss": 0.262, "step": 73912 }, { "epoch": 4.46, "learning_rate": 6.755390581383067e-06, "loss": 0.172, "step": 73914 }, { "epoch": 4.46, "learning_rate": 6.753881671268843e-06, "loss": 0.1996, "step": 73916 }, { "epoch": 4.46, "learning_rate": 6.752372761154619e-06, "loss": 0.2953, "step": 73918 }, { "epoch": 4.46, "learning_rate": 6.7508638510403944e-06, "loss": 0.2072, "step": 73920 }, { "epoch": 4.46, "learning_rate": 6.74935494092617e-06, "loss": 0.3183, "step": 73922 }, { "epoch": 4.46, "learning_rate": 6.747846030811946e-06, "loss": 0.1987, "step": 73924 }, { "epoch": 4.46, "learning_rate": 6.74633712069772e-06, "loss": 0.2051, "step": 73926 }, { "epoch": 4.46, "learning_rate": 6.744828210583495e-06, "loss": 0.2774, "step": 73928 }, { "epoch": 4.46, "learning_rate": 6.743319300469271e-06, "loss": 0.272, "step": 73930 }, { "epoch": 4.46, "learning_rate": 6.741810390355047e-06, "loss": 0.194, "step": 73932 }, { "epoch": 4.46, "learning_rate": 6.740301480240822e-06, "loss": 0.3068, "step": 73934 }, { "epoch": 4.46, "learning_rate": 6.738792570126598e-06, "loss": 0.331, "step": 73936 }, { "epoch": 4.46, "learning_rate": 6.737283660012374e-06, "loss": 0.174, "step": 73938 }, { "epoch": 4.46, "learning_rate": 6.735774749898149e-06, "loss": 0.2568, "step": 73940 }, { "epoch": 4.46, "learning_rate": 6.734265839783925e-06, "loss": 0.1558, "step": 73942 }, { "epoch": 4.46, "learning_rate": 6.732756929669699e-06, "loss": 0.19, "step": 73944 }, { "epoch": 4.46, "learning_rate": 6.731248019555475e-06, "loss": 0.2878, "step": 73946 }, { "epoch": 4.46, "learning_rate": 6.72973910944125e-06, "loss": 0.2652, "step": 73948 }, { "epoch": 4.46, "learning_rate": 6.728230199327026e-06, "loss": 0.2447, "step": 73950 }, { "epoch": 4.46, "learning_rate": 6.726721289212802e-06, "loss": 0.2701, "step": 73952 }, { "epoch": 4.46, "learning_rate": 6.725212379098577e-06, "loss": 0.1534, "step": 73954 }, { "epoch": 4.46, "learning_rate": 6.723703468984353e-06, "loss": 0.1337, "step": 73956 }, { "epoch": 4.46, "learning_rate": 6.722194558870129e-06, "loss": 0.2843, "step": 73958 }, { "epoch": 4.46, "learning_rate": 6.7206856487559035e-06, "loss": 0.2449, "step": 73960 }, { "epoch": 4.46, "learning_rate": 6.719176738641679e-06, "loss": 0.1824, "step": 73962 }, { "epoch": 4.46, "learning_rate": 6.717667828527455e-06, "loss": 0.1599, "step": 73964 }, { "epoch": 4.46, "learning_rate": 6.7161589184132305e-06, "loss": 0.2256, "step": 73966 }, { "epoch": 4.46, "learning_rate": 6.714650008299006e-06, "loss": 0.1721, "step": 73968 }, { "epoch": 4.46, "learning_rate": 6.713141098184782e-06, "loss": 0.1831, "step": 73970 }, { "epoch": 4.46, "learning_rate": 6.7116321880705575e-06, "loss": 0.3822, "step": 73972 }, { "epoch": 4.46, "learning_rate": 6.710123277956333e-06, "loss": 0.1241, "step": 73974 }, { "epoch": 4.46, "learning_rate": 6.708614367842107e-06, "loss": 0.2693, "step": 73976 }, { "epoch": 4.47, "learning_rate": 6.707105457727883e-06, "loss": 0.2301, "step": 73978 }, { "epoch": 4.47, "learning_rate": 6.7055965476136585e-06, "loss": 0.2451, "step": 73980 }, { "epoch": 4.47, "learning_rate": 6.704087637499434e-06, "loss": 0.203, "step": 73982 }, { "epoch": 4.47, "learning_rate": 6.70257872738521e-06, "loss": 0.1706, "step": 73984 }, { "epoch": 4.47, "learning_rate": 6.7010698172709855e-06, "loss": 0.2648, "step": 73986 }, { "epoch": 4.47, "learning_rate": 6.699560907156761e-06, "loss": 0.3402, "step": 73988 }, { "epoch": 4.47, "learning_rate": 6.698051997042537e-06, "loss": 0.3847, "step": 73990 }, { "epoch": 4.47, "learning_rate": 6.696543086928312e-06, "loss": 0.2454, "step": 73992 }, { "epoch": 4.47, "learning_rate": 6.695034176814087e-06, "loss": 0.2008, "step": 73994 }, { "epoch": 4.47, "learning_rate": 6.693525266699863e-06, "loss": 0.2269, "step": 73996 }, { "epoch": 4.47, "learning_rate": 6.692016356585639e-06, "loss": 0.3231, "step": 73998 }, { "epoch": 4.47, "learning_rate": 6.690507446471414e-06, "loss": 0.1953, "step": 74000 }, { "epoch": 4.47, "learning_rate": 6.68899853635719e-06, "loss": 0.2596, "step": 74002 }, { "epoch": 4.47, "learning_rate": 6.687489626242966e-06, "loss": 0.3185, "step": 74004 }, { "epoch": 4.47, "learning_rate": 6.685980716128741e-06, "loss": 0.2727, "step": 74006 }, { "epoch": 4.47, "learning_rate": 6.684471806014517e-06, "loss": 0.195, "step": 74008 }, { "epoch": 4.47, "learning_rate": 6.682962895900291e-06, "loss": 0.292, "step": 74010 }, { "epoch": 4.47, "learning_rate": 6.681453985786067e-06, "loss": 0.125, "step": 74012 }, { "epoch": 4.47, "learning_rate": 6.679945075671842e-06, "loss": 0.1322, "step": 74014 }, { "epoch": 4.47, "learning_rate": 6.678436165557618e-06, "loss": 0.2442, "step": 74016 }, { "epoch": 4.47, "learning_rate": 6.676927255443394e-06, "loss": 0.2037, "step": 74018 }, { "epoch": 4.47, "learning_rate": 6.675418345329169e-06, "loss": 0.3495, "step": 74020 }, { "epoch": 4.47, "learning_rate": 6.673909435214945e-06, "loss": 0.1401, "step": 74022 }, { "epoch": 4.47, "learning_rate": 6.672400525100721e-06, "loss": 0.3126, "step": 74024 }, { "epoch": 4.47, "learning_rate": 6.670891614986495e-06, "loss": 0.3367, "step": 74026 }, { "epoch": 4.47, "learning_rate": 6.66938270487227e-06, "loss": 0.2208, "step": 74028 }, { "epoch": 4.47, "learning_rate": 6.667873794758046e-06, "loss": 0.2374, "step": 74030 }, { "epoch": 4.47, "learning_rate": 6.666364884643822e-06, "loss": 0.2084, "step": 74032 }, { "epoch": 4.47, "learning_rate": 6.664855974529598e-06, "loss": 0.2861, "step": 74034 }, { "epoch": 4.47, "learning_rate": 6.663347064415374e-06, "loss": 0.1457, "step": 74036 }, { "epoch": 4.47, "learning_rate": 6.6618381543011495e-06, "loss": 0.1858, "step": 74038 }, { "epoch": 4.47, "learning_rate": 6.660329244186925e-06, "loss": 0.3088, "step": 74040 }, { "epoch": 4.47, "learning_rate": 6.658820334072699e-06, "loss": 0.3078, "step": 74042 }, { "epoch": 4.47, "learning_rate": 6.657311423958475e-06, "loss": 0.4273, "step": 74044 }, { "epoch": 4.47, "learning_rate": 6.6558025138442504e-06, "loss": 0.2745, "step": 74046 }, { "epoch": 4.47, "learning_rate": 6.654293603730026e-06, "loss": 0.2233, "step": 74048 }, { "epoch": 4.47, "learning_rate": 6.652784693615802e-06, "loss": 0.2923, "step": 74050 }, { "epoch": 4.47, "learning_rate": 6.6512757835015774e-06, "loss": 0.1763, "step": 74052 }, { "epoch": 4.47, "learning_rate": 6.649766873387353e-06, "loss": 0.1478, "step": 74054 }, { "epoch": 4.47, "learning_rate": 6.648257963273129e-06, "loss": 0.275, "step": 74056 }, { "epoch": 4.47, "learning_rate": 6.646749053158903e-06, "loss": 0.2496, "step": 74058 }, { "epoch": 4.47, "learning_rate": 6.645240143044678e-06, "loss": 0.2349, "step": 74060 }, { "epoch": 4.47, "learning_rate": 6.643731232930454e-06, "loss": 0.209, "step": 74062 }, { "epoch": 4.47, "learning_rate": 6.64222232281623e-06, "loss": 0.2266, "step": 74064 }, { "epoch": 4.47, "learning_rate": 6.640713412702005e-06, "loss": 0.1788, "step": 74066 }, { "epoch": 4.47, "learning_rate": 6.639204502587781e-06, "loss": 0.141, "step": 74068 }, { "epoch": 4.47, "learning_rate": 6.637695592473557e-06, "loss": 0.2051, "step": 74070 }, { "epoch": 4.47, "learning_rate": 6.636186682359332e-06, "loss": 0.3032, "step": 74072 }, { "epoch": 4.47, "learning_rate": 6.634677772245107e-06, "loss": 0.2357, "step": 74074 }, { "epoch": 4.47, "learning_rate": 6.633168862130883e-06, "loss": 0.2188, "step": 74076 }, { "epoch": 4.47, "learning_rate": 6.6316599520166586e-06, "loss": 0.2456, "step": 74078 }, { "epoch": 4.47, "learning_rate": 6.630151041902434e-06, "loss": 0.187, "step": 74080 }, { "epoch": 4.47, "learning_rate": 6.62864213178821e-06, "loss": 0.0942, "step": 74082 }, { "epoch": 4.47, "learning_rate": 6.6271332216739856e-06, "loss": 0.3151, "step": 74084 }, { "epoch": 4.47, "learning_rate": 6.625624311559761e-06, "loss": 0.1606, "step": 74086 }, { "epoch": 4.47, "learning_rate": 6.624115401445537e-06, "loss": 0.2558, "step": 74088 }, { "epoch": 4.47, "learning_rate": 6.6226064913313126e-06, "loss": 0.3181, "step": 74090 }, { "epoch": 4.47, "learning_rate": 6.6210975812170865e-06, "loss": 0.1305, "step": 74092 }, { "epoch": 4.47, "learning_rate": 6.619588671102862e-06, "loss": 0.2819, "step": 74094 }, { "epoch": 4.47, "learning_rate": 6.618079760988638e-06, "loss": 0.237, "step": 74096 }, { "epoch": 4.47, "learning_rate": 6.6165708508744135e-06, "loss": 0.2597, "step": 74098 }, { "epoch": 4.47, "learning_rate": 6.615061940760189e-06, "loss": 0.1666, "step": 74100 }, { "epoch": 4.47, "learning_rate": 6.613553030645965e-06, "loss": 0.2937, "step": 74102 }, { "epoch": 4.47, "learning_rate": 6.6120441205317405e-06, "loss": 0.1961, "step": 74104 }, { "epoch": 4.47, "learning_rate": 6.610535210417516e-06, "loss": 0.1849, "step": 74106 }, { "epoch": 4.47, "learning_rate": 6.609026300303291e-06, "loss": 0.3492, "step": 74108 }, { "epoch": 4.47, "learning_rate": 6.607517390189067e-06, "loss": 0.2716, "step": 74110 }, { "epoch": 4.47, "learning_rate": 6.606008480074842e-06, "loss": 0.1978, "step": 74112 }, { "epoch": 4.47, "learning_rate": 6.604499569960618e-06, "loss": 0.2839, "step": 74114 }, { "epoch": 4.47, "learning_rate": 6.602990659846394e-06, "loss": 0.2023, "step": 74116 }, { "epoch": 4.47, "learning_rate": 6.601481749732169e-06, "loss": 0.2185, "step": 74118 }, { "epoch": 4.47, "learning_rate": 6.599972839617945e-06, "loss": 0.2337, "step": 74120 }, { "epoch": 4.47, "learning_rate": 6.598463929503721e-06, "loss": 0.1596, "step": 74122 }, { "epoch": 4.47, "learning_rate": 6.596955019389495e-06, "loss": 0.2094, "step": 74124 }, { "epoch": 4.47, "learning_rate": 6.59544610927527e-06, "loss": 0.3221, "step": 74126 }, { "epoch": 4.47, "learning_rate": 6.593937199161046e-06, "loss": 0.298, "step": 74128 }, { "epoch": 4.47, "learning_rate": 6.592428289046822e-06, "loss": 0.1455, "step": 74130 }, { "epoch": 4.47, "learning_rate": 6.590919378932597e-06, "loss": 0.1809, "step": 74132 }, { "epoch": 4.47, "learning_rate": 6.589410468818373e-06, "loss": 0.1791, "step": 74134 }, { "epoch": 4.47, "learning_rate": 6.587901558704149e-06, "loss": 0.1676, "step": 74136 }, { "epoch": 4.47, "learning_rate": 6.586392648589924e-06, "loss": 0.2488, "step": 74138 }, { "epoch": 4.47, "learning_rate": 6.584883738475698e-06, "loss": 0.2238, "step": 74140 }, { "epoch": 4.48, "learning_rate": 6.583374828361474e-06, "loss": 0.1762, "step": 74142 }, { "epoch": 4.48, "learning_rate": 6.58186591824725e-06, "loss": 0.2163, "step": 74144 }, { "epoch": 4.48, "learning_rate": 6.580357008133025e-06, "loss": 0.1786, "step": 74146 }, { "epoch": 4.48, "learning_rate": 6.578848098018801e-06, "loss": 0.1989, "step": 74148 }, { "epoch": 4.48, "learning_rate": 6.577339187904577e-06, "loss": 0.1785, "step": 74150 }, { "epoch": 4.48, "learning_rate": 6.575830277790352e-06, "loss": 0.311, "step": 74152 }, { "epoch": 4.48, "learning_rate": 6.574321367676129e-06, "loss": 0.3317, "step": 74154 }, { "epoch": 4.48, "learning_rate": 6.572812457561903e-06, "loss": 0.2721, "step": 74156 }, { "epoch": 4.48, "learning_rate": 6.5713035474476784e-06, "loss": 0.2049, "step": 74158 }, { "epoch": 4.48, "learning_rate": 6.569794637333454e-06, "loss": 0.2751, "step": 74160 }, { "epoch": 4.48, "learning_rate": 6.56828572721923e-06, "loss": 0.2055, "step": 74162 }, { "epoch": 4.48, "learning_rate": 6.5667768171050054e-06, "loss": 0.1449, "step": 74164 }, { "epoch": 4.48, "learning_rate": 6.565267906990781e-06, "loss": 0.2977, "step": 74166 }, { "epoch": 4.48, "learning_rate": 6.563758996876557e-06, "loss": 0.2178, "step": 74168 }, { "epoch": 4.48, "learning_rate": 6.5622500867623324e-06, "loss": 0.4012, "step": 74170 }, { "epoch": 4.48, "learning_rate": 6.560741176648108e-06, "loss": 0.2765, "step": 74172 }, { "epoch": 4.48, "learning_rate": 6.559232266533882e-06, "loss": 0.3066, "step": 74174 }, { "epoch": 4.48, "learning_rate": 6.557723356419658e-06, "loss": 0.1899, "step": 74176 }, { "epoch": 4.48, "learning_rate": 6.556214446305433e-06, "loss": 0.1768, "step": 74178 }, { "epoch": 4.48, "learning_rate": 6.554705536191209e-06, "loss": 0.1746, "step": 74180 }, { "epoch": 4.48, "learning_rate": 6.553196626076985e-06, "loss": 0.3164, "step": 74182 }, { "epoch": 4.48, "learning_rate": 6.55168771596276e-06, "loss": 0.2233, "step": 74184 }, { "epoch": 4.48, "learning_rate": 6.550178805848536e-06, "loss": 0.2282, "step": 74186 }, { "epoch": 4.48, "learning_rate": 6.548669895734312e-06, "loss": 0.2082, "step": 74188 }, { "epoch": 4.48, "learning_rate": 6.5471609856200866e-06, "loss": 0.2359, "step": 74190 }, { "epoch": 4.48, "learning_rate": 6.545652075505862e-06, "loss": 0.2337, "step": 74192 }, { "epoch": 4.48, "learning_rate": 6.544143165391638e-06, "loss": 0.2642, "step": 74194 }, { "epoch": 4.48, "learning_rate": 6.5426342552774136e-06, "loss": 0.1408, "step": 74196 }, { "epoch": 4.48, "learning_rate": 6.541125345163189e-06, "loss": 0.2302, "step": 74198 }, { "epoch": 4.48, "learning_rate": 6.539616435048965e-06, "loss": 0.2286, "step": 74200 }, { "epoch": 4.48, "learning_rate": 6.5381075249347406e-06, "loss": 0.4095, "step": 74202 }, { "epoch": 4.48, "learning_rate": 6.536598614820516e-06, "loss": 0.3063, "step": 74204 }, { "epoch": 4.48, "learning_rate": 6.53508970470629e-06, "loss": 0.2132, "step": 74206 }, { "epoch": 4.48, "learning_rate": 6.533580794592066e-06, "loss": 0.3066, "step": 74208 }, { "epoch": 4.48, "learning_rate": 6.5320718844778415e-06, "loss": 0.2318, "step": 74210 }, { "epoch": 4.48, "learning_rate": 6.530562974363617e-06, "loss": 0.1304, "step": 74212 }, { "epoch": 4.48, "learning_rate": 6.529054064249393e-06, "loss": 0.1806, "step": 74214 }, { "epoch": 4.48, "learning_rate": 6.5275451541351685e-06, "loss": 0.2081, "step": 74216 }, { "epoch": 4.48, "learning_rate": 6.526036244020944e-06, "loss": 0.2426, "step": 74218 }, { "epoch": 4.48, "learning_rate": 6.52452733390672e-06, "loss": 0.268, "step": 74220 }, { "epoch": 4.48, "learning_rate": 6.523018423792495e-06, "loss": 0.1846, "step": 74222 }, { "epoch": 4.48, "learning_rate": 6.52150951367827e-06, "loss": 0.2372, "step": 74224 }, { "epoch": 4.48, "learning_rate": 6.520000603564046e-06, "loss": 0.2484, "step": 74226 }, { "epoch": 4.48, "learning_rate": 6.518491693449822e-06, "loss": 0.2847, "step": 74228 }, { "epoch": 4.48, "learning_rate": 6.516982783335597e-06, "loss": 0.1675, "step": 74230 }, { "epoch": 4.48, "learning_rate": 6.515473873221373e-06, "loss": 0.2709, "step": 74232 }, { "epoch": 4.48, "learning_rate": 6.513964963107149e-06, "loss": 0.3037, "step": 74234 }, { "epoch": 4.48, "learning_rate": 6.512456052992924e-06, "loss": 0.2347, "step": 74236 }, { "epoch": 4.48, "learning_rate": 6.510947142878698e-06, "loss": 0.1571, "step": 74238 }, { "epoch": 4.48, "learning_rate": 6.509438232764474e-06, "loss": 0.2198, "step": 74240 }, { "epoch": 4.48, "learning_rate": 6.50792932265025e-06, "loss": 0.2638, "step": 74242 }, { "epoch": 4.48, "learning_rate": 6.506420412536025e-06, "loss": 0.1797, "step": 74244 }, { "epoch": 4.48, "learning_rate": 6.504911502421801e-06, "loss": 0.1842, "step": 74246 }, { "epoch": 4.48, "learning_rate": 6.503402592307577e-06, "loss": 0.2783, "step": 74248 }, { "epoch": 4.48, "learning_rate": 6.501893682193352e-06, "loss": 0.2089, "step": 74250 }, { "epoch": 4.48, "learning_rate": 6.500384772079128e-06, "loss": 0.3846, "step": 74252 }, { "epoch": 4.48, "learning_rate": 6.498875861964904e-06, "loss": 0.2383, "step": 74254 }, { "epoch": 4.48, "learning_rate": 6.497366951850678e-06, "loss": 0.1961, "step": 74256 }, { "epoch": 4.48, "learning_rate": 6.495858041736453e-06, "loss": 0.1707, "step": 74258 }, { "epoch": 4.48, "learning_rate": 6.494349131622229e-06, "loss": 0.2482, "step": 74260 }, { "epoch": 4.48, "learning_rate": 6.492840221508005e-06, "loss": 0.2461, "step": 74262 }, { "epoch": 4.48, "learning_rate": 6.49133131139378e-06, "loss": 0.1074, "step": 74264 }, { "epoch": 4.48, "learning_rate": 6.489822401279556e-06, "loss": 0.2558, "step": 74266 }, { "epoch": 4.48, "learning_rate": 6.488313491165332e-06, "loss": 0.2101, "step": 74268 }, { "epoch": 4.48, "learning_rate": 6.486804581051107e-06, "loss": 0.2082, "step": 74270 }, { "epoch": 4.48, "learning_rate": 6.485295670936882e-06, "loss": 0.2853, "step": 74272 }, { "epoch": 4.48, "learning_rate": 6.483786760822658e-06, "loss": 0.2916, "step": 74274 }, { "epoch": 4.48, "learning_rate": 6.4822778507084335e-06, "loss": 0.2268, "step": 74276 }, { "epoch": 4.48, "learning_rate": 6.480768940594209e-06, "loss": 0.2659, "step": 74278 }, { "epoch": 4.48, "learning_rate": 6.479260030479985e-06, "loss": 0.2554, "step": 74280 }, { "epoch": 4.48, "learning_rate": 6.4777511203657605e-06, "loss": 0.1296, "step": 74282 }, { "epoch": 4.48, "learning_rate": 6.476242210251536e-06, "loss": 0.1787, "step": 74284 }, { "epoch": 4.48, "learning_rate": 6.474733300137312e-06, "loss": 0.2085, "step": 74286 }, { "epoch": 4.48, "learning_rate": 6.473224390023086e-06, "loss": 0.2333, "step": 74288 }, { "epoch": 4.48, "learning_rate": 6.4717154799088614e-06, "loss": 0.2239, "step": 74290 }, { "epoch": 4.48, "learning_rate": 6.470206569794637e-06, "loss": 0.188, "step": 74292 }, { "epoch": 4.48, "learning_rate": 6.468697659680413e-06, "loss": 0.4054, "step": 74294 }, { "epoch": 4.48, "learning_rate": 6.4671887495661884e-06, "loss": 0.1719, "step": 74296 }, { "epoch": 4.48, "learning_rate": 6.465679839451964e-06, "loss": 0.2246, "step": 74298 }, { "epoch": 4.48, "learning_rate": 6.46417092933774e-06, "loss": 0.2945, "step": 74300 }, { "epoch": 4.48, "learning_rate": 6.4626620192235154e-06, "loss": 0.2388, "step": 74302 }, { "epoch": 4.48, "learning_rate": 6.46115310910929e-06, "loss": 0.1995, "step": 74304 }, { "epoch": 4.48, "learning_rate": 6.459644198995066e-06, "loss": 0.2774, "step": 74306 }, { "epoch": 4.49, "learning_rate": 6.458135288880842e-06, "loss": 0.2038, "step": 74308 }, { "epoch": 4.49, "learning_rate": 6.456626378766617e-06, "loss": 0.2491, "step": 74310 }, { "epoch": 4.49, "learning_rate": 6.455117468652393e-06, "loss": 0.2709, "step": 74312 }, { "epoch": 4.49, "learning_rate": 6.453608558538169e-06, "loss": 0.3384, "step": 74314 }, { "epoch": 4.49, "learning_rate": 6.452099648423944e-06, "loss": 0.2456, "step": 74316 }, { "epoch": 4.49, "learning_rate": 6.45059073830972e-06, "loss": 0.1998, "step": 74318 }, { "epoch": 4.49, "learning_rate": 6.449081828195496e-06, "loss": 0.1607, "step": 74320 }, { "epoch": 4.49, "learning_rate": 6.4475729180812695e-06, "loss": 0.233, "step": 74322 }, { "epoch": 4.49, "learning_rate": 6.446064007967045e-06, "loss": 0.1653, "step": 74324 }, { "epoch": 4.49, "learning_rate": 6.444555097852821e-06, "loss": 0.2564, "step": 74326 }, { "epoch": 4.49, "learning_rate": 6.4430461877385965e-06, "loss": 0.1816, "step": 74328 }, { "epoch": 4.49, "learning_rate": 6.441537277624372e-06, "loss": 0.2669, "step": 74330 }, { "epoch": 4.49, "learning_rate": 6.440028367510148e-06, "loss": 0.1899, "step": 74332 }, { "epoch": 4.49, "learning_rate": 6.4385194573959235e-06, "loss": 0.3113, "step": 74334 }, { "epoch": 4.49, "learning_rate": 6.437010547281699e-06, "loss": 0.2059, "step": 74336 }, { "epoch": 4.49, "learning_rate": 6.435501637167474e-06, "loss": 0.2064, "step": 74338 }, { "epoch": 4.49, "learning_rate": 6.43399272705325e-06, "loss": 0.3264, "step": 74340 }, { "epoch": 4.49, "learning_rate": 6.432483816939025e-06, "loss": 0.2147, "step": 74342 }, { "epoch": 4.49, "learning_rate": 6.430974906824801e-06, "loss": 0.1789, "step": 74344 }, { "epoch": 4.49, "learning_rate": 6.429465996710577e-06, "loss": 0.2005, "step": 74346 }, { "epoch": 4.49, "learning_rate": 6.427957086596352e-06, "loss": 0.2995, "step": 74348 }, { "epoch": 4.49, "learning_rate": 6.426448176482128e-06, "loss": 0.1684, "step": 74350 }, { "epoch": 4.49, "learning_rate": 6.424939266367904e-06, "loss": 0.2293, "step": 74352 }, { "epoch": 4.49, "learning_rate": 6.423430356253678e-06, "loss": 0.2285, "step": 74354 }, { "epoch": 4.49, "learning_rate": 6.421921446139453e-06, "loss": 0.2274, "step": 74356 }, { "epoch": 4.49, "learning_rate": 6.420412536025229e-06, "loss": 0.2224, "step": 74358 }, { "epoch": 4.49, "learning_rate": 6.418903625911005e-06, "loss": 0.2031, "step": 74360 }, { "epoch": 4.49, "learning_rate": 6.41739471579678e-06, "loss": 0.2036, "step": 74362 }, { "epoch": 4.49, "learning_rate": 6.415885805682556e-06, "loss": 0.2861, "step": 74364 }, { "epoch": 4.49, "learning_rate": 6.414376895568332e-06, "loss": 0.1127, "step": 74366 }, { "epoch": 4.49, "learning_rate": 6.412867985454107e-06, "loss": 0.0969, "step": 74368 }, { "epoch": 4.49, "learning_rate": 6.411359075339881e-06, "loss": 0.2555, "step": 74370 }, { "epoch": 4.49, "learning_rate": 6.409850165225657e-06, "loss": 0.2827, "step": 74372 }, { "epoch": 4.49, "learning_rate": 6.408341255111433e-06, "loss": 0.2401, "step": 74374 }, { "epoch": 4.49, "learning_rate": 6.406832344997208e-06, "loss": 0.2062, "step": 74376 }, { "epoch": 4.49, "learning_rate": 6.405323434882984e-06, "loss": 0.1784, "step": 74378 }, { "epoch": 4.49, "learning_rate": 6.40381452476876e-06, "loss": 0.1724, "step": 74380 }, { "epoch": 4.49, "learning_rate": 6.402305614654535e-06, "loss": 0.1651, "step": 74382 }, { "epoch": 4.49, "learning_rate": 6.400796704540311e-06, "loss": 0.3337, "step": 74384 }, { "epoch": 4.49, "learning_rate": 6.399287794426086e-06, "loss": 0.2733, "step": 74386 }, { "epoch": 4.49, "learning_rate": 6.3977788843118615e-06, "loss": 0.3594, "step": 74388 }, { "epoch": 4.49, "learning_rate": 6.396269974197637e-06, "loss": 0.2134, "step": 74390 }, { "epoch": 4.49, "learning_rate": 6.394761064083413e-06, "loss": 0.2055, "step": 74392 }, { "epoch": 4.49, "learning_rate": 6.3932521539691885e-06, "loss": 0.2637, "step": 74394 }, { "epoch": 4.49, "learning_rate": 6.391743243854964e-06, "loss": 0.2201, "step": 74396 }, { "epoch": 4.49, "learning_rate": 6.39023433374074e-06, "loss": 0.2956, "step": 74398 }, { "epoch": 4.49, "learning_rate": 6.3887254236265155e-06, "loss": 0.2047, "step": 74400 }, { "epoch": 4.49, "learning_rate": 6.387216513512291e-06, "loss": 0.1691, "step": 74402 }, { "epoch": 4.49, "learning_rate": 6.385707603398065e-06, "loss": 0.229, "step": 74404 }, { "epoch": 4.49, "learning_rate": 6.384198693283841e-06, "loss": 0.1321, "step": 74406 }, { "epoch": 4.49, "learning_rate": 6.3826897831696164e-06, "loss": 0.3357, "step": 74408 }, { "epoch": 4.49, "learning_rate": 6.381180873055392e-06, "loss": 0.2406, "step": 74410 }, { "epoch": 4.49, "learning_rate": 6.379671962941168e-06, "loss": 0.1741, "step": 74412 }, { "epoch": 4.49, "learning_rate": 6.3781630528269434e-06, "loss": 0.3126, "step": 74414 }, { "epoch": 4.49, "learning_rate": 6.376654142712719e-06, "loss": 0.3026, "step": 74416 }, { "epoch": 4.49, "learning_rate": 6.375145232598495e-06, "loss": 0.2503, "step": 74418 }, { "epoch": 4.49, "learning_rate": 6.37363632248427e-06, "loss": 0.3023, "step": 74420 }, { "epoch": 4.49, "learning_rate": 6.372127412370045e-06, "loss": 0.1802, "step": 74422 }, { "epoch": 4.49, "learning_rate": 6.370618502255821e-06, "loss": 0.226, "step": 74424 }, { "epoch": 4.49, "learning_rate": 6.369109592141597e-06, "loss": 0.3259, "step": 74426 }, { "epoch": 4.49, "learning_rate": 6.367600682027372e-06, "loss": 0.3035, "step": 74428 }, { "epoch": 4.49, "learning_rate": 6.366091771913148e-06, "loss": 0.2712, "step": 74430 }, { "epoch": 4.49, "learning_rate": 6.364582861798924e-06, "loss": 0.1938, "step": 74432 }, { "epoch": 4.49, "learning_rate": 6.363073951684699e-06, "loss": 0.2063, "step": 74434 }, { "epoch": 4.49, "learning_rate": 6.362319496627586e-06, "loss": 0.1707, "step": 74436 }, { "epoch": 4.49, "learning_rate": 6.3608105865133615e-06, "loss": 0.1784, "step": 74438 }, { "epoch": 4.49, "learning_rate": 6.359301676399137e-06, "loss": 0.2325, "step": 74440 }, { "epoch": 4.49, "learning_rate": 6.357792766284913e-06, "loss": 0.2869, "step": 74442 }, { "epoch": 4.49, "learning_rate": 6.3562838561706885e-06, "loss": 0.3664, "step": 74444 }, { "epoch": 4.49, "learning_rate": 6.354774946056463e-06, "loss": 0.1285, "step": 74446 }, { "epoch": 4.49, "learning_rate": 6.353266035942239e-06, "loss": 0.2361, "step": 74448 }, { "epoch": 4.49, "learning_rate": 6.351757125828015e-06, "loss": 0.2008, "step": 74450 }, { "epoch": 4.49, "learning_rate": 6.35024821571379e-06, "loss": 0.2026, "step": 74452 }, { "epoch": 4.49, "learning_rate": 6.348739305599566e-06, "loss": 0.2509, "step": 74454 }, { "epoch": 4.49, "learning_rate": 6.347230395485342e-06, "loss": 0.2128, "step": 74456 }, { "epoch": 4.49, "learning_rate": 6.345721485371117e-06, "loss": 0.3688, "step": 74458 }, { "epoch": 4.49, "learning_rate": 6.344212575256893e-06, "loss": 0.2102, "step": 74460 }, { "epoch": 4.49, "learning_rate": 6.342703665142667e-06, "loss": 0.278, "step": 74462 }, { "epoch": 4.49, "learning_rate": 6.341194755028443e-06, "loss": 0.2058, "step": 74464 }, { "epoch": 4.49, "learning_rate": 6.339685844914218e-06, "loss": 0.2085, "step": 74466 }, { "epoch": 4.49, "learning_rate": 6.338176934799994e-06, "loss": 0.2792, "step": 74468 }, { "epoch": 4.49, "learning_rate": 6.33666802468577e-06, "loss": 0.1416, "step": 74470 }, { "epoch": 4.49, "learning_rate": 6.335159114571545e-06, "loss": 0.1945, "step": 74472 }, { "epoch": 4.5, "learning_rate": 6.333650204457321e-06, "loss": 0.2235, "step": 74474 }, { "epoch": 4.5, "learning_rate": 6.332141294343097e-06, "loss": 0.3082, "step": 74476 }, { "epoch": 4.5, "learning_rate": 6.3306323842288714e-06, "loss": 0.2319, "step": 74478 }, { "epoch": 4.5, "learning_rate": 6.329123474114647e-06, "loss": 0.3005, "step": 74480 }, { "epoch": 4.5, "learning_rate": 6.327614564000423e-06, "loss": 0.2066, "step": 74482 }, { "epoch": 4.5, "learning_rate": 6.3261056538861984e-06, "loss": 0.1802, "step": 74484 }, { "epoch": 4.5, "learning_rate": 6.324596743771974e-06, "loss": 0.2493, "step": 74486 }, { "epoch": 4.5, "learning_rate": 6.32308783365775e-06, "loss": 0.2052, "step": 74488 }, { "epoch": 4.5, "learning_rate": 6.3215789235435254e-06, "loss": 0.2129, "step": 74490 }, { "epoch": 4.5, "learning_rate": 6.320070013429301e-06, "loss": 0.3053, "step": 74492 }, { "epoch": 4.5, "learning_rate": 6.318561103315075e-06, "loss": 0.2248, "step": 74494 }, { "epoch": 4.5, "learning_rate": 6.317052193200851e-06, "loss": 0.2203, "step": 74496 }, { "epoch": 4.5, "learning_rate": 6.315543283086626e-06, "loss": 0.2301, "step": 74498 }, { "epoch": 4.5, "learning_rate": 6.314034372972402e-06, "loss": 0.2074, "step": 74500 }, { "epoch": 4.5, "learning_rate": 6.312525462858178e-06, "loss": 0.2171, "step": 74502 }, { "epoch": 4.5, "learning_rate": 6.311016552743953e-06, "loss": 0.3013, "step": 74504 }, { "epoch": 4.5, "learning_rate": 6.309507642629729e-06, "loss": 0.2155, "step": 74506 }, { "epoch": 4.5, "learning_rate": 6.307998732515505e-06, "loss": 0.1986, "step": 74508 }, { "epoch": 4.5, "learning_rate": 6.3064898224012796e-06, "loss": 0.2706, "step": 74510 }, { "epoch": 4.5, "learning_rate": 6.304980912287055e-06, "loss": 0.1881, "step": 74512 }, { "epoch": 4.5, "learning_rate": 6.303472002172831e-06, "loss": 0.1323, "step": 74514 }, { "epoch": 4.5, "learning_rate": 6.3019630920586066e-06, "loss": 0.2849, "step": 74516 }, { "epoch": 4.5, "learning_rate": 6.300454181944382e-06, "loss": 0.2783, "step": 74518 }, { "epoch": 4.5, "learning_rate": 6.298945271830158e-06, "loss": 0.2409, "step": 74520 }, { "epoch": 4.5, "learning_rate": 6.2974363617159336e-06, "loss": 0.3349, "step": 74522 }, { "epoch": 4.5, "learning_rate": 6.295927451601709e-06, "loss": 0.3265, "step": 74524 }, { "epoch": 4.5, "learning_rate": 6.294418541487485e-06, "loss": 0.1571, "step": 74526 }, { "epoch": 4.5, "learning_rate": 6.292909631373259e-06, "loss": 0.1844, "step": 74528 }, { "epoch": 4.5, "learning_rate": 6.2914007212590345e-06, "loss": 0.2265, "step": 74530 }, { "epoch": 4.5, "learning_rate": 6.28989181114481e-06, "loss": 0.2504, "step": 74532 }, { "epoch": 4.5, "learning_rate": 6.288382901030586e-06, "loss": 0.1569, "step": 74534 }, { "epoch": 4.5, "learning_rate": 6.2868739909163615e-06, "loss": 0.1782, "step": 74536 }, { "epoch": 4.5, "learning_rate": 6.285365080802137e-06, "loss": 0.2445, "step": 74538 }, { "epoch": 4.5, "learning_rate": 6.283856170687913e-06, "loss": 0.2245, "step": 74540 }, { "epoch": 4.5, "learning_rate": 6.2823472605736885e-06, "loss": 0.2063, "step": 74542 }, { "epoch": 4.5, "learning_rate": 6.2808383504594625e-06, "loss": 0.2125, "step": 74544 }, { "epoch": 4.5, "learning_rate": 6.279329440345238e-06, "loss": 0.2274, "step": 74546 }, { "epoch": 4.5, "learning_rate": 6.277820530231014e-06, "loss": 0.2672, "step": 74548 }, { "epoch": 4.5, "learning_rate": 6.2763116201167895e-06, "loss": 0.3282, "step": 74550 }, { "epoch": 4.5, "learning_rate": 6.274802710002565e-06, "loss": 0.1497, "step": 74552 }, { "epoch": 4.5, "learning_rate": 6.273293799888341e-06, "loss": 0.2354, "step": 74554 }, { "epoch": 4.5, "learning_rate": 6.2717848897741165e-06, "loss": 0.1624, "step": 74556 }, { "epoch": 4.5, "learning_rate": 6.270275979659892e-06, "loss": 0.2447, "step": 74558 }, { "epoch": 4.5, "learning_rate": 6.268767069545667e-06, "loss": 0.1588, "step": 74560 }, { "epoch": 4.5, "learning_rate": 6.267258159431443e-06, "loss": 0.1923, "step": 74562 }, { "epoch": 4.5, "learning_rate": 6.265749249317218e-06, "loss": 0.2406, "step": 74564 }, { "epoch": 4.5, "learning_rate": 6.264240339202994e-06, "loss": 0.3445, "step": 74566 }, { "epoch": 4.5, "learning_rate": 6.26273142908877e-06, "loss": 0.1454, "step": 74568 }, { "epoch": 4.5, "learning_rate": 6.261222518974545e-06, "loss": 0.2389, "step": 74570 }, { "epoch": 4.5, "learning_rate": 6.259713608860321e-06, "loss": 0.2056, "step": 74572 }, { "epoch": 4.5, "learning_rate": 6.258204698746097e-06, "loss": 0.2098, "step": 74574 }, { "epoch": 4.5, "learning_rate": 6.256695788631871e-06, "loss": 0.2004, "step": 74576 }, { "epoch": 4.5, "learning_rate": 6.255186878517646e-06, "loss": 0.1771, "step": 74578 }, { "epoch": 4.5, "learning_rate": 6.253677968403422e-06, "loss": 0.3493, "step": 74580 }, { "epoch": 4.5, "learning_rate": 6.252169058289198e-06, "loss": 0.233, "step": 74582 }, { "epoch": 4.5, "learning_rate": 6.250660148174973e-06, "loss": 0.3113, "step": 74584 }, { "epoch": 4.5, "learning_rate": 6.249151238060749e-06, "loss": 0.1413, "step": 74586 }, { "epoch": 4.5, "learning_rate": 6.247642327946524e-06, "loss": 0.1946, "step": 74588 }, { "epoch": 4.5, "learning_rate": 6.2461334178322994e-06, "loss": 0.2692, "step": 74590 }, { "epoch": 4.5, "learning_rate": 6.244624507718075e-06, "loss": 0.2368, "step": 74592 }, { "epoch": 4.5, "learning_rate": 6.243115597603852e-06, "loss": 0.2639, "step": 74594 }, { "epoch": 4.5, "learning_rate": 6.241606687489627e-06, "loss": 0.1929, "step": 74596 }, { "epoch": 4.5, "learning_rate": 6.240097777375402e-06, "loss": 0.152, "step": 74598 }, { "epoch": 4.5, "learning_rate": 6.238588867261178e-06, "loss": 0.163, "step": 74600 }, { "epoch": 4.5, "learning_rate": 6.2370799571469534e-06, "loss": 0.2388, "step": 74602 }, { "epoch": 4.5, "learning_rate": 6.235571047032729e-06, "loss": 0.1147, "step": 74604 }, { "epoch": 4.5, "learning_rate": 6.234062136918504e-06, "loss": 0.1635, "step": 74606 }, { "epoch": 4.5, "learning_rate": 6.23255322680428e-06, "loss": 0.2481, "step": 74608 }, { "epoch": 4.5, "learning_rate": 6.231044316690055e-06, "loss": 0.3518, "step": 74610 }, { "epoch": 4.5, "learning_rate": 6.229535406575831e-06, "loss": 0.2923, "step": 74612 }, { "epoch": 4.5, "learning_rate": 6.228026496461606e-06, "loss": 0.284, "step": 74614 }, { "epoch": 4.5, "learning_rate": 6.226517586347381e-06, "loss": 0.2139, "step": 74616 }, { "epoch": 4.5, "learning_rate": 6.225008676233157e-06, "loss": 0.2044, "step": 74618 }, { "epoch": 4.5, "learning_rate": 6.223499766118933e-06, "loss": 0.2498, "step": 74620 }, { "epoch": 4.5, "learning_rate": 6.2219908560047076e-06, "loss": 0.1596, "step": 74622 }, { "epoch": 4.5, "learning_rate": 6.220481945890483e-06, "loss": 0.3221, "step": 74624 }, { "epoch": 4.5, "learning_rate": 6.218973035776259e-06, "loss": 0.1155, "step": 74626 }, { "epoch": 4.5, "learning_rate": 6.2174641256620346e-06, "loss": 0.1564, "step": 74628 }, { "epoch": 4.5, "learning_rate": 6.21595521554781e-06, "loss": 0.2014, "step": 74630 }, { "epoch": 4.5, "learning_rate": 6.214446305433586e-06, "loss": 0.1857, "step": 74632 }, { "epoch": 4.5, "learning_rate": 6.2129373953193616e-06, "loss": 0.2582, "step": 74634 }, { "epoch": 4.5, "learning_rate": 6.211428485205137e-06, "loss": 0.3272, "step": 74636 }, { "epoch": 4.5, "learning_rate": 6.209919575090912e-06, "loss": 0.2344, "step": 74638 }, { "epoch": 4.51, "learning_rate": 6.208410664976688e-06, "loss": 0.16, "step": 74640 }, { "epoch": 4.51, "learning_rate": 6.206901754862463e-06, "loss": 0.324, "step": 74642 }, { "epoch": 4.51, "learning_rate": 6.205392844748239e-06, "loss": 0.2152, "step": 74644 }, { "epoch": 4.51, "learning_rate": 6.203883934634014e-06, "loss": 0.2032, "step": 74646 }, { "epoch": 4.51, "learning_rate": 6.2023750245197895e-06, "loss": 0.2712, "step": 74648 }, { "epoch": 4.51, "learning_rate": 6.200866114405565e-06, "loss": 0.1704, "step": 74650 }, { "epoch": 4.51, "learning_rate": 6.199357204291341e-06, "loss": 0.1051, "step": 74652 }, { "epoch": 4.51, "learning_rate": 6.197848294177116e-06, "loss": 0.1548, "step": 74654 }, { "epoch": 4.51, "learning_rate": 6.196339384062891e-06, "loss": 0.1531, "step": 74656 }, { "epoch": 4.51, "learning_rate": 6.194830473948667e-06, "loss": 0.1926, "step": 74658 }, { "epoch": 4.51, "learning_rate": 6.193321563834443e-06, "loss": 0.193, "step": 74660 }, { "epoch": 4.51, "learning_rate": 6.1918126537202175e-06, "loss": 0.163, "step": 74662 }, { "epoch": 4.51, "learning_rate": 6.190303743605993e-06, "loss": 0.2556, "step": 74664 }, { "epoch": 4.51, "learning_rate": 6.188794833491769e-06, "loss": 0.2499, "step": 74666 }, { "epoch": 4.51, "learning_rate": 6.1872859233775445e-06, "loss": 0.2274, "step": 74668 }, { "epoch": 4.51, "learning_rate": 6.18577701326332e-06, "loss": 0.2504, "step": 74670 }, { "epoch": 4.51, "learning_rate": 6.184268103149096e-06, "loss": 0.3071, "step": 74672 }, { "epoch": 4.51, "learning_rate": 6.1827591930348715e-06, "loss": 0.2314, "step": 74674 }, { "epoch": 4.51, "learning_rate": 6.181250282920647e-06, "loss": 0.2251, "step": 74676 }, { "epoch": 4.51, "learning_rate": 6.179741372806423e-06, "loss": 0.2065, "step": 74678 }, { "epoch": 4.51, "learning_rate": 6.178232462692198e-06, "loss": 0.16, "step": 74680 }, { "epoch": 4.51, "learning_rate": 6.176723552577973e-06, "loss": 0.2216, "step": 74682 }, { "epoch": 4.51, "learning_rate": 6.175214642463749e-06, "loss": 0.1794, "step": 74684 }, { "epoch": 4.51, "learning_rate": 6.173705732349525e-06, "loss": 0.1589, "step": 74686 }, { "epoch": 4.51, "learning_rate": 6.1721968222352995e-06, "loss": 0.184, "step": 74688 }, { "epoch": 4.51, "learning_rate": 6.170687912121075e-06, "loss": 0.1717, "step": 74690 }, { "epoch": 4.51, "learning_rate": 6.169179002006851e-06, "loss": 0.3214, "step": 74692 }, { "epoch": 4.51, "learning_rate": 6.1676700918926265e-06, "loss": 0.1769, "step": 74694 }, { "epoch": 4.51, "learning_rate": 6.166161181778401e-06, "loss": 0.2813, "step": 74696 }, { "epoch": 4.51, "learning_rate": 6.164652271664177e-06, "loss": 0.2417, "step": 74698 }, { "epoch": 4.51, "learning_rate": 6.163143361549953e-06, "loss": 0.2343, "step": 74700 }, { "epoch": 4.51, "learning_rate": 6.161634451435728e-06, "loss": 0.2551, "step": 74702 }, { "epoch": 4.51, "learning_rate": 6.160125541321503e-06, "loss": 0.2458, "step": 74704 }, { "epoch": 4.51, "learning_rate": 6.158616631207279e-06, "loss": 0.2088, "step": 74706 }, { "epoch": 4.51, "learning_rate": 6.1571077210930545e-06, "loss": 0.1881, "step": 74708 }, { "epoch": 4.51, "learning_rate": 6.15559881097883e-06, "loss": 0.3686, "step": 74710 }, { "epoch": 4.51, "learning_rate": 6.154089900864606e-06, "loss": 0.2465, "step": 74712 }, { "epoch": 4.51, "learning_rate": 6.1525809907503815e-06, "loss": 0.1933, "step": 74714 }, { "epoch": 4.51, "learning_rate": 6.151072080636157e-06, "loss": 0.1145, "step": 74716 }, { "epoch": 4.51, "learning_rate": 6.149563170521933e-06, "loss": 0.1731, "step": 74718 }, { "epoch": 4.51, "learning_rate": 6.148054260407708e-06, "loss": 0.2545, "step": 74720 }, { "epoch": 4.51, "learning_rate": 6.146545350293483e-06, "loss": 0.2, "step": 74722 }, { "epoch": 4.51, "learning_rate": 6.145036440179259e-06, "loss": 0.3411, "step": 74724 }, { "epoch": 4.51, "learning_rate": 6.143527530065035e-06, "loss": 0.2166, "step": 74726 }, { "epoch": 4.51, "learning_rate": 6.142018619950809e-06, "loss": 0.1538, "step": 74728 }, { "epoch": 4.51, "learning_rate": 6.140509709836585e-06, "loss": 0.2009, "step": 74730 }, { "epoch": 4.51, "learning_rate": 6.139000799722361e-06, "loss": 0.1683, "step": 74732 }, { "epoch": 4.51, "learning_rate": 6.137491889608136e-06, "loss": 0.253, "step": 74734 }, { "epoch": 4.51, "learning_rate": 6.135982979493911e-06, "loss": 0.2112, "step": 74736 }, { "epoch": 4.51, "learning_rate": 6.134474069379687e-06, "loss": 0.2809, "step": 74738 }, { "epoch": 4.51, "learning_rate": 6.1329651592654626e-06, "loss": 0.1465, "step": 74740 }, { "epoch": 4.51, "learning_rate": 6.131456249151238e-06, "loss": 0.1945, "step": 74742 }, { "epoch": 4.51, "learning_rate": 6.129947339037014e-06, "loss": 0.1696, "step": 74744 }, { "epoch": 4.51, "learning_rate": 6.1284384289227896e-06, "loss": 0.0994, "step": 74746 }, { "epoch": 4.51, "learning_rate": 6.126929518808565e-06, "loss": 0.192, "step": 74748 }, { "epoch": 4.51, "learning_rate": 6.125420608694341e-06, "loss": 0.2585, "step": 74750 }, { "epoch": 4.51, "learning_rate": 6.1239116985801166e-06, "loss": 0.4547, "step": 74752 }, { "epoch": 4.51, "learning_rate": 6.122402788465891e-06, "loss": 0.1885, "step": 74754 }, { "epoch": 4.51, "learning_rate": 6.120893878351667e-06, "loss": 0.2631, "step": 74756 }, { "epoch": 4.51, "learning_rate": 6.119384968237443e-06, "loss": 0.285, "step": 74758 }, { "epoch": 4.51, "learning_rate": 6.117876058123218e-06, "loss": 0.2452, "step": 74760 }, { "epoch": 4.51, "learning_rate": 6.116367148008993e-06, "loss": 0.2997, "step": 74762 }, { "epoch": 4.51, "learning_rate": 6.114858237894769e-06, "loss": 0.1932, "step": 74764 }, { "epoch": 4.51, "learning_rate": 6.1133493277805445e-06, "loss": 0.1701, "step": 74766 }, { "epoch": 4.51, "learning_rate": 6.11184041766632e-06, "loss": 0.2579, "step": 74768 }, { "epoch": 4.51, "learning_rate": 6.110331507552095e-06, "loss": 0.1858, "step": 74770 }, { "epoch": 4.51, "learning_rate": 6.108822597437871e-06, "loss": 0.2218, "step": 74772 }, { "epoch": 4.51, "learning_rate": 6.107313687323646e-06, "loss": 0.2576, "step": 74774 }, { "epoch": 4.51, "learning_rate": 6.105804777209422e-06, "loss": 0.2071, "step": 74776 }, { "epoch": 4.51, "learning_rate": 6.104295867095197e-06, "loss": 0.29, "step": 74778 }, { "epoch": 4.51, "learning_rate": 6.1027869569809725e-06, "loss": 0.1588, "step": 74780 }, { "epoch": 4.51, "learning_rate": 6.101278046866748e-06, "loss": 0.2154, "step": 74782 }, { "epoch": 4.51, "learning_rate": 6.099769136752524e-06, "loss": 0.2122, "step": 74784 }, { "epoch": 4.51, "learning_rate": 6.0982602266382995e-06, "loss": 0.2321, "step": 74786 }, { "epoch": 4.51, "learning_rate": 6.096751316524075e-06, "loss": 0.2509, "step": 74788 }, { "epoch": 4.51, "learning_rate": 6.095242406409851e-06, "loss": 0.2432, "step": 74790 }, { "epoch": 4.51, "learning_rate": 6.0937334962956265e-06, "loss": 0.2153, "step": 74792 }, { "epoch": 4.51, "learning_rate": 6.092224586181401e-06, "loss": 0.3048, "step": 74794 }, { "epoch": 4.51, "learning_rate": 6.090715676067177e-06, "loss": 0.2385, "step": 74796 }, { "epoch": 4.51, "learning_rate": 6.089206765952953e-06, "loss": 0.2133, "step": 74798 }, { "epoch": 4.51, "learning_rate": 6.087697855838728e-06, "loss": 0.2599, "step": 74800 }, { "epoch": 4.51, "learning_rate": 6.086188945724503e-06, "loss": 0.1404, "step": 74802 }, { "epoch": 4.51, "learning_rate": 6.084680035610279e-06, "loss": 0.2147, "step": 74804 }, { "epoch": 4.52, "learning_rate": 6.0831711254960545e-06, "loss": 0.2795, "step": 74806 }, { "epoch": 4.52, "learning_rate": 6.08166221538183e-06, "loss": 0.1359, "step": 74808 }, { "epoch": 4.52, "learning_rate": 6.080153305267605e-06, "loss": 0.2739, "step": 74810 }, { "epoch": 4.52, "learning_rate": 6.078644395153381e-06, "loss": 0.1602, "step": 74812 }, { "epoch": 4.52, "learning_rate": 6.077135485039156e-06, "loss": 0.1918, "step": 74814 }, { "epoch": 4.52, "learning_rate": 6.075626574924932e-06, "loss": 0.2217, "step": 74816 }, { "epoch": 4.52, "learning_rate": 6.074117664810707e-06, "loss": 0.1966, "step": 74818 }, { "epoch": 4.52, "learning_rate": 6.0726087546964825e-06, "loss": 0.1943, "step": 74820 }, { "epoch": 4.52, "learning_rate": 6.071099844582258e-06, "loss": 0.2351, "step": 74822 }, { "epoch": 4.52, "learning_rate": 6.069590934468034e-06, "loss": 0.0837, "step": 74824 }, { "epoch": 4.52, "learning_rate": 6.0680820243538095e-06, "loss": 0.2978, "step": 74826 }, { "epoch": 4.52, "learning_rate": 6.066573114239585e-06, "loss": 0.2541, "step": 74828 }, { "epoch": 4.52, "learning_rate": 6.065064204125361e-06, "loss": 0.2245, "step": 74830 }, { "epoch": 4.52, "learning_rate": 6.0635552940111365e-06, "loss": 0.2206, "step": 74832 }, { "epoch": 4.52, "learning_rate": 6.062046383896912e-06, "loss": 0.2323, "step": 74834 }, { "epoch": 4.52, "learning_rate": 6.060537473782687e-06, "loss": 0.3497, "step": 74836 }, { "epoch": 4.52, "learning_rate": 6.059028563668463e-06, "loss": 0.3367, "step": 74838 }, { "epoch": 4.52, "learning_rate": 6.057519653554238e-06, "loss": 0.2398, "step": 74840 }, { "epoch": 4.52, "learning_rate": 6.056010743440014e-06, "loss": 0.1625, "step": 74842 }, { "epoch": 4.52, "learning_rate": 6.054501833325789e-06, "loss": 0.3007, "step": 74844 }, { "epoch": 4.52, "learning_rate": 6.0529929232115644e-06, "loss": 0.1926, "step": 74846 }, { "epoch": 4.52, "learning_rate": 6.05148401309734e-06, "loss": 0.3651, "step": 74848 }, { "epoch": 4.52, "learning_rate": 6.049975102983116e-06, "loss": 0.1975, "step": 74850 }, { "epoch": 4.52, "learning_rate": 6.048466192868891e-06, "loss": 0.1887, "step": 74852 }, { "epoch": 4.52, "learning_rate": 6.046957282754666e-06, "loss": 0.2777, "step": 74854 }, { "epoch": 4.52, "learning_rate": 6.045448372640442e-06, "loss": 0.1575, "step": 74856 }, { "epoch": 4.52, "learning_rate": 6.043939462526218e-06, "loss": 0.1426, "step": 74858 }, { "epoch": 4.52, "learning_rate": 6.042430552411993e-06, "loss": 0.3065, "step": 74860 }, { "epoch": 4.52, "learning_rate": 6.040921642297769e-06, "loss": 0.2541, "step": 74862 }, { "epoch": 4.52, "learning_rate": 6.039412732183545e-06, "loss": 0.2524, "step": 74864 }, { "epoch": 4.52, "learning_rate": 6.03790382206932e-06, "loss": 0.2183, "step": 74866 }, { "epoch": 4.52, "learning_rate": 6.036394911955095e-06, "loss": 0.4205, "step": 74868 }, { "epoch": 4.52, "learning_rate": 6.034886001840871e-06, "loss": 0.179, "step": 74870 }, { "epoch": 4.52, "learning_rate": 6.033377091726646e-06, "loss": 0.2867, "step": 74872 }, { "epoch": 4.52, "learning_rate": 6.031868181612422e-06, "loss": 0.2482, "step": 74874 }, { "epoch": 4.52, "learning_rate": 6.030359271498197e-06, "loss": 0.281, "step": 74876 }, { "epoch": 4.52, "learning_rate": 6.0288503613839726e-06, "loss": 0.2714, "step": 74878 }, { "epoch": 4.52, "learning_rate": 6.027341451269748e-06, "loss": 0.1871, "step": 74880 }, { "epoch": 4.52, "learning_rate": 6.025832541155524e-06, "loss": 0.2406, "step": 74882 }, { "epoch": 4.52, "learning_rate": 6.024323631041299e-06, "loss": 0.2127, "step": 74884 }, { "epoch": 4.52, "learning_rate": 6.022814720927074e-06, "loss": 0.2271, "step": 74886 }, { "epoch": 4.52, "learning_rate": 6.02130581081285e-06, "loss": 0.2296, "step": 74888 }, { "epoch": 4.52, "learning_rate": 6.019796900698626e-06, "loss": 0.1767, "step": 74890 }, { "epoch": 4.52, "learning_rate": 6.0182879905844005e-06, "loss": 0.158, "step": 74892 }, { "epoch": 4.52, "learning_rate": 6.016779080470176e-06, "loss": 0.2721, "step": 74894 }, { "epoch": 4.52, "learning_rate": 6.015270170355952e-06, "loss": 0.1308, "step": 74896 }, { "epoch": 4.52, "learning_rate": 6.0137612602417275e-06, "loss": 0.1856, "step": 74898 }, { "epoch": 4.52, "learning_rate": 6.012252350127503e-06, "loss": 0.1673, "step": 74900 }, { "epoch": 4.52, "learning_rate": 6.010743440013279e-06, "loss": 0.5575, "step": 74902 }, { "epoch": 4.52, "learning_rate": 6.0092345298990545e-06, "loss": 0.1897, "step": 74904 }, { "epoch": 4.52, "learning_rate": 6.00772561978483e-06, "loss": 0.1529, "step": 74906 }, { "epoch": 4.52, "learning_rate": 6.006216709670606e-06, "loss": 0.2263, "step": 74908 }, { "epoch": 4.52, "learning_rate": 6.004707799556381e-06, "loss": 0.1872, "step": 74910 }, { "epoch": 4.52, "learning_rate": 6.003198889442156e-06, "loss": 0.2429, "step": 74912 }, { "epoch": 4.52, "learning_rate": 6.001689979327932e-06, "loss": 0.1611, "step": 74914 }, { "epoch": 4.52, "learning_rate": 6.000181069213708e-06, "loss": 0.3571, "step": 74916 }, { "epoch": 4.52, "learning_rate": 5.9986721590994825e-06, "loss": 0.228, "step": 74918 }, { "epoch": 4.52, "learning_rate": 5.997163248985258e-06, "loss": 0.2469, "step": 74920 }, { "epoch": 4.52, "learning_rate": 5.995654338871034e-06, "loss": 0.1919, "step": 74922 }, { "epoch": 4.52, "learning_rate": 5.9941454287568095e-06, "loss": 0.2123, "step": 74924 }, { "epoch": 4.52, "learning_rate": 5.992636518642584e-06, "loss": 0.3171, "step": 74926 }, { "epoch": 4.52, "learning_rate": 5.99112760852836e-06, "loss": 0.2965, "step": 74928 }, { "epoch": 4.52, "learning_rate": 5.989618698414136e-06, "loss": 0.1936, "step": 74930 }, { "epoch": 4.52, "learning_rate": 5.988109788299911e-06, "loss": 0.2647, "step": 74932 }, { "epoch": 4.52, "learning_rate": 5.986600878185686e-06, "loss": 0.3813, "step": 74934 }, { "epoch": 4.52, "learning_rate": 5.985091968071462e-06, "loss": 0.299, "step": 74936 }, { "epoch": 4.52, "learning_rate": 5.9835830579572375e-06, "loss": 0.2277, "step": 74938 }, { "epoch": 4.52, "learning_rate": 5.982074147843013e-06, "loss": 0.1982, "step": 74940 }, { "epoch": 4.52, "learning_rate": 5.980565237728789e-06, "loss": 0.366, "step": 74942 }, { "epoch": 4.52, "learning_rate": 5.9790563276145645e-06, "loss": 0.2115, "step": 74944 }, { "epoch": 4.52, "learning_rate": 5.97754741750034e-06, "loss": 0.1797, "step": 74946 }, { "epoch": 4.52, "learning_rate": 5.976038507386116e-06, "loss": 0.2534, "step": 74948 }, { "epoch": 4.52, "learning_rate": 5.974529597271891e-06, "loss": 0.3348, "step": 74950 }, { "epoch": 4.52, "learning_rate": 5.973020687157666e-06, "loss": 0.2999, "step": 74952 }, { "epoch": 4.52, "learning_rate": 5.971511777043442e-06, "loss": 0.2487, "step": 74954 }, { "epoch": 4.52, "learning_rate": 5.970002866929218e-06, "loss": 0.1947, "step": 74956 }, { "epoch": 4.52, "learning_rate": 5.9684939568149924e-06, "loss": 0.1752, "step": 74958 }, { "epoch": 4.52, "learning_rate": 5.966985046700768e-06, "loss": 0.2248, "step": 74960 }, { "epoch": 4.52, "learning_rate": 5.965476136586544e-06, "loss": 0.1398, "step": 74962 }, { "epoch": 4.52, "learning_rate": 5.9639672264723194e-06, "loss": 0.2681, "step": 74964 }, { "epoch": 4.52, "learning_rate": 5.962458316358094e-06, "loss": 0.1872, "step": 74966 }, { "epoch": 4.52, "learning_rate": 5.96094940624387e-06, "loss": 0.1418, "step": 74968 }, { "epoch": 4.52, "learning_rate": 5.959440496129646e-06, "loss": 0.3788, "step": 74970 }, { "epoch": 4.53, "learning_rate": 5.957931586015421e-06, "loss": 0.2324, "step": 74972 }, { "epoch": 4.53, "learning_rate": 5.956422675901197e-06, "loss": 0.2637, "step": 74974 }, { "epoch": 4.53, "learning_rate": 5.954913765786973e-06, "loss": 0.31, "step": 74976 }, { "epoch": 4.53, "learning_rate": 5.953404855672748e-06, "loss": 0.2005, "step": 74978 }, { "epoch": 4.53, "learning_rate": 5.951895945558524e-06, "loss": 0.198, "step": 74980 }, { "epoch": 4.53, "learning_rate": 5.9503870354443e-06, "loss": 0.2166, "step": 74982 }, { "epoch": 4.53, "learning_rate": 5.948878125330074e-06, "loss": 0.267, "step": 74984 }, { "epoch": 4.53, "learning_rate": 5.94736921521585e-06, "loss": 0.2453, "step": 74986 }, { "epoch": 4.53, "learning_rate": 5.945860305101626e-06, "loss": 0.3719, "step": 74988 }, { "epoch": 4.53, "learning_rate": 5.944351394987401e-06, "loss": 0.2198, "step": 74990 }, { "epoch": 4.53, "learning_rate": 5.942842484873176e-06, "loss": 0.1887, "step": 74992 }, { "epoch": 4.53, "learning_rate": 5.941333574758952e-06, "loss": 0.3202, "step": 74994 }, { "epoch": 4.53, "learning_rate": 5.9398246646447276e-06, "loss": 0.2297, "step": 74996 }, { "epoch": 4.53, "learning_rate": 5.938315754530503e-06, "loss": 0.2804, "step": 74998 }, { "epoch": 4.53, "learning_rate": 5.936806844416278e-06, "loss": 0.2173, "step": 75000 }, { "epoch": 4.53, "learning_rate": 5.935297934302054e-06, "loss": 0.2306, "step": 75002 }, { "epoch": 4.53, "learning_rate": 5.933789024187829e-06, "loss": 0.3477, "step": 75004 }, { "epoch": 4.53, "learning_rate": 5.932280114073605e-06, "loss": 0.2373, "step": 75006 }, { "epoch": 4.53, "learning_rate": 5.93077120395938e-06, "loss": 0.2497, "step": 75008 }, { "epoch": 4.53, "learning_rate": 5.9292622938451555e-06, "loss": 0.2095, "step": 75010 }, { "epoch": 4.53, "learning_rate": 5.927753383730931e-06, "loss": 0.2214, "step": 75012 }, { "epoch": 4.53, "learning_rate": 5.926244473616707e-06, "loss": 0.2334, "step": 75014 }, { "epoch": 4.53, "learning_rate": 5.9247355635024825e-06, "loss": 0.3218, "step": 75016 }, { "epoch": 4.53, "learning_rate": 5.923226653388258e-06, "loss": 0.1321, "step": 75018 }, { "epoch": 4.53, "learning_rate": 5.921717743274034e-06, "loss": 0.235, "step": 75020 }, { "epoch": 4.53, "learning_rate": 5.9202088331598095e-06, "loss": 0.2004, "step": 75022 }, { "epoch": 4.53, "learning_rate": 5.918699923045584e-06, "loss": 0.1999, "step": 75024 }, { "epoch": 4.53, "learning_rate": 5.91719101293136e-06, "loss": 0.2237, "step": 75026 }, { "epoch": 4.53, "learning_rate": 5.915682102817136e-06, "loss": 0.248, "step": 75028 }, { "epoch": 4.53, "learning_rate": 5.914173192702911e-06, "loss": 0.2618, "step": 75030 }, { "epoch": 4.53, "learning_rate": 5.912664282588686e-06, "loss": 0.3253, "step": 75032 }, { "epoch": 4.53, "learning_rate": 5.911155372474462e-06, "loss": 0.2615, "step": 75034 }, { "epoch": 4.53, "learning_rate": 5.9096464623602375e-06, "loss": 0.1853, "step": 75036 }, { "epoch": 4.53, "learning_rate": 5.908137552246013e-06, "loss": 0.2396, "step": 75038 }, { "epoch": 4.53, "learning_rate": 5.906628642131788e-06, "loss": 0.202, "step": 75040 }, { "epoch": 4.53, "learning_rate": 5.905119732017564e-06, "loss": 0.1364, "step": 75042 }, { "epoch": 4.53, "learning_rate": 5.903610821903339e-06, "loss": 0.3628, "step": 75044 }, { "epoch": 4.53, "learning_rate": 5.902101911789115e-06, "loss": 0.1731, "step": 75046 }, { "epoch": 4.53, "learning_rate": 5.90059300167489e-06, "loss": 0.4479, "step": 75048 }, { "epoch": 4.53, "learning_rate": 5.8990840915606655e-06, "loss": 0.2568, "step": 75050 }, { "epoch": 4.53, "learning_rate": 5.897575181446441e-06, "loss": 0.2809, "step": 75052 }, { "epoch": 4.53, "learning_rate": 5.896066271332217e-06, "loss": 0.302, "step": 75054 }, { "epoch": 4.53, "learning_rate": 5.8945573612179925e-06, "loss": 0.2594, "step": 75056 }, { "epoch": 4.53, "learning_rate": 5.893048451103768e-06, "loss": 0.2103, "step": 75058 }, { "epoch": 4.53, "learning_rate": 5.891539540989544e-06, "loss": 0.2887, "step": 75060 }, { "epoch": 4.53, "learning_rate": 5.8900306308753195e-06, "loss": 0.2659, "step": 75062 }, { "epoch": 4.53, "learning_rate": 5.888521720761095e-06, "loss": 0.1841, "step": 75064 }, { "epoch": 4.53, "learning_rate": 5.88701281064687e-06, "loss": 0.1975, "step": 75066 }, { "epoch": 4.53, "learning_rate": 5.885503900532646e-06, "loss": 0.2181, "step": 75068 }, { "epoch": 4.53, "learning_rate": 5.883994990418421e-06, "loss": 0.215, "step": 75070 }, { "epoch": 4.53, "learning_rate": 5.882486080304197e-06, "loss": 0.3372, "step": 75072 }, { "epoch": 4.53, "learning_rate": 5.880977170189972e-06, "loss": 0.2453, "step": 75074 }, { "epoch": 4.53, "learning_rate": 5.8794682600757475e-06, "loss": 0.2453, "step": 75076 }, { "epoch": 4.53, "learning_rate": 5.877959349961523e-06, "loss": 0.1927, "step": 75078 }, { "epoch": 4.53, "learning_rate": 5.876450439847299e-06, "loss": 0.2291, "step": 75080 }, { "epoch": 4.53, "learning_rate": 5.874941529733074e-06, "loss": 0.2781, "step": 75082 }, { "epoch": 4.53, "learning_rate": 5.873432619618849e-06, "loss": 0.1915, "step": 75084 }, { "epoch": 4.53, "learning_rate": 5.871923709504625e-06, "loss": 0.1955, "step": 75086 }, { "epoch": 4.53, "learning_rate": 5.870414799390401e-06, "loss": 0.2599, "step": 75088 }, { "epoch": 4.53, "learning_rate": 5.8689058892761754e-06, "loss": 0.3173, "step": 75090 }, { "epoch": 4.53, "learning_rate": 5.867396979161952e-06, "loss": 0.2981, "step": 75092 }, { "epoch": 4.53, "learning_rate": 5.865888069047728e-06, "loss": 0.1947, "step": 75094 }, { "epoch": 4.53, "learning_rate": 5.864379158933503e-06, "loss": 0.2773, "step": 75096 }, { "epoch": 4.53, "learning_rate": 5.862870248819278e-06, "loss": 0.3127, "step": 75098 }, { "epoch": 4.53, "learning_rate": 5.861361338705054e-06, "loss": 0.4668, "step": 75100 }, { "epoch": 4.53, "learning_rate": 5.8598524285908294e-06, "loss": 0.1649, "step": 75102 }, { "epoch": 4.53, "learning_rate": 5.858343518476605e-06, "loss": 0.253, "step": 75104 }, { "epoch": 4.53, "learning_rate": 5.85683460836238e-06, "loss": 0.183, "step": 75106 }, { "epoch": 4.53, "learning_rate": 5.855325698248156e-06, "loss": 0.271, "step": 75108 }, { "epoch": 4.53, "learning_rate": 5.853816788133931e-06, "loss": 0.2633, "step": 75110 }, { "epoch": 4.53, "learning_rate": 5.852307878019707e-06, "loss": 0.192, "step": 75112 }, { "epoch": 4.53, "learning_rate": 5.850798967905482e-06, "loss": 0.1484, "step": 75114 }, { "epoch": 4.53, "learning_rate": 5.849290057791257e-06, "loss": 0.4194, "step": 75116 }, { "epoch": 4.53, "learning_rate": 5.847781147677033e-06, "loss": 0.2349, "step": 75118 }, { "epoch": 4.53, "learning_rate": 5.846272237562809e-06, "loss": 0.1815, "step": 75120 }, { "epoch": 4.53, "learning_rate": 5.8447633274485836e-06, "loss": 0.1901, "step": 75122 }, { "epoch": 4.53, "learning_rate": 5.843254417334359e-06, "loss": 0.3084, "step": 75124 }, { "epoch": 4.53, "learning_rate": 5.841745507220135e-06, "loss": 0.1845, "step": 75126 }, { "epoch": 4.53, "learning_rate": 5.8402365971059106e-06, "loss": 0.1487, "step": 75128 }, { "epoch": 4.53, "learning_rate": 5.838727686991686e-06, "loss": 0.4321, "step": 75130 }, { "epoch": 4.53, "learning_rate": 5.837218776877462e-06, "loss": 0.1462, "step": 75132 }, { "epoch": 4.53, "learning_rate": 5.8357098667632376e-06, "loss": 0.3039, "step": 75134 }, { "epoch": 4.54, "learning_rate": 5.834200956649013e-06, "loss": 0.1922, "step": 75136 }, { "epoch": 4.54, "learning_rate": 5.832692046534789e-06, "loss": 0.2259, "step": 75138 }, { "epoch": 4.54, "learning_rate": 5.831183136420564e-06, "loss": 0.2988, "step": 75140 }, { "epoch": 4.54, "learning_rate": 5.829674226306339e-06, "loss": 0.2742, "step": 75142 }, { "epoch": 4.54, "learning_rate": 5.828165316192115e-06, "loss": 0.1776, "step": 75144 }, { "epoch": 4.54, "learning_rate": 5.826656406077891e-06, "loss": 0.1908, "step": 75146 }, { "epoch": 4.54, "learning_rate": 5.8251474959636655e-06, "loss": 0.223, "step": 75148 }, { "epoch": 4.54, "learning_rate": 5.823638585849441e-06, "loss": 0.2477, "step": 75150 }, { "epoch": 4.54, "learning_rate": 5.822129675735217e-06, "loss": 0.1721, "step": 75152 }, { "epoch": 4.54, "learning_rate": 5.8206207656209925e-06, "loss": 0.3004, "step": 75154 }, { "epoch": 4.54, "learning_rate": 5.819111855506767e-06, "loss": 0.2355, "step": 75156 }, { "epoch": 4.54, "learning_rate": 5.817602945392543e-06, "loss": 0.2151, "step": 75158 }, { "epoch": 4.54, "learning_rate": 5.816094035278319e-06, "loss": 0.242, "step": 75160 }, { "epoch": 4.54, "learning_rate": 5.814585125164094e-06, "loss": 0.1408, "step": 75162 }, { "epoch": 4.54, "learning_rate": 5.813076215049869e-06, "loss": 0.2526, "step": 75164 }, { "epoch": 4.54, "learning_rate": 5.811567304935645e-06, "loss": 0.2149, "step": 75166 }, { "epoch": 4.54, "learning_rate": 5.8100583948214205e-06, "loss": 0.1999, "step": 75168 }, { "epoch": 4.54, "learning_rate": 5.808549484707196e-06, "loss": 0.2227, "step": 75170 }, { "epoch": 4.54, "learning_rate": 5.807040574592972e-06, "loss": 0.2703, "step": 75172 }, { "epoch": 4.54, "learning_rate": 5.8055316644787475e-06, "loss": 0.2074, "step": 75174 }, { "epoch": 4.54, "learning_rate": 5.804022754364523e-06, "loss": 0.1851, "step": 75176 }, { "epoch": 4.54, "learning_rate": 5.802513844250299e-06, "loss": 0.169, "step": 75178 }, { "epoch": 4.54, "learning_rate": 5.801004934136074e-06, "loss": 0.2307, "step": 75180 }, { "epoch": 4.54, "learning_rate": 5.799496024021849e-06, "loss": 0.2205, "step": 75182 }, { "epoch": 4.54, "learning_rate": 5.797987113907625e-06, "loss": 0.2859, "step": 75184 }, { "epoch": 4.54, "learning_rate": 5.796478203793401e-06, "loss": 0.1824, "step": 75186 }, { "epoch": 4.54, "learning_rate": 5.7949692936791755e-06, "loss": 0.2413, "step": 75188 }, { "epoch": 4.54, "learning_rate": 5.793460383564951e-06, "loss": 0.2984, "step": 75190 }, { "epoch": 4.54, "learning_rate": 5.791951473450727e-06, "loss": 0.2301, "step": 75192 }, { "epoch": 4.54, "learning_rate": 5.7904425633365025e-06, "loss": 0.1699, "step": 75194 }, { "epoch": 4.54, "learning_rate": 5.788933653222277e-06, "loss": 0.2635, "step": 75196 }, { "epoch": 4.54, "learning_rate": 5.787424743108053e-06, "loss": 0.2167, "step": 75198 }, { "epoch": 4.54, "learning_rate": 5.785915832993829e-06, "loss": 0.2157, "step": 75200 }, { "epoch": 4.54, "learning_rate": 5.784406922879604e-06, "loss": 0.2357, "step": 75202 }, { "epoch": 4.54, "learning_rate": 5.782898012765379e-06, "loss": 0.377, "step": 75204 }, { "epoch": 4.54, "learning_rate": 5.781389102651155e-06, "loss": 0.1975, "step": 75206 }, { "epoch": 4.54, "learning_rate": 5.7798801925369304e-06, "loss": 0.2187, "step": 75208 }, { "epoch": 4.54, "learning_rate": 5.778371282422707e-06, "loss": 0.1762, "step": 75210 }, { "epoch": 4.54, "learning_rate": 5.776862372308483e-06, "loss": 0.3338, "step": 75212 }, { "epoch": 4.54, "learning_rate": 5.7753534621942574e-06, "loss": 0.2348, "step": 75214 }, { "epoch": 4.54, "learning_rate": 5.773844552080033e-06, "loss": 0.2107, "step": 75216 }, { "epoch": 4.54, "learning_rate": 5.772335641965809e-06, "loss": 0.2298, "step": 75218 }, { "epoch": 4.54, "learning_rate": 5.7708267318515844e-06, "loss": 0.2332, "step": 75220 }, { "epoch": 4.54, "learning_rate": 5.769317821737359e-06, "loss": 0.2051, "step": 75222 }, { "epoch": 4.54, "learning_rate": 5.767808911623135e-06, "loss": 0.261, "step": 75224 }, { "epoch": 4.54, "learning_rate": 5.766300001508911e-06, "loss": 0.3081, "step": 75226 }, { "epoch": 4.54, "learning_rate": 5.764791091394686e-06, "loss": 0.2567, "step": 75228 }, { "epoch": 4.54, "learning_rate": 5.763282181280461e-06, "loss": 0.2233, "step": 75230 }, { "epoch": 4.54, "learning_rate": 5.761773271166237e-06, "loss": 0.1593, "step": 75232 }, { "epoch": 4.54, "learning_rate": 5.760264361052012e-06, "loss": 0.3101, "step": 75234 }, { "epoch": 4.54, "learning_rate": 5.758755450937788e-06, "loss": 0.1993, "step": 75236 }, { "epoch": 4.54, "learning_rate": 5.757246540823563e-06, "loss": 0.2259, "step": 75238 }, { "epoch": 4.54, "learning_rate": 5.7557376307093386e-06, "loss": 0.2984, "step": 75240 }, { "epoch": 4.54, "learning_rate": 5.754228720595114e-06, "loss": 0.2573, "step": 75242 }, { "epoch": 4.54, "learning_rate": 5.75271981048089e-06, "loss": 0.1757, "step": 75244 }, { "epoch": 4.54, "learning_rate": 5.7512109003666656e-06, "loss": 0.1254, "step": 75246 }, { "epoch": 4.54, "learning_rate": 5.749701990252441e-06, "loss": 0.223, "step": 75248 }, { "epoch": 4.54, "learning_rate": 5.748193080138217e-06, "loss": 0.2571, "step": 75250 }, { "epoch": 4.54, "learning_rate": 5.7466841700239926e-06, "loss": 0.165, "step": 75252 }, { "epoch": 4.54, "learning_rate": 5.745175259909767e-06, "loss": 0.1738, "step": 75254 }, { "epoch": 4.54, "learning_rate": 5.743666349795543e-06, "loss": 0.2298, "step": 75256 }, { "epoch": 4.54, "learning_rate": 5.742157439681319e-06, "loss": 0.2251, "step": 75258 }, { "epoch": 4.54, "learning_rate": 5.740648529567094e-06, "loss": 0.3217, "step": 75260 }, { "epoch": 4.54, "learning_rate": 5.739139619452869e-06, "loss": 0.1611, "step": 75262 }, { "epoch": 4.54, "learning_rate": 5.737630709338645e-06, "loss": 0.2089, "step": 75264 }, { "epoch": 4.54, "learning_rate": 5.7361217992244205e-06, "loss": 0.2052, "step": 75266 }, { "epoch": 4.54, "learning_rate": 5.734612889110196e-06, "loss": 0.175, "step": 75268 }, { "epoch": 4.54, "learning_rate": 5.733103978995971e-06, "loss": 0.2697, "step": 75270 }, { "epoch": 4.54, "learning_rate": 5.731595068881747e-06, "loss": 0.1348, "step": 75272 }, { "epoch": 4.54, "learning_rate": 5.730086158767522e-06, "loss": 0.219, "step": 75274 }, { "epoch": 4.54, "learning_rate": 5.728577248653298e-06, "loss": 0.1885, "step": 75276 }, { "epoch": 4.54, "learning_rate": 5.727068338539073e-06, "loss": 0.3056, "step": 75278 }, { "epoch": 4.54, "learning_rate": 5.7255594284248485e-06, "loss": 0.3209, "step": 75280 }, { "epoch": 4.54, "learning_rate": 5.724050518310624e-06, "loss": 0.1854, "step": 75282 }, { "epoch": 4.54, "learning_rate": 5.7225416081964e-06, "loss": 0.23, "step": 75284 }, { "epoch": 4.54, "learning_rate": 5.7210326980821755e-06, "loss": 0.2454, "step": 75286 }, { "epoch": 4.54, "learning_rate": 5.719523787967951e-06, "loss": 0.3016, "step": 75288 }, { "epoch": 4.54, "learning_rate": 5.718014877853727e-06, "loss": 0.2208, "step": 75290 }, { "epoch": 4.54, "learning_rate": 5.7165059677395025e-06, "loss": 0.25, "step": 75292 }, { "epoch": 4.54, "learning_rate": 5.714997057625278e-06, "loss": 0.1965, "step": 75294 }, { "epoch": 4.54, "learning_rate": 5.713488147511053e-06, "loss": 0.2859, "step": 75296 }, { "epoch": 4.54, "learning_rate": 5.711979237396829e-06, "loss": 0.2801, "step": 75298 }, { "epoch": 4.54, "learning_rate": 5.710470327282604e-06, "loss": 0.1321, "step": 75300 }, { "epoch": 4.55, "learning_rate": 5.70896141716838e-06, "loss": 0.3032, "step": 75302 }, { "epoch": 4.55, "learning_rate": 5.707452507054155e-06, "loss": 0.1683, "step": 75304 }, { "epoch": 4.55, "learning_rate": 5.7059435969399305e-06, "loss": 0.2262, "step": 75306 }, { "epoch": 4.55, "learning_rate": 5.704434686825706e-06, "loss": 0.1696, "step": 75308 }, { "epoch": 4.55, "learning_rate": 5.702925776711482e-06, "loss": 0.2093, "step": 75310 }, { "epoch": 4.55, "learning_rate": 5.701416866597257e-06, "loss": 0.1844, "step": 75312 }, { "epoch": 4.55, "learning_rate": 5.699907956483032e-06, "loss": 0.2947, "step": 75314 }, { "epoch": 4.55, "learning_rate": 5.698399046368808e-06, "loss": 0.2315, "step": 75316 }, { "epoch": 4.55, "learning_rate": 5.696890136254584e-06, "loss": 0.2695, "step": 75318 }, { "epoch": 4.55, "learning_rate": 5.6953812261403585e-06, "loss": 0.209, "step": 75320 }, { "epoch": 4.55, "learning_rate": 5.693872316026134e-06, "loss": 0.2685, "step": 75322 }, { "epoch": 4.55, "learning_rate": 5.69236340591191e-06, "loss": 0.2456, "step": 75324 }, { "epoch": 4.55, "learning_rate": 5.6908544957976855e-06, "loss": 0.2281, "step": 75326 }, { "epoch": 4.55, "learning_rate": 5.689345585683461e-06, "loss": 0.2336, "step": 75328 }, { "epoch": 4.55, "learning_rate": 5.687836675569237e-06, "loss": 0.2086, "step": 75330 }, { "epoch": 4.55, "learning_rate": 5.6863277654550125e-06, "loss": 0.2697, "step": 75332 }, { "epoch": 4.55, "learning_rate": 5.684818855340788e-06, "loss": 0.1953, "step": 75334 }, { "epoch": 4.55, "learning_rate": 5.683309945226563e-06, "loss": 0.4764, "step": 75336 }, { "epoch": 4.55, "learning_rate": 5.681801035112339e-06, "loss": 0.2276, "step": 75338 }, { "epoch": 4.55, "learning_rate": 5.680292124998114e-06, "loss": 0.285, "step": 75340 }, { "epoch": 4.55, "learning_rate": 5.67878321488389e-06, "loss": 0.2623, "step": 75342 }, { "epoch": 4.55, "learning_rate": 5.677274304769665e-06, "loss": 0.2821, "step": 75344 }, { "epoch": 4.55, "learning_rate": 5.67576539465544e-06, "loss": 0.23, "step": 75346 }, { "epoch": 4.55, "learning_rate": 5.674256484541216e-06, "loss": 0.2349, "step": 75348 }, { "epoch": 4.55, "learning_rate": 5.672747574426992e-06, "loss": 0.2586, "step": 75350 }, { "epoch": 4.55, "learning_rate": 5.6712386643127666e-06, "loss": 0.1906, "step": 75352 }, { "epoch": 4.55, "learning_rate": 5.669729754198542e-06, "loss": 0.2154, "step": 75354 }, { "epoch": 4.55, "learning_rate": 5.668220844084318e-06, "loss": 0.3237, "step": 75356 }, { "epoch": 4.55, "learning_rate": 5.6667119339700936e-06, "loss": 0.201, "step": 75358 }, { "epoch": 4.55, "learning_rate": 5.665203023855869e-06, "loss": 0.183, "step": 75360 }, { "epoch": 4.55, "learning_rate": 5.663694113741645e-06, "loss": 0.2012, "step": 75362 }, { "epoch": 4.55, "learning_rate": 5.6621852036274206e-06, "loss": 0.1666, "step": 75364 }, { "epoch": 4.55, "learning_rate": 5.660676293513196e-06, "loss": 0.2832, "step": 75366 }, { "epoch": 4.55, "learning_rate": 5.659167383398972e-06, "loss": 0.2923, "step": 75368 }, { "epoch": 4.55, "learning_rate": 5.657658473284747e-06, "loss": 0.2198, "step": 75370 }, { "epoch": 4.55, "learning_rate": 5.656149563170522e-06, "loss": 0.3351, "step": 75372 }, { "epoch": 4.55, "learning_rate": 5.654640653056298e-06, "loss": 0.1393, "step": 75374 }, { "epoch": 4.55, "learning_rate": 5.653131742942074e-06, "loss": 0.2697, "step": 75376 }, { "epoch": 4.55, "learning_rate": 5.6516228328278485e-06, "loss": 0.2498, "step": 75378 }, { "epoch": 4.55, "learning_rate": 5.650113922713624e-06, "loss": 0.3684, "step": 75380 }, { "epoch": 4.55, "learning_rate": 5.6486050125994e-06, "loss": 0.2365, "step": 75382 }, { "epoch": 4.55, "learning_rate": 5.6470961024851755e-06, "loss": 0.2866, "step": 75384 }, { "epoch": 4.55, "learning_rate": 5.64558719237095e-06, "loss": 0.3521, "step": 75386 }, { "epoch": 4.55, "learning_rate": 5.644078282256726e-06, "loss": 0.205, "step": 75388 }, { "epoch": 4.55, "learning_rate": 5.642569372142502e-06, "loss": 0.2556, "step": 75390 }, { "epoch": 4.55, "learning_rate": 5.641060462028277e-06, "loss": 0.256, "step": 75392 }, { "epoch": 4.55, "learning_rate": 5.639551551914052e-06, "loss": 0.302, "step": 75394 }, { "epoch": 4.55, "learning_rate": 5.638042641799828e-06, "loss": 0.2334, "step": 75396 }, { "epoch": 4.55, "learning_rate": 5.6365337316856035e-06, "loss": 0.2341, "step": 75398 }, { "epoch": 4.55, "learning_rate": 5.635024821571379e-06, "loss": 0.197, "step": 75400 }, { "epoch": 4.55, "learning_rate": 5.633515911457155e-06, "loss": 0.2835, "step": 75402 }, { "epoch": 4.55, "learning_rate": 5.6320070013429305e-06, "loss": 0.2112, "step": 75404 }, { "epoch": 4.55, "learning_rate": 5.630498091228706e-06, "loss": 0.1686, "step": 75406 }, { "epoch": 4.55, "learning_rate": 5.628989181114482e-06, "loss": 0.1864, "step": 75408 }, { "epoch": 4.55, "learning_rate": 5.627480271000257e-06, "loss": 0.2585, "step": 75410 }, { "epoch": 4.55, "learning_rate": 5.625971360886032e-06, "loss": 0.3236, "step": 75412 }, { "epoch": 4.55, "learning_rate": 5.624462450771808e-06, "loss": 0.2324, "step": 75414 }, { "epoch": 4.55, "learning_rate": 5.622953540657584e-06, "loss": 0.242, "step": 75416 }, { "epoch": 4.55, "learning_rate": 5.6214446305433585e-06, "loss": 0.2952, "step": 75418 }, { "epoch": 4.55, "learning_rate": 5.619935720429134e-06, "loss": 0.1813, "step": 75420 }, { "epoch": 4.55, "learning_rate": 5.61842681031491e-06, "loss": 0.3011, "step": 75422 }, { "epoch": 4.55, "learning_rate": 5.6169179002006855e-06, "loss": 0.3303, "step": 75424 }, { "epoch": 4.55, "learning_rate": 5.61540899008646e-06, "loss": 0.2232, "step": 75426 }, { "epoch": 4.55, "learning_rate": 5.613900079972236e-06, "loss": 0.2865, "step": 75428 }, { "epoch": 4.55, "learning_rate": 5.612391169858012e-06, "loss": 0.2438, "step": 75430 }, { "epoch": 4.55, "learning_rate": 5.610882259743787e-06, "loss": 0.2806, "step": 75432 }, { "epoch": 4.55, "learning_rate": 5.609373349629562e-06, "loss": 0.2156, "step": 75434 }, { "epoch": 4.55, "learning_rate": 5.607864439515338e-06, "loss": 0.2808, "step": 75436 }, { "epoch": 4.55, "learning_rate": 5.6063555294011135e-06, "loss": 0.271, "step": 75438 }, { "epoch": 4.55, "learning_rate": 5.604846619286889e-06, "loss": 0.2025, "step": 75440 }, { "epoch": 4.55, "learning_rate": 5.603337709172665e-06, "loss": 0.1346, "step": 75442 }, { "epoch": 4.55, "learning_rate": 5.6018287990584405e-06, "loss": 0.1737, "step": 75444 }, { "epoch": 4.55, "learning_rate": 5.600319888944216e-06, "loss": 0.2662, "step": 75446 }, { "epoch": 4.55, "learning_rate": 5.598810978829992e-06, "loss": 0.2276, "step": 75448 }, { "epoch": 4.55, "learning_rate": 5.5973020687157675e-06, "loss": 0.1895, "step": 75450 }, { "epoch": 4.55, "learning_rate": 5.595793158601542e-06, "loss": 0.28, "step": 75452 }, { "epoch": 4.55, "learning_rate": 5.594284248487318e-06, "loss": 0.2338, "step": 75454 }, { "epoch": 4.55, "learning_rate": 5.592775338373094e-06, "loss": 0.1542, "step": 75456 }, { "epoch": 4.55, "learning_rate": 5.591266428258869e-06, "loss": 0.2396, "step": 75458 }, { "epoch": 4.55, "learning_rate": 5.589757518144644e-06, "loss": 0.1191, "step": 75460 }, { "epoch": 4.55, "learning_rate": 5.58824860803042e-06, "loss": 0.2365, "step": 75462 }, { "epoch": 4.55, "learning_rate": 5.5867396979161954e-06, "loss": 0.3472, "step": 75464 }, { "epoch": 4.55, "learning_rate": 5.585230787801971e-06, "loss": 0.2403, "step": 75466 }, { "epoch": 4.56, "learning_rate": 5.583721877687746e-06, "loss": 0.259, "step": 75468 }, { "epoch": 4.56, "learning_rate": 5.582212967573522e-06, "loss": 0.3187, "step": 75470 }, { "epoch": 4.56, "learning_rate": 5.580704057459297e-06, "loss": 0.2977, "step": 75472 }, { "epoch": 4.56, "learning_rate": 5.579195147345073e-06, "loss": 0.2184, "step": 75474 }, { "epoch": 4.56, "learning_rate": 5.577686237230849e-06, "loss": 0.273, "step": 75476 }, { "epoch": 4.56, "learning_rate": 5.576177327116624e-06, "loss": 0.2923, "step": 75478 }, { "epoch": 4.56, "learning_rate": 5.5746684170024e-06, "loss": 0.3407, "step": 75480 }, { "epoch": 4.56, "learning_rate": 5.573159506888176e-06, "loss": 0.2331, "step": 75482 }, { "epoch": 4.56, "learning_rate": 5.57165059677395e-06, "loss": 0.2666, "step": 75484 }, { "epoch": 4.56, "learning_rate": 5.570141686659726e-06, "loss": 0.1626, "step": 75486 }, { "epoch": 4.56, "learning_rate": 5.568632776545502e-06, "loss": 0.1767, "step": 75488 }, { "epoch": 4.56, "learning_rate": 5.567123866431277e-06, "loss": 0.2538, "step": 75490 }, { "epoch": 4.56, "learning_rate": 5.565614956317052e-06, "loss": 0.1114, "step": 75492 }, { "epoch": 4.56, "learning_rate": 5.564106046202828e-06, "loss": 0.2538, "step": 75494 }, { "epoch": 4.56, "learning_rate": 5.5625971360886036e-06, "loss": 0.2018, "step": 75496 }, { "epoch": 4.56, "learning_rate": 5.561088225974379e-06, "loss": 0.1649, "step": 75498 }, { "epoch": 4.56, "learning_rate": 5.559579315860154e-06, "loss": 0.1819, "step": 75500 }, { "epoch": 4.56, "learning_rate": 5.55807040574593e-06, "loss": 0.2313, "step": 75502 }, { "epoch": 4.56, "learning_rate": 5.556561495631705e-06, "loss": 0.204, "step": 75504 }, { "epoch": 4.56, "learning_rate": 5.555052585517481e-06, "loss": 0.2646, "step": 75506 }, { "epoch": 4.56, "learning_rate": 5.553543675403256e-06, "loss": 0.18, "step": 75508 }, { "epoch": 4.56, "learning_rate": 5.5520347652890315e-06, "loss": 0.2927, "step": 75510 }, { "epoch": 4.56, "learning_rate": 5.550525855174807e-06, "loss": 0.1775, "step": 75512 }, { "epoch": 4.56, "learning_rate": 5.549016945060583e-06, "loss": 0.1485, "step": 75514 }, { "epoch": 4.56, "learning_rate": 5.5475080349463585e-06, "loss": 0.1938, "step": 75516 }, { "epoch": 4.56, "learning_rate": 5.545999124832134e-06, "loss": 0.3067, "step": 75518 }, { "epoch": 4.56, "learning_rate": 5.54449021471791e-06, "loss": 0.1915, "step": 75520 }, { "epoch": 4.56, "learning_rate": 5.5429813046036855e-06, "loss": 0.2268, "step": 75522 }, { "epoch": 4.56, "learning_rate": 5.541472394489461e-06, "loss": 0.2691, "step": 75524 }, { "epoch": 4.56, "learning_rate": 5.539963484375236e-06, "loss": 0.198, "step": 75526 }, { "epoch": 4.56, "learning_rate": 5.538454574261012e-06, "loss": 0.2585, "step": 75528 }, { "epoch": 4.56, "learning_rate": 5.536945664146787e-06, "loss": 0.3271, "step": 75530 }, { "epoch": 4.56, "learning_rate": 5.535436754032563e-06, "loss": 0.1657, "step": 75532 }, { "epoch": 4.56, "learning_rate": 5.533927843918338e-06, "loss": 0.2278, "step": 75534 }, { "epoch": 4.56, "learning_rate": 5.5324189338041135e-06, "loss": 0.1635, "step": 75536 }, { "epoch": 4.56, "learning_rate": 5.530910023689889e-06, "loss": 0.1568, "step": 75538 }, { "epoch": 4.56, "learning_rate": 5.529401113575665e-06, "loss": 0.2331, "step": 75540 }, { "epoch": 4.56, "learning_rate": 5.52789220346144e-06, "loss": 0.189, "step": 75542 }, { "epoch": 4.56, "learning_rate": 5.526383293347215e-06, "loss": 0.2167, "step": 75544 }, { "epoch": 4.56, "learning_rate": 5.524874383232991e-06, "loss": 0.2831, "step": 75546 }, { "epoch": 4.56, "learning_rate": 5.523365473118767e-06, "loss": 0.1586, "step": 75548 }, { "epoch": 4.56, "learning_rate": 5.5218565630045415e-06, "loss": 0.2519, "step": 75550 }, { "epoch": 4.56, "learning_rate": 5.520347652890317e-06, "loss": 0.2947, "step": 75552 }, { "epoch": 4.56, "learning_rate": 5.518838742776093e-06, "loss": 0.2473, "step": 75554 }, { "epoch": 4.56, "learning_rate": 5.5173298326618685e-06, "loss": 0.2489, "step": 75556 }, { "epoch": 4.56, "learning_rate": 5.515820922547644e-06, "loss": 0.2928, "step": 75558 }, { "epoch": 4.56, "learning_rate": 5.51431201243342e-06, "loss": 0.2754, "step": 75560 }, { "epoch": 4.56, "learning_rate": 5.5128031023191955e-06, "loss": 0.2891, "step": 75562 }, { "epoch": 4.56, "learning_rate": 5.511294192204971e-06, "loss": 0.2789, "step": 75564 }, { "epoch": 4.56, "learning_rate": 5.509785282090746e-06, "loss": 0.2742, "step": 75566 }, { "epoch": 4.56, "learning_rate": 5.508276371976522e-06, "loss": 0.1914, "step": 75568 }, { "epoch": 4.56, "learning_rate": 5.506767461862297e-06, "loss": 0.2952, "step": 75570 }, { "epoch": 4.56, "learning_rate": 5.505258551748073e-06, "loss": 0.3003, "step": 75572 }, { "epoch": 4.56, "learning_rate": 5.503749641633848e-06, "loss": 0.2335, "step": 75574 }, { "epoch": 4.56, "learning_rate": 5.5022407315196234e-06, "loss": 0.1824, "step": 75576 }, { "epoch": 4.56, "learning_rate": 5.500731821405399e-06, "loss": 0.2187, "step": 75578 }, { "epoch": 4.56, "learning_rate": 5.499222911291175e-06, "loss": 0.2407, "step": 75580 }, { "epoch": 4.56, "learning_rate": 5.49771400117695e-06, "loss": 0.1804, "step": 75582 }, { "epoch": 4.56, "learning_rate": 5.496205091062725e-06, "loss": 0.2194, "step": 75584 }, { "epoch": 4.56, "learning_rate": 5.494696180948501e-06, "loss": 0.2224, "step": 75586 }, { "epoch": 4.56, "learning_rate": 5.493187270834277e-06, "loss": 0.2234, "step": 75588 }, { "epoch": 4.56, "learning_rate": 5.491678360720052e-06, "loss": 0.238, "step": 75590 }, { "epoch": 4.56, "learning_rate": 5.490169450605828e-06, "loss": 0.2873, "step": 75592 }, { "epoch": 4.56, "learning_rate": 5.488660540491604e-06, "loss": 0.2137, "step": 75594 }, { "epoch": 4.56, "learning_rate": 5.487151630377379e-06, "loss": 0.3096, "step": 75596 }, { "epoch": 4.56, "learning_rate": 5.485642720263154e-06, "loss": 0.1949, "step": 75598 }, { "epoch": 4.56, "learning_rate": 5.48413381014893e-06, "loss": 0.2162, "step": 75600 }, { "epoch": 4.56, "learning_rate": 5.482624900034705e-06, "loss": 0.3121, "step": 75602 }, { "epoch": 4.56, "learning_rate": 5.481115989920481e-06, "loss": 0.307, "step": 75604 }, { "epoch": 4.56, "learning_rate": 5.479607079806257e-06, "loss": 0.1521, "step": 75606 }, { "epoch": 4.56, "learning_rate": 5.4780981696920316e-06, "loss": 0.2591, "step": 75608 }, { "epoch": 4.56, "learning_rate": 5.476589259577807e-06, "loss": 0.1899, "step": 75610 }, { "epoch": 4.56, "learning_rate": 5.475080349463583e-06, "loss": 0.2334, "step": 75612 }, { "epoch": 4.56, "learning_rate": 5.4735714393493586e-06, "loss": 0.1119, "step": 75614 }, { "epoch": 4.56, "learning_rate": 5.472062529235133e-06, "loss": 0.3145, "step": 75616 }, { "epoch": 4.56, "learning_rate": 5.470553619120909e-06, "loss": 0.2635, "step": 75618 }, { "epoch": 4.56, "learning_rate": 5.469044709006685e-06, "loss": 0.23, "step": 75620 }, { "epoch": 4.56, "learning_rate": 5.46753579889246e-06, "loss": 0.1846, "step": 75622 }, { "epoch": 4.56, "learning_rate": 5.466026888778235e-06, "loss": 0.208, "step": 75624 }, { "epoch": 4.56, "learning_rate": 5.464517978664011e-06, "loss": 0.1723, "step": 75626 }, { "epoch": 4.56, "learning_rate": 5.4630090685497865e-06, "loss": 0.2465, "step": 75628 }, { "epoch": 4.56, "learning_rate": 5.461500158435562e-06, "loss": 0.2522, "step": 75630 }, { "epoch": 4.56, "learning_rate": 5.459991248321338e-06, "loss": 0.1779, "step": 75632 }, { "epoch": 4.57, "learning_rate": 5.4584823382071135e-06, "loss": 0.1767, "step": 75634 }, { "epoch": 4.57, "learning_rate": 5.456973428092889e-06, "loss": 0.2511, "step": 75636 }, { "epoch": 4.57, "learning_rate": 5.455464517978665e-06, "loss": 0.2313, "step": 75638 }, { "epoch": 4.57, "learning_rate": 5.45395560786444e-06, "loss": 0.2528, "step": 75640 }, { "epoch": 4.57, "learning_rate": 5.452446697750215e-06, "loss": 0.2179, "step": 75642 }, { "epoch": 4.57, "learning_rate": 5.450937787635991e-06, "loss": 0.2458, "step": 75644 }, { "epoch": 4.57, "learning_rate": 5.449428877521767e-06, "loss": 0.3382, "step": 75646 }, { "epoch": 4.57, "learning_rate": 5.4479199674075415e-06, "loss": 0.368, "step": 75648 }, { "epoch": 4.57, "learning_rate": 5.446411057293317e-06, "loss": 0.1302, "step": 75650 }, { "epoch": 4.57, "learning_rate": 5.444902147179093e-06, "loss": 0.2709, "step": 75652 }, { "epoch": 4.57, "learning_rate": 5.4433932370648685e-06, "loss": 0.2634, "step": 75654 }, { "epoch": 4.57, "learning_rate": 5.441884326950643e-06, "loss": 0.2395, "step": 75656 }, { "epoch": 4.57, "learning_rate": 5.440375416836419e-06, "loss": 0.2953, "step": 75658 }, { "epoch": 4.57, "learning_rate": 5.438866506722195e-06, "loss": 0.1781, "step": 75660 }, { "epoch": 4.57, "learning_rate": 5.43735759660797e-06, "loss": 0.3095, "step": 75662 }, { "epoch": 4.57, "learning_rate": 5.435848686493745e-06, "loss": 0.3014, "step": 75664 }, { "epoch": 4.57, "learning_rate": 5.434339776379521e-06, "loss": 0.2358, "step": 75666 }, { "epoch": 4.57, "learning_rate": 5.4328308662652965e-06, "loss": 0.2404, "step": 75668 }, { "epoch": 4.57, "learning_rate": 5.431321956151072e-06, "loss": 0.291, "step": 75670 }, { "epoch": 4.57, "learning_rate": 5.429813046036848e-06, "loss": 0.2609, "step": 75672 }, { "epoch": 4.57, "learning_rate": 5.4283041359226235e-06, "loss": 0.1805, "step": 75674 }, { "epoch": 4.57, "learning_rate": 5.426795225808399e-06, "loss": 0.2461, "step": 75676 }, { "epoch": 4.57, "learning_rate": 5.425286315694175e-06, "loss": 0.3018, "step": 75678 }, { "epoch": 4.57, "learning_rate": 5.4237774055799505e-06, "loss": 0.1022, "step": 75680 }, { "epoch": 4.57, "learning_rate": 5.422268495465725e-06, "loss": 0.2617, "step": 75682 }, { "epoch": 4.57, "learning_rate": 5.420759585351501e-06, "loss": 0.1926, "step": 75684 }, { "epoch": 4.57, "learning_rate": 5.419250675237277e-06, "loss": 0.2134, "step": 75686 }, { "epoch": 4.57, "learning_rate": 5.417741765123052e-06, "loss": 0.155, "step": 75688 }, { "epoch": 4.57, "learning_rate": 5.416232855008827e-06, "loss": 0.187, "step": 75690 }, { "epoch": 4.57, "learning_rate": 5.414723944894603e-06, "loss": 0.2122, "step": 75692 }, { "epoch": 4.57, "learning_rate": 5.4132150347803785e-06, "loss": 0.1931, "step": 75694 }, { "epoch": 4.57, "learning_rate": 5.411706124666154e-06, "loss": 0.2385, "step": 75696 }, { "epoch": 4.57, "learning_rate": 5.410197214551929e-06, "loss": 0.2087, "step": 75698 }, { "epoch": 4.57, "learning_rate": 5.408688304437705e-06, "loss": 0.1936, "step": 75700 }, { "epoch": 4.57, "learning_rate": 5.40717939432348e-06, "loss": 0.2655, "step": 75702 }, { "epoch": 4.57, "learning_rate": 5.405670484209256e-06, "loss": 0.1376, "step": 75704 }, { "epoch": 4.57, "learning_rate": 5.404161574095031e-06, "loss": 0.279, "step": 75706 }, { "epoch": 4.57, "learning_rate": 5.402652663980807e-06, "loss": 0.2504, "step": 75708 }, { "epoch": 4.57, "learning_rate": 5.401143753866583e-06, "loss": 0.2522, "step": 75710 }, { "epoch": 4.57, "learning_rate": 5.399634843752359e-06, "loss": 0.2281, "step": 75712 }, { "epoch": 4.57, "learning_rate": 5.3981259336381334e-06, "loss": 0.1186, "step": 75714 }, { "epoch": 4.57, "learning_rate": 5.396617023523909e-06, "loss": 0.0894, "step": 75716 }, { "epoch": 4.57, "learning_rate": 5.395108113409685e-06, "loss": 0.2032, "step": 75718 }, { "epoch": 4.57, "learning_rate": 5.3935992032954604e-06, "loss": 0.1896, "step": 75720 }, { "epoch": 4.57, "learning_rate": 5.392090293181235e-06, "loss": 0.1868, "step": 75722 }, { "epoch": 4.57, "learning_rate": 5.390581383067011e-06, "loss": 0.1922, "step": 75724 }, { "epoch": 4.57, "learning_rate": 5.389072472952787e-06, "loss": 0.1857, "step": 75726 }, { "epoch": 4.57, "learning_rate": 5.387563562838562e-06, "loss": 0.2281, "step": 75728 }, { "epoch": 4.57, "learning_rate": 5.386054652724337e-06, "loss": 0.1554, "step": 75730 }, { "epoch": 4.57, "learning_rate": 5.384545742610113e-06, "loss": 0.1747, "step": 75732 }, { "epoch": 4.57, "learning_rate": 5.383036832495888e-06, "loss": 0.237, "step": 75734 }, { "epoch": 4.57, "learning_rate": 5.381527922381664e-06, "loss": 0.36, "step": 75736 }, { "epoch": 4.57, "learning_rate": 5.380019012267439e-06, "loss": 0.1555, "step": 75738 }, { "epoch": 4.57, "learning_rate": 5.3785101021532146e-06, "loss": 0.1948, "step": 75740 }, { "epoch": 4.57, "learning_rate": 5.37700119203899e-06, "loss": 0.1586, "step": 75742 }, { "epoch": 4.57, "learning_rate": 5.375492281924766e-06, "loss": 0.2702, "step": 75744 }, { "epoch": 4.57, "learning_rate": 5.3739833718105416e-06, "loss": 0.219, "step": 75746 }, { "epoch": 4.57, "learning_rate": 5.372474461696317e-06, "loss": 0.2284, "step": 75748 }, { "epoch": 4.57, "learning_rate": 5.370965551582093e-06, "loss": 0.3236, "step": 75750 }, { "epoch": 4.57, "learning_rate": 5.3694566414678686e-06, "loss": 0.136, "step": 75752 }, { "epoch": 4.57, "learning_rate": 5.367947731353643e-06, "loss": 0.19, "step": 75754 }, { "epoch": 4.57, "learning_rate": 5.366438821239419e-06, "loss": 0.2208, "step": 75756 }, { "epoch": 4.57, "learning_rate": 5.364929911125195e-06, "loss": 0.3018, "step": 75758 }, { "epoch": 4.57, "learning_rate": 5.36342100101097e-06, "loss": 0.2808, "step": 75760 }, { "epoch": 4.57, "learning_rate": 5.361912090896746e-06, "loss": 0.2527, "step": 75762 }, { "epoch": 4.57, "learning_rate": 5.360403180782521e-06, "loss": 0.2713, "step": 75764 }, { "epoch": 4.57, "learning_rate": 5.3588942706682965e-06, "loss": 0.1941, "step": 75766 }, { "epoch": 4.57, "learning_rate": 5.357385360554072e-06, "loss": 0.1479, "step": 75768 }, { "epoch": 4.57, "learning_rate": 5.355876450439848e-06, "loss": 0.2313, "step": 75770 }, { "epoch": 4.57, "learning_rate": 5.354367540325623e-06, "loss": 0.1912, "step": 75772 }, { "epoch": 4.57, "learning_rate": 5.352858630211398e-06, "loss": 0.1862, "step": 75774 }, { "epoch": 4.57, "learning_rate": 5.351349720097174e-06, "loss": 0.2084, "step": 75776 }, { "epoch": 4.57, "learning_rate": 5.34984080998295e-06, "loss": 0.2225, "step": 75778 }, { "epoch": 4.57, "learning_rate": 5.3483318998687245e-06, "loss": 0.2003, "step": 75780 }, { "epoch": 4.57, "learning_rate": 5.3468229897545e-06, "loss": 0.141, "step": 75782 }, { "epoch": 4.57, "learning_rate": 5.345314079640276e-06, "loss": 0.3661, "step": 75784 }, { "epoch": 4.57, "learning_rate": 5.3438051695260515e-06, "loss": 0.245, "step": 75786 }, { "epoch": 4.57, "learning_rate": 5.342296259411827e-06, "loss": 0.1747, "step": 75788 }, { "epoch": 4.57, "learning_rate": 5.340787349297603e-06, "loss": 0.2195, "step": 75790 }, { "epoch": 4.57, "learning_rate": 5.3392784391833785e-06, "loss": 0.242, "step": 75792 }, { "epoch": 4.57, "learning_rate": 5.337769529069154e-06, "loss": 0.2931, "step": 75794 }, { "epoch": 4.57, "learning_rate": 5.336260618954929e-06, "loss": 0.3652, "step": 75796 }, { "epoch": 4.57, "learning_rate": 5.334751708840705e-06, "loss": 0.1505, "step": 75798 }, { "epoch": 4.58, "learning_rate": 5.33324279872648e-06, "loss": 0.1938, "step": 75800 }, { "epoch": 4.58, "learning_rate": 5.331733888612256e-06, "loss": 0.1812, "step": 75802 }, { "epoch": 4.58, "learning_rate": 5.330224978498031e-06, "loss": 0.1538, "step": 75804 }, { "epoch": 4.58, "learning_rate": 5.3287160683838065e-06, "loss": 0.2002, "step": 75806 }, { "epoch": 4.58, "learning_rate": 5.327207158269582e-06, "loss": 0.1932, "step": 75808 }, { "epoch": 4.58, "learning_rate": 5.325698248155358e-06, "loss": 0.2842, "step": 75810 }, { "epoch": 4.58, "learning_rate": 5.324189338041133e-06, "loss": 0.1901, "step": 75812 }, { "epoch": 4.58, "learning_rate": 5.322680427926908e-06, "loss": 0.3054, "step": 75814 }, { "epoch": 4.58, "learning_rate": 5.321171517812684e-06, "loss": 0.1838, "step": 75816 }, { "epoch": 4.58, "learning_rate": 5.31966260769846e-06, "loss": 0.1781, "step": 75818 }, { "epoch": 4.58, "learning_rate": 5.3181536975842344e-06, "loss": 0.1756, "step": 75820 }, { "epoch": 4.58, "learning_rate": 5.31664478747001e-06, "loss": 0.1765, "step": 75822 }, { "epoch": 4.58, "learning_rate": 5.315135877355786e-06, "loss": 0.23, "step": 75824 }, { "epoch": 4.58, "learning_rate": 5.313626967241562e-06, "loss": 0.2783, "step": 75826 }, { "epoch": 4.58, "learning_rate": 5.312118057127337e-06, "loss": 0.2825, "step": 75828 }, { "epoch": 4.58, "learning_rate": 5.310609147013113e-06, "loss": 0.3018, "step": 75830 }, { "epoch": 4.58, "learning_rate": 5.3091002368988884e-06, "loss": 0.2471, "step": 75832 }, { "epoch": 4.58, "learning_rate": 5.307591326784664e-06, "loss": 0.2899, "step": 75834 }, { "epoch": 4.58, "learning_rate": 5.30608241667044e-06, "loss": 0.2338, "step": 75836 }, { "epoch": 4.58, "learning_rate": 5.304573506556215e-06, "loss": 0.1713, "step": 75838 }, { "epoch": 4.58, "learning_rate": 5.30306459644199e-06, "loss": 0.2876, "step": 75840 }, { "epoch": 4.58, "learning_rate": 5.301555686327766e-06, "loss": 0.2371, "step": 75842 }, { "epoch": 4.58, "learning_rate": 5.300046776213542e-06, "loss": 0.1404, "step": 75844 }, { "epoch": 4.58, "learning_rate": 5.298537866099316e-06, "loss": 0.1928, "step": 75846 }, { "epoch": 4.58, "learning_rate": 5.297028955985092e-06, "loss": 0.193, "step": 75848 }, { "epoch": 4.58, "learning_rate": 5.295520045870868e-06, "loss": 0.2954, "step": 75850 }, { "epoch": 4.58, "learning_rate": 5.294011135756643e-06, "loss": 0.1565, "step": 75852 }, { "epoch": 4.58, "learning_rate": 5.292502225642418e-06, "loss": 0.2033, "step": 75854 }, { "epoch": 4.58, "learning_rate": 5.290993315528194e-06, "loss": 0.1946, "step": 75856 }, { "epoch": 4.58, "learning_rate": 5.2894844054139696e-06, "loss": 0.2433, "step": 75858 }, { "epoch": 4.58, "learning_rate": 5.287975495299745e-06, "loss": 0.2277, "step": 75860 }, { "epoch": 4.58, "learning_rate": 5.286466585185521e-06, "loss": 0.2395, "step": 75862 }, { "epoch": 4.58, "learning_rate": 5.2849576750712966e-06, "loss": 0.2784, "step": 75864 }, { "epoch": 4.58, "learning_rate": 5.283448764957072e-06, "loss": 0.1998, "step": 75866 }, { "epoch": 4.58, "learning_rate": 5.281939854842848e-06, "loss": 0.2613, "step": 75868 }, { "epoch": 4.58, "learning_rate": 5.280430944728623e-06, "loss": 0.1865, "step": 75870 }, { "epoch": 4.58, "learning_rate": 5.278922034614398e-06, "loss": 0.3181, "step": 75872 }, { "epoch": 4.58, "learning_rate": 5.277413124500174e-06, "loss": 0.1364, "step": 75874 }, { "epoch": 4.58, "learning_rate": 5.27590421438595e-06, "loss": 0.2419, "step": 75876 }, { "epoch": 4.58, "learning_rate": 5.2743953042717245e-06, "loss": 0.2267, "step": 75878 }, { "epoch": 4.58, "learning_rate": 5.2728863941575e-06, "loss": 0.2064, "step": 75880 }, { "epoch": 4.58, "learning_rate": 5.271377484043276e-06, "loss": 0.2858, "step": 75882 }, { "epoch": 4.58, "learning_rate": 5.2698685739290515e-06, "loss": 0.3489, "step": 75884 }, { "epoch": 4.58, "learning_rate": 5.268359663814826e-06, "loss": 0.2603, "step": 75886 }, { "epoch": 4.58, "learning_rate": 5.266850753700602e-06, "loss": 0.2677, "step": 75888 }, { "epoch": 4.58, "learning_rate": 5.265341843586378e-06, "loss": 0.3264, "step": 75890 }, { "epoch": 4.58, "learning_rate": 5.263832933472153e-06, "loss": 0.2724, "step": 75892 }, { "epoch": 4.58, "learning_rate": 5.262324023357928e-06, "loss": 0.2184, "step": 75894 }, { "epoch": 4.58, "learning_rate": 5.260815113243704e-06, "loss": 0.22, "step": 75896 }, { "epoch": 4.58, "learning_rate": 5.2593062031294795e-06, "loss": 0.1692, "step": 75898 }, { "epoch": 4.58, "learning_rate": 5.257797293015255e-06, "loss": 0.2648, "step": 75900 }, { "epoch": 4.58, "learning_rate": 5.256288382901031e-06, "loss": 0.1816, "step": 75902 }, { "epoch": 4.58, "learning_rate": 5.2547794727868065e-06, "loss": 0.2167, "step": 75904 }, { "epoch": 4.58, "learning_rate": 5.253270562672582e-06, "loss": 0.1747, "step": 75906 }, { "epoch": 4.58, "learning_rate": 5.251761652558358e-06, "loss": 0.1813, "step": 75908 }, { "epoch": 4.58, "learning_rate": 5.250252742444133e-06, "loss": 0.1888, "step": 75910 }, { "epoch": 4.58, "learning_rate": 5.248743832329908e-06, "loss": 0.2118, "step": 75912 }, { "epoch": 4.58, "learning_rate": 5.247234922215684e-06, "loss": 0.2036, "step": 75914 }, { "epoch": 4.58, "learning_rate": 5.24572601210146e-06, "loss": 0.1498, "step": 75916 }, { "epoch": 4.58, "learning_rate": 5.244217101987235e-06, "loss": 0.1986, "step": 75918 }, { "epoch": 4.58, "learning_rate": 5.24270819187301e-06, "loss": 0.2534, "step": 75920 }, { "epoch": 4.58, "learning_rate": 5.241199281758786e-06, "loss": 0.1744, "step": 75922 }, { "epoch": 4.58, "learning_rate": 5.2396903716445615e-06, "loss": 0.1675, "step": 75924 }, { "epoch": 4.58, "learning_rate": 5.238181461530337e-06, "loss": 0.3182, "step": 75926 }, { "epoch": 4.58, "learning_rate": 5.236672551416112e-06, "loss": 0.2017, "step": 75928 }, { "epoch": 4.58, "learning_rate": 5.235163641301888e-06, "loss": 0.1568, "step": 75930 }, { "epoch": 4.58, "learning_rate": 5.233654731187663e-06, "loss": 0.1565, "step": 75932 }, { "epoch": 4.58, "learning_rate": 5.232145821073439e-06, "loss": 0.2193, "step": 75934 }, { "epoch": 4.58, "learning_rate": 5.230636910959214e-06, "loss": 0.2142, "step": 75936 }, { "epoch": 4.58, "learning_rate": 5.2291280008449895e-06, "loss": 0.2303, "step": 75938 }, { "epoch": 4.58, "learning_rate": 5.227619090730765e-06, "loss": 0.1726, "step": 75940 }, { "epoch": 4.58, "learning_rate": 5.226110180616541e-06, "loss": 0.1395, "step": 75942 }, { "epoch": 4.58, "learning_rate": 5.2246012705023165e-06, "loss": 0.3334, "step": 75944 }, { "epoch": 4.58, "learning_rate": 5.223092360388092e-06, "loss": 0.2594, "step": 75946 }, { "epoch": 4.58, "learning_rate": 5.221583450273868e-06, "loss": 0.2259, "step": 75948 }, { "epoch": 4.58, "learning_rate": 5.2200745401596435e-06, "loss": 0.2767, "step": 75950 }, { "epoch": 4.58, "learning_rate": 5.218565630045418e-06, "loss": 0.1573, "step": 75952 }, { "epoch": 4.58, "learning_rate": 5.217056719931194e-06, "loss": 0.2579, "step": 75954 }, { "epoch": 4.58, "learning_rate": 5.21554780981697e-06, "loss": 0.2439, "step": 75956 }, { "epoch": 4.58, "learning_rate": 5.214038899702745e-06, "loss": 0.2346, "step": 75958 }, { "epoch": 4.58, "learning_rate": 5.21252998958852e-06, "loss": 0.1695, "step": 75960 }, { "epoch": 4.58, "learning_rate": 5.211021079474296e-06, "loss": 0.3061, "step": 75962 }, { "epoch": 4.58, "learning_rate": 5.209512169360071e-06, "loss": 0.262, "step": 75964 }, { "epoch": 4.59, "learning_rate": 5.208003259245847e-06, "loss": 0.2287, "step": 75966 }, { "epoch": 4.59, "learning_rate": 5.206494349131622e-06, "loss": 0.3405, "step": 75968 }, { "epoch": 4.59, "learning_rate": 5.2049854390173976e-06, "loss": 0.1413, "step": 75970 }, { "epoch": 4.59, "learning_rate": 5.203476528903173e-06, "loss": 0.2775, "step": 75972 }, { "epoch": 4.59, "learning_rate": 5.201967618788949e-06, "loss": 0.3057, "step": 75974 }, { "epoch": 4.59, "learning_rate": 5.2004587086747246e-06, "loss": 0.2013, "step": 75976 }, { "epoch": 4.59, "learning_rate": 5.1989497985605e-06, "loss": 0.2346, "step": 75978 }, { "epoch": 4.59, "learning_rate": 5.197440888446276e-06, "loss": 0.2171, "step": 75980 }, { "epoch": 4.59, "learning_rate": 5.1959319783320516e-06, "loss": 0.1612, "step": 75982 }, { "epoch": 4.59, "learning_rate": 5.194423068217826e-06, "loss": 0.1827, "step": 75984 }, { "epoch": 4.59, "learning_rate": 5.192914158103602e-06, "loss": 0.4117, "step": 75986 }, { "epoch": 4.59, "learning_rate": 5.191405247989378e-06, "loss": 0.256, "step": 75988 }, { "epoch": 4.59, "learning_rate": 5.189896337875153e-06, "loss": 0.2568, "step": 75990 }, { "epoch": 4.59, "learning_rate": 5.188387427760929e-06, "loss": 0.2707, "step": 75992 }, { "epoch": 4.59, "learning_rate": 5.186878517646704e-06, "loss": 0.1723, "step": 75994 }, { "epoch": 4.59, "learning_rate": 5.1853696075324795e-06, "loss": 0.286, "step": 75996 }, { "epoch": 4.59, "learning_rate": 5.183860697418255e-06, "loss": 0.1756, "step": 75998 }, { "epoch": 4.59, "learning_rate": 5.182351787304031e-06, "loss": 0.1756, "step": 76000 }, { "epoch": 4.59, "learning_rate": 5.180842877189806e-06, "loss": 0.2587, "step": 76002 }, { "epoch": 4.59, "learning_rate": 5.179333967075581e-06, "loss": 0.2543, "step": 76004 }, { "epoch": 4.59, "learning_rate": 5.177825056961357e-06, "loss": 0.1554, "step": 76006 }, { "epoch": 4.59, "learning_rate": 5.176316146847133e-06, "loss": 0.1656, "step": 76008 }, { "epoch": 4.59, "learning_rate": 5.1748072367329075e-06, "loss": 0.3388, "step": 76010 }, { "epoch": 4.59, "learning_rate": 5.173298326618683e-06, "loss": 0.2049, "step": 76012 }, { "epoch": 4.59, "learning_rate": 5.171789416504459e-06, "loss": 0.1861, "step": 76014 }, { "epoch": 4.59, "learning_rate": 5.1702805063902345e-06, "loss": 0.276, "step": 76016 }, { "epoch": 4.59, "learning_rate": 5.16877159627601e-06, "loss": 0.2181, "step": 76018 }, { "epoch": 4.59, "learning_rate": 5.167262686161786e-06, "loss": 0.2906, "step": 76020 }, { "epoch": 4.59, "learning_rate": 5.1657537760475615e-06, "loss": 0.2574, "step": 76022 }, { "epoch": 4.59, "learning_rate": 5.164244865933337e-06, "loss": 0.3343, "step": 76024 }, { "epoch": 4.59, "learning_rate": 5.162735955819112e-06, "loss": 0.2681, "step": 76026 }, { "epoch": 4.59, "learning_rate": 5.161227045704888e-06, "loss": 0.2314, "step": 76028 }, { "epoch": 4.59, "learning_rate": 5.159718135590663e-06, "loss": 0.0987, "step": 76030 }, { "epoch": 4.59, "learning_rate": 5.158209225476439e-06, "loss": 0.1723, "step": 76032 }, { "epoch": 4.59, "learning_rate": 5.156700315362214e-06, "loss": 0.1383, "step": 76034 }, { "epoch": 4.59, "learning_rate": 5.1551914052479895e-06, "loss": 0.1857, "step": 76036 }, { "epoch": 4.59, "learning_rate": 5.153682495133765e-06, "loss": 0.2924, "step": 76038 }, { "epoch": 4.59, "learning_rate": 5.152173585019541e-06, "loss": 0.3382, "step": 76040 }, { "epoch": 4.59, "learning_rate": 5.150664674905316e-06, "loss": 0.1806, "step": 76042 }, { "epoch": 4.59, "learning_rate": 5.149155764791091e-06, "loss": 0.4744, "step": 76044 }, { "epoch": 4.59, "learning_rate": 5.147646854676867e-06, "loss": 0.2171, "step": 76046 }, { "epoch": 4.59, "learning_rate": 5.146137944562643e-06, "loss": 0.211, "step": 76048 }, { "epoch": 4.59, "learning_rate": 5.1446290344484175e-06, "loss": 0.2442, "step": 76050 }, { "epoch": 4.59, "learning_rate": 5.143120124334193e-06, "loss": 0.145, "step": 76052 }, { "epoch": 4.59, "learning_rate": 5.141611214219969e-06, "loss": 0.192, "step": 76054 }, { "epoch": 4.59, "learning_rate": 5.1401023041057445e-06, "loss": 0.2903, "step": 76056 }, { "epoch": 4.59, "learning_rate": 5.13859339399152e-06, "loss": 0.1614, "step": 76058 }, { "epoch": 4.59, "learning_rate": 5.137084483877296e-06, "loss": 0.26, "step": 76060 }, { "epoch": 4.59, "learning_rate": 5.1355755737630715e-06, "loss": 0.2762, "step": 76062 }, { "epoch": 4.59, "learning_rate": 5.134066663648847e-06, "loss": 0.2943, "step": 76064 }, { "epoch": 4.59, "learning_rate": 5.132557753534622e-06, "loss": 0.3174, "step": 76066 }, { "epoch": 4.59, "learning_rate": 5.131048843420398e-06, "loss": 0.3327, "step": 76068 }, { "epoch": 4.59, "learning_rate": 5.129539933306173e-06, "loss": 0.1989, "step": 76070 }, { "epoch": 4.59, "learning_rate": 5.128031023191949e-06, "loss": 0.2693, "step": 76072 }, { "epoch": 4.59, "learning_rate": 5.126522113077725e-06, "loss": 0.2314, "step": 76074 }, { "epoch": 4.59, "learning_rate": 5.1250132029634994e-06, "loss": 0.1626, "step": 76076 }, { "epoch": 4.59, "learning_rate": 5.123504292849275e-06, "loss": 0.2358, "step": 76078 }, { "epoch": 4.59, "learning_rate": 5.121995382735051e-06, "loss": 0.1655, "step": 76080 }, { "epoch": 4.59, "learning_rate": 5.1204864726208264e-06, "loss": 0.3041, "step": 76082 }, { "epoch": 4.59, "learning_rate": 5.118977562506601e-06, "loss": 0.4013, "step": 76084 }, { "epoch": 4.59, "learning_rate": 5.117468652392377e-06, "loss": 0.2685, "step": 76086 }, { "epoch": 4.59, "learning_rate": 5.115959742278153e-06, "loss": 0.3138, "step": 76088 }, { "epoch": 4.59, "learning_rate": 5.114450832163928e-06, "loss": 0.1424, "step": 76090 }, { "epoch": 4.59, "learning_rate": 5.112941922049704e-06, "loss": 0.2518, "step": 76092 }, { "epoch": 4.59, "learning_rate": 5.11143301193548e-06, "loss": 0.2069, "step": 76094 }, { "epoch": 4.59, "learning_rate": 5.109924101821255e-06, "loss": 0.2764, "step": 76096 }, { "epoch": 4.59, "learning_rate": 5.108415191707031e-06, "loss": 0.2302, "step": 76098 }, { "epoch": 4.59, "learning_rate": 5.106906281592806e-06, "loss": 0.1588, "step": 76100 }, { "epoch": 4.59, "learning_rate": 5.105397371478581e-06, "loss": 0.2552, "step": 76102 }, { "epoch": 4.59, "learning_rate": 5.103888461364357e-06, "loss": 0.2772, "step": 76104 }, { "epoch": 4.59, "learning_rate": 5.102379551250133e-06, "loss": 0.2473, "step": 76106 }, { "epoch": 4.59, "learning_rate": 5.1008706411359076e-06, "loss": 0.2352, "step": 76108 }, { "epoch": 4.59, "learning_rate": 5.099361731021683e-06, "loss": 0.1564, "step": 76110 }, { "epoch": 4.59, "learning_rate": 5.097852820907459e-06, "loss": 0.1654, "step": 76112 }, { "epoch": 4.59, "learning_rate": 5.0963439107932346e-06, "loss": 0.2729, "step": 76114 }, { "epoch": 4.59, "learning_rate": 5.094835000679009e-06, "loss": 0.243, "step": 76116 }, { "epoch": 4.59, "learning_rate": 5.093326090564785e-06, "loss": 0.3304, "step": 76118 }, { "epoch": 4.59, "learning_rate": 5.091817180450561e-06, "loss": 0.3056, "step": 76120 }, { "epoch": 4.59, "learning_rate": 5.090308270336336e-06, "loss": 0.2254, "step": 76122 }, { "epoch": 4.59, "learning_rate": 5.088799360222111e-06, "loss": 0.183, "step": 76124 }, { "epoch": 4.59, "learning_rate": 5.087290450107887e-06, "loss": 0.1835, "step": 76126 }, { "epoch": 4.59, "learning_rate": 5.0857815399936625e-06, "loss": 0.237, "step": 76128 }, { "epoch": 4.6, "learning_rate": 5.084272629879438e-06, "loss": 0.1555, "step": 76130 }, { "epoch": 4.6, "learning_rate": 5.082763719765214e-06, "loss": 0.1992, "step": 76132 }, { "epoch": 4.6, "learning_rate": 5.0812548096509895e-06, "loss": 0.139, "step": 76134 }, { "epoch": 4.6, "learning_rate": 5.079745899536765e-06, "loss": 0.2664, "step": 76136 }, { "epoch": 4.6, "learning_rate": 5.078236989422541e-06, "loss": 0.1365, "step": 76138 }, { "epoch": 4.6, "learning_rate": 5.076728079308316e-06, "loss": 0.2371, "step": 76140 }, { "epoch": 4.6, "learning_rate": 5.075219169194091e-06, "loss": 0.1698, "step": 76142 }, { "epoch": 4.6, "learning_rate": 5.073710259079867e-06, "loss": 0.1988, "step": 76144 }, { "epoch": 4.6, "learning_rate": 5.072201348965643e-06, "loss": 0.257, "step": 76146 }, { "epoch": 4.6, "learning_rate": 5.070692438851418e-06, "loss": 0.1964, "step": 76148 }, { "epoch": 4.6, "learning_rate": 5.069183528737193e-06, "loss": 0.2044, "step": 76150 }, { "epoch": 4.6, "learning_rate": 5.067674618622969e-06, "loss": 0.1909, "step": 76152 }, { "epoch": 4.6, "learning_rate": 5.0661657085087445e-06, "loss": 0.1827, "step": 76154 }, { "epoch": 4.6, "learning_rate": 5.06465679839452e-06, "loss": 0.242, "step": 76156 }, { "epoch": 4.6, "learning_rate": 5.063147888280295e-06, "loss": 0.3443, "step": 76158 }, { "epoch": 4.6, "learning_rate": 5.061638978166071e-06, "loss": 0.2111, "step": 76160 }, { "epoch": 4.6, "learning_rate": 5.060130068051846e-06, "loss": 0.1525, "step": 76162 }, { "epoch": 4.6, "learning_rate": 5.058621157937622e-06, "loss": 0.261, "step": 76164 }, { "epoch": 4.6, "learning_rate": 5.057112247823397e-06, "loss": 0.1899, "step": 76166 }, { "epoch": 4.6, "learning_rate": 5.0556033377091725e-06, "loss": 0.2592, "step": 76168 }, { "epoch": 4.6, "learning_rate": 5.054094427594948e-06, "loss": 0.2652, "step": 76170 }, { "epoch": 4.6, "learning_rate": 5.052585517480724e-06, "loss": 0.3591, "step": 76172 }, { "epoch": 4.6, "learning_rate": 5.0510766073664995e-06, "loss": 0.3466, "step": 76174 }, { "epoch": 4.6, "learning_rate": 5.049567697252275e-06, "loss": 0.2439, "step": 76176 }, { "epoch": 4.6, "learning_rate": 5.048058787138051e-06, "loss": 0.241, "step": 76178 }, { "epoch": 4.6, "learning_rate": 5.0465498770238265e-06, "loss": 0.1883, "step": 76180 }, { "epoch": 4.6, "learning_rate": 5.045040966909601e-06, "loss": 0.206, "step": 76182 }, { "epoch": 4.6, "learning_rate": 5.043532056795377e-06, "loss": 0.1505, "step": 76184 }, { "epoch": 4.6, "learning_rate": 5.042023146681153e-06, "loss": 0.2566, "step": 76186 }, { "epoch": 4.6, "learning_rate": 5.040514236566928e-06, "loss": 0.2575, "step": 76188 }, { "epoch": 4.6, "learning_rate": 5.039005326452703e-06, "loss": 0.3442, "step": 76190 }, { "epoch": 4.6, "learning_rate": 5.037496416338479e-06, "loss": 0.2116, "step": 76192 }, { "epoch": 4.6, "learning_rate": 5.0359875062242544e-06, "loss": 0.2833, "step": 76194 }, { "epoch": 4.6, "learning_rate": 5.03447859611003e-06, "loss": 0.2755, "step": 76196 }, { "epoch": 4.6, "learning_rate": 5.032969685995805e-06, "loss": 0.1706, "step": 76198 }, { "epoch": 4.6, "learning_rate": 5.031460775881581e-06, "loss": 0.3266, "step": 76200 }, { "epoch": 4.6, "learning_rate": 5.029951865767356e-06, "loss": 0.2621, "step": 76202 }, { "epoch": 4.6, "learning_rate": 5.028442955653132e-06, "loss": 0.2779, "step": 76204 }, { "epoch": 4.6, "learning_rate": 5.026934045538908e-06, "loss": 0.2304, "step": 76206 }, { "epoch": 4.6, "learning_rate": 5.025425135424683e-06, "loss": 0.1593, "step": 76208 }, { "epoch": 4.6, "learning_rate": 5.023916225310459e-06, "loss": 0.1964, "step": 76210 }, { "epoch": 4.6, "learning_rate": 5.022407315196235e-06, "loss": 0.1495, "step": 76212 }, { "epoch": 4.6, "learning_rate": 5.020898405082009e-06, "loss": 0.1848, "step": 76214 }, { "epoch": 4.6, "learning_rate": 5.019389494967785e-06, "loss": 0.1765, "step": 76216 }, { "epoch": 4.6, "learning_rate": 5.017880584853561e-06, "loss": 0.226, "step": 76218 }, { "epoch": 4.6, "learning_rate": 5.016371674739336e-06, "loss": 0.2246, "step": 76220 }, { "epoch": 4.6, "learning_rate": 5.014862764625111e-06, "loss": 0.216, "step": 76222 }, { "epoch": 4.6, "learning_rate": 5.013353854510887e-06, "loss": 0.1795, "step": 76224 }, { "epoch": 4.6, "learning_rate": 5.0118449443966626e-06, "loss": 0.2939, "step": 76226 }, { "epoch": 4.6, "learning_rate": 5.010336034282438e-06, "loss": 0.148, "step": 76228 }, { "epoch": 4.6, "learning_rate": 5.008827124168214e-06, "loss": 0.2111, "step": 76230 }, { "epoch": 4.6, "learning_rate": 5.007318214053989e-06, "loss": 0.1634, "step": 76232 }, { "epoch": 4.6, "learning_rate": 5.005809303939764e-06, "loss": 0.1979, "step": 76234 }, { "epoch": 4.6, "learning_rate": 5.00430039382554e-06, "loss": 0.2078, "step": 76236 }, { "epoch": 4.6, "learning_rate": 5.002791483711316e-06, "loss": 0.2196, "step": 76238 }, { "epoch": 4.6, "learning_rate": 5.0012825735970905e-06, "loss": 0.2734, "step": 76240 }, { "epoch": 4.6, "learning_rate": 4.999773663482866e-06, "loss": 0.1573, "step": 76242 }, { "epoch": 4.6, "learning_rate": 4.998264753368642e-06, "loss": 0.2114, "step": 76244 }, { "epoch": 4.6, "learning_rate": 4.9967558432544175e-06, "loss": 0.1755, "step": 76246 }, { "epoch": 4.6, "learning_rate": 4.995246933140193e-06, "loss": 0.1699, "step": 76248 }, { "epoch": 4.6, "learning_rate": 4.993738023025969e-06, "loss": 0.2543, "step": 76250 }, { "epoch": 4.6, "learning_rate": 4.9922291129117445e-06, "loss": 0.2148, "step": 76252 }, { "epoch": 4.6, "learning_rate": 4.99072020279752e-06, "loss": 0.2, "step": 76254 }, { "epoch": 4.6, "learning_rate": 4.989211292683295e-06, "loss": 0.1357, "step": 76256 }, { "epoch": 4.6, "learning_rate": 4.987702382569071e-06, "loss": 0.2525, "step": 76258 }, { "epoch": 4.6, "learning_rate": 4.986193472454846e-06, "loss": 0.2163, "step": 76260 }, { "epoch": 4.6, "learning_rate": 4.984684562340622e-06, "loss": 0.2144, "step": 76262 }, { "epoch": 4.6, "learning_rate": 4.983175652226397e-06, "loss": 0.153, "step": 76264 }, { "epoch": 4.6, "learning_rate": 4.9816667421121725e-06, "loss": 0.2412, "step": 76266 }, { "epoch": 4.6, "learning_rate": 4.980157831997948e-06, "loss": 0.2911, "step": 76268 }, { "epoch": 4.6, "learning_rate": 4.978648921883724e-06, "loss": 0.3031, "step": 76270 }, { "epoch": 4.6, "learning_rate": 4.977140011769499e-06, "loss": 0.2081, "step": 76272 }, { "epoch": 4.6, "learning_rate": 4.975631101655274e-06, "loss": 0.207, "step": 76274 }, { "epoch": 4.6, "learning_rate": 4.97412219154105e-06, "loss": 0.1329, "step": 76276 }, { "epoch": 4.6, "learning_rate": 4.972613281426826e-06, "loss": 0.1894, "step": 76278 }, { "epoch": 4.6, "learning_rate": 4.9711043713126005e-06, "loss": 0.2941, "step": 76280 }, { "epoch": 4.6, "learning_rate": 4.969595461198376e-06, "loss": 0.3507, "step": 76282 }, { "epoch": 4.6, "learning_rate": 4.968086551084152e-06, "loss": 0.1734, "step": 76284 }, { "epoch": 4.6, "learning_rate": 4.9665776409699275e-06, "loss": 0.2436, "step": 76286 }, { "epoch": 4.6, "learning_rate": 4.965068730855703e-06, "loss": 0.2315, "step": 76288 }, { "epoch": 4.6, "learning_rate": 4.963559820741479e-06, "loss": 0.1536, "step": 76290 }, { "epoch": 4.6, "learning_rate": 4.9620509106272545e-06, "loss": 0.2057, "step": 76292 }, { "epoch": 4.6, "learning_rate": 4.96054200051303e-06, "loss": 0.1868, "step": 76294 }, { "epoch": 4.61, "learning_rate": 4.959033090398805e-06, "loss": 0.1799, "step": 76296 }, { "epoch": 4.61, "learning_rate": 4.957524180284581e-06, "loss": 0.2441, "step": 76298 }, { "epoch": 4.61, "learning_rate": 4.956015270170356e-06, "loss": 0.2443, "step": 76300 }, { "epoch": 4.61, "learning_rate": 4.954506360056132e-06, "loss": 0.2652, "step": 76302 }, { "epoch": 4.61, "learning_rate": 4.952997449941908e-06, "loss": 0.2798, "step": 76304 }, { "epoch": 4.61, "learning_rate": 4.9514885398276825e-06, "loss": 0.2379, "step": 76306 }, { "epoch": 4.61, "learning_rate": 4.949979629713458e-06, "loss": 0.213, "step": 76308 }, { "epoch": 4.61, "learning_rate": 4.948470719599234e-06, "loss": 0.3032, "step": 76310 }, { "epoch": 4.61, "learning_rate": 4.9469618094850095e-06, "loss": 0.1467, "step": 76312 }, { "epoch": 4.61, "learning_rate": 4.945452899370784e-06, "loss": 0.1684, "step": 76314 }, { "epoch": 4.61, "learning_rate": 4.94394398925656e-06, "loss": 0.2374, "step": 76316 }, { "epoch": 4.61, "learning_rate": 4.942435079142336e-06, "loss": 0.1563, "step": 76318 }, { "epoch": 4.61, "learning_rate": 4.940926169028111e-06, "loss": 0.1954, "step": 76320 }, { "epoch": 4.61, "learning_rate": 4.939417258913886e-06, "loss": 0.257, "step": 76322 }, { "epoch": 4.61, "learning_rate": 4.937908348799663e-06, "loss": 0.2051, "step": 76324 }, { "epoch": 4.61, "learning_rate": 4.936399438685438e-06, "loss": 0.2041, "step": 76326 }, { "epoch": 4.61, "learning_rate": 4.934890528571214e-06, "loss": 0.1622, "step": 76328 }, { "epoch": 4.61, "learning_rate": 4.933381618456989e-06, "loss": 0.2917, "step": 76330 }, { "epoch": 4.61, "learning_rate": 4.9318727083427644e-06, "loss": 0.231, "step": 76332 }, { "epoch": 4.61, "learning_rate": 4.93036379822854e-06, "loss": 0.2394, "step": 76334 }, { "epoch": 4.61, "learning_rate": 4.928854888114316e-06, "loss": 0.0876, "step": 76336 }, { "epoch": 4.61, "learning_rate": 4.927345978000091e-06, "loss": 0.3472, "step": 76338 }, { "epoch": 4.61, "learning_rate": 4.925837067885866e-06, "loss": 0.2456, "step": 76340 }, { "epoch": 4.61, "learning_rate": 4.924328157771642e-06, "loss": 0.1814, "step": 76342 }, { "epoch": 4.61, "learning_rate": 4.922819247657418e-06, "loss": 0.1237, "step": 76344 }, { "epoch": 4.61, "learning_rate": 4.921310337543192e-06, "loss": 0.2488, "step": 76346 }, { "epoch": 4.61, "learning_rate": 4.919801427428968e-06, "loss": 0.2563, "step": 76348 }, { "epoch": 4.61, "learning_rate": 4.918292517314744e-06, "loss": 0.2479, "step": 76350 }, { "epoch": 4.61, "learning_rate": 4.916783607200519e-06, "loss": 0.1397, "step": 76352 }, { "epoch": 4.61, "learning_rate": 4.915274697086294e-06, "loss": 0.2514, "step": 76354 }, { "epoch": 4.61, "learning_rate": 4.91376578697207e-06, "loss": 0.2822, "step": 76356 }, { "epoch": 4.61, "learning_rate": 4.9122568768578456e-06, "loss": 0.23, "step": 76358 }, { "epoch": 4.61, "learning_rate": 4.910747966743621e-06, "loss": 0.1298, "step": 76360 }, { "epoch": 4.61, "learning_rate": 4.909239056629397e-06, "loss": 0.1577, "step": 76362 }, { "epoch": 4.61, "learning_rate": 4.9077301465151726e-06, "loss": 0.2159, "step": 76364 }, { "epoch": 4.61, "learning_rate": 4.906221236400948e-06, "loss": 0.144, "step": 76366 }, { "epoch": 4.61, "learning_rate": 4.904712326286724e-06, "loss": 0.2313, "step": 76368 }, { "epoch": 4.61, "learning_rate": 4.903203416172499e-06, "loss": 0.274, "step": 76370 }, { "epoch": 4.61, "learning_rate": 4.901694506058274e-06, "loss": 0.1788, "step": 76372 }, { "epoch": 4.61, "learning_rate": 4.90018559594405e-06, "loss": 0.2042, "step": 76374 }, { "epoch": 4.61, "learning_rate": 4.898676685829826e-06, "loss": 0.2653, "step": 76376 }, { "epoch": 4.61, "learning_rate": 4.897167775715601e-06, "loss": 0.1414, "step": 76378 }, { "epoch": 4.61, "learning_rate": 4.895658865601376e-06, "loss": 0.1672, "step": 76380 }, { "epoch": 4.61, "learning_rate": 4.894149955487152e-06, "loss": 0.267, "step": 76382 }, { "epoch": 4.61, "learning_rate": 4.8926410453729275e-06, "loss": 0.1894, "step": 76384 }, { "epoch": 4.61, "learning_rate": 4.891132135258703e-06, "loss": 0.0967, "step": 76386 }, { "epoch": 4.61, "learning_rate": 4.889623225144478e-06, "loss": 0.3031, "step": 76388 }, { "epoch": 4.61, "learning_rate": 4.888114315030254e-06, "loss": 0.1788, "step": 76390 }, { "epoch": 4.61, "learning_rate": 4.886605404916029e-06, "loss": 0.2107, "step": 76392 }, { "epoch": 4.61, "learning_rate": 4.885096494801805e-06, "loss": 0.2497, "step": 76394 }, { "epoch": 4.61, "learning_rate": 4.88358758468758e-06, "loss": 0.2275, "step": 76396 }, { "epoch": 4.61, "learning_rate": 4.8820786745733555e-06, "loss": 0.2096, "step": 76398 }, { "epoch": 4.61, "learning_rate": 4.880569764459131e-06, "loss": 0.2302, "step": 76400 }, { "epoch": 4.61, "learning_rate": 4.879060854344907e-06, "loss": 0.2174, "step": 76402 }, { "epoch": 4.61, "learning_rate": 4.8775519442306825e-06, "loss": 0.2345, "step": 76404 }, { "epoch": 4.61, "learning_rate": 4.876043034116458e-06, "loss": 0.2795, "step": 76406 }, { "epoch": 4.61, "learning_rate": 4.874534124002234e-06, "loss": 0.1616, "step": 76408 }, { "epoch": 4.61, "learning_rate": 4.8730252138880095e-06, "loss": 0.2035, "step": 76410 }, { "epoch": 4.61, "learning_rate": 4.871516303773784e-06, "loss": 0.2447, "step": 76412 }, { "epoch": 4.61, "learning_rate": 4.87000739365956e-06, "loss": 0.2766, "step": 76414 }, { "epoch": 4.61, "learning_rate": 4.868498483545336e-06, "loss": 0.3249, "step": 76416 }, { "epoch": 4.61, "learning_rate": 4.866989573431111e-06, "loss": 0.1154, "step": 76418 }, { "epoch": 4.61, "learning_rate": 4.865480663316886e-06, "loss": 0.1681, "step": 76420 }, { "epoch": 4.61, "learning_rate": 4.863971753202662e-06, "loss": 0.2452, "step": 76422 }, { "epoch": 4.61, "learning_rate": 4.8624628430884375e-06, "loss": 0.3921, "step": 76424 }, { "epoch": 4.61, "learning_rate": 4.860953932974213e-06, "loss": 0.2579, "step": 76426 }, { "epoch": 4.61, "learning_rate": 4.859445022859988e-06, "loss": 0.2469, "step": 76428 }, { "epoch": 4.61, "learning_rate": 4.857936112745764e-06, "loss": 0.2098, "step": 76430 }, { "epoch": 4.61, "learning_rate": 4.856427202631539e-06, "loss": 0.2353, "step": 76432 }, { "epoch": 4.61, "learning_rate": 4.854918292517315e-06, "loss": 0.2066, "step": 76434 }, { "epoch": 4.61, "learning_rate": 4.85340938240309e-06, "loss": 0.1643, "step": 76436 }, { "epoch": 4.61, "learning_rate": 4.8519004722888654e-06, "loss": 0.1172, "step": 76438 }, { "epoch": 4.61, "learning_rate": 4.850391562174641e-06, "loss": 0.193, "step": 76440 }, { "epoch": 4.61, "learning_rate": 4.848882652060417e-06, "loss": 0.1496, "step": 76442 }, { "epoch": 4.61, "learning_rate": 4.8473737419461924e-06, "loss": 0.1356, "step": 76444 }, { "epoch": 4.61, "learning_rate": 4.845864831831968e-06, "loss": 0.21, "step": 76446 }, { "epoch": 4.61, "learning_rate": 4.844355921717744e-06, "loss": 0.2214, "step": 76448 }, { "epoch": 4.61, "learning_rate": 4.8428470116035194e-06, "loss": 0.3597, "step": 76450 }, { "epoch": 4.61, "learning_rate": 4.841338101489294e-06, "loss": 0.1949, "step": 76452 }, { "epoch": 4.61, "learning_rate": 4.83982919137507e-06, "loss": 0.2503, "step": 76454 }, { "epoch": 4.61, "learning_rate": 4.838320281260846e-06, "loss": 0.2886, "step": 76456 }, { "epoch": 4.61, "learning_rate": 4.836811371146621e-06, "loss": 0.2577, "step": 76458 }, { "epoch": 4.61, "learning_rate": 4.835302461032397e-06, "loss": 0.1972, "step": 76460 }, { "epoch": 4.62, "learning_rate": 4.833793550918172e-06, "loss": 0.2602, "step": 76462 }, { "epoch": 4.62, "learning_rate": 4.832284640803947e-06, "loss": 0.2223, "step": 76464 }, { "epoch": 4.62, "learning_rate": 4.830775730689723e-06, "loss": 0.2383, "step": 76466 }, { "epoch": 4.62, "learning_rate": 4.829266820575499e-06, "loss": 0.2021, "step": 76468 }, { "epoch": 4.62, "learning_rate": 4.8277579104612736e-06, "loss": 0.3222, "step": 76470 }, { "epoch": 4.62, "learning_rate": 4.826249000347049e-06, "loss": 0.2755, "step": 76472 }, { "epoch": 4.62, "learning_rate": 4.824740090232825e-06, "loss": 0.4263, "step": 76474 }, { "epoch": 4.62, "learning_rate": 4.8232311801186006e-06, "loss": 0.1635, "step": 76476 }, { "epoch": 4.62, "learning_rate": 4.821722270004376e-06, "loss": 0.1255, "step": 76478 }, { "epoch": 4.62, "learning_rate": 4.820213359890152e-06, "loss": 0.2552, "step": 76480 }, { "epoch": 4.62, "learning_rate": 4.8187044497759276e-06, "loss": 0.3899, "step": 76482 }, { "epoch": 4.62, "learning_rate": 4.817195539661703e-06, "loss": 0.2573, "step": 76484 }, { "epoch": 4.62, "learning_rate": 4.815686629547478e-06, "loss": 0.2103, "step": 76486 }, { "epoch": 4.62, "learning_rate": 4.814177719433254e-06, "loss": 0.291, "step": 76488 }, { "epoch": 4.62, "learning_rate": 4.812668809319029e-06, "loss": 0.2019, "step": 76490 }, { "epoch": 4.62, "learning_rate": 4.811159899204805e-06, "loss": 0.1841, "step": 76492 }, { "epoch": 4.62, "learning_rate": 4.80965098909058e-06, "loss": 0.1903, "step": 76494 }, { "epoch": 4.62, "learning_rate": 4.8081420789763555e-06, "loss": 0.1754, "step": 76496 }, { "epoch": 4.62, "learning_rate": 4.806633168862131e-06, "loss": 0.3126, "step": 76498 }, { "epoch": 4.62, "learning_rate": 4.805124258747907e-06, "loss": 0.1867, "step": 76500 }, { "epoch": 4.62, "learning_rate": 4.803615348633682e-06, "loss": 0.1788, "step": 76502 }, { "epoch": 4.62, "learning_rate": 4.802106438519457e-06, "loss": 0.217, "step": 76504 }, { "epoch": 4.62, "learning_rate": 4.800597528405233e-06, "loss": 0.2715, "step": 76506 }, { "epoch": 4.62, "learning_rate": 4.799088618291009e-06, "loss": 0.1516, "step": 76508 }, { "epoch": 4.62, "learning_rate": 4.7975797081767835e-06, "loss": 0.2066, "step": 76510 }, { "epoch": 4.62, "learning_rate": 4.796070798062559e-06, "loss": 0.2479, "step": 76512 }, { "epoch": 4.62, "learning_rate": 4.794561887948335e-06, "loss": 0.2094, "step": 76514 }, { "epoch": 4.62, "learning_rate": 4.7930529778341105e-06, "loss": 0.2559, "step": 76516 }, { "epoch": 4.62, "learning_rate": 4.791544067719886e-06, "loss": 0.1897, "step": 76518 }, { "epoch": 4.62, "learning_rate": 4.790035157605662e-06, "loss": 0.2986, "step": 76520 }, { "epoch": 4.62, "learning_rate": 4.7885262474914375e-06, "loss": 0.1995, "step": 76522 }, { "epoch": 4.62, "learning_rate": 4.787017337377213e-06, "loss": 0.2181, "step": 76524 }, { "epoch": 4.62, "learning_rate": 4.785508427262988e-06, "loss": 0.2015, "step": 76526 }, { "epoch": 4.62, "learning_rate": 4.783999517148764e-06, "loss": 0.2585, "step": 76528 }, { "epoch": 4.62, "learning_rate": 4.782490607034539e-06, "loss": 0.1931, "step": 76530 }, { "epoch": 4.62, "learning_rate": 4.780981696920315e-06, "loss": 0.1226, "step": 76532 }, { "epoch": 4.62, "learning_rate": 4.779472786806091e-06, "loss": 0.2503, "step": 76534 }, { "epoch": 4.62, "learning_rate": 4.7779638766918655e-06, "loss": 0.1218, "step": 76536 }, { "epoch": 4.62, "learning_rate": 4.776454966577641e-06, "loss": 0.2663, "step": 76538 }, { "epoch": 4.62, "learning_rate": 4.774946056463417e-06, "loss": 0.2758, "step": 76540 }, { "epoch": 4.62, "learning_rate": 4.7734371463491925e-06, "loss": 0.2193, "step": 76542 }, { "epoch": 4.62, "learning_rate": 4.771928236234967e-06, "loss": 0.2008, "step": 76544 }, { "epoch": 4.62, "learning_rate": 4.770419326120743e-06, "loss": 0.1872, "step": 76546 }, { "epoch": 4.62, "learning_rate": 4.768910416006519e-06, "loss": 0.2029, "step": 76548 }, { "epoch": 4.62, "learning_rate": 4.767401505892294e-06, "loss": 0.1655, "step": 76550 }, { "epoch": 4.62, "learning_rate": 4.765892595778069e-06, "loss": 0.3463, "step": 76552 }, { "epoch": 4.62, "learning_rate": 4.764383685663845e-06, "loss": 0.1701, "step": 76554 }, { "epoch": 4.62, "learning_rate": 4.7628747755496205e-06, "loss": 0.1896, "step": 76556 }, { "epoch": 4.62, "learning_rate": 4.761365865435396e-06, "loss": 0.2276, "step": 76558 }, { "epoch": 4.62, "learning_rate": 4.759856955321172e-06, "loss": 0.182, "step": 76560 }, { "epoch": 4.62, "learning_rate": 4.7583480452069475e-06, "loss": 0.192, "step": 76562 }, { "epoch": 4.62, "learning_rate": 4.756839135092723e-06, "loss": 0.191, "step": 76564 }, { "epoch": 4.62, "learning_rate": 4.755330224978499e-06, "loss": 0.181, "step": 76566 }, { "epoch": 4.62, "learning_rate": 4.753821314864274e-06, "loss": 0.2303, "step": 76568 }, { "epoch": 4.62, "learning_rate": 4.752312404750049e-06, "loss": 0.2718, "step": 76570 }, { "epoch": 4.62, "learning_rate": 4.750803494635825e-06, "loss": 0.2753, "step": 76572 }, { "epoch": 4.62, "learning_rate": 4.749294584521601e-06, "loss": 0.2001, "step": 76574 }, { "epoch": 4.62, "learning_rate": 4.747785674407375e-06, "loss": 0.1806, "step": 76576 }, { "epoch": 4.62, "learning_rate": 4.746276764293151e-06, "loss": 0.1163, "step": 76578 }, { "epoch": 4.62, "learning_rate": 4.744767854178927e-06, "loss": 0.2734, "step": 76580 }, { "epoch": 4.62, "learning_rate": 4.743258944064702e-06, "loss": 0.2541, "step": 76582 }, { "epoch": 4.62, "learning_rate": 4.741750033950477e-06, "loss": 0.2332, "step": 76584 }, { "epoch": 4.62, "learning_rate": 4.740241123836253e-06, "loss": 0.177, "step": 76586 }, { "epoch": 4.62, "learning_rate": 4.7387322137220286e-06, "loss": 0.231, "step": 76588 }, { "epoch": 4.62, "learning_rate": 4.737223303607804e-06, "loss": 0.2745, "step": 76590 }, { "epoch": 4.62, "learning_rate": 4.73571439349358e-06, "loss": 0.2189, "step": 76592 }, { "epoch": 4.62, "learning_rate": 4.7342054833793556e-06, "loss": 0.1804, "step": 76594 }, { "epoch": 4.62, "learning_rate": 4.732696573265131e-06, "loss": 0.1953, "step": 76596 }, { "epoch": 4.62, "learning_rate": 4.731187663150907e-06, "loss": 0.2925, "step": 76598 }, { "epoch": 4.62, "learning_rate": 4.729678753036682e-06, "loss": 0.1738, "step": 76600 }, { "epoch": 4.62, "learning_rate": 4.728169842922457e-06, "loss": 0.1876, "step": 76602 }, { "epoch": 4.62, "learning_rate": 4.726660932808233e-06, "loss": 0.2723, "step": 76604 }, { "epoch": 4.62, "learning_rate": 4.725152022694009e-06, "loss": 0.1691, "step": 76606 }, { "epoch": 4.62, "learning_rate": 4.7236431125797835e-06, "loss": 0.2017, "step": 76608 }, { "epoch": 4.62, "learning_rate": 4.722134202465559e-06, "loss": 0.2407, "step": 76610 }, { "epoch": 4.62, "learning_rate": 4.720625292351335e-06, "loss": 0.1706, "step": 76612 }, { "epoch": 4.62, "learning_rate": 4.7191163822371105e-06, "loss": 0.2895, "step": 76614 }, { "epoch": 4.62, "learning_rate": 4.717607472122886e-06, "loss": 0.3222, "step": 76616 }, { "epoch": 4.62, "learning_rate": 4.716098562008661e-06, "loss": 0.2138, "step": 76618 }, { "epoch": 4.62, "learning_rate": 4.714589651894437e-06, "loss": 0.2649, "step": 76620 }, { "epoch": 4.62, "learning_rate": 4.713080741780212e-06, "loss": 0.178, "step": 76622 }, { "epoch": 4.62, "learning_rate": 4.711571831665988e-06, "loss": 0.3044, "step": 76624 }, { "epoch": 4.62, "learning_rate": 4.710062921551763e-06, "loss": 0.2328, "step": 76626 }, { "epoch": 4.63, "learning_rate": 4.7085540114375385e-06, "loss": 0.2292, "step": 76628 }, { "epoch": 4.63, "learning_rate": 4.707045101323314e-06, "loss": 0.1521, "step": 76630 }, { "epoch": 4.63, "learning_rate": 4.70553619120909e-06, "loss": 0.2812, "step": 76632 }, { "epoch": 4.63, "learning_rate": 4.7040272810948655e-06, "loss": 0.2811, "step": 76634 }, { "epoch": 4.63, "learning_rate": 4.702518370980641e-06, "loss": 0.2852, "step": 76636 }, { "epoch": 4.63, "learning_rate": 4.701009460866417e-06, "loss": 0.2134, "step": 76638 }, { "epoch": 4.63, "learning_rate": 4.6995005507521925e-06, "loss": 0.2313, "step": 76640 }, { "epoch": 4.63, "learning_rate": 4.697991640637967e-06, "loss": 0.2175, "step": 76642 }, { "epoch": 4.63, "learning_rate": 4.696482730523743e-06, "loss": 0.2493, "step": 76644 }, { "epoch": 4.63, "learning_rate": 4.694973820409519e-06, "loss": 0.2434, "step": 76646 }, { "epoch": 4.63, "learning_rate": 4.693464910295294e-06, "loss": 0.2338, "step": 76648 }, { "epoch": 4.63, "learning_rate": 4.691956000181069e-06, "loss": 0.237, "step": 76650 }, { "epoch": 4.63, "learning_rate": 4.690447090066845e-06, "loss": 0.2137, "step": 76652 }, { "epoch": 4.63, "learning_rate": 4.6889381799526205e-06, "loss": 0.2094, "step": 76654 }, { "epoch": 4.63, "learning_rate": 4.687429269838396e-06, "loss": 0.2083, "step": 76656 }, { "epoch": 4.63, "learning_rate": 4.685920359724171e-06, "loss": 0.3411, "step": 76658 }, { "epoch": 4.63, "learning_rate": 4.684411449609947e-06, "loss": 0.3313, "step": 76660 }, { "epoch": 4.63, "learning_rate": 4.682902539495722e-06, "loss": 0.1984, "step": 76662 }, { "epoch": 4.63, "learning_rate": 4.681393629381498e-06, "loss": 0.2769, "step": 76664 }, { "epoch": 4.63, "learning_rate": 4.679884719267273e-06, "loss": 0.1838, "step": 76666 }, { "epoch": 4.63, "learning_rate": 4.6783758091530485e-06, "loss": 0.2349, "step": 76668 }, { "epoch": 4.63, "learning_rate": 4.676866899038824e-06, "loss": 0.1817, "step": 76670 }, { "epoch": 4.63, "learning_rate": 4.6753579889246e-06, "loss": 0.1956, "step": 76672 }, { "epoch": 4.63, "learning_rate": 4.6738490788103755e-06, "loss": 0.2243, "step": 76674 }, { "epoch": 4.63, "learning_rate": 4.672340168696151e-06, "loss": 0.3632, "step": 76676 }, { "epoch": 4.63, "learning_rate": 4.670831258581927e-06, "loss": 0.1949, "step": 76678 }, { "epoch": 4.63, "learning_rate": 4.6693223484677025e-06, "loss": 0.2325, "step": 76680 }, { "epoch": 4.63, "learning_rate": 4.667813438353477e-06, "loss": 0.1569, "step": 76682 }, { "epoch": 4.63, "learning_rate": 4.666304528239253e-06, "loss": 0.1454, "step": 76684 }, { "epoch": 4.63, "learning_rate": 4.664795618125029e-06, "loss": 0.1478, "step": 76686 }, { "epoch": 4.63, "learning_rate": 4.663286708010804e-06, "loss": 0.1954, "step": 76688 }, { "epoch": 4.63, "learning_rate": 4.66177779789658e-06, "loss": 0.2419, "step": 76690 }, { "epoch": 4.63, "learning_rate": 4.660268887782355e-06, "loss": 0.1379, "step": 76692 }, { "epoch": 4.63, "learning_rate": 4.6587599776681304e-06, "loss": 0.2978, "step": 76694 }, { "epoch": 4.63, "learning_rate": 4.657251067553906e-06, "loss": 0.2093, "step": 76696 }, { "epoch": 4.63, "learning_rate": 4.655742157439682e-06, "loss": 0.2653, "step": 76698 }, { "epoch": 4.63, "learning_rate": 4.654233247325457e-06, "loss": 0.1683, "step": 76700 }, { "epoch": 4.63, "learning_rate": 4.652724337211232e-06, "loss": 0.4028, "step": 76702 }, { "epoch": 4.63, "learning_rate": 4.651215427097008e-06, "loss": 0.3137, "step": 76704 }, { "epoch": 4.63, "learning_rate": 4.649706516982784e-06, "loss": 0.2648, "step": 76706 }, { "epoch": 4.63, "learning_rate": 4.648197606868559e-06, "loss": 0.1739, "step": 76708 }, { "epoch": 4.63, "learning_rate": 4.646688696754335e-06, "loss": 0.2553, "step": 76710 }, { "epoch": 4.63, "learning_rate": 4.645179786640111e-06, "loss": 0.1146, "step": 76712 }, { "epoch": 4.63, "learning_rate": 4.643670876525886e-06, "loss": 0.1808, "step": 76714 }, { "epoch": 4.63, "learning_rate": 4.642161966411661e-06, "loss": 0.2608, "step": 76716 }, { "epoch": 4.63, "learning_rate": 4.640653056297437e-06, "loss": 0.2475, "step": 76718 }, { "epoch": 4.63, "learning_rate": 4.639144146183212e-06, "loss": 0.2264, "step": 76720 }, { "epoch": 4.63, "learning_rate": 4.637635236068988e-06, "loss": 0.2245, "step": 76722 }, { "epoch": 4.63, "learning_rate": 4.636126325954763e-06, "loss": 0.1903, "step": 76724 }, { "epoch": 4.63, "learning_rate": 4.6346174158405386e-06, "loss": 0.2256, "step": 76726 }, { "epoch": 4.63, "learning_rate": 4.633108505726314e-06, "loss": 0.227, "step": 76728 }, { "epoch": 4.63, "learning_rate": 4.63159959561209e-06, "loss": 0.2025, "step": 76730 }, { "epoch": 4.63, "learning_rate": 4.630090685497865e-06, "loss": 0.2467, "step": 76732 }, { "epoch": 4.63, "learning_rate": 4.62858177538364e-06, "loss": 0.2147, "step": 76734 }, { "epoch": 4.63, "learning_rate": 4.627072865269416e-06, "loss": 0.2723, "step": 76736 }, { "epoch": 4.63, "learning_rate": 4.625563955155192e-06, "loss": 0.2706, "step": 76738 }, { "epoch": 4.63, "learning_rate": 4.6240550450409665e-06, "loss": 0.1786, "step": 76740 }, { "epoch": 4.63, "learning_rate": 4.622546134926742e-06, "loss": 0.1748, "step": 76742 }, { "epoch": 4.63, "learning_rate": 4.621037224812518e-06, "loss": 0.3183, "step": 76744 }, { "epoch": 4.63, "learning_rate": 4.6195283146982935e-06, "loss": 0.1532, "step": 76746 }, { "epoch": 4.63, "learning_rate": 4.618019404584069e-06, "loss": 0.2287, "step": 76748 }, { "epoch": 4.63, "learning_rate": 4.616510494469845e-06, "loss": 0.3297, "step": 76750 }, { "epoch": 4.63, "learning_rate": 4.6150015843556205e-06, "loss": 0.2323, "step": 76752 }, { "epoch": 4.63, "learning_rate": 4.613492674241396e-06, "loss": 0.157, "step": 76754 }, { "epoch": 4.63, "learning_rate": 4.611983764127171e-06, "loss": 0.2526, "step": 76756 }, { "epoch": 4.63, "learning_rate": 4.610474854012947e-06, "loss": 0.2433, "step": 76758 }, { "epoch": 4.63, "learning_rate": 4.608965943898722e-06, "loss": 0.27, "step": 76760 }, { "epoch": 4.63, "learning_rate": 4.607457033784498e-06, "loss": 0.1767, "step": 76762 }, { "epoch": 4.63, "learning_rate": 4.605948123670273e-06, "loss": 0.2314, "step": 76764 }, { "epoch": 4.63, "learning_rate": 4.6044392135560485e-06, "loss": 0.2309, "step": 76766 }, { "epoch": 4.63, "learning_rate": 4.602930303441824e-06, "loss": 0.2387, "step": 76768 }, { "epoch": 4.63, "learning_rate": 4.6014213933276e-06, "loss": 0.3044, "step": 76770 }, { "epoch": 4.63, "learning_rate": 4.5999124832133755e-06, "loss": 0.298, "step": 76772 }, { "epoch": 4.63, "learning_rate": 4.59840357309915e-06, "loss": 0.1259, "step": 76774 }, { "epoch": 4.63, "learning_rate": 4.596894662984926e-06, "loss": 0.2427, "step": 76776 }, { "epoch": 4.63, "learning_rate": 4.595385752870702e-06, "loss": 0.2164, "step": 76778 }, { "epoch": 4.63, "learning_rate": 4.593876842756477e-06, "loss": 0.2029, "step": 76780 }, { "epoch": 4.63, "learning_rate": 4.592367932642252e-06, "loss": 0.2977, "step": 76782 }, { "epoch": 4.63, "learning_rate": 4.590859022528028e-06, "loss": 0.4489, "step": 76784 }, { "epoch": 4.63, "learning_rate": 4.5893501124138035e-06, "loss": 0.1587, "step": 76786 }, { "epoch": 4.63, "learning_rate": 4.587841202299579e-06, "loss": 0.1399, "step": 76788 }, { "epoch": 4.63, "learning_rate": 4.586332292185355e-06, "loss": 0.3766, "step": 76790 }, { "epoch": 4.63, "learning_rate": 4.5848233820711305e-06, "loss": 0.1578, "step": 76792 }, { "epoch": 4.64, "learning_rate": 4.583314471956906e-06, "loss": 0.2456, "step": 76794 }, { "epoch": 4.64, "learning_rate": 4.581805561842682e-06, "loss": 0.2303, "step": 76796 }, { "epoch": 4.64, "learning_rate": 4.580296651728457e-06, "loss": 0.2327, "step": 76798 }, { "epoch": 4.64, "learning_rate": 4.578787741614232e-06, "loss": 0.1993, "step": 76800 }, { "epoch": 4.64, "learning_rate": 4.577278831500008e-06, "loss": 0.2187, "step": 76802 }, { "epoch": 4.64, "learning_rate": 4.575769921385784e-06, "loss": 0.2768, "step": 76804 }, { "epoch": 4.64, "learning_rate": 4.5742610112715584e-06, "loss": 0.3002, "step": 76806 }, { "epoch": 4.64, "learning_rate": 4.572752101157334e-06, "loss": 0.1508, "step": 76808 }, { "epoch": 4.64, "learning_rate": 4.57124319104311e-06, "loss": 0.2541, "step": 76810 }, { "epoch": 4.64, "learning_rate": 4.5697342809288854e-06, "loss": 0.3301, "step": 76812 }, { "epoch": 4.64, "learning_rate": 4.56822537081466e-06, "loss": 0.1424, "step": 76814 }, { "epoch": 4.64, "learning_rate": 4.566716460700436e-06, "loss": 0.2436, "step": 76816 }, { "epoch": 4.64, "learning_rate": 4.565207550586212e-06, "loss": 0.22, "step": 76818 }, { "epoch": 4.64, "learning_rate": 4.563698640471987e-06, "loss": 0.1946, "step": 76820 }, { "epoch": 4.64, "learning_rate": 4.562189730357763e-06, "loss": 0.254, "step": 76822 }, { "epoch": 4.64, "learning_rate": 4.560680820243539e-06, "loss": 0.1886, "step": 76824 }, { "epoch": 4.64, "learning_rate": 4.559171910129314e-06, "loss": 0.1809, "step": 76826 }, { "epoch": 4.64, "learning_rate": 4.55766300001509e-06, "loss": 0.2548, "step": 76828 }, { "epoch": 4.64, "learning_rate": 4.556154089900865e-06, "loss": 0.2072, "step": 76830 }, { "epoch": 4.64, "learning_rate": 4.55464517978664e-06, "loss": 0.228, "step": 76832 }, { "epoch": 4.64, "learning_rate": 4.553136269672416e-06, "loss": 0.2586, "step": 76834 }, { "epoch": 4.64, "learning_rate": 4.551627359558192e-06, "loss": 0.1729, "step": 76836 }, { "epoch": 4.64, "learning_rate": 4.5501184494439666e-06, "loss": 0.2414, "step": 76838 }, { "epoch": 4.64, "learning_rate": 4.548609539329742e-06, "loss": 0.34, "step": 76840 }, { "epoch": 4.64, "learning_rate": 4.547100629215518e-06, "loss": 0.2811, "step": 76842 }, { "epoch": 4.64, "learning_rate": 4.5455917191012936e-06, "loss": 0.182, "step": 76844 }, { "epoch": 4.64, "learning_rate": 4.544082808987069e-06, "loss": 0.2312, "step": 76846 }, { "epoch": 4.64, "learning_rate": 4.542573898872844e-06, "loss": 0.2892, "step": 76848 }, { "epoch": 4.64, "learning_rate": 4.54106498875862e-06, "loss": 0.2931, "step": 76850 }, { "epoch": 4.64, "learning_rate": 4.539556078644395e-06, "loss": 0.2021, "step": 76852 }, { "epoch": 4.64, "learning_rate": 4.538047168530171e-06, "loss": 0.1748, "step": 76854 }, { "epoch": 4.64, "learning_rate": 4.536538258415946e-06, "loss": 0.2447, "step": 76856 }, { "epoch": 4.64, "learning_rate": 4.5350293483017215e-06, "loss": 0.1861, "step": 76858 }, { "epoch": 4.64, "learning_rate": 4.533520438187497e-06, "loss": 0.1788, "step": 76860 }, { "epoch": 4.64, "learning_rate": 4.532011528073273e-06, "loss": 0.1925, "step": 76862 }, { "epoch": 4.64, "learning_rate": 4.5305026179590485e-06, "loss": 0.2698, "step": 76864 }, { "epoch": 4.64, "learning_rate": 4.528993707844824e-06, "loss": 0.3038, "step": 76866 }, { "epoch": 4.64, "learning_rate": 4.5274847977306e-06, "loss": 0.2059, "step": 76868 }, { "epoch": 4.64, "learning_rate": 4.5259758876163755e-06, "loss": 0.3284, "step": 76870 }, { "epoch": 4.64, "learning_rate": 4.525221432559262e-06, "loss": 0.2068, "step": 76872 }, { "epoch": 4.64, "learning_rate": 4.523712522445038e-06, "loss": 0.3007, "step": 76874 }, { "epoch": 4.64, "learning_rate": 4.5222036123308134e-06, "loss": 0.187, "step": 76876 }, { "epoch": 4.64, "learning_rate": 4.520694702216589e-06, "loss": 0.2669, "step": 76878 }, { "epoch": 4.64, "learning_rate": 4.519185792102364e-06, "loss": 0.1663, "step": 76880 }, { "epoch": 4.64, "learning_rate": 4.5176768819881404e-06, "loss": 0.2202, "step": 76882 }, { "epoch": 4.64, "learning_rate": 4.516167971873916e-06, "loss": 0.2341, "step": 76884 }, { "epoch": 4.64, "learning_rate": 4.514659061759692e-06, "loss": 0.2155, "step": 76886 }, { "epoch": 4.64, "learning_rate": 4.5131501516454674e-06, "loss": 0.161, "step": 76888 }, { "epoch": 4.64, "learning_rate": 4.511641241531242e-06, "loss": 0.3128, "step": 76890 }, { "epoch": 4.64, "learning_rate": 4.510132331417018e-06, "loss": 0.2505, "step": 76892 }, { "epoch": 4.64, "learning_rate": 4.508623421302794e-06, "loss": 0.2144, "step": 76894 }, { "epoch": 4.64, "learning_rate": 4.507114511188569e-06, "loss": 0.2258, "step": 76896 }, { "epoch": 4.64, "learning_rate": 4.505605601074344e-06, "loss": 0.3202, "step": 76898 }, { "epoch": 4.64, "learning_rate": 4.50409669096012e-06, "loss": 0.1813, "step": 76900 }, { "epoch": 4.64, "learning_rate": 4.502587780845895e-06, "loss": 0.1325, "step": 76902 }, { "epoch": 4.64, "learning_rate": 4.501078870731671e-06, "loss": 0.1468, "step": 76904 }, { "epoch": 4.64, "learning_rate": 4.499569960617446e-06, "loss": 0.2704, "step": 76906 }, { "epoch": 4.64, "learning_rate": 4.4980610505032216e-06, "loss": 0.1615, "step": 76908 }, { "epoch": 4.64, "learning_rate": 4.496552140388997e-06, "loss": 0.2005, "step": 76910 }, { "epoch": 4.64, "learning_rate": 4.495043230274773e-06, "loss": 0.2341, "step": 76912 }, { "epoch": 4.64, "learning_rate": 4.493534320160548e-06, "loss": 0.2385, "step": 76914 }, { "epoch": 4.64, "learning_rate": 4.492025410046323e-06, "loss": 0.1762, "step": 76916 }, { "epoch": 4.64, "learning_rate": 4.490516499932099e-06, "loss": 0.2714, "step": 76918 }, { "epoch": 4.64, "learning_rate": 4.489007589817875e-06, "loss": 0.317, "step": 76920 }, { "epoch": 4.64, "learning_rate": 4.48749867970365e-06, "loss": 0.305, "step": 76922 }, { "epoch": 4.64, "learning_rate": 4.485989769589426e-06, "loss": 0.19, "step": 76924 }, { "epoch": 4.64, "learning_rate": 4.484480859475202e-06, "loss": 0.2684, "step": 76926 }, { "epoch": 4.64, "learning_rate": 4.482971949360977e-06, "loss": 0.1488, "step": 76928 }, { "epoch": 4.64, "learning_rate": 4.481463039246752e-06, "loss": 0.1637, "step": 76930 }, { "epoch": 4.64, "learning_rate": 4.479954129132528e-06, "loss": 0.3347, "step": 76932 }, { "epoch": 4.64, "learning_rate": 4.4784452190183035e-06, "loss": 0.1132, "step": 76934 }, { "epoch": 4.64, "learning_rate": 4.476936308904079e-06, "loss": 0.2679, "step": 76936 }, { "epoch": 4.64, "learning_rate": 4.475427398789854e-06, "loss": 0.2305, "step": 76938 }, { "epoch": 4.64, "learning_rate": 4.47391848867563e-06, "loss": 0.1607, "step": 76940 }, { "epoch": 4.64, "learning_rate": 4.472409578561405e-06, "loss": 0.1795, "step": 76942 }, { "epoch": 4.64, "learning_rate": 4.470900668447181e-06, "loss": 0.148, "step": 76944 }, { "epoch": 4.64, "learning_rate": 4.469391758332956e-06, "loss": 0.4211, "step": 76946 }, { "epoch": 4.64, "learning_rate": 4.4678828482187315e-06, "loss": 0.3177, "step": 76948 }, { "epoch": 4.64, "learning_rate": 4.466373938104507e-06, "loss": 0.2518, "step": 76950 }, { "epoch": 4.64, "learning_rate": 4.464865027990283e-06, "loss": 0.4138, "step": 76952 }, { "epoch": 4.64, "learning_rate": 4.463356117876058e-06, "loss": 0.2052, "step": 76954 }, { "epoch": 4.64, "learning_rate": 4.461847207761833e-06, "loss": 0.2062, "step": 76956 }, { "epoch": 4.64, "learning_rate": 4.460338297647609e-06, "loss": 0.1873, "step": 76958 }, { "epoch": 4.65, "learning_rate": 4.458829387533385e-06, "loss": 0.1773, "step": 76960 }, { "epoch": 4.65, "learning_rate": 4.45732047741916e-06, "loss": 0.2664, "step": 76962 }, { "epoch": 4.65, "learning_rate": 4.455811567304936e-06, "loss": 0.1843, "step": 76964 }, { "epoch": 4.65, "learning_rate": 4.454302657190712e-06, "loss": 0.2067, "step": 76966 }, { "epoch": 4.65, "learning_rate": 4.452793747076487e-06, "loss": 0.2308, "step": 76968 }, { "epoch": 4.65, "learning_rate": 4.451284836962263e-06, "loss": 0.2679, "step": 76970 }, { "epoch": 4.65, "learning_rate": 4.449775926848038e-06, "loss": 0.357, "step": 76972 }, { "epoch": 4.65, "learning_rate": 4.4482670167338135e-06, "loss": 0.1935, "step": 76974 }, { "epoch": 4.65, "learning_rate": 4.446758106619589e-06, "loss": 0.2434, "step": 76976 }, { "epoch": 4.65, "learning_rate": 4.445249196505365e-06, "loss": 0.2393, "step": 76978 }, { "epoch": 4.65, "learning_rate": 4.44374028639114e-06, "loss": 0.1434, "step": 76980 }, { "epoch": 4.65, "learning_rate": 4.442231376276915e-06, "loss": 0.1753, "step": 76982 }, { "epoch": 4.65, "learning_rate": 4.440722466162691e-06, "loss": 0.2371, "step": 76984 }, { "epoch": 4.65, "learning_rate": 4.439213556048467e-06, "loss": 0.2765, "step": 76986 }, { "epoch": 4.65, "learning_rate": 4.4377046459342414e-06, "loss": 0.1715, "step": 76988 }, { "epoch": 4.65, "learning_rate": 4.436195735820017e-06, "loss": 0.1765, "step": 76990 }, { "epoch": 4.65, "learning_rate": 4.434686825705793e-06, "loss": 0.2037, "step": 76992 }, { "epoch": 4.65, "learning_rate": 4.4331779155915684e-06, "loss": 0.2082, "step": 76994 }, { "epoch": 4.65, "learning_rate": 4.431669005477343e-06, "loss": 0.2309, "step": 76996 }, { "epoch": 4.65, "learning_rate": 4.430160095363119e-06, "loss": 0.2345, "step": 76998 }, { "epoch": 4.65, "learning_rate": 4.428651185248895e-06, "loss": 0.1315, "step": 77000 }, { "epoch": 4.65, "learning_rate": 4.427142275134671e-06, "loss": 0.2746, "step": 77002 }, { "epoch": 4.65, "learning_rate": 4.425633365020446e-06, "loss": 0.2275, "step": 77004 }, { "epoch": 4.65, "learning_rate": 4.424124454906222e-06, "loss": 0.1843, "step": 77006 }, { "epoch": 4.65, "learning_rate": 4.422615544791997e-06, "loss": 0.1209, "step": 77008 }, { "epoch": 4.65, "learning_rate": 4.421106634677773e-06, "loss": 0.1522, "step": 77010 }, { "epoch": 4.65, "learning_rate": 4.419597724563548e-06, "loss": 0.3355, "step": 77012 }, { "epoch": 4.65, "learning_rate": 4.418088814449323e-06, "loss": 0.285, "step": 77014 }, { "epoch": 4.65, "learning_rate": 4.416579904335099e-06, "loss": 0.2451, "step": 77016 }, { "epoch": 4.65, "learning_rate": 4.415070994220875e-06, "loss": 0.2536, "step": 77018 }, { "epoch": 4.65, "learning_rate": 4.4135620841066496e-06, "loss": 0.1537, "step": 77020 }, { "epoch": 4.65, "learning_rate": 4.412053173992425e-06, "loss": 0.2158, "step": 77022 }, { "epoch": 4.65, "learning_rate": 4.410544263878201e-06, "loss": 0.2841, "step": 77024 }, { "epoch": 4.65, "learning_rate": 4.4090353537639766e-06, "loss": 0.2364, "step": 77026 }, { "epoch": 4.65, "learning_rate": 4.407526443649751e-06, "loss": 0.2667, "step": 77028 }, { "epoch": 4.65, "learning_rate": 4.406017533535527e-06, "loss": 0.24, "step": 77030 }, { "epoch": 4.65, "learning_rate": 4.404508623421303e-06, "loss": 0.2485, "step": 77032 }, { "epoch": 4.65, "learning_rate": 4.402999713307078e-06, "loss": 0.3084, "step": 77034 }, { "epoch": 4.65, "learning_rate": 4.401490803192854e-06, "loss": 0.3017, "step": 77036 }, { "epoch": 4.65, "learning_rate": 4.39998189307863e-06, "loss": 0.2239, "step": 77038 }, { "epoch": 4.65, "learning_rate": 4.398472982964405e-06, "loss": 0.2329, "step": 77040 }, { "epoch": 4.65, "learning_rate": 4.396964072850181e-06, "loss": 0.222, "step": 77042 }, { "epoch": 4.65, "learning_rate": 4.395455162735957e-06, "loss": 0.3122, "step": 77044 }, { "epoch": 4.65, "learning_rate": 4.3939462526217315e-06, "loss": 0.2809, "step": 77046 }, { "epoch": 4.65, "learning_rate": 4.392437342507507e-06, "loss": 0.2283, "step": 77048 }, { "epoch": 4.65, "learning_rate": 4.390928432393283e-06, "loss": 0.2019, "step": 77050 }, { "epoch": 4.65, "learning_rate": 4.3894195222790585e-06, "loss": 0.3153, "step": 77052 }, { "epoch": 4.65, "learning_rate": 4.387910612164833e-06, "loss": 0.1933, "step": 77054 }, { "epoch": 4.65, "learning_rate": 4.386401702050609e-06, "loss": 0.171, "step": 77056 }, { "epoch": 4.65, "learning_rate": 4.384892791936385e-06, "loss": 0.2495, "step": 77058 }, { "epoch": 4.65, "learning_rate": 4.38338388182216e-06, "loss": 0.1698, "step": 77060 }, { "epoch": 4.65, "learning_rate": 4.381874971707935e-06, "loss": 0.2938, "step": 77062 }, { "epoch": 4.65, "learning_rate": 4.380366061593711e-06, "loss": 0.2771, "step": 77064 }, { "epoch": 4.65, "learning_rate": 4.3788571514794865e-06, "loss": 0.3835, "step": 77066 }, { "epoch": 4.65, "learning_rate": 4.377348241365262e-06, "loss": 0.2948, "step": 77068 }, { "epoch": 4.65, "learning_rate": 4.375839331251037e-06, "loss": 0.1815, "step": 77070 }, { "epoch": 4.65, "learning_rate": 4.374330421136813e-06, "loss": 0.1634, "step": 77072 }, { "epoch": 4.65, "learning_rate": 4.372821511022588e-06, "loss": 0.2294, "step": 77074 }, { "epoch": 4.65, "learning_rate": 4.371312600908364e-06, "loss": 0.1688, "step": 77076 }, { "epoch": 4.65, "learning_rate": 4.36980369079414e-06, "loss": 0.1506, "step": 77078 }, { "epoch": 4.65, "learning_rate": 4.368294780679915e-06, "loss": 0.2619, "step": 77080 }, { "epoch": 4.65, "learning_rate": 4.366785870565691e-06, "loss": 0.2441, "step": 77082 }, { "epoch": 4.65, "learning_rate": 4.365276960451467e-06, "loss": 0.2157, "step": 77084 }, { "epoch": 4.65, "learning_rate": 4.3637680503372415e-06, "loss": 0.3712, "step": 77086 }, { "epoch": 4.65, "learning_rate": 4.362259140223017e-06, "loss": 0.1717, "step": 77088 }, { "epoch": 4.65, "learning_rate": 4.360750230108793e-06, "loss": 0.219, "step": 77090 }, { "epoch": 4.65, "learning_rate": 4.3592413199945685e-06, "loss": 0.2031, "step": 77092 }, { "epoch": 4.65, "learning_rate": 4.357732409880343e-06, "loss": 0.203, "step": 77094 }, { "epoch": 4.65, "learning_rate": 4.356223499766119e-06, "loss": 0.2203, "step": 77096 }, { "epoch": 4.65, "learning_rate": 4.354714589651895e-06, "loss": 0.2355, "step": 77098 }, { "epoch": 4.65, "learning_rate": 4.35320567953767e-06, "loss": 0.3389, "step": 77100 }, { "epoch": 4.65, "learning_rate": 4.351696769423445e-06, "loss": 0.2266, "step": 77102 }, { "epoch": 4.65, "learning_rate": 4.350187859309221e-06, "loss": 0.1148, "step": 77104 }, { "epoch": 4.65, "learning_rate": 4.3486789491949965e-06, "loss": 0.1783, "step": 77106 }, { "epoch": 4.65, "learning_rate": 4.347170039080772e-06, "loss": 0.2053, "step": 77108 }, { "epoch": 4.65, "learning_rate": 4.345661128966547e-06, "loss": 0.1817, "step": 77110 }, { "epoch": 4.65, "learning_rate": 4.344152218852323e-06, "loss": 0.2106, "step": 77112 }, { "epoch": 4.65, "learning_rate": 4.342643308738098e-06, "loss": 0.1897, "step": 77114 }, { "epoch": 4.65, "learning_rate": 4.341134398623874e-06, "loss": 0.3473, "step": 77116 }, { "epoch": 4.65, "learning_rate": 4.33962548850965e-06, "loss": 0.2208, "step": 77118 }, { "epoch": 4.65, "learning_rate": 4.338116578395425e-06, "loss": 0.2621, "step": 77120 }, { "epoch": 4.65, "learning_rate": 4.336607668281201e-06, "loss": 0.1641, "step": 77122 }, { "epoch": 4.65, "learning_rate": 4.335098758166977e-06, "loss": 0.3376, "step": 77124 }, { "epoch": 4.66, "learning_rate": 4.333589848052752e-06, "loss": 0.1985, "step": 77126 }, { "epoch": 4.66, "learning_rate": 4.332080937938527e-06, "loss": 0.18, "step": 77128 }, { "epoch": 4.66, "learning_rate": 4.330572027824303e-06, "loss": 0.1755, "step": 77130 }, { "epoch": 4.66, "learning_rate": 4.3290631177100784e-06, "loss": 0.1619, "step": 77132 }, { "epoch": 4.66, "learning_rate": 4.327554207595854e-06, "loss": 0.2074, "step": 77134 }, { "epoch": 4.66, "learning_rate": 4.326045297481629e-06, "loss": 0.4084, "step": 77136 }, { "epoch": 4.66, "learning_rate": 4.324536387367405e-06, "loss": 0.216, "step": 77138 }, { "epoch": 4.66, "learning_rate": 4.32302747725318e-06, "loss": 0.1129, "step": 77140 }, { "epoch": 4.66, "learning_rate": 4.321518567138956e-06, "loss": 0.2551, "step": 77142 }, { "epoch": 4.66, "learning_rate": 4.320009657024731e-06, "loss": 0.1616, "step": 77144 }, { "epoch": 4.66, "learning_rate": 4.318500746910506e-06, "loss": 0.1182, "step": 77146 }, { "epoch": 4.66, "learning_rate": 4.316991836796282e-06, "loss": 0.2075, "step": 77148 }, { "epoch": 4.66, "learning_rate": 4.315482926682058e-06, "loss": 0.1916, "step": 77150 }, { "epoch": 4.66, "learning_rate": 4.313974016567833e-06, "loss": 0.221, "step": 77152 }, { "epoch": 4.66, "learning_rate": 4.312465106453609e-06, "loss": 0.1933, "step": 77154 }, { "epoch": 4.66, "learning_rate": 4.310956196339385e-06, "loss": 0.2849, "step": 77156 }, { "epoch": 4.66, "learning_rate": 4.30944728622516e-06, "loss": 0.2115, "step": 77158 }, { "epoch": 4.66, "learning_rate": 4.307938376110935e-06, "loss": 0.1077, "step": 77160 }, { "epoch": 4.66, "learning_rate": 4.306429465996711e-06, "loss": 0.2061, "step": 77162 }, { "epoch": 4.66, "learning_rate": 4.3049205558824866e-06, "loss": 0.2574, "step": 77164 }, { "epoch": 4.66, "learning_rate": 4.303411645768262e-06, "loss": 0.2644, "step": 77166 }, { "epoch": 4.66, "learning_rate": 4.301902735654037e-06, "loss": 0.3348, "step": 77168 }, { "epoch": 4.66, "learning_rate": 4.300393825539813e-06, "loss": 0.277, "step": 77170 }, { "epoch": 4.66, "learning_rate": 4.298884915425588e-06, "loss": 0.1695, "step": 77172 }, { "epoch": 4.66, "learning_rate": 4.297376005311364e-06, "loss": 0.2114, "step": 77174 }, { "epoch": 4.66, "learning_rate": 4.295867095197139e-06, "loss": 0.1895, "step": 77176 }, { "epoch": 4.66, "learning_rate": 4.2943581850829145e-06, "loss": 0.1573, "step": 77178 }, { "epoch": 4.66, "learning_rate": 4.29284927496869e-06, "loss": 0.1898, "step": 77180 }, { "epoch": 4.66, "learning_rate": 4.291340364854466e-06, "loss": 0.2015, "step": 77182 }, { "epoch": 4.66, "learning_rate": 4.289831454740241e-06, "loss": 0.1512, "step": 77184 }, { "epoch": 4.66, "learning_rate": 4.288322544626016e-06, "loss": 0.3082, "step": 77186 }, { "epoch": 4.66, "learning_rate": 4.286813634511792e-06, "loss": 0.2456, "step": 77188 }, { "epoch": 4.66, "learning_rate": 4.285304724397568e-06, "loss": 0.3013, "step": 77190 }, { "epoch": 4.66, "learning_rate": 4.283795814283343e-06, "loss": 0.1624, "step": 77192 }, { "epoch": 4.66, "learning_rate": 4.282286904169119e-06, "loss": 0.4272, "step": 77194 }, { "epoch": 4.66, "learning_rate": 4.280777994054895e-06, "loss": 0.1682, "step": 77196 }, { "epoch": 4.66, "learning_rate": 4.27926908394067e-06, "loss": 0.1582, "step": 77198 }, { "epoch": 4.66, "learning_rate": 4.277760173826446e-06, "loss": 0.255, "step": 77200 }, { "epoch": 4.66, "learning_rate": 4.276251263712221e-06, "loss": 0.2572, "step": 77202 }, { "epoch": 4.66, "learning_rate": 4.2747423535979965e-06, "loss": 0.2728, "step": 77204 }, { "epoch": 4.66, "learning_rate": 4.273233443483772e-06, "loss": 0.3004, "step": 77206 }, { "epoch": 4.66, "learning_rate": 4.271724533369548e-06, "loss": 0.2851, "step": 77208 }, { "epoch": 4.66, "learning_rate": 4.270215623255323e-06, "loss": 0.2129, "step": 77210 }, { "epoch": 4.66, "learning_rate": 4.268706713141098e-06, "loss": 0.1389, "step": 77212 }, { "epoch": 4.66, "learning_rate": 4.267197803026874e-06, "loss": 0.2472, "step": 77214 }, { "epoch": 4.66, "learning_rate": 4.26568889291265e-06, "loss": 0.2955, "step": 77216 }, { "epoch": 4.66, "learning_rate": 4.2641799827984245e-06, "loss": 0.2167, "step": 77218 }, { "epoch": 4.66, "learning_rate": 4.2626710726842e-06, "loss": 0.1555, "step": 77220 }, { "epoch": 4.66, "learning_rate": 4.261162162569976e-06, "loss": 0.2576, "step": 77222 }, { "epoch": 4.66, "learning_rate": 4.2596532524557515e-06, "loss": 0.2451, "step": 77224 }, { "epoch": 4.66, "learning_rate": 4.258144342341526e-06, "loss": 0.2914, "step": 77226 }, { "epoch": 4.66, "learning_rate": 4.256635432227302e-06, "loss": 0.2017, "step": 77228 }, { "epoch": 4.66, "learning_rate": 4.255126522113078e-06, "loss": 0.259, "step": 77230 }, { "epoch": 4.66, "learning_rate": 4.253617611998853e-06, "loss": 0.2979, "step": 77232 }, { "epoch": 4.66, "learning_rate": 4.252108701884629e-06, "loss": 0.1795, "step": 77234 }, { "epoch": 4.66, "learning_rate": 4.250599791770405e-06, "loss": 0.2355, "step": 77236 }, { "epoch": 4.66, "learning_rate": 4.24909088165618e-06, "loss": 0.2163, "step": 77238 }, { "epoch": 4.66, "learning_rate": 4.247581971541956e-06, "loss": 0.2134, "step": 77240 }, { "epoch": 4.66, "learning_rate": 4.246073061427731e-06, "loss": 0.2837, "step": 77242 }, { "epoch": 4.66, "learning_rate": 4.2445641513135064e-06, "loss": 0.2328, "step": 77244 }, { "epoch": 4.66, "learning_rate": 4.243055241199282e-06, "loss": 0.2836, "step": 77246 }, { "epoch": 4.66, "learning_rate": 4.241546331085058e-06, "loss": 0.2425, "step": 77248 }, { "epoch": 4.66, "learning_rate": 4.240037420970833e-06, "loss": 0.2046, "step": 77250 }, { "epoch": 4.66, "learning_rate": 4.238528510856608e-06, "loss": 0.2665, "step": 77252 }, { "epoch": 4.66, "learning_rate": 4.237019600742384e-06, "loss": 0.2905, "step": 77254 }, { "epoch": 4.66, "learning_rate": 4.23551069062816e-06, "loss": 0.2189, "step": 77256 }, { "epoch": 4.66, "learning_rate": 4.234001780513934e-06, "loss": 0.3364, "step": 77258 }, { "epoch": 4.66, "learning_rate": 4.23249287039971e-06, "loss": 0.1768, "step": 77260 }, { "epoch": 4.66, "learning_rate": 4.230983960285486e-06, "loss": 0.161, "step": 77262 }, { "epoch": 4.66, "learning_rate": 4.229475050171261e-06, "loss": 0.1567, "step": 77264 }, { "epoch": 4.66, "learning_rate": 4.227966140057037e-06, "loss": 0.2176, "step": 77266 }, { "epoch": 4.66, "learning_rate": 4.226457229942813e-06, "loss": 0.2629, "step": 77268 }, { "epoch": 4.66, "learning_rate": 4.224948319828588e-06, "loss": 0.2449, "step": 77270 }, { "epoch": 4.66, "learning_rate": 4.223439409714364e-06, "loss": 0.1587, "step": 77272 }, { "epoch": 4.66, "learning_rate": 4.22193049960014e-06, "loss": 0.2749, "step": 77274 }, { "epoch": 4.66, "learning_rate": 4.2204215894859146e-06, "loss": 0.2513, "step": 77276 }, { "epoch": 4.66, "learning_rate": 4.21891267937169e-06, "loss": 0.1983, "step": 77278 }, { "epoch": 4.66, "learning_rate": 4.217403769257466e-06, "loss": 0.2702, "step": 77280 }, { "epoch": 4.66, "learning_rate": 4.2158948591432416e-06, "loss": 0.2406, "step": 77282 }, { "epoch": 4.66, "learning_rate": 4.214385949029016e-06, "loss": 0.2509, "step": 77284 }, { "epoch": 4.66, "learning_rate": 4.212877038914792e-06, "loss": 0.2129, "step": 77286 }, { "epoch": 4.66, "learning_rate": 4.211368128800568e-06, "loss": 0.2083, "step": 77288 }, { "epoch": 4.67, "learning_rate": 4.209859218686343e-06, "loss": 0.2822, "step": 77290 }, { "epoch": 4.67, "learning_rate": 4.208350308572118e-06, "loss": 0.2073, "step": 77292 }, { "epoch": 4.67, "learning_rate": 4.206841398457894e-06, "loss": 0.289, "step": 77294 }, { "epoch": 4.67, "learning_rate": 4.2053324883436695e-06, "loss": 0.1843, "step": 77296 }, { "epoch": 4.67, "learning_rate": 4.203823578229445e-06, "loss": 0.2753, "step": 77298 }, { "epoch": 4.67, "learning_rate": 4.20231466811522e-06, "loss": 0.2306, "step": 77300 }, { "epoch": 4.67, "learning_rate": 4.200805758000996e-06, "loss": 0.2453, "step": 77302 }, { "epoch": 4.67, "learning_rate": 4.199296847886771e-06, "loss": 0.1901, "step": 77304 }, { "epoch": 4.67, "learning_rate": 4.197787937772547e-06, "loss": 0.2311, "step": 77306 }, { "epoch": 4.67, "learning_rate": 4.196279027658323e-06, "loss": 0.1876, "step": 77308 }, { "epoch": 4.67, "learning_rate": 4.194770117544098e-06, "loss": 0.1668, "step": 77310 }, { "epoch": 4.67, "learning_rate": 4.193261207429874e-06, "loss": 0.255, "step": 77312 }, { "epoch": 4.67, "learning_rate": 4.19175229731565e-06, "loss": 0.1644, "step": 77314 }, { "epoch": 4.67, "learning_rate": 4.1902433872014245e-06, "loss": 0.243, "step": 77316 }, { "epoch": 4.67, "learning_rate": 4.1887344770872e-06, "loss": 0.2597, "step": 77318 }, { "epoch": 4.67, "learning_rate": 4.187225566972976e-06, "loss": 0.2268, "step": 77320 }, { "epoch": 4.67, "learning_rate": 4.1857166568587515e-06, "loss": 0.1855, "step": 77322 }, { "epoch": 4.67, "learning_rate": 4.184207746744526e-06, "loss": 0.2298, "step": 77324 }, { "epoch": 4.67, "learning_rate": 4.182698836630302e-06, "loss": 0.2791, "step": 77326 }, { "epoch": 4.67, "learning_rate": 4.181189926516078e-06, "loss": 0.303, "step": 77328 }, { "epoch": 4.67, "learning_rate": 4.179681016401853e-06, "loss": 0.2944, "step": 77330 }, { "epoch": 4.67, "learning_rate": 4.178172106287628e-06, "loss": 0.2109, "step": 77332 }, { "epoch": 4.67, "learning_rate": 4.176663196173404e-06, "loss": 0.3037, "step": 77334 }, { "epoch": 4.67, "learning_rate": 4.1751542860591795e-06, "loss": 0.3039, "step": 77336 }, { "epoch": 4.67, "learning_rate": 4.173645375944955e-06, "loss": 0.1791, "step": 77338 }, { "epoch": 4.67, "learning_rate": 4.17213646583073e-06, "loss": 0.2324, "step": 77340 }, { "epoch": 4.67, "learning_rate": 4.170627555716506e-06, "loss": 0.2651, "step": 77342 }, { "epoch": 4.67, "learning_rate": 4.169118645602281e-06, "loss": 0.0964, "step": 77344 }, { "epoch": 4.67, "learning_rate": 4.167609735488057e-06, "loss": 0.1768, "step": 77346 }, { "epoch": 4.67, "learning_rate": 4.166100825373833e-06, "loss": 0.2039, "step": 77348 }, { "epoch": 4.67, "learning_rate": 4.164591915259608e-06, "loss": 0.1509, "step": 77350 }, { "epoch": 4.67, "learning_rate": 4.163083005145384e-06, "loss": 0.1688, "step": 77352 }, { "epoch": 4.67, "learning_rate": 4.16157409503116e-06, "loss": 0.2077, "step": 77354 }, { "epoch": 4.67, "learning_rate": 4.160065184916935e-06, "loss": 0.1827, "step": 77356 }, { "epoch": 4.67, "learning_rate": 4.15855627480271e-06, "loss": 0.2745, "step": 77358 }, { "epoch": 4.67, "learning_rate": 4.157047364688486e-06, "loss": 0.2343, "step": 77360 }, { "epoch": 4.67, "learning_rate": 4.1555384545742615e-06, "loss": 0.2863, "step": 77362 }, { "epoch": 4.67, "learning_rate": 4.154029544460037e-06, "loss": 0.2302, "step": 77364 }, { "epoch": 4.67, "learning_rate": 4.152520634345812e-06, "loss": 0.3109, "step": 77366 }, { "epoch": 4.67, "learning_rate": 4.151011724231588e-06, "loss": 0.174, "step": 77368 }, { "epoch": 4.67, "learning_rate": 4.149502814117363e-06, "loss": 0.2377, "step": 77370 }, { "epoch": 4.67, "learning_rate": 4.147993904003139e-06, "loss": 0.2158, "step": 77372 }, { "epoch": 4.67, "learning_rate": 4.146484993888914e-06, "loss": 0.3102, "step": 77374 }, { "epoch": 4.67, "learning_rate": 4.1449760837746894e-06, "loss": 0.137, "step": 77376 }, { "epoch": 4.67, "learning_rate": 4.143467173660465e-06, "loss": 0.1705, "step": 77378 }, { "epoch": 4.67, "learning_rate": 4.141958263546241e-06, "loss": 0.2058, "step": 77380 }, { "epoch": 4.67, "learning_rate": 4.1404493534320164e-06, "loss": 0.2051, "step": 77382 }, { "epoch": 4.67, "learning_rate": 4.138940443317792e-06, "loss": 0.1911, "step": 77384 }, { "epoch": 4.67, "learning_rate": 4.137431533203568e-06, "loss": 0.2211, "step": 77386 }, { "epoch": 4.67, "learning_rate": 4.1359226230893434e-06, "loss": 0.1733, "step": 77388 }, { "epoch": 4.67, "learning_rate": 4.134413712975118e-06, "loss": 0.1867, "step": 77390 }, { "epoch": 4.67, "learning_rate": 4.132904802860894e-06, "loss": 0.1658, "step": 77392 }, { "epoch": 4.67, "learning_rate": 4.13139589274667e-06, "loss": 0.2469, "step": 77394 }, { "epoch": 4.67, "learning_rate": 4.129886982632445e-06, "loss": 0.1979, "step": 77396 }, { "epoch": 4.67, "learning_rate": 4.12837807251822e-06, "loss": 0.2711, "step": 77398 }, { "epoch": 4.67, "learning_rate": 4.126869162403996e-06, "loss": 0.2765, "step": 77400 }, { "epoch": 4.67, "learning_rate": 4.125360252289771e-06, "loss": 0.2216, "step": 77402 }, { "epoch": 4.67, "learning_rate": 4.123851342175547e-06, "loss": 0.2287, "step": 77404 }, { "epoch": 4.67, "learning_rate": 4.122342432061322e-06, "loss": 0.2644, "step": 77406 }, { "epoch": 4.67, "learning_rate": 4.1208335219470976e-06, "loss": 0.1836, "step": 77408 }, { "epoch": 4.67, "learning_rate": 4.119324611832873e-06, "loss": 0.2083, "step": 77410 }, { "epoch": 4.67, "learning_rate": 4.117815701718649e-06, "loss": 0.2302, "step": 77412 }, { "epoch": 4.67, "learning_rate": 4.116306791604424e-06, "loss": 0.1792, "step": 77414 }, { "epoch": 4.67, "learning_rate": 4.114797881490199e-06, "loss": 0.1686, "step": 77416 }, { "epoch": 4.67, "learning_rate": 4.113288971375975e-06, "loss": 0.1357, "step": 77418 }, { "epoch": 4.67, "learning_rate": 4.111780061261751e-06, "loss": 0.3025, "step": 77420 }, { "epoch": 4.67, "learning_rate": 4.110271151147526e-06, "loss": 0.2199, "step": 77422 }, { "epoch": 4.67, "learning_rate": 4.108762241033302e-06, "loss": 0.1298, "step": 77424 }, { "epoch": 4.67, "learning_rate": 4.107253330919078e-06, "loss": 0.2607, "step": 77426 }, { "epoch": 4.67, "learning_rate": 4.105744420804853e-06, "loss": 0.2819, "step": 77428 }, { "epoch": 4.67, "learning_rate": 4.104235510690629e-06, "loss": 0.2216, "step": 77430 }, { "epoch": 4.67, "learning_rate": 4.102726600576404e-06, "loss": 0.2476, "step": 77432 }, { "epoch": 4.67, "learning_rate": 4.1012176904621795e-06, "loss": 0.2399, "step": 77434 }, { "epoch": 4.67, "learning_rate": 4.099708780347955e-06, "loss": 0.2716, "step": 77436 }, { "epoch": 4.67, "learning_rate": 4.098199870233731e-06, "loss": 0.1904, "step": 77438 }, { "epoch": 4.67, "learning_rate": 4.096690960119506e-06, "loss": 0.1584, "step": 77440 }, { "epoch": 4.67, "learning_rate": 4.095182050005281e-06, "loss": 0.1735, "step": 77442 }, { "epoch": 4.67, "learning_rate": 4.093673139891057e-06, "loss": 0.193, "step": 77444 }, { "epoch": 4.67, "learning_rate": 4.092164229776833e-06, "loss": 0.1437, "step": 77446 }, { "epoch": 4.67, "learning_rate": 4.0906553196626075e-06, "loss": 0.1558, "step": 77448 }, { "epoch": 4.67, "learning_rate": 4.089146409548383e-06, "loss": 0.1636, "step": 77450 }, { "epoch": 4.67, "learning_rate": 4.087637499434159e-06, "loss": 0.193, "step": 77452 }, { "epoch": 4.67, "learning_rate": 4.0861285893199345e-06, "loss": 0.1958, "step": 77454 }, { "epoch": 4.68, "learning_rate": 4.084619679205709e-06, "loss": 0.2795, "step": 77456 }, { "epoch": 4.68, "learning_rate": 4.083110769091485e-06, "loss": 0.2464, "step": 77458 }, { "epoch": 4.68, "learning_rate": 4.081601858977261e-06, "loss": 0.1813, "step": 77460 }, { "epoch": 4.68, "learning_rate": 4.080092948863036e-06, "loss": 0.3505, "step": 77462 }, { "epoch": 4.68, "learning_rate": 4.078584038748812e-06, "loss": 0.2107, "step": 77464 }, { "epoch": 4.68, "learning_rate": 4.077075128634588e-06, "loss": 0.2301, "step": 77466 }, { "epoch": 4.68, "learning_rate": 4.075566218520363e-06, "loss": 0.1711, "step": 77468 }, { "epoch": 4.68, "learning_rate": 4.074057308406139e-06, "loss": 0.2373, "step": 77470 }, { "epoch": 4.68, "learning_rate": 4.072548398291914e-06, "loss": 0.2713, "step": 77472 }, { "epoch": 4.68, "learning_rate": 4.0710394881776895e-06, "loss": 0.2069, "step": 77474 }, { "epoch": 4.68, "learning_rate": 4.069530578063465e-06, "loss": 0.2694, "step": 77476 }, { "epoch": 4.68, "learning_rate": 4.068021667949241e-06, "loss": 0.1264, "step": 77478 }, { "epoch": 4.68, "learning_rate": 4.066512757835016e-06, "loss": 0.1785, "step": 77480 }, { "epoch": 4.68, "learning_rate": 4.065003847720791e-06, "loss": 0.2933, "step": 77482 }, { "epoch": 4.68, "learning_rate": 4.063494937606567e-06, "loss": 0.1614, "step": 77484 }, { "epoch": 4.68, "learning_rate": 4.061986027492343e-06, "loss": 0.2178, "step": 77486 }, { "epoch": 4.68, "learning_rate": 4.0604771173781174e-06, "loss": 0.1682, "step": 77488 }, { "epoch": 4.68, "learning_rate": 4.058968207263893e-06, "loss": 0.2467, "step": 77490 }, { "epoch": 4.68, "learning_rate": 4.057459297149669e-06, "loss": 0.2825, "step": 77492 }, { "epoch": 4.68, "learning_rate": 4.0559503870354444e-06, "loss": 0.3873, "step": 77494 }, { "epoch": 4.68, "learning_rate": 4.054441476921219e-06, "loss": 0.2952, "step": 77496 }, { "epoch": 4.68, "learning_rate": 4.052932566806995e-06, "loss": 0.2991, "step": 77498 }, { "epoch": 4.68, "learning_rate": 4.0514236566927714e-06, "loss": 0.1907, "step": 77500 }, { "epoch": 4.68, "learning_rate": 4.049914746578547e-06, "loss": 0.1758, "step": 77502 }, { "epoch": 4.68, "learning_rate": 4.048405836464322e-06, "loss": 0.2344, "step": 77504 }, { "epoch": 4.68, "learning_rate": 4.046896926350098e-06, "loss": 0.2278, "step": 77506 }, { "epoch": 4.68, "learning_rate": 4.045388016235873e-06, "loss": 0.1538, "step": 77508 }, { "epoch": 4.68, "learning_rate": 4.043879106121649e-06, "loss": 0.1831, "step": 77510 }, { "epoch": 4.68, "learning_rate": 4.042370196007425e-06, "loss": 0.34, "step": 77512 }, { "epoch": 4.68, "learning_rate": 4.040861285893199e-06, "loss": 0.1441, "step": 77514 }, { "epoch": 4.68, "learning_rate": 4.039352375778975e-06, "loss": 0.2557, "step": 77516 }, { "epoch": 4.68, "learning_rate": 4.037843465664751e-06, "loss": 0.3284, "step": 77518 }, { "epoch": 4.68, "learning_rate": 4.036334555550526e-06, "loss": 0.2545, "step": 77520 }, { "epoch": 4.68, "learning_rate": 4.034825645436301e-06, "loss": 0.1637, "step": 77522 }, { "epoch": 4.68, "learning_rate": 4.033316735322077e-06, "loss": 0.2964, "step": 77524 }, { "epoch": 4.68, "learning_rate": 4.0318078252078526e-06, "loss": 0.2045, "step": 77526 }, { "epoch": 4.68, "learning_rate": 4.030298915093628e-06, "loss": 0.1599, "step": 77528 }, { "epoch": 4.68, "learning_rate": 4.028790004979403e-06, "loss": 0.3323, "step": 77530 }, { "epoch": 4.68, "learning_rate": 4.027281094865179e-06, "loss": 0.2681, "step": 77532 }, { "epoch": 4.68, "learning_rate": 4.025772184750954e-06, "loss": 0.1641, "step": 77534 }, { "epoch": 4.68, "learning_rate": 4.02426327463673e-06, "loss": 0.2427, "step": 77536 }, { "epoch": 4.68, "learning_rate": 4.022754364522506e-06, "loss": 0.2557, "step": 77538 }, { "epoch": 4.68, "learning_rate": 4.021245454408281e-06, "loss": 0.2404, "step": 77540 }, { "epoch": 4.68, "learning_rate": 4.019736544294057e-06, "loss": 0.1904, "step": 77542 }, { "epoch": 4.68, "learning_rate": 4.018227634179833e-06, "loss": 0.318, "step": 77544 }, { "epoch": 4.68, "learning_rate": 4.0167187240656075e-06, "loss": 0.2678, "step": 77546 }, { "epoch": 4.68, "learning_rate": 4.015209813951383e-06, "loss": 0.265, "step": 77548 }, { "epoch": 4.68, "learning_rate": 4.013700903837159e-06, "loss": 0.2424, "step": 77550 }, { "epoch": 4.68, "learning_rate": 4.0121919937229345e-06, "loss": 0.2903, "step": 77552 }, { "epoch": 4.68, "learning_rate": 4.010683083608709e-06, "loss": 0.2259, "step": 77554 }, { "epoch": 4.68, "learning_rate": 4.009174173494485e-06, "loss": 0.2041, "step": 77556 }, { "epoch": 4.68, "learning_rate": 4.007665263380261e-06, "loss": 0.2426, "step": 77558 }, { "epoch": 4.68, "learning_rate": 4.006156353266036e-06, "loss": 0.2322, "step": 77560 }, { "epoch": 4.68, "learning_rate": 4.004647443151811e-06, "loss": 0.2528, "step": 77562 }, { "epoch": 4.68, "learning_rate": 4.003138533037587e-06, "loss": 0.1564, "step": 77564 }, { "epoch": 4.68, "learning_rate": 4.0016296229233625e-06, "loss": 0.2816, "step": 77566 }, { "epoch": 4.68, "learning_rate": 4.000120712809138e-06, "loss": 0.1945, "step": 77568 }, { "epoch": 4.68, "learning_rate": 3.998611802694913e-06, "loss": 0.2072, "step": 77570 }, { "epoch": 4.68, "learning_rate": 3.997102892580689e-06, "loss": 0.2828, "step": 77572 }, { "epoch": 4.68, "learning_rate": 3.995593982466464e-06, "loss": 0.2361, "step": 77574 }, { "epoch": 4.68, "learning_rate": 3.99408507235224e-06, "loss": 0.1298, "step": 77576 }, { "epoch": 4.68, "learning_rate": 3.992576162238016e-06, "loss": 0.1368, "step": 77578 }, { "epoch": 4.68, "learning_rate": 3.991067252123791e-06, "loss": 0.2509, "step": 77580 }, { "epoch": 4.68, "learning_rate": 3.989558342009567e-06, "loss": 0.2354, "step": 77582 }, { "epoch": 4.68, "learning_rate": 3.988049431895343e-06, "loss": 0.3011, "step": 77584 }, { "epoch": 4.68, "learning_rate": 3.986540521781118e-06, "loss": 0.2577, "step": 77586 }, { "epoch": 4.68, "learning_rate": 3.985031611666893e-06, "loss": 0.1131, "step": 77588 }, { "epoch": 4.68, "learning_rate": 3.983522701552669e-06, "loss": 0.1588, "step": 77590 }, { "epoch": 4.68, "learning_rate": 3.9820137914384445e-06, "loss": 0.3071, "step": 77592 }, { "epoch": 4.68, "learning_rate": 3.98050488132422e-06, "loss": 0.187, "step": 77594 }, { "epoch": 4.68, "learning_rate": 3.978995971209995e-06, "loss": 0.2568, "step": 77596 }, { "epoch": 4.68, "learning_rate": 3.977487061095771e-06, "loss": 0.254, "step": 77598 }, { "epoch": 4.68, "learning_rate": 3.975978150981546e-06, "loss": 0.164, "step": 77600 }, { "epoch": 4.68, "learning_rate": 3.974469240867322e-06, "loss": 0.2866, "step": 77602 }, { "epoch": 4.68, "learning_rate": 3.972960330753097e-06, "loss": 0.3058, "step": 77604 }, { "epoch": 4.68, "learning_rate": 3.9714514206388724e-06, "loss": 0.1623, "step": 77606 }, { "epoch": 4.68, "learning_rate": 3.969942510524648e-06, "loss": 0.1654, "step": 77608 }, { "epoch": 4.68, "learning_rate": 3.968433600410424e-06, "loss": 0.1297, "step": 77610 }, { "epoch": 4.68, "learning_rate": 3.966924690296199e-06, "loss": 0.2419, "step": 77612 }, { "epoch": 4.68, "learning_rate": 3.965415780181974e-06, "loss": 0.3006, "step": 77614 }, { "epoch": 4.68, "learning_rate": 3.96390687006775e-06, "loss": 0.3126, "step": 77616 }, { "epoch": 4.68, "learning_rate": 3.9623979599535264e-06, "loss": 0.1545, "step": 77618 }, { "epoch": 4.68, "learning_rate": 3.960889049839301e-06, "loss": 0.2383, "step": 77620 }, { "epoch": 4.69, "learning_rate": 3.959380139725077e-06, "loss": 0.1917, "step": 77622 }, { "epoch": 4.69, "learning_rate": 3.957871229610853e-06, "loss": 0.223, "step": 77624 }, { "epoch": 4.69, "learning_rate": 3.956362319496628e-06, "loss": 0.187, "step": 77626 }, { "epoch": 4.69, "learning_rate": 3.954853409382403e-06, "loss": 0.1598, "step": 77628 }, { "epoch": 4.69, "learning_rate": 3.953344499268179e-06, "loss": 0.2552, "step": 77630 }, { "epoch": 4.69, "learning_rate": 3.951835589153954e-06, "loss": 0.2616, "step": 77632 }, { "epoch": 4.69, "learning_rate": 3.95032667903973e-06, "loss": 0.2011, "step": 77634 }, { "epoch": 4.69, "learning_rate": 3.948817768925505e-06, "loss": 0.2012, "step": 77636 }, { "epoch": 4.69, "learning_rate": 3.9473088588112806e-06, "loss": 0.2393, "step": 77638 }, { "epoch": 4.69, "learning_rate": 3.945799948697056e-06, "loss": 0.1646, "step": 77640 }, { "epoch": 4.69, "learning_rate": 3.944291038582832e-06, "loss": 0.2519, "step": 77642 }, { "epoch": 4.69, "learning_rate": 3.942782128468607e-06, "loss": 0.1978, "step": 77644 }, { "epoch": 4.69, "learning_rate": 3.941273218354382e-06, "loss": 0.1776, "step": 77646 }, { "epoch": 4.69, "learning_rate": 3.939764308240158e-06, "loss": 0.2104, "step": 77648 }, { "epoch": 4.69, "learning_rate": 3.938255398125934e-06, "loss": 0.1562, "step": 77650 }, { "epoch": 4.69, "learning_rate": 3.936746488011709e-06, "loss": 0.1339, "step": 77652 }, { "epoch": 4.69, "learning_rate": 3.935237577897485e-06, "loss": 0.2586, "step": 77654 }, { "epoch": 4.69, "learning_rate": 3.933728667783261e-06, "loss": 0.2096, "step": 77656 }, { "epoch": 4.69, "learning_rate": 3.932219757669036e-06, "loss": 0.1533, "step": 77658 }, { "epoch": 4.69, "learning_rate": 3.930710847554811e-06, "loss": 0.2094, "step": 77660 }, { "epoch": 4.69, "learning_rate": 3.929201937440587e-06, "loss": 0.2603, "step": 77662 }, { "epoch": 4.69, "learning_rate": 3.9276930273263625e-06, "loss": 0.2322, "step": 77664 }, { "epoch": 4.69, "learning_rate": 3.926184117212138e-06, "loss": 0.1682, "step": 77666 }, { "epoch": 4.69, "learning_rate": 3.924675207097914e-06, "loss": 0.218, "step": 77668 }, { "epoch": 4.69, "learning_rate": 3.923166296983689e-06, "loss": 0.2496, "step": 77670 }, { "epoch": 4.69, "learning_rate": 3.921657386869464e-06, "loss": 0.2448, "step": 77672 }, { "epoch": 4.69, "learning_rate": 3.92014847675524e-06, "loss": 0.2069, "step": 77674 }, { "epoch": 4.69, "learning_rate": 3.918639566641016e-06, "loss": 0.1737, "step": 77676 }, { "epoch": 4.69, "learning_rate": 3.9171306565267905e-06, "loss": 0.1898, "step": 77678 }, { "epoch": 4.69, "learning_rate": 3.915621746412566e-06, "loss": 0.2492, "step": 77680 }, { "epoch": 4.69, "learning_rate": 3.914112836298342e-06, "loss": 0.3487, "step": 77682 }, { "epoch": 4.69, "learning_rate": 3.9126039261841175e-06, "loss": 0.1808, "step": 77684 }, { "epoch": 4.69, "learning_rate": 3.911095016069892e-06, "loss": 0.1595, "step": 77686 }, { "epoch": 4.69, "learning_rate": 3.909586105955668e-06, "loss": 0.1859, "step": 77688 }, { "epoch": 4.69, "learning_rate": 3.908077195841444e-06, "loss": 0.3833, "step": 77690 }, { "epoch": 4.69, "learning_rate": 3.906568285727219e-06, "loss": 0.2624, "step": 77692 }, { "epoch": 4.69, "learning_rate": 3.905059375612995e-06, "loss": 0.2991, "step": 77694 }, { "epoch": 4.69, "learning_rate": 3.903550465498771e-06, "loss": 0.1733, "step": 77696 }, { "epoch": 4.69, "learning_rate": 3.902041555384546e-06, "loss": 0.285, "step": 77698 }, { "epoch": 4.69, "learning_rate": 3.900532645270322e-06, "loss": 0.1955, "step": 77700 }, { "epoch": 4.69, "learning_rate": 3.899023735156097e-06, "loss": 0.3093, "step": 77702 }, { "epoch": 4.69, "learning_rate": 3.8975148250418725e-06, "loss": 0.2379, "step": 77704 }, { "epoch": 4.69, "learning_rate": 3.896005914927648e-06, "loss": 0.2574, "step": 77706 }, { "epoch": 4.69, "learning_rate": 3.894497004813424e-06, "loss": 0.2382, "step": 77708 }, { "epoch": 4.69, "learning_rate": 3.892988094699199e-06, "loss": 0.1176, "step": 77710 }, { "epoch": 4.69, "learning_rate": 3.891479184584974e-06, "loss": 0.2029, "step": 77712 }, { "epoch": 4.69, "learning_rate": 3.88997027447075e-06, "loss": 0.1767, "step": 77714 }, { "epoch": 4.69, "learning_rate": 3.888461364356526e-06, "loss": 0.1417, "step": 77716 }, { "epoch": 4.69, "learning_rate": 3.8869524542423005e-06, "loss": 0.2381, "step": 77718 }, { "epoch": 4.69, "learning_rate": 3.885443544128076e-06, "loss": 0.2396, "step": 77720 }, { "epoch": 4.69, "learning_rate": 3.883934634013852e-06, "loss": 0.197, "step": 77722 }, { "epoch": 4.69, "learning_rate": 3.8824257238996275e-06, "loss": 0.2209, "step": 77724 }, { "epoch": 4.69, "learning_rate": 3.880916813785402e-06, "loss": 0.172, "step": 77726 }, { "epoch": 4.69, "learning_rate": 3.879407903671178e-06, "loss": 0.184, "step": 77728 }, { "epoch": 4.69, "learning_rate": 3.877898993556954e-06, "loss": 0.2547, "step": 77730 }, { "epoch": 4.69, "learning_rate": 3.876390083442729e-06, "loss": 0.1708, "step": 77732 }, { "epoch": 4.69, "learning_rate": 3.874881173328505e-06, "loss": 0.3015, "step": 77734 }, { "epoch": 4.69, "learning_rate": 3.873372263214281e-06, "loss": 0.2325, "step": 77736 }, { "epoch": 4.69, "learning_rate": 3.871863353100056e-06, "loss": 0.2063, "step": 77738 }, { "epoch": 4.69, "learning_rate": 3.870354442985832e-06, "loss": 0.3309, "step": 77740 }, { "epoch": 4.69, "learning_rate": 3.868845532871608e-06, "loss": 0.1992, "step": 77742 }, { "epoch": 4.69, "learning_rate": 3.8673366227573824e-06, "loss": 0.1812, "step": 77744 }, { "epoch": 4.69, "learning_rate": 3.865827712643158e-06, "loss": 0.1657, "step": 77746 }, { "epoch": 4.69, "learning_rate": 3.864318802528934e-06, "loss": 0.3055, "step": 77748 }, { "epoch": 4.69, "learning_rate": 3.8628098924147094e-06, "loss": 0.2342, "step": 77750 }, { "epoch": 4.69, "learning_rate": 3.861300982300484e-06, "loss": 0.2257, "step": 77752 }, { "epoch": 4.69, "learning_rate": 3.85979207218626e-06, "loss": 0.1438, "step": 77754 }, { "epoch": 4.69, "learning_rate": 3.858283162072036e-06, "loss": 0.1877, "step": 77756 }, { "epoch": 4.69, "learning_rate": 3.856774251957811e-06, "loss": 0.1658, "step": 77758 }, { "epoch": 4.69, "learning_rate": 3.855265341843586e-06, "loss": 0.2238, "step": 77760 }, { "epoch": 4.69, "learning_rate": 3.853756431729362e-06, "loss": 0.1572, "step": 77762 }, { "epoch": 4.69, "learning_rate": 3.852247521615137e-06, "loss": 0.2056, "step": 77764 }, { "epoch": 4.69, "learning_rate": 3.850738611500913e-06, "loss": 0.2373, "step": 77766 }, { "epoch": 4.69, "learning_rate": 3.849229701386689e-06, "loss": 0.123, "step": 77768 }, { "epoch": 4.69, "learning_rate": 3.847720791272464e-06, "loss": 0.2566, "step": 77770 }, { "epoch": 4.69, "learning_rate": 3.84621188115824e-06, "loss": 0.0836, "step": 77772 }, { "epoch": 4.69, "learning_rate": 3.844702971044016e-06, "loss": 0.2778, "step": 77774 }, { "epoch": 4.69, "learning_rate": 3.8431940609297906e-06, "loss": 0.2092, "step": 77776 }, { "epoch": 4.69, "learning_rate": 3.841685150815566e-06, "loss": 0.3031, "step": 77778 }, { "epoch": 4.69, "learning_rate": 3.840176240701342e-06, "loss": 0.1221, "step": 77780 }, { "epoch": 4.69, "learning_rate": 3.8386673305871176e-06, "loss": 0.264, "step": 77782 }, { "epoch": 4.69, "learning_rate": 3.837158420472892e-06, "loss": 0.2327, "step": 77784 }, { "epoch": 4.69, "learning_rate": 3.835649510358668e-06, "loss": 0.2996, "step": 77786 }, { "epoch": 4.7, "learning_rate": 3.834140600244444e-06, "loss": 0.1964, "step": 77788 }, { "epoch": 4.7, "learning_rate": 3.832631690130219e-06, "loss": 0.2356, "step": 77790 }, { "epoch": 4.7, "learning_rate": 3.831122780015994e-06, "loss": 0.1954, "step": 77792 }, { "epoch": 4.7, "learning_rate": 3.82961386990177e-06, "loss": 0.3344, "step": 77794 }, { "epoch": 4.7, "learning_rate": 3.8281049597875455e-06, "loss": 0.2507, "step": 77796 }, { "epoch": 4.7, "learning_rate": 3.826596049673321e-06, "loss": 0.1226, "step": 77798 }, { "epoch": 4.7, "learning_rate": 3.825087139559096e-06, "loss": 0.1809, "step": 77800 }, { "epoch": 4.7, "learning_rate": 3.823578229444872e-06, "loss": 0.1717, "step": 77802 }, { "epoch": 4.7, "learning_rate": 3.822069319330647e-06, "loss": 0.2425, "step": 77804 }, { "epoch": 4.7, "learning_rate": 3.820560409216423e-06, "loss": 0.3536, "step": 77806 }, { "epoch": 4.7, "learning_rate": 3.819051499102199e-06, "loss": 0.2628, "step": 77808 }, { "epoch": 4.7, "learning_rate": 3.817542588987974e-06, "loss": 0.2193, "step": 77810 }, { "epoch": 4.7, "learning_rate": 3.81603367887375e-06, "loss": 0.2719, "step": 77812 }, { "epoch": 4.7, "learning_rate": 3.8145247687595253e-06, "loss": 0.2564, "step": 77814 }, { "epoch": 4.7, "learning_rate": 3.8130158586453005e-06, "loss": 0.1845, "step": 77816 }, { "epoch": 4.7, "learning_rate": 3.811506948531076e-06, "loss": 0.1585, "step": 77818 }, { "epoch": 4.7, "learning_rate": 3.809998038416852e-06, "loss": 0.2265, "step": 77820 }, { "epoch": 4.7, "learning_rate": 3.8084891283026275e-06, "loss": 0.299, "step": 77822 }, { "epoch": 4.7, "learning_rate": 3.806980218188403e-06, "loss": 0.1865, "step": 77824 }, { "epoch": 4.7, "learning_rate": 3.805471308074178e-06, "loss": 0.2461, "step": 77826 }, { "epoch": 4.7, "learning_rate": 3.8039623979599537e-06, "loss": 0.1785, "step": 77828 }, { "epoch": 4.7, "learning_rate": 3.8024534878457293e-06, "loss": 0.1361, "step": 77830 }, { "epoch": 4.7, "learning_rate": 3.800944577731505e-06, "loss": 0.3634, "step": 77832 }, { "epoch": 4.7, "learning_rate": 3.79943566761728e-06, "loss": 0.1964, "step": 77834 }, { "epoch": 4.7, "learning_rate": 3.7979267575030555e-06, "loss": 0.1466, "step": 77836 }, { "epoch": 4.7, "learning_rate": 3.7964178473888316e-06, "loss": 0.204, "step": 77838 }, { "epoch": 4.7, "learning_rate": 3.7949089372746072e-06, "loss": 0.2231, "step": 77840 }, { "epoch": 4.7, "learning_rate": 3.793400027160382e-06, "loss": 0.3129, "step": 77842 }, { "epoch": 4.7, "learning_rate": 3.7918911170461577e-06, "loss": 0.2043, "step": 77844 }, { "epoch": 4.7, "learning_rate": 3.7903822069319334e-06, "loss": 0.307, "step": 77846 }, { "epoch": 4.7, "learning_rate": 3.788873296817709e-06, "loss": 0.2011, "step": 77848 }, { "epoch": 4.7, "learning_rate": 3.787364386703484e-06, "loss": 0.1875, "step": 77850 }, { "epoch": 4.7, "learning_rate": 3.7858554765892595e-06, "loss": 0.3183, "step": 77852 }, { "epoch": 4.7, "learning_rate": 3.784346566475035e-06, "loss": 0.2506, "step": 77854 }, { "epoch": 4.7, "learning_rate": 3.782837656360811e-06, "loss": 0.4282, "step": 77856 }, { "epoch": 4.7, "learning_rate": 3.781328746246586e-06, "loss": 0.1748, "step": 77858 }, { "epoch": 4.7, "learning_rate": 3.7798198361323618e-06, "loss": 0.1516, "step": 77860 }, { "epoch": 4.7, "learning_rate": 3.7783109260181374e-06, "loss": 0.4129, "step": 77862 }, { "epoch": 4.7, "learning_rate": 3.776802015903913e-06, "loss": 0.1926, "step": 77864 }, { "epoch": 4.7, "learning_rate": 3.775293105789688e-06, "loss": 0.1611, "step": 77866 }, { "epoch": 4.7, "learning_rate": 3.7737841956754636e-06, "loss": 0.2474, "step": 77868 }, { "epoch": 4.7, "learning_rate": 3.7722752855612393e-06, "loss": 0.1869, "step": 77870 }, { "epoch": 4.7, "learning_rate": 3.770766375447015e-06, "loss": 0.2804, "step": 77872 }, { "epoch": 4.7, "learning_rate": 3.76925746533279e-06, "loss": 0.2683, "step": 77874 }, { "epoch": 4.7, "learning_rate": 3.767748555218566e-06, "loss": 0.2455, "step": 77876 }, { "epoch": 4.7, "learning_rate": 3.7662396451043415e-06, "loss": 0.2467, "step": 77878 }, { "epoch": 4.7, "learning_rate": 3.764730734990117e-06, "loss": 0.1159, "step": 77880 }, { "epoch": 4.7, "learning_rate": 3.763221824875892e-06, "loss": 0.1414, "step": 77882 }, { "epoch": 4.7, "learning_rate": 3.7617129147616677e-06, "loss": 0.3253, "step": 77884 }, { "epoch": 4.7, "learning_rate": 3.7602040046474433e-06, "loss": 0.2246, "step": 77886 }, { "epoch": 4.7, "learning_rate": 3.758695094533219e-06, "loss": 0.2428, "step": 77888 }, { "epoch": 4.7, "learning_rate": 3.757186184418994e-06, "loss": 0.2742, "step": 77890 }, { "epoch": 4.7, "learning_rate": 3.7556772743047695e-06, "loss": 0.1971, "step": 77892 }, { "epoch": 4.7, "learning_rate": 3.754168364190545e-06, "loss": 0.1611, "step": 77894 }, { "epoch": 4.7, "learning_rate": 3.7526594540763212e-06, "loss": 0.2333, "step": 77896 }, { "epoch": 4.7, "learning_rate": 3.751904999019209e-06, "loss": 0.2887, "step": 77898 }, { "epoch": 4.7, "learning_rate": 3.750396088904984e-06, "loss": 0.2122, "step": 77900 }, { "epoch": 4.7, "learning_rate": 3.7488871787907596e-06, "loss": 0.1674, "step": 77902 }, { "epoch": 4.7, "learning_rate": 3.7473782686765352e-06, "loss": 0.195, "step": 77904 }, { "epoch": 4.7, "learning_rate": 3.745869358562311e-06, "loss": 0.2188, "step": 77906 }, { "epoch": 4.7, "learning_rate": 3.7443604484480857e-06, "loss": 0.2635, "step": 77908 }, { "epoch": 4.7, "learning_rate": 3.7428515383338614e-06, "loss": 0.3865, "step": 77910 }, { "epoch": 4.7, "learning_rate": 3.741342628219637e-06, "loss": 0.2242, "step": 77912 }, { "epoch": 4.7, "learning_rate": 3.7398337181054127e-06, "loss": 0.1663, "step": 77914 }, { "epoch": 4.7, "learning_rate": 3.738324807991188e-06, "loss": 0.2474, "step": 77916 }, { "epoch": 4.7, "learning_rate": 3.7368158978769636e-06, "loss": 0.203, "step": 77918 }, { "epoch": 4.7, "learning_rate": 3.7353069877627393e-06, "loss": 0.2436, "step": 77920 }, { "epoch": 4.7, "learning_rate": 3.733798077648515e-06, "loss": 0.2384, "step": 77922 }, { "epoch": 4.7, "learning_rate": 3.7322891675342898e-06, "loss": 0.3369, "step": 77924 }, { "epoch": 4.7, "learning_rate": 3.7307802574200654e-06, "loss": 0.1471, "step": 77926 }, { "epoch": 4.7, "learning_rate": 3.729271347305841e-06, "loss": 0.153, "step": 77928 }, { "epoch": 4.7, "learning_rate": 3.7277624371916168e-06, "loss": 0.2655, "step": 77930 }, { "epoch": 4.7, "learning_rate": 3.726253527077392e-06, "loss": 0.2212, "step": 77932 }, { "epoch": 4.7, "learning_rate": 3.7247446169631677e-06, "loss": 0.2164, "step": 77934 }, { "epoch": 4.7, "learning_rate": 3.7232357068489433e-06, "loss": 0.1959, "step": 77936 }, { "epoch": 4.7, "learning_rate": 3.721726796734719e-06, "loss": 0.1928, "step": 77938 }, { "epoch": 4.7, "learning_rate": 3.7202178866204947e-06, "loss": 0.2001, "step": 77940 }, { "epoch": 4.7, "learning_rate": 3.7187089765062695e-06, "loss": 0.1688, "step": 77942 }, { "epoch": 4.7, "learning_rate": 3.717200066392045e-06, "loss": 0.2558, "step": 77944 }, { "epoch": 4.7, "learning_rate": 3.715691156277821e-06, "loss": 0.2565, "step": 77946 }, { "epoch": 4.7, "learning_rate": 3.7141822461635965e-06, "loss": 0.1888, "step": 77948 }, { "epoch": 4.7, "learning_rate": 3.7126733360493713e-06, "loss": 0.3143, "step": 77950 }, { "epoch": 4.7, "learning_rate": 3.711164425935147e-06, "loss": 0.1851, "step": 77952 }, { "epoch": 4.71, "learning_rate": 3.7096555158209226e-06, "loss": 0.2046, "step": 77954 }, { "epoch": 4.71, "learning_rate": 3.7081466057066987e-06, "loss": 0.3908, "step": 77956 }, { "epoch": 4.71, "learning_rate": 3.7066376955924736e-06, "loss": 0.2142, "step": 77958 }, { "epoch": 4.71, "learning_rate": 3.7051287854782492e-06, "loss": 0.1626, "step": 77960 }, { "epoch": 4.71, "learning_rate": 3.703619875364025e-06, "loss": 0.2604, "step": 77962 }, { "epoch": 4.71, "learning_rate": 3.7021109652498006e-06, "loss": 0.3742, "step": 77964 }, { "epoch": 4.71, "learning_rate": 3.7006020551355754e-06, "loss": 0.1711, "step": 77966 }, { "epoch": 4.71, "learning_rate": 3.699093145021351e-06, "loss": 0.1657, "step": 77968 }, { "epoch": 4.71, "learning_rate": 3.6975842349071267e-06, "loss": 0.2707, "step": 77970 }, { "epoch": 4.71, "learning_rate": 3.6960753247929024e-06, "loss": 0.2627, "step": 77972 }, { "epoch": 4.71, "learning_rate": 3.6945664146786776e-06, "loss": 0.175, "step": 77974 }, { "epoch": 4.71, "learning_rate": 3.6930575045644533e-06, "loss": 0.2268, "step": 77976 }, { "epoch": 4.71, "learning_rate": 3.691548594450229e-06, "loss": 0.1894, "step": 77978 }, { "epoch": 4.71, "learning_rate": 3.6900396843360046e-06, "loss": 0.3392, "step": 77980 }, { "epoch": 4.71, "learning_rate": 3.6885307742217794e-06, "loss": 0.2024, "step": 77982 }, { "epoch": 4.71, "learning_rate": 3.687021864107555e-06, "loss": 0.1155, "step": 77984 }, { "epoch": 4.71, "learning_rate": 3.6855129539933308e-06, "loss": 0.2321, "step": 77986 }, { "epoch": 4.71, "learning_rate": 3.6840040438791064e-06, "loss": 0.2405, "step": 77988 }, { "epoch": 4.71, "learning_rate": 3.6824951337648817e-06, "loss": 0.1603, "step": 77990 }, { "epoch": 4.71, "learning_rate": 3.6809862236506574e-06, "loss": 0.1417, "step": 77992 }, { "epoch": 4.71, "learning_rate": 3.679477313536433e-06, "loss": 0.1611, "step": 77994 }, { "epoch": 4.71, "learning_rate": 3.6779684034222087e-06, "loss": 0.204, "step": 77996 }, { "epoch": 4.71, "learning_rate": 3.6764594933079835e-06, "loss": 0.2196, "step": 77998 }, { "epoch": 4.71, "learning_rate": 3.674950583193759e-06, "loss": 0.2638, "step": 78000 }, { "epoch": 4.71, "learning_rate": 3.673441673079535e-06, "loss": 0.269, "step": 78002 }, { "epoch": 4.71, "learning_rate": 3.6719327629653105e-06, "loss": 0.1665, "step": 78004 }, { "epoch": 4.71, "learning_rate": 3.6704238528510853e-06, "loss": 0.1244, "step": 78006 }, { "epoch": 4.71, "learning_rate": 3.668914942736861e-06, "loss": 0.2031, "step": 78008 }, { "epoch": 4.71, "learning_rate": 3.6674060326226367e-06, "loss": 0.1754, "step": 78010 }, { "epoch": 4.71, "learning_rate": 3.6658971225084123e-06, "loss": 0.2788, "step": 78012 }, { "epoch": 4.71, "learning_rate": 3.664388212394188e-06, "loss": 0.2055, "step": 78014 }, { "epoch": 4.71, "learning_rate": 3.6628793022799632e-06, "loss": 0.2387, "step": 78016 }, { "epoch": 4.71, "learning_rate": 3.661370392165739e-06, "loss": 0.3318, "step": 78018 }, { "epoch": 4.71, "learning_rate": 3.6598614820515146e-06, "loss": 0.1592, "step": 78020 }, { "epoch": 4.71, "learning_rate": 3.6583525719372902e-06, "loss": 0.1907, "step": 78022 }, { "epoch": 4.71, "learning_rate": 3.656843661823065e-06, "loss": 0.23, "step": 78024 }, { "epoch": 4.71, "learning_rate": 3.6553347517088407e-06, "loss": 0.2841, "step": 78026 }, { "epoch": 4.71, "learning_rate": 3.6538258415946164e-06, "loss": 0.2427, "step": 78028 }, { "epoch": 4.71, "learning_rate": 3.652316931480392e-06, "loss": 0.1489, "step": 78030 }, { "epoch": 4.71, "learning_rate": 3.6508080213661673e-06, "loss": 0.0875, "step": 78032 }, { "epoch": 4.71, "learning_rate": 3.649299111251943e-06, "loss": 0.1705, "step": 78034 }, { "epoch": 4.71, "learning_rate": 3.6477902011377186e-06, "loss": 0.2136, "step": 78036 }, { "epoch": 4.71, "learning_rate": 3.6462812910234943e-06, "loss": 0.1791, "step": 78038 }, { "epoch": 4.71, "learning_rate": 3.644772380909269e-06, "loss": 0.2486, "step": 78040 }, { "epoch": 4.71, "learning_rate": 3.6432634707950448e-06, "loss": 0.3791, "step": 78042 }, { "epoch": 4.71, "learning_rate": 3.6417545606808204e-06, "loss": 0.1694, "step": 78044 }, { "epoch": 4.71, "learning_rate": 3.640245650566596e-06, "loss": 0.3923, "step": 78046 }, { "epoch": 4.71, "learning_rate": 3.6387367404523714e-06, "loss": 0.2218, "step": 78048 }, { "epoch": 4.71, "learning_rate": 3.637227830338147e-06, "loss": 0.2153, "step": 78050 }, { "epoch": 4.71, "learning_rate": 3.6357189202239227e-06, "loss": 0.1607, "step": 78052 }, { "epoch": 4.71, "learning_rate": 3.6342100101096984e-06, "loss": 0.2694, "step": 78054 }, { "epoch": 4.71, "learning_rate": 3.632701099995473e-06, "loss": 0.2156, "step": 78056 }, { "epoch": 4.71, "learning_rate": 3.631192189881249e-06, "loss": 0.2075, "step": 78058 }, { "epoch": 4.71, "learning_rate": 3.6296832797670245e-06, "loss": 0.1932, "step": 78060 }, { "epoch": 4.71, "learning_rate": 3.6281743696528e-06, "loss": 0.3008, "step": 78062 }, { "epoch": 4.71, "learning_rate": 3.626665459538575e-06, "loss": 0.1953, "step": 78064 }, { "epoch": 4.71, "learning_rate": 3.6251565494243507e-06, "loss": 0.2276, "step": 78066 }, { "epoch": 4.71, "learning_rate": 3.6236476393101263e-06, "loss": 0.212, "step": 78068 }, { "epoch": 4.71, "learning_rate": 3.622138729195902e-06, "loss": 0.1793, "step": 78070 }, { "epoch": 4.71, "learning_rate": 3.6206298190816772e-06, "loss": 0.2137, "step": 78072 }, { "epoch": 4.71, "learning_rate": 3.619120908967453e-06, "loss": 0.1609, "step": 78074 }, { "epoch": 4.71, "learning_rate": 3.6176119988532286e-06, "loss": 0.1304, "step": 78076 }, { "epoch": 4.71, "learning_rate": 3.6161030887390042e-06, "loss": 0.1106, "step": 78078 }, { "epoch": 4.71, "learning_rate": 3.614594178624779e-06, "loss": 0.2542, "step": 78080 }, { "epoch": 4.71, "learning_rate": 3.6130852685105547e-06, "loss": 0.4116, "step": 78082 }, { "epoch": 4.71, "learning_rate": 3.6115763583963304e-06, "loss": 0.2227, "step": 78084 }, { "epoch": 4.71, "learning_rate": 3.610067448282106e-06, "loss": 0.2032, "step": 78086 }, { "epoch": 4.71, "learning_rate": 3.6085585381678813e-06, "loss": 0.2384, "step": 78088 }, { "epoch": 4.71, "learning_rate": 3.607049628053657e-06, "loss": 0.1714, "step": 78090 }, { "epoch": 4.71, "learning_rate": 3.6055407179394326e-06, "loss": 0.3154, "step": 78092 }, { "epoch": 4.71, "learning_rate": 3.6040318078252083e-06, "loss": 0.1661, "step": 78094 }, { "epoch": 4.71, "learning_rate": 3.602522897710984e-06, "loss": 0.3061, "step": 78096 }, { "epoch": 4.71, "learning_rate": 3.6010139875967588e-06, "loss": 0.148, "step": 78098 }, { "epoch": 4.71, "learning_rate": 3.5995050774825345e-06, "loss": 0.1181, "step": 78100 }, { "epoch": 4.71, "learning_rate": 3.59799616736831e-06, "loss": 0.2977, "step": 78102 }, { "epoch": 4.71, "learning_rate": 3.5964872572540858e-06, "loss": 0.1937, "step": 78104 }, { "epoch": 4.71, "learning_rate": 3.594978347139861e-06, "loss": 0.1339, "step": 78106 }, { "epoch": 4.71, "learning_rate": 3.5934694370256367e-06, "loss": 0.171, "step": 78108 }, { "epoch": 4.71, "learning_rate": 3.5919605269114124e-06, "loss": 0.3469, "step": 78110 }, { "epoch": 4.71, "learning_rate": 3.590451616797188e-06, "loss": 0.1975, "step": 78112 }, { "epoch": 4.71, "learning_rate": 3.588942706682963e-06, "loss": 0.2819, "step": 78114 }, { "epoch": 4.71, "learning_rate": 3.5874337965687385e-06, "loss": 0.1381, "step": 78116 }, { "epoch": 4.71, "learning_rate": 3.585924886454514e-06, "loss": 0.2131, "step": 78118 }, { "epoch": 4.72, "learning_rate": 3.58441597634029e-06, "loss": 0.2183, "step": 78120 }, { "epoch": 4.72, "learning_rate": 3.5829070662260647e-06, "loss": 0.1979, "step": 78122 }, { "epoch": 4.72, "learning_rate": 3.5813981561118403e-06, "loss": 0.1785, "step": 78124 }, { "epoch": 4.72, "learning_rate": 3.579889245997616e-06, "loss": 0.1591, "step": 78126 }, { "epoch": 4.72, "learning_rate": 3.5783803358833917e-06, "loss": 0.2629, "step": 78128 }, { "epoch": 4.72, "learning_rate": 3.576871425769167e-06, "loss": 0.2408, "step": 78130 }, { "epoch": 4.72, "learning_rate": 3.5753625156549426e-06, "loss": 0.2374, "step": 78132 }, { "epoch": 4.72, "learning_rate": 3.5738536055407182e-06, "loss": 0.2929, "step": 78134 }, { "epoch": 4.72, "learning_rate": 3.572344695426494e-06, "loss": 0.2345, "step": 78136 }, { "epoch": 4.72, "learning_rate": 3.5708357853122687e-06, "loss": 0.191, "step": 78138 }, { "epoch": 4.72, "learning_rate": 3.5693268751980444e-06, "loss": 0.1554, "step": 78140 }, { "epoch": 4.72, "learning_rate": 3.56781796508382e-06, "loss": 0.2324, "step": 78142 }, { "epoch": 4.72, "learning_rate": 3.5663090549695957e-06, "loss": 0.222, "step": 78144 }, { "epoch": 4.72, "learning_rate": 3.564800144855371e-06, "loss": 0.2138, "step": 78146 }, { "epoch": 4.72, "learning_rate": 3.5632912347411466e-06, "loss": 0.184, "step": 78148 }, { "epoch": 4.72, "learning_rate": 3.5617823246269223e-06, "loss": 0.1538, "step": 78150 }, { "epoch": 4.72, "learning_rate": 3.560273414512698e-06, "loss": 0.2479, "step": 78152 }, { "epoch": 4.72, "learning_rate": 3.558764504398473e-06, "loss": 0.3152, "step": 78154 }, { "epoch": 4.72, "learning_rate": 3.5572555942842485e-06, "loss": 0.2167, "step": 78156 }, { "epoch": 4.72, "learning_rate": 3.555746684170024e-06, "loss": 0.1547, "step": 78158 }, { "epoch": 4.72, "learning_rate": 3.5542377740558e-06, "loss": 0.1483, "step": 78160 }, { "epoch": 4.72, "learning_rate": 3.552728863941575e-06, "loss": 0.1753, "step": 78162 }, { "epoch": 4.72, "learning_rate": 3.5512199538273507e-06, "loss": 0.2219, "step": 78164 }, { "epoch": 4.72, "learning_rate": 3.5497110437131264e-06, "loss": 0.2774, "step": 78166 }, { "epoch": 4.72, "learning_rate": 3.548202133598902e-06, "loss": 0.2115, "step": 78168 }, { "epoch": 4.72, "learning_rate": 3.5466932234846777e-06, "loss": 0.3296, "step": 78170 }, { "epoch": 4.72, "learning_rate": 3.5451843133704525e-06, "loss": 0.3136, "step": 78172 }, { "epoch": 4.72, "learning_rate": 3.543675403256228e-06, "loss": 0.1616, "step": 78174 }, { "epoch": 4.72, "learning_rate": 3.542166493142004e-06, "loss": 0.2394, "step": 78176 }, { "epoch": 4.72, "learning_rate": 3.5406575830277795e-06, "loss": 0.1711, "step": 78178 }, { "epoch": 4.72, "learning_rate": 3.5391486729135543e-06, "loss": 0.2304, "step": 78180 }, { "epoch": 4.72, "learning_rate": 3.53763976279933e-06, "loss": 0.2307, "step": 78182 }, { "epoch": 4.72, "learning_rate": 3.5361308526851057e-06, "loss": 0.1504, "step": 78184 }, { "epoch": 4.72, "learning_rate": 3.5346219425708813e-06, "loss": 0.2576, "step": 78186 }, { "epoch": 4.72, "learning_rate": 3.5331130324566566e-06, "loss": 0.2364, "step": 78188 }, { "epoch": 4.72, "learning_rate": 3.5316041223424322e-06, "loss": 0.1692, "step": 78190 }, { "epoch": 4.72, "learning_rate": 3.530095212228208e-06, "loss": 0.2176, "step": 78192 }, { "epoch": 4.72, "learning_rate": 3.5285863021139836e-06, "loss": 0.1019, "step": 78194 }, { "epoch": 4.72, "learning_rate": 3.5270773919997584e-06, "loss": 0.235, "step": 78196 }, { "epoch": 4.72, "learning_rate": 3.525568481885534e-06, "loss": 0.25, "step": 78198 }, { "epoch": 4.72, "learning_rate": 3.5240595717713097e-06, "loss": 0.3034, "step": 78200 }, { "epoch": 4.72, "learning_rate": 3.5225506616570854e-06, "loss": 0.2438, "step": 78202 }, { "epoch": 4.72, "learning_rate": 3.5210417515428606e-06, "loss": 0.2354, "step": 78204 }, { "epoch": 4.72, "learning_rate": 3.5195328414286363e-06, "loss": 0.2802, "step": 78206 }, { "epoch": 4.72, "learning_rate": 3.518023931314412e-06, "loss": 0.269, "step": 78208 }, { "epoch": 4.72, "learning_rate": 3.5165150212001876e-06, "loss": 0.2692, "step": 78210 }, { "epoch": 4.72, "learning_rate": 3.5150061110859625e-06, "loss": 0.1912, "step": 78212 }, { "epoch": 4.72, "learning_rate": 3.513497200971738e-06, "loss": 0.1293, "step": 78214 }, { "epoch": 4.72, "learning_rate": 3.511988290857514e-06, "loss": 0.273, "step": 78216 }, { "epoch": 4.72, "learning_rate": 3.5104793807432895e-06, "loss": 0.1336, "step": 78218 }, { "epoch": 4.72, "learning_rate": 3.5089704706290647e-06, "loss": 0.1826, "step": 78220 }, { "epoch": 4.72, "learning_rate": 3.5074615605148404e-06, "loss": 0.2701, "step": 78222 }, { "epoch": 4.72, "learning_rate": 3.505952650400616e-06, "loss": 0.2333, "step": 78224 }, { "epoch": 4.72, "learning_rate": 3.5044437402863917e-06, "loss": 0.1783, "step": 78226 }, { "epoch": 4.72, "learning_rate": 3.5029348301721665e-06, "loss": 0.2594, "step": 78228 }, { "epoch": 4.72, "learning_rate": 3.501425920057942e-06, "loss": 0.3617, "step": 78230 }, { "epoch": 4.72, "learning_rate": 3.499917009943718e-06, "loss": 0.1456, "step": 78232 }, { "epoch": 4.72, "learning_rate": 3.4984080998294935e-06, "loss": 0.2128, "step": 78234 }, { "epoch": 4.72, "learning_rate": 3.4968991897152683e-06, "loss": 0.1845, "step": 78236 }, { "epoch": 4.72, "learning_rate": 3.495390279601044e-06, "loss": 0.1282, "step": 78238 }, { "epoch": 4.72, "learning_rate": 3.4938813694868197e-06, "loss": 0.2799, "step": 78240 }, { "epoch": 4.72, "learning_rate": 3.4923724593725953e-06, "loss": 0.347, "step": 78242 }, { "epoch": 4.72, "learning_rate": 3.4908635492583706e-06, "loss": 0.1419, "step": 78244 }, { "epoch": 4.72, "learning_rate": 3.4893546391441463e-06, "loss": 0.2616, "step": 78246 }, { "epoch": 4.72, "learning_rate": 3.487845729029922e-06, "loss": 0.2798, "step": 78248 }, { "epoch": 4.72, "learning_rate": 3.4863368189156976e-06, "loss": 0.2644, "step": 78250 }, { "epoch": 4.72, "learning_rate": 3.4848279088014733e-06, "loss": 0.1913, "step": 78252 }, { "epoch": 4.72, "learning_rate": 3.483318998687248e-06, "loss": 0.2299, "step": 78254 }, { "epoch": 4.72, "learning_rate": 3.4818100885730237e-06, "loss": 0.1828, "step": 78256 }, { "epoch": 4.72, "learning_rate": 3.4803011784587994e-06, "loss": 0.288, "step": 78258 }, { "epoch": 4.72, "learning_rate": 3.478792268344575e-06, "loss": 0.2134, "step": 78260 }, { "epoch": 4.72, "learning_rate": 3.4772833582303503e-06, "loss": 0.195, "step": 78262 }, { "epoch": 4.72, "learning_rate": 3.475774448116126e-06, "loss": 0.2545, "step": 78264 }, { "epoch": 4.72, "learning_rate": 3.4742655380019016e-06, "loss": 0.2662, "step": 78266 }, { "epoch": 4.72, "learning_rate": 3.4727566278876773e-06, "loss": 0.2474, "step": 78268 }, { "epoch": 4.72, "learning_rate": 3.471247717773452e-06, "loss": 0.1935, "step": 78270 }, { "epoch": 4.72, "learning_rate": 3.469738807659228e-06, "loss": 0.2002, "step": 78272 }, { "epoch": 4.72, "learning_rate": 3.4682298975450035e-06, "loss": 0.2642, "step": 78274 }, { "epoch": 4.72, "learning_rate": 3.466720987430779e-06, "loss": 0.196, "step": 78276 }, { "epoch": 4.72, "learning_rate": 3.4652120773165544e-06, "loss": 0.0833, "step": 78278 }, { "epoch": 4.72, "learning_rate": 3.46370316720233e-06, "loss": 0.1817, "step": 78280 }, { "epoch": 4.72, "learning_rate": 3.4621942570881057e-06, "loss": 0.2381, "step": 78282 }, { "epoch": 4.73, "learning_rate": 3.4606853469738814e-06, "loss": 0.1976, "step": 78284 }, { "epoch": 4.73, "learning_rate": 3.459176436859656e-06, "loss": 0.1785, "step": 78286 }, { "epoch": 4.73, "learning_rate": 3.457667526745432e-06, "loss": 0.2432, "step": 78288 }, { "epoch": 4.73, "learning_rate": 3.4561586166312075e-06, "loss": 0.1874, "step": 78290 }, { "epoch": 4.73, "learning_rate": 3.454649706516983e-06, "loss": 0.1846, "step": 78292 }, { "epoch": 4.73, "learning_rate": 3.453140796402758e-06, "loss": 0.224, "step": 78294 }, { "epoch": 4.73, "learning_rate": 3.4516318862885337e-06, "loss": 0.1996, "step": 78296 }, { "epoch": 4.73, "learning_rate": 3.4501229761743094e-06, "loss": 0.2081, "step": 78298 }, { "epoch": 4.73, "learning_rate": 3.448614066060085e-06, "loss": 0.2474, "step": 78300 }, { "epoch": 4.73, "learning_rate": 3.4471051559458603e-06, "loss": 0.2176, "step": 78302 }, { "epoch": 4.73, "learning_rate": 3.445596245831636e-06, "loss": 0.1434, "step": 78304 }, { "epoch": 4.73, "learning_rate": 3.4440873357174116e-06, "loss": 0.1349, "step": 78306 }, { "epoch": 4.73, "learning_rate": 3.4425784256031873e-06, "loss": 0.2184, "step": 78308 }, { "epoch": 4.73, "learning_rate": 3.441069515488962e-06, "loss": 0.0843, "step": 78310 }, { "epoch": 4.73, "learning_rate": 3.4395606053747377e-06, "loss": 0.238, "step": 78312 }, { "epoch": 4.73, "learning_rate": 3.4380516952605134e-06, "loss": 0.2369, "step": 78314 }, { "epoch": 4.73, "learning_rate": 3.436542785146289e-06, "loss": 0.3237, "step": 78316 }, { "epoch": 4.73, "learning_rate": 3.4350338750320643e-06, "loss": 0.1563, "step": 78318 }, { "epoch": 4.73, "learning_rate": 3.43352496491784e-06, "loss": 0.3013, "step": 78320 }, { "epoch": 4.73, "learning_rate": 3.4320160548036157e-06, "loss": 0.1691, "step": 78322 }, { "epoch": 4.73, "learning_rate": 3.4305071446893913e-06, "loss": 0.2514, "step": 78324 }, { "epoch": 4.73, "learning_rate": 3.428998234575167e-06, "loss": 0.1427, "step": 78326 }, { "epoch": 4.73, "learning_rate": 3.427489324460942e-06, "loss": 0.2345, "step": 78328 }, { "epoch": 4.73, "learning_rate": 3.4259804143467175e-06, "loss": 0.2818, "step": 78330 }, { "epoch": 4.73, "learning_rate": 3.424471504232493e-06, "loss": 0.2158, "step": 78332 }, { "epoch": 4.73, "learning_rate": 3.422962594118269e-06, "loss": 0.2642, "step": 78334 }, { "epoch": 4.73, "learning_rate": 3.421453684004044e-06, "loss": 0.2085, "step": 78336 }, { "epoch": 4.73, "learning_rate": 3.4199447738898197e-06, "loss": 0.162, "step": 78338 }, { "epoch": 4.73, "learning_rate": 3.4184358637755954e-06, "loss": 0.2935, "step": 78340 }, { "epoch": 4.73, "learning_rate": 3.416926953661371e-06, "loss": 0.2062, "step": 78342 }, { "epoch": 4.73, "learning_rate": 3.415418043547146e-06, "loss": 0.1328, "step": 78344 }, { "epoch": 4.73, "learning_rate": 3.4139091334329215e-06, "loss": 0.1362, "step": 78346 }, { "epoch": 4.73, "learning_rate": 3.412400223318697e-06, "loss": 0.1695, "step": 78348 }, { "epoch": 4.73, "learning_rate": 3.410891313204473e-06, "loss": 0.2147, "step": 78350 }, { "epoch": 4.73, "learning_rate": 3.4093824030902477e-06, "loss": 0.1869, "step": 78352 }, { "epoch": 4.73, "learning_rate": 3.4078734929760234e-06, "loss": 0.2726, "step": 78354 }, { "epoch": 4.73, "learning_rate": 3.406364582861799e-06, "loss": 0.192, "step": 78356 }, { "epoch": 4.73, "learning_rate": 3.4048556727475747e-06, "loss": 0.195, "step": 78358 }, { "epoch": 4.73, "learning_rate": 3.40334676263335e-06, "loss": 0.2232, "step": 78360 }, { "epoch": 4.73, "learning_rate": 3.4018378525191256e-06, "loss": 0.321, "step": 78362 }, { "epoch": 4.73, "learning_rate": 3.4003289424049013e-06, "loss": 0.2405, "step": 78364 }, { "epoch": 4.73, "learning_rate": 3.398820032290677e-06, "loss": 0.1452, "step": 78366 }, { "epoch": 4.73, "learning_rate": 3.3973111221764518e-06, "loss": 0.2374, "step": 78368 }, { "epoch": 4.73, "learning_rate": 3.3958022120622274e-06, "loss": 0.2533, "step": 78370 }, { "epoch": 4.73, "learning_rate": 3.394293301948003e-06, "loss": 0.2289, "step": 78372 }, { "epoch": 4.73, "learning_rate": 3.3927843918337788e-06, "loss": 0.2965, "step": 78374 }, { "epoch": 4.73, "learning_rate": 3.391275481719554e-06, "loss": 0.2349, "step": 78376 }, { "epoch": 4.73, "learning_rate": 3.3897665716053297e-06, "loss": 0.2256, "step": 78378 }, { "epoch": 4.73, "learning_rate": 3.3882576614911053e-06, "loss": 0.2675, "step": 78380 }, { "epoch": 4.73, "learning_rate": 3.386748751376881e-06, "loss": 0.1864, "step": 78382 }, { "epoch": 4.73, "learning_rate": 3.385239841262656e-06, "loss": 0.322, "step": 78384 }, { "epoch": 4.73, "learning_rate": 3.3837309311484315e-06, "loss": 0.2155, "step": 78386 }, { "epoch": 4.73, "learning_rate": 3.382222021034207e-06, "loss": 0.2442, "step": 78388 }, { "epoch": 4.73, "learning_rate": 3.380713110919983e-06, "loss": 0.241, "step": 78390 }, { "epoch": 4.73, "learning_rate": 3.3792042008057576e-06, "loss": 0.3122, "step": 78392 }, { "epoch": 4.73, "learning_rate": 3.3776952906915333e-06, "loss": 0.2846, "step": 78394 }, { "epoch": 4.73, "learning_rate": 3.3761863805773094e-06, "loss": 0.2453, "step": 78396 }, { "epoch": 4.73, "learning_rate": 3.374677470463085e-06, "loss": 0.2434, "step": 78398 }, { "epoch": 4.73, "learning_rate": 3.37316856034886e-06, "loss": 0.2384, "step": 78400 }, { "epoch": 4.73, "learning_rate": 3.3716596502346355e-06, "loss": 0.1883, "step": 78402 }, { "epoch": 4.73, "learning_rate": 3.370150740120411e-06, "loss": 0.1704, "step": 78404 }, { "epoch": 4.73, "learning_rate": 3.368641830006187e-06, "loss": 0.1997, "step": 78406 }, { "epoch": 4.73, "learning_rate": 3.3671329198919625e-06, "loss": 0.1287, "step": 78408 }, { "epoch": 4.73, "learning_rate": 3.3656240097777374e-06, "loss": 0.2897, "step": 78410 }, { "epoch": 4.73, "learning_rate": 3.364115099663513e-06, "loss": 0.1713, "step": 78412 }, { "epoch": 4.73, "learning_rate": 3.3626061895492887e-06, "loss": 0.2804, "step": 78414 }, { "epoch": 4.73, "learning_rate": 3.3610972794350644e-06, "loss": 0.2069, "step": 78416 }, { "epoch": 4.73, "learning_rate": 3.3595883693208396e-06, "loss": 0.3244, "step": 78418 }, { "epoch": 4.73, "learning_rate": 3.3580794592066153e-06, "loss": 0.2572, "step": 78420 }, { "epoch": 4.73, "learning_rate": 3.356570549092391e-06, "loss": 0.1851, "step": 78422 }, { "epoch": 4.73, "learning_rate": 3.3550616389781666e-06, "loss": 0.2934, "step": 78424 }, { "epoch": 4.73, "learning_rate": 3.3535527288639414e-06, "loss": 0.2778, "step": 78426 }, { "epoch": 4.73, "learning_rate": 3.352043818749717e-06, "loss": 0.1903, "step": 78428 }, { "epoch": 4.73, "learning_rate": 3.3505349086354928e-06, "loss": 0.163, "step": 78430 }, { "epoch": 4.73, "learning_rate": 3.3490259985212684e-06, "loss": 0.277, "step": 78432 }, { "epoch": 4.73, "learning_rate": 3.3475170884070437e-06, "loss": 0.2091, "step": 78434 }, { "epoch": 4.73, "learning_rate": 3.3460081782928193e-06, "loss": 0.2395, "step": 78436 }, { "epoch": 4.73, "learning_rate": 3.344499268178595e-06, "loss": 0.3729, "step": 78438 }, { "epoch": 4.73, "learning_rate": 3.3429903580643707e-06, "loss": 0.2822, "step": 78440 }, { "epoch": 4.73, "learning_rate": 3.3414814479501455e-06, "loss": 0.192, "step": 78442 }, { "epoch": 4.73, "learning_rate": 3.339972537835921e-06, "loss": 0.261, "step": 78444 }, { "epoch": 4.73, "learning_rate": 3.338463627721697e-06, "loss": 0.1836, "step": 78446 }, { "epoch": 4.73, "learning_rate": 3.3369547176074725e-06, "loss": 0.2516, "step": 78448 }, { "epoch": 4.74, "learning_rate": 3.3354458074932473e-06, "loss": 0.1802, "step": 78450 }, { "epoch": 4.74, "learning_rate": 3.333936897379023e-06, "loss": 0.1516, "step": 78452 }, { "epoch": 4.74, "learning_rate": 3.332427987264799e-06, "loss": 0.0994, "step": 78454 }, { "epoch": 4.74, "learning_rate": 3.3309190771505747e-06, "loss": 0.2626, "step": 78456 }, { "epoch": 4.74, "learning_rate": 3.3294101670363495e-06, "loss": 0.2536, "step": 78458 }, { "epoch": 4.74, "learning_rate": 3.3279012569221252e-06, "loss": 0.2724, "step": 78460 }, { "epoch": 4.74, "learning_rate": 3.326392346807901e-06, "loss": 0.1486, "step": 78462 }, { "epoch": 4.74, "learning_rate": 3.3248834366936765e-06, "loss": 0.1982, "step": 78464 }, { "epoch": 4.74, "learning_rate": 3.3233745265794514e-06, "loss": 0.1575, "step": 78466 }, { "epoch": 4.74, "learning_rate": 3.321865616465227e-06, "loss": 0.139, "step": 78468 }, { "epoch": 4.74, "learning_rate": 3.3203567063510027e-06, "loss": 0.2032, "step": 78470 }, { "epoch": 4.74, "learning_rate": 3.3188477962367784e-06, "loss": 0.2516, "step": 78472 }, { "epoch": 4.74, "learning_rate": 3.3173388861225536e-06, "loss": 0.2039, "step": 78474 }, { "epoch": 4.74, "learning_rate": 3.3158299760083293e-06, "loss": 0.2552, "step": 78476 }, { "epoch": 4.74, "learning_rate": 3.314321065894105e-06, "loss": 0.2612, "step": 78478 }, { "epoch": 4.74, "learning_rate": 3.3128121557798806e-06, "loss": 0.1671, "step": 78480 }, { "epoch": 4.74, "learning_rate": 3.3113032456656563e-06, "loss": 0.1246, "step": 78482 }, { "epoch": 4.74, "learning_rate": 3.309794335551431e-06, "loss": 0.1782, "step": 78484 }, { "epoch": 4.74, "learning_rate": 3.3082854254372068e-06, "loss": 0.1925, "step": 78486 }, { "epoch": 4.74, "learning_rate": 3.3067765153229824e-06, "loss": 0.1829, "step": 78488 }, { "epoch": 4.74, "learning_rate": 3.305267605208758e-06, "loss": 0.1318, "step": 78490 }, { "epoch": 4.74, "learning_rate": 3.3037586950945333e-06, "loss": 0.2042, "step": 78492 }, { "epoch": 4.74, "learning_rate": 3.302249784980309e-06, "loss": 0.2009, "step": 78494 }, { "epoch": 4.74, "learning_rate": 3.3007408748660847e-06, "loss": 0.2339, "step": 78496 }, { "epoch": 4.74, "learning_rate": 3.2992319647518603e-06, "loss": 0.1524, "step": 78498 }, { "epoch": 4.74, "learning_rate": 3.297723054637635e-06, "loss": 0.2432, "step": 78500 }, { "epoch": 4.74, "learning_rate": 3.296214144523411e-06, "loss": 0.2982, "step": 78502 }, { "epoch": 4.74, "learning_rate": 3.2947052344091865e-06, "loss": 0.2223, "step": 78504 }, { "epoch": 4.74, "learning_rate": 3.293196324294962e-06, "loss": 0.1839, "step": 78506 }, { "epoch": 4.74, "learning_rate": 3.291687414180737e-06, "loss": 0.2084, "step": 78508 }, { "epoch": 4.74, "learning_rate": 3.2901785040665126e-06, "loss": 0.1686, "step": 78510 }, { "epoch": 4.74, "learning_rate": 3.2886695939522883e-06, "loss": 0.239, "step": 78512 }, { "epoch": 4.74, "learning_rate": 3.2871606838380644e-06, "loss": 0.1179, "step": 78514 }, { "epoch": 4.74, "learning_rate": 3.2856517737238392e-06, "loss": 0.1757, "step": 78516 }, { "epoch": 4.74, "learning_rate": 3.284142863609615e-06, "loss": 0.3103, "step": 78518 }, { "epoch": 4.74, "learning_rate": 3.2826339534953906e-06, "loss": 0.1657, "step": 78520 }, { "epoch": 4.74, "learning_rate": 3.2811250433811662e-06, "loss": 0.205, "step": 78522 }, { "epoch": 4.74, "learning_rate": 3.279616133266941e-06, "loss": 0.1319, "step": 78524 }, { "epoch": 4.74, "learning_rate": 3.2781072231527167e-06, "loss": 0.185, "step": 78526 }, { "epoch": 4.74, "learning_rate": 3.2765983130384924e-06, "loss": 0.2378, "step": 78528 }, { "epoch": 4.74, "learning_rate": 3.275089402924268e-06, "loss": 0.1754, "step": 78530 }, { "epoch": 4.74, "learning_rate": 3.2735804928100433e-06, "loss": 0.1575, "step": 78532 }, { "epoch": 4.74, "learning_rate": 3.272071582695819e-06, "loss": 0.1927, "step": 78534 }, { "epoch": 4.74, "learning_rate": 3.2705626725815946e-06, "loss": 0.2424, "step": 78536 }, { "epoch": 4.74, "learning_rate": 3.2690537624673703e-06, "loss": 0.1758, "step": 78538 }, { "epoch": 4.74, "learning_rate": 3.267544852353145e-06, "loss": 0.1466, "step": 78540 }, { "epoch": 4.74, "learning_rate": 3.2660359422389208e-06, "loss": 0.1944, "step": 78542 }, { "epoch": 4.74, "learning_rate": 3.2645270321246964e-06, "loss": 0.2596, "step": 78544 }, { "epoch": 4.74, "learning_rate": 3.263018122010472e-06, "loss": 0.2537, "step": 78546 }, { "epoch": 4.74, "learning_rate": 3.2615092118962473e-06, "loss": 0.2038, "step": 78548 }, { "epoch": 4.74, "learning_rate": 3.260000301782023e-06, "loss": 0.1631, "step": 78550 }, { "epoch": 4.74, "learning_rate": 3.2584913916677987e-06, "loss": 0.2607, "step": 78552 }, { "epoch": 4.74, "learning_rate": 3.2569824815535743e-06, "loss": 0.3286, "step": 78554 }, { "epoch": 4.74, "learning_rate": 3.255473571439349e-06, "loss": 0.2824, "step": 78556 }, { "epoch": 4.74, "learning_rate": 3.253964661325125e-06, "loss": 0.2328, "step": 78558 }, { "epoch": 4.74, "learning_rate": 3.2524557512109005e-06, "loss": 0.263, "step": 78560 }, { "epoch": 4.74, "learning_rate": 3.250946841096676e-06, "loss": 0.2457, "step": 78562 }, { "epoch": 4.74, "learning_rate": 3.249437930982452e-06, "loss": 0.2192, "step": 78564 }, { "epoch": 4.74, "learning_rate": 3.2479290208682267e-06, "loss": 0.2147, "step": 78566 }, { "epoch": 4.74, "learning_rate": 3.2464201107540023e-06, "loss": 0.3217, "step": 78568 }, { "epoch": 4.74, "learning_rate": 3.244911200639778e-06, "loss": 0.1624, "step": 78570 }, { "epoch": 4.74, "learning_rate": 3.2434022905255537e-06, "loss": 0.1506, "step": 78572 }, { "epoch": 4.74, "learning_rate": 3.241893380411329e-06, "loss": 0.1801, "step": 78574 }, { "epoch": 4.74, "learning_rate": 3.2403844702971046e-06, "loss": 0.1698, "step": 78576 }, { "epoch": 4.74, "learning_rate": 3.2388755601828802e-06, "loss": 0.2577, "step": 78578 }, { "epoch": 4.74, "learning_rate": 3.237366650068656e-06, "loss": 0.2856, "step": 78580 }, { "epoch": 4.74, "learning_rate": 3.2358577399544307e-06, "loss": 0.2243, "step": 78582 }, { "epoch": 4.74, "learning_rate": 3.2343488298402064e-06, "loss": 0.1839, "step": 78584 }, { "epoch": 4.74, "learning_rate": 3.232839919725982e-06, "loss": 0.1927, "step": 78586 }, { "epoch": 4.74, "learning_rate": 3.2313310096117577e-06, "loss": 0.1712, "step": 78588 }, { "epoch": 4.74, "learning_rate": 3.229822099497533e-06, "loss": 0.172, "step": 78590 }, { "epoch": 4.74, "learning_rate": 3.2283131893833086e-06, "loss": 0.1172, "step": 78592 }, { "epoch": 4.74, "learning_rate": 3.2268042792690843e-06, "loss": 0.1868, "step": 78594 }, { "epoch": 4.74, "learning_rate": 3.22529536915486e-06, "loss": 0.189, "step": 78596 }, { "epoch": 4.74, "learning_rate": 3.2237864590406348e-06, "loss": 0.2014, "step": 78598 }, { "epoch": 4.74, "learning_rate": 3.2222775489264104e-06, "loss": 0.2008, "step": 78600 }, { "epoch": 4.74, "learning_rate": 3.220768638812186e-06, "loss": 0.2075, "step": 78602 }, { "epoch": 4.74, "learning_rate": 3.2192597286979618e-06, "loss": 0.1708, "step": 78604 }, { "epoch": 4.74, "learning_rate": 3.217750818583737e-06, "loss": 0.2041, "step": 78606 }, { "epoch": 4.74, "learning_rate": 3.2162419084695127e-06, "loss": 0.2216, "step": 78608 }, { "epoch": 4.74, "learning_rate": 3.2147329983552884e-06, "loss": 0.1425, "step": 78610 }, { "epoch": 4.74, "learning_rate": 3.213224088241064e-06, "loss": 0.1863, "step": 78612 }, { "epoch": 4.74, "learning_rate": 3.211715178126839e-06, "loss": 0.2747, "step": 78614 }, { "epoch": 4.75, "learning_rate": 3.2102062680126145e-06, "loss": 0.1501, "step": 78616 }, { "epoch": 4.75, "learning_rate": 3.20869735789839e-06, "loss": 0.1794, "step": 78618 }, { "epoch": 4.75, "learning_rate": 3.207188447784166e-06, "loss": 0.2796, "step": 78620 }, { "epoch": 4.75, "learning_rate": 3.2056795376699407e-06, "loss": 0.2615, "step": 78622 }, { "epoch": 4.75, "learning_rate": 3.2041706275557163e-06, "loss": 0.1549, "step": 78624 }, { "epoch": 4.75, "learning_rate": 3.202661717441492e-06, "loss": 0.1795, "step": 78626 }, { "epoch": 4.75, "learning_rate": 3.2011528073272677e-06, "loss": 0.2042, "step": 78628 }, { "epoch": 4.75, "learning_rate": 3.199643897213043e-06, "loss": 0.2106, "step": 78630 }, { "epoch": 4.75, "learning_rate": 3.1981349870988186e-06, "loss": 0.2436, "step": 78632 }, { "epoch": 4.75, "learning_rate": 3.1966260769845942e-06, "loss": 0.1725, "step": 78634 }, { "epoch": 4.75, "learning_rate": 3.19511716687037e-06, "loss": 0.2785, "step": 78636 }, { "epoch": 4.75, "learning_rate": 3.1936082567561456e-06, "loss": 0.2307, "step": 78638 }, { "epoch": 4.75, "learning_rate": 3.1920993466419204e-06, "loss": 0.2679, "step": 78640 }, { "epoch": 4.75, "learning_rate": 3.190590436527696e-06, "loss": 0.2142, "step": 78642 }, { "epoch": 4.75, "learning_rate": 3.1890815264134717e-06, "loss": 0.2831, "step": 78644 }, { "epoch": 4.75, "learning_rate": 3.1875726162992474e-06, "loss": 0.1647, "step": 78646 }, { "epoch": 4.75, "learning_rate": 3.1860637061850226e-06, "loss": 0.301, "step": 78648 }, { "epoch": 4.75, "learning_rate": 3.1845547960707983e-06, "loss": 0.282, "step": 78650 }, { "epoch": 4.75, "learning_rate": 3.183045885956574e-06, "loss": 0.2566, "step": 78652 }, { "epoch": 4.75, "learning_rate": 3.1815369758423496e-06, "loss": 0.1977, "step": 78654 }, { "epoch": 4.75, "learning_rate": 3.1800280657281244e-06, "loss": 0.2048, "step": 78656 }, { "epoch": 4.75, "learning_rate": 3.1785191556139e-06, "loss": 0.3379, "step": 78658 }, { "epoch": 4.75, "learning_rate": 3.1770102454996758e-06, "loss": 0.2314, "step": 78660 }, { "epoch": 4.75, "learning_rate": 3.1755013353854514e-06, "loss": 0.2446, "step": 78662 }, { "epoch": 4.75, "learning_rate": 3.1739924252712267e-06, "loss": 0.2209, "step": 78664 }, { "epoch": 4.75, "learning_rate": 3.1724835151570024e-06, "loss": 0.2981, "step": 78666 }, { "epoch": 4.75, "learning_rate": 3.170974605042778e-06, "loss": 0.1715, "step": 78668 }, { "epoch": 4.75, "learning_rate": 3.1694656949285537e-06, "loss": 0.1409, "step": 78670 }, { "epoch": 4.75, "learning_rate": 3.1679567848143285e-06, "loss": 0.3001, "step": 78672 }, { "epoch": 4.75, "learning_rate": 3.166447874700104e-06, "loss": 0.2098, "step": 78674 }, { "epoch": 4.75, "learning_rate": 3.16493896458588e-06, "loss": 0.2657, "step": 78676 }, { "epoch": 4.75, "learning_rate": 3.1634300544716555e-06, "loss": 0.1985, "step": 78678 }, { "epoch": 4.75, "learning_rate": 3.1619211443574303e-06, "loss": 0.2708, "step": 78680 }, { "epoch": 4.75, "learning_rate": 3.160412234243206e-06, "loss": 0.2666, "step": 78682 }, { "epoch": 4.75, "learning_rate": 3.1589033241289817e-06, "loss": 0.2708, "step": 78684 }, { "epoch": 4.75, "learning_rate": 3.1573944140147573e-06, "loss": 0.2858, "step": 78686 }, { "epoch": 4.75, "learning_rate": 3.1558855039005326e-06, "loss": 0.2332, "step": 78688 }, { "epoch": 4.75, "learning_rate": 3.1543765937863082e-06, "loss": 0.2174, "step": 78690 }, { "epoch": 4.75, "learning_rate": 3.152867683672084e-06, "loss": 0.1879, "step": 78692 }, { "epoch": 4.75, "learning_rate": 3.1513587735578596e-06, "loss": 0.1667, "step": 78694 }, { "epoch": 4.75, "learning_rate": 3.1498498634436344e-06, "loss": 0.2588, "step": 78696 }, { "epoch": 4.75, "learning_rate": 3.14834095332941e-06, "loss": 0.2092, "step": 78698 }, { "epoch": 4.75, "learning_rate": 3.1468320432151857e-06, "loss": 0.2787, "step": 78700 }, { "epoch": 4.75, "learning_rate": 3.1453231331009614e-06, "loss": 0.1726, "step": 78702 }, { "epoch": 4.75, "learning_rate": 3.1438142229867366e-06, "loss": 0.1961, "step": 78704 }, { "epoch": 4.75, "learning_rate": 3.1423053128725123e-06, "loss": 0.1349, "step": 78706 }, { "epoch": 4.75, "learning_rate": 3.140796402758288e-06, "loss": 0.142, "step": 78708 }, { "epoch": 4.75, "learning_rate": 3.1392874926440636e-06, "loss": 0.318, "step": 78710 }, { "epoch": 4.75, "learning_rate": 3.1377785825298393e-06, "loss": 0.198, "step": 78712 }, { "epoch": 4.75, "learning_rate": 3.136269672415614e-06, "loss": 0.3111, "step": 78714 }, { "epoch": 4.75, "learning_rate": 3.1347607623013898e-06, "loss": 0.204, "step": 78716 }, { "epoch": 4.75, "learning_rate": 3.1332518521871655e-06, "loss": 0.2815, "step": 78718 }, { "epoch": 4.75, "learning_rate": 3.131742942072941e-06, "loss": 0.2265, "step": 78720 }, { "epoch": 4.75, "learning_rate": 3.1302340319587164e-06, "loss": 0.2735, "step": 78722 }, { "epoch": 4.75, "learning_rate": 3.128725121844492e-06, "loss": 0.2921, "step": 78724 }, { "epoch": 4.75, "learning_rate": 3.1272162117302677e-06, "loss": 0.2878, "step": 78726 }, { "epoch": 4.75, "learning_rate": 3.1257073016160434e-06, "loss": 0.2394, "step": 78728 }, { "epoch": 4.75, "learning_rate": 3.1241983915018186e-06, "loss": 0.2316, "step": 78730 }, { "epoch": 4.75, "learning_rate": 3.122689481387594e-06, "loss": 0.3054, "step": 78732 }, { "epoch": 4.75, "learning_rate": 3.1211805712733695e-06, "loss": 0.1868, "step": 78734 }, { "epoch": 4.75, "learning_rate": 3.1196716611591448e-06, "loss": 0.2907, "step": 78736 }, { "epoch": 4.75, "learning_rate": 3.1181627510449204e-06, "loss": 0.2525, "step": 78738 }, { "epoch": 4.75, "learning_rate": 3.1166538409306957e-06, "loss": 0.2303, "step": 78740 }, { "epoch": 4.75, "learning_rate": 3.1151449308164713e-06, "loss": 0.1953, "step": 78742 }, { "epoch": 4.75, "learning_rate": 3.113636020702247e-06, "loss": 0.2291, "step": 78744 }, { "epoch": 4.75, "learning_rate": 3.1121271105880227e-06, "loss": 0.1819, "step": 78746 }, { "epoch": 4.75, "learning_rate": 3.110618200473798e-06, "loss": 0.1863, "step": 78748 }, { "epoch": 4.75, "learning_rate": 3.1091092903595736e-06, "loss": 0.1528, "step": 78750 }, { "epoch": 4.75, "learning_rate": 3.107600380245349e-06, "loss": 0.1684, "step": 78752 }, { "epoch": 4.75, "learning_rate": 3.1060914701311245e-06, "loss": 0.2975, "step": 78754 }, { "epoch": 4.75, "learning_rate": 3.1045825600168997e-06, "loss": 0.2425, "step": 78756 }, { "epoch": 4.75, "learning_rate": 3.1030736499026754e-06, "loss": 0.1807, "step": 78758 }, { "epoch": 4.75, "learning_rate": 3.1015647397884506e-06, "loss": 0.2871, "step": 78760 }, { "epoch": 4.75, "learning_rate": 3.1000558296742263e-06, "loss": 0.1891, "step": 78762 }, { "epoch": 4.75, "learning_rate": 3.098546919560002e-06, "loss": 0.1854, "step": 78764 }, { "epoch": 4.75, "learning_rate": 3.0970380094457776e-06, "loss": 0.2067, "step": 78766 }, { "epoch": 4.75, "learning_rate": 3.095529099331553e-06, "loss": 0.1997, "step": 78768 }, { "epoch": 4.75, "learning_rate": 3.0940201892173285e-06, "loss": 0.1905, "step": 78770 }, { "epoch": 4.75, "learning_rate": 3.092511279103104e-06, "loss": 0.2664, "step": 78772 }, { "epoch": 4.75, "learning_rate": 3.0910023689888795e-06, "loss": 0.2578, "step": 78774 }, { "epoch": 4.75, "learning_rate": 3.0894934588746547e-06, "loss": 0.2223, "step": 78776 }, { "epoch": 4.75, "learning_rate": 3.0879845487604304e-06, "loss": 0.1595, "step": 78778 }, { "epoch": 4.75, "learning_rate": 3.086475638646206e-06, "loss": 0.1581, "step": 78780 }, { "epoch": 4.76, "learning_rate": 3.0849667285319817e-06, "loss": 0.1468, "step": 78782 }, { "epoch": 4.76, "learning_rate": 3.083457818417757e-06, "loss": 0.3203, "step": 78784 }, { "epoch": 4.76, "learning_rate": 3.0819489083035326e-06, "loss": 0.2811, "step": 78786 }, { "epoch": 4.76, "learning_rate": 3.080439998189308e-06, "loss": 0.2754, "step": 78788 }, { "epoch": 4.76, "learning_rate": 3.0789310880750835e-06, "loss": 0.1761, "step": 78790 }, { "epoch": 4.76, "learning_rate": 3.077422177960859e-06, "loss": 0.1971, "step": 78792 }, { "epoch": 4.76, "learning_rate": 3.0759132678466344e-06, "loss": 0.2216, "step": 78794 }, { "epoch": 4.76, "learning_rate": 3.07440435773241e-06, "loss": 0.1666, "step": 78796 }, { "epoch": 4.76, "learning_rate": 3.0728954476181853e-06, "loss": 0.2047, "step": 78798 }, { "epoch": 4.76, "learning_rate": 3.071386537503961e-06, "loss": 0.2264, "step": 78800 }, { "epoch": 4.76, "learning_rate": 3.0698776273897367e-06, "loss": 0.2737, "step": 78802 }, { "epoch": 4.76, "learning_rate": 3.0683687172755123e-06, "loss": 0.1674, "step": 78804 }, { "epoch": 4.76, "learning_rate": 3.0668598071612876e-06, "loss": 0.1321, "step": 78806 }, { "epoch": 4.76, "learning_rate": 3.0653508970470632e-06, "loss": 0.2495, "step": 78808 }, { "epoch": 4.76, "learning_rate": 3.0638419869328385e-06, "loss": 0.4153, "step": 78810 }, { "epoch": 4.76, "learning_rate": 3.062333076818614e-06, "loss": 0.2117, "step": 78812 }, { "epoch": 4.76, "learning_rate": 3.0608241667043894e-06, "loss": 0.1796, "step": 78814 }, { "epoch": 4.76, "learning_rate": 3.059315256590165e-06, "loss": 0.1853, "step": 78816 }, { "epoch": 4.76, "learning_rate": 3.0578063464759403e-06, "loss": 0.237, "step": 78818 }, { "epoch": 4.76, "learning_rate": 3.056297436361716e-06, "loss": 0.2285, "step": 78820 }, { "epoch": 4.76, "learning_rate": 3.0547885262474916e-06, "loss": 0.2196, "step": 78822 }, { "epoch": 4.76, "learning_rate": 3.0532796161332673e-06, "loss": 0.2186, "step": 78824 }, { "epoch": 4.76, "learning_rate": 3.0517707060190426e-06, "loss": 0.1922, "step": 78826 }, { "epoch": 4.76, "learning_rate": 3.0502617959048182e-06, "loss": 0.2223, "step": 78828 }, { "epoch": 4.76, "learning_rate": 3.0487528857905935e-06, "loss": 0.1507, "step": 78830 }, { "epoch": 4.76, "learning_rate": 3.047243975676369e-06, "loss": 0.1953, "step": 78832 }, { "epoch": 4.76, "learning_rate": 3.0457350655621444e-06, "loss": 0.2314, "step": 78834 }, { "epoch": 4.76, "learning_rate": 3.04422615544792e-06, "loss": 0.1396, "step": 78836 }, { "epoch": 4.76, "learning_rate": 3.0427172453336957e-06, "loss": 0.2586, "step": 78838 }, { "epoch": 4.76, "learning_rate": 3.0412083352194714e-06, "loss": 0.3262, "step": 78840 }, { "epoch": 4.76, "learning_rate": 3.0396994251052466e-06, "loss": 0.3106, "step": 78842 }, { "epoch": 4.76, "learning_rate": 3.0381905149910223e-06, "loss": 0.2954, "step": 78844 }, { "epoch": 4.76, "learning_rate": 3.0366816048767975e-06, "loss": 0.3023, "step": 78846 }, { "epoch": 4.76, "learning_rate": 3.035172694762573e-06, "loss": 0.2502, "step": 78848 }, { "epoch": 4.76, "learning_rate": 3.0336637846483484e-06, "loss": 0.2334, "step": 78850 }, { "epoch": 4.76, "learning_rate": 3.032154874534124e-06, "loss": 0.3366, "step": 78852 }, { "epoch": 4.76, "learning_rate": 3.0306459644198993e-06, "loss": 0.2694, "step": 78854 }, { "epoch": 4.76, "learning_rate": 3.029137054305675e-06, "loss": 0.1973, "step": 78856 }, { "epoch": 4.76, "learning_rate": 3.0276281441914507e-06, "loss": 0.209, "step": 78858 }, { "epoch": 4.76, "learning_rate": 3.0261192340772263e-06, "loss": 0.1951, "step": 78860 }, { "epoch": 4.76, "learning_rate": 3.0246103239630016e-06, "loss": 0.2027, "step": 78862 }, { "epoch": 4.76, "learning_rate": 3.0231014138487773e-06, "loss": 0.1876, "step": 78864 }, { "epoch": 4.76, "learning_rate": 3.0215925037345525e-06, "loss": 0.2233, "step": 78866 }, { "epoch": 4.76, "learning_rate": 3.020083593620328e-06, "loss": 0.2582, "step": 78868 }, { "epoch": 4.76, "learning_rate": 3.018574683506104e-06, "loss": 0.1406, "step": 78870 }, { "epoch": 4.76, "learning_rate": 3.017065773391879e-06, "loss": 0.2026, "step": 78872 }, { "epoch": 4.76, "learning_rate": 3.0155568632776547e-06, "loss": 0.3224, "step": 78874 }, { "epoch": 4.76, "learning_rate": 3.01404795316343e-06, "loss": 0.1716, "step": 78876 }, { "epoch": 4.76, "learning_rate": 3.0125390430492057e-06, "loss": 0.202, "step": 78878 }, { "epoch": 4.76, "learning_rate": 3.0110301329349813e-06, "loss": 0.3168, "step": 78880 }, { "epoch": 4.76, "learning_rate": 3.009521222820757e-06, "loss": 0.1388, "step": 78882 }, { "epoch": 4.76, "learning_rate": 3.0080123127065322e-06, "loss": 0.2439, "step": 78884 }, { "epoch": 4.76, "learning_rate": 3.006503402592308e-06, "loss": 0.1777, "step": 78886 }, { "epoch": 4.76, "learning_rate": 3.004994492478083e-06, "loss": 0.3372, "step": 78888 }, { "epoch": 4.76, "learning_rate": 3.003485582363859e-06, "loss": 0.3122, "step": 78890 }, { "epoch": 4.76, "learning_rate": 3.001976672249634e-06, "loss": 0.2102, "step": 78892 }, { "epoch": 4.76, "learning_rate": 3.0004677621354097e-06, "loss": 0.2365, "step": 78894 }, { "epoch": 4.76, "learning_rate": 2.9989588520211854e-06, "loss": 0.1962, "step": 78896 }, { "epoch": 4.76, "learning_rate": 2.997449941906961e-06, "loss": 0.2332, "step": 78898 }, { "epoch": 4.76, "learning_rate": 2.9959410317927363e-06, "loss": 0.1403, "step": 78900 }, { "epoch": 4.76, "learning_rate": 2.994432121678512e-06, "loss": 0.1465, "step": 78902 }, { "epoch": 4.76, "learning_rate": 2.992923211564287e-06, "loss": 0.2834, "step": 78904 }, { "epoch": 4.76, "learning_rate": 2.991414301450063e-06, "loss": 0.2824, "step": 78906 }, { "epoch": 4.76, "learning_rate": 2.989905391335838e-06, "loss": 0.1806, "step": 78908 }, { "epoch": 4.76, "learning_rate": 2.9883964812216138e-06, "loss": 0.2974, "step": 78910 }, { "epoch": 4.76, "learning_rate": 2.986887571107389e-06, "loss": 0.1885, "step": 78912 }, { "epoch": 4.76, "learning_rate": 2.9853786609931647e-06, "loss": 0.2886, "step": 78914 }, { "epoch": 4.76, "learning_rate": 2.9838697508789404e-06, "loss": 0.2102, "step": 78916 }, { "epoch": 4.76, "learning_rate": 2.982360840764716e-06, "loss": 0.2477, "step": 78918 }, { "epoch": 4.76, "learning_rate": 2.9808519306504913e-06, "loss": 0.3309, "step": 78920 }, { "epoch": 4.76, "learning_rate": 2.979343020536267e-06, "loss": 0.2639, "step": 78922 }, { "epoch": 4.76, "learning_rate": 2.977834110422042e-06, "loss": 0.201, "step": 78924 }, { "epoch": 4.76, "learning_rate": 2.976325200307818e-06, "loss": 0.1515, "step": 78926 }, { "epoch": 4.76, "learning_rate": 2.974816290193593e-06, "loss": 0.2341, "step": 78928 }, { "epoch": 4.76, "learning_rate": 2.9733073800793687e-06, "loss": 0.1254, "step": 78930 }, { "epoch": 4.76, "learning_rate": 2.971798469965144e-06, "loss": 0.223, "step": 78932 }, { "epoch": 4.76, "learning_rate": 2.9702895598509197e-06, "loss": 0.1629, "step": 78934 }, { "epoch": 4.76, "learning_rate": 2.9687806497366953e-06, "loss": 0.2722, "step": 78936 }, { "epoch": 4.76, "learning_rate": 2.967271739622471e-06, "loss": 0.26, "step": 78938 }, { "epoch": 4.76, "learning_rate": 2.9657628295082462e-06, "loss": 0.1229, "step": 78940 }, { "epoch": 4.76, "learning_rate": 2.964253919394022e-06, "loss": 0.2739, "step": 78942 }, { "epoch": 4.76, "learning_rate": 2.962745009279797e-06, "loss": 0.2409, "step": 78944 }, { "epoch": 4.76, "learning_rate": 2.961236099165573e-06, "loss": 0.2061, "step": 78946 }, { "epoch": 4.77, "learning_rate": 2.9597271890513485e-06, "loss": 0.183, "step": 78948 }, { "epoch": 4.77, "learning_rate": 2.9582182789371237e-06, "loss": 0.2377, "step": 78950 }, { "epoch": 4.77, "learning_rate": 2.9567093688228994e-06, "loss": 0.152, "step": 78952 }, { "epoch": 4.77, "learning_rate": 2.955200458708675e-06, "loss": 0.2676, "step": 78954 }, { "epoch": 4.77, "learning_rate": 2.9536915485944507e-06, "loss": 0.3218, "step": 78956 }, { "epoch": 4.77, "learning_rate": 2.952182638480226e-06, "loss": 0.2481, "step": 78958 }, { "epoch": 4.77, "learning_rate": 2.9506737283660016e-06, "loss": 0.2252, "step": 78960 }, { "epoch": 4.77, "learning_rate": 2.949164818251777e-06, "loss": 0.2192, "step": 78962 }, { "epoch": 4.77, "learning_rate": 2.9476559081375525e-06, "loss": 0.1483, "step": 78964 }, { "epoch": 4.77, "learning_rate": 2.9461469980233278e-06, "loss": 0.1497, "step": 78966 }, { "epoch": 4.77, "learning_rate": 2.9446380879091034e-06, "loss": 0.285, "step": 78968 }, { "epoch": 4.77, "learning_rate": 2.9431291777948787e-06, "loss": 0.2129, "step": 78970 }, { "epoch": 4.77, "learning_rate": 2.9416202676806544e-06, "loss": 0.2159, "step": 78972 }, { "epoch": 4.77, "learning_rate": 2.94011135756643e-06, "loss": 0.2683, "step": 78974 }, { "epoch": 4.77, "learning_rate": 2.9386024474522057e-06, "loss": 0.3753, "step": 78976 }, { "epoch": 4.77, "learning_rate": 2.937093537337981e-06, "loss": 0.2496, "step": 78978 }, { "epoch": 4.77, "learning_rate": 2.9355846272237566e-06, "loss": 0.2739, "step": 78980 }, { "epoch": 4.77, "learning_rate": 2.934075717109532e-06, "loss": 0.2297, "step": 78982 }, { "epoch": 4.77, "learning_rate": 2.9325668069953075e-06, "loss": 0.1773, "step": 78984 }, { "epoch": 4.77, "learning_rate": 2.9310578968810828e-06, "loss": 0.2762, "step": 78986 }, { "epoch": 4.77, "learning_rate": 2.9295489867668584e-06, "loss": 0.127, "step": 78988 }, { "epoch": 4.77, "learning_rate": 2.9280400766526337e-06, "loss": 0.2504, "step": 78990 }, { "epoch": 4.77, "learning_rate": 2.9265311665384093e-06, "loss": 0.1606, "step": 78992 }, { "epoch": 4.77, "learning_rate": 2.925022256424185e-06, "loss": 0.2133, "step": 78994 }, { "epoch": 4.77, "learning_rate": 2.9235133463099607e-06, "loss": 0.1593, "step": 78996 }, { "epoch": 4.77, "learning_rate": 2.922004436195736e-06, "loss": 0.2838, "step": 78998 }, { "epoch": 4.77, "learning_rate": 2.9204955260815116e-06, "loss": 0.2431, "step": 79000 }, { "epoch": 4.77, "learning_rate": 2.918986615967287e-06, "loss": 0.2061, "step": 79002 }, { "epoch": 4.77, "learning_rate": 2.9174777058530625e-06, "loss": 0.1973, "step": 79004 }, { "epoch": 4.77, "learning_rate": 2.9159687957388377e-06, "loss": 0.2099, "step": 79006 }, { "epoch": 4.77, "learning_rate": 2.9144598856246134e-06, "loss": 0.2415, "step": 79008 }, { "epoch": 4.77, "learning_rate": 2.9129509755103886e-06, "loss": 0.1861, "step": 79010 }, { "epoch": 4.77, "learning_rate": 2.9114420653961647e-06, "loss": 0.1525, "step": 79012 }, { "epoch": 4.77, "learning_rate": 2.90993315528194e-06, "loss": 0.2543, "step": 79014 }, { "epoch": 4.77, "learning_rate": 2.9084242451677156e-06, "loss": 0.1885, "step": 79016 }, { "epoch": 4.77, "learning_rate": 2.906915335053491e-06, "loss": 0.1038, "step": 79018 }, { "epoch": 4.77, "learning_rate": 2.9054064249392665e-06, "loss": 0.2177, "step": 79020 }, { "epoch": 4.77, "learning_rate": 2.9038975148250418e-06, "loss": 0.2656, "step": 79022 }, { "epoch": 4.77, "learning_rate": 2.9023886047108175e-06, "loss": 0.2358, "step": 79024 }, { "epoch": 4.77, "learning_rate": 2.900879694596593e-06, "loss": 0.1305, "step": 79026 }, { "epoch": 4.77, "learning_rate": 2.8993707844823684e-06, "loss": 0.1938, "step": 79028 }, { "epoch": 4.77, "learning_rate": 2.897861874368144e-06, "loss": 0.1704, "step": 79030 }, { "epoch": 4.77, "learning_rate": 2.8963529642539197e-06, "loss": 0.1333, "step": 79032 }, { "epoch": 4.77, "learning_rate": 2.8948440541396954e-06, "loss": 0.2713, "step": 79034 }, { "epoch": 4.77, "learning_rate": 2.8933351440254706e-06, "loss": 0.2844, "step": 79036 }, { "epoch": 4.77, "learning_rate": 2.8918262339112463e-06, "loss": 0.1928, "step": 79038 }, { "epoch": 4.77, "learning_rate": 2.8903173237970215e-06, "loss": 0.2161, "step": 79040 }, { "epoch": 4.77, "learning_rate": 2.888808413682797e-06, "loss": 0.2777, "step": 79042 }, { "epoch": 4.77, "learning_rate": 2.8872995035685724e-06, "loss": 0.2356, "step": 79044 }, { "epoch": 4.77, "learning_rate": 2.885790593454348e-06, "loss": 0.2049, "step": 79046 }, { "epoch": 4.77, "learning_rate": 2.8842816833401233e-06, "loss": 0.2702, "step": 79048 }, { "epoch": 4.77, "learning_rate": 2.882772773225899e-06, "loss": 0.2298, "step": 79050 }, { "epoch": 4.77, "learning_rate": 2.8812638631116747e-06, "loss": 0.2643, "step": 79052 }, { "epoch": 4.77, "learning_rate": 2.8797549529974503e-06, "loss": 0.287, "step": 79054 }, { "epoch": 4.77, "learning_rate": 2.8782460428832256e-06, "loss": 0.2444, "step": 79056 }, { "epoch": 4.77, "learning_rate": 2.8767371327690012e-06, "loss": 0.2757, "step": 79058 }, { "epoch": 4.77, "learning_rate": 2.8752282226547765e-06, "loss": 0.2141, "step": 79060 }, { "epoch": 4.77, "learning_rate": 2.873719312540552e-06, "loss": 0.3108, "step": 79062 }, { "epoch": 4.77, "learning_rate": 2.8722104024263274e-06, "loss": 0.2528, "step": 79064 }, { "epoch": 4.77, "learning_rate": 2.870701492312103e-06, "loss": 0.1952, "step": 79066 }, { "epoch": 4.77, "learning_rate": 2.8691925821978783e-06, "loss": 0.2409, "step": 79068 }, { "epoch": 4.77, "learning_rate": 2.867683672083654e-06, "loss": 0.2084, "step": 79070 }, { "epoch": 4.77, "learning_rate": 2.8661747619694296e-06, "loss": 0.2933, "step": 79072 }, { "epoch": 4.77, "learning_rate": 2.8646658518552053e-06, "loss": 0.2684, "step": 79074 }, { "epoch": 4.77, "learning_rate": 2.8631569417409805e-06, "loss": 0.2128, "step": 79076 }, { "epoch": 4.77, "learning_rate": 2.8616480316267562e-06, "loss": 0.2394, "step": 79078 }, { "epoch": 4.77, "learning_rate": 2.8601391215125315e-06, "loss": 0.1308, "step": 79080 }, { "epoch": 4.77, "learning_rate": 2.858630211398307e-06, "loss": 0.1934, "step": 79082 }, { "epoch": 4.77, "learning_rate": 2.8571213012840824e-06, "loss": 0.3697, "step": 79084 }, { "epoch": 4.77, "learning_rate": 2.855612391169858e-06, "loss": 0.3122, "step": 79086 }, { "epoch": 4.77, "learning_rate": 2.8541034810556337e-06, "loss": 0.214, "step": 79088 }, { "epoch": 4.77, "learning_rate": 2.8525945709414094e-06, "loss": 0.2552, "step": 79090 }, { "epoch": 4.77, "learning_rate": 2.8510856608271846e-06, "loss": 0.2165, "step": 79092 }, { "epoch": 4.77, "learning_rate": 2.8495767507129603e-06, "loss": 0.1598, "step": 79094 }, { "epoch": 4.77, "learning_rate": 2.8480678405987355e-06, "loss": 0.255, "step": 79096 }, { "epoch": 4.77, "learning_rate": 2.846558930484511e-06, "loss": 0.2182, "step": 79098 }, { "epoch": 4.77, "learning_rate": 2.8450500203702864e-06, "loss": 0.2524, "step": 79100 }, { "epoch": 4.77, "learning_rate": 2.843541110256062e-06, "loss": 0.2761, "step": 79102 }, { "epoch": 4.77, "learning_rate": 2.8420322001418378e-06, "loss": 0.2579, "step": 79104 }, { "epoch": 4.77, "learning_rate": 2.840523290027613e-06, "loss": 0.1684, "step": 79106 }, { "epoch": 4.77, "learning_rate": 2.8390143799133887e-06, "loss": 0.2436, "step": 79108 }, { "epoch": 4.77, "learning_rate": 2.8375054697991643e-06, "loss": 0.1851, "step": 79110 }, { "epoch": 4.77, "learning_rate": 2.83599655968494e-06, "loss": 0.2608, "step": 79112 }, { "epoch": 4.78, "learning_rate": 2.8344876495707152e-06, "loss": 0.3077, "step": 79114 }, { "epoch": 4.78, "learning_rate": 2.832978739456491e-06, "loss": 0.2052, "step": 79116 }, { "epoch": 4.78, "learning_rate": 2.831469829342266e-06, "loss": 0.2301, "step": 79118 }, { "epoch": 4.78, "learning_rate": 2.829960919228042e-06, "loss": 0.1631, "step": 79120 }, { "epoch": 4.78, "learning_rate": 2.828452009113817e-06, "loss": 0.2568, "step": 79122 }, { "epoch": 4.78, "learning_rate": 2.8269430989995927e-06, "loss": 0.2182, "step": 79124 }, { "epoch": 4.78, "learning_rate": 2.825434188885368e-06, "loss": 0.2042, "step": 79126 }, { "epoch": 4.78, "learning_rate": 2.8239252787711436e-06, "loss": 0.2369, "step": 79128 }, { "epoch": 4.78, "learning_rate": 2.8224163686569193e-06, "loss": 0.2, "step": 79130 }, { "epoch": 4.78, "learning_rate": 2.820907458542695e-06, "loss": 0.2421, "step": 79132 }, { "epoch": 4.78, "learning_rate": 2.8193985484284702e-06, "loss": 0.1277, "step": 79134 }, { "epoch": 4.78, "learning_rate": 2.817889638314246e-06, "loss": 0.2278, "step": 79136 }, { "epoch": 4.78, "learning_rate": 2.816380728200021e-06, "loss": 0.2503, "step": 79138 }, { "epoch": 4.78, "learning_rate": 2.814871818085797e-06, "loss": 0.2724, "step": 79140 }, { "epoch": 4.78, "learning_rate": 2.813362907971572e-06, "loss": 0.2091, "step": 79142 }, { "epoch": 4.78, "learning_rate": 2.8118539978573477e-06, "loss": 0.2607, "step": 79144 }, { "epoch": 4.78, "learning_rate": 2.8103450877431234e-06, "loss": 0.1742, "step": 79146 }, { "epoch": 4.78, "learning_rate": 2.808836177628899e-06, "loss": 0.2299, "step": 79148 }, { "epoch": 4.78, "learning_rate": 2.8073272675146743e-06, "loss": 0.1946, "step": 79150 }, { "epoch": 4.78, "learning_rate": 2.80581835740045e-06, "loss": 0.181, "step": 79152 }, { "epoch": 4.78, "learning_rate": 2.804309447286225e-06, "loss": 0.2495, "step": 79154 }, { "epoch": 4.78, "learning_rate": 2.802800537172001e-06, "loss": 0.2736, "step": 79156 }, { "epoch": 4.78, "learning_rate": 2.801291627057776e-06, "loss": 0.2321, "step": 79158 }, { "epoch": 4.78, "learning_rate": 2.7997827169435518e-06, "loss": 0.2984, "step": 79160 }, { "epoch": 4.78, "learning_rate": 2.798273806829327e-06, "loss": 0.209, "step": 79162 }, { "epoch": 4.78, "learning_rate": 2.7967648967151027e-06, "loss": 0.1839, "step": 79164 }, { "epoch": 4.78, "learning_rate": 2.7952559866008783e-06, "loss": 0.3233, "step": 79166 }, { "epoch": 4.78, "learning_rate": 2.793747076486654e-06, "loss": 0.2693, "step": 79168 }, { "epoch": 4.78, "learning_rate": 2.7922381663724293e-06, "loss": 0.2504, "step": 79170 }, { "epoch": 4.78, "learning_rate": 2.790729256258205e-06, "loss": 0.226, "step": 79172 }, { "epoch": 4.78, "learning_rate": 2.78922034614398e-06, "loss": 0.1965, "step": 79174 }, { "epoch": 4.78, "learning_rate": 2.787711436029756e-06, "loss": 0.2046, "step": 79176 }, { "epoch": 4.78, "learning_rate": 2.786202525915531e-06, "loss": 0.2051, "step": 79178 }, { "epoch": 4.78, "learning_rate": 2.7846936158013067e-06, "loss": 0.3059, "step": 79180 }, { "epoch": 4.78, "learning_rate": 2.7831847056870824e-06, "loss": 0.2767, "step": 79182 }, { "epoch": 4.78, "learning_rate": 2.7816757955728577e-06, "loss": 0.1949, "step": 79184 }, { "epoch": 4.78, "learning_rate": 2.7801668854586333e-06, "loss": 0.2196, "step": 79186 }, { "epoch": 4.78, "learning_rate": 2.778657975344409e-06, "loss": 0.2088, "step": 79188 }, { "epoch": 4.78, "learning_rate": 2.7771490652301847e-06, "loss": 0.2978, "step": 79190 }, { "epoch": 4.78, "learning_rate": 2.77564015511596e-06, "loss": 0.1838, "step": 79192 }, { "epoch": 4.78, "learning_rate": 2.7741312450017356e-06, "loss": 0.4122, "step": 79194 }, { "epoch": 4.78, "learning_rate": 2.772622334887511e-06, "loss": 0.3263, "step": 79196 }, { "epoch": 4.78, "learning_rate": 2.7711134247732865e-06, "loss": 0.2203, "step": 79198 }, { "epoch": 4.78, "learning_rate": 2.7696045146590617e-06, "loss": 0.3262, "step": 79200 }, { "epoch": 4.78, "learning_rate": 2.7680956045448374e-06, "loss": 0.3203, "step": 79202 }, { "epoch": 4.78, "learning_rate": 2.766586694430613e-06, "loss": 0.1328, "step": 79204 }, { "epoch": 4.78, "learning_rate": 2.7650777843163887e-06, "loss": 0.2746, "step": 79206 }, { "epoch": 4.78, "learning_rate": 2.763568874202164e-06, "loss": 0.1867, "step": 79208 }, { "epoch": 4.78, "learning_rate": 2.7620599640879396e-06, "loss": 0.1281, "step": 79210 }, { "epoch": 4.78, "learning_rate": 2.760551053973715e-06, "loss": 0.1953, "step": 79212 }, { "epoch": 4.78, "learning_rate": 2.7590421438594905e-06, "loss": 0.3217, "step": 79214 }, { "epoch": 4.78, "learning_rate": 2.7575332337452658e-06, "loss": 0.219, "step": 79216 }, { "epoch": 4.78, "learning_rate": 2.7560243236310414e-06, "loss": 0.3103, "step": 79218 }, { "epoch": 4.78, "learning_rate": 2.7545154135168167e-06, "loss": 0.1892, "step": 79220 }, { "epoch": 4.78, "learning_rate": 2.7530065034025924e-06, "loss": 0.2758, "step": 79222 }, { "epoch": 4.78, "learning_rate": 2.751497593288368e-06, "loss": 0.2033, "step": 79224 }, { "epoch": 4.78, "learning_rate": 2.7499886831741437e-06, "loss": 0.2402, "step": 79226 }, { "epoch": 4.78, "learning_rate": 2.748479773059919e-06, "loss": 0.2892, "step": 79228 }, { "epoch": 4.78, "learning_rate": 2.7469708629456946e-06, "loss": 0.2565, "step": 79230 }, { "epoch": 4.78, "learning_rate": 2.74546195283147e-06, "loss": 0.2016, "step": 79232 }, { "epoch": 4.78, "learning_rate": 2.7439530427172455e-06, "loss": 0.1994, "step": 79234 }, { "epoch": 4.78, "learning_rate": 2.7424441326030207e-06, "loss": 0.2505, "step": 79236 }, { "epoch": 4.78, "learning_rate": 2.7409352224887964e-06, "loss": 0.2104, "step": 79238 }, { "epoch": 4.78, "learning_rate": 2.7394263123745717e-06, "loss": 0.2738, "step": 79240 }, { "epoch": 4.78, "learning_rate": 2.7379174022603473e-06, "loss": 0.1798, "step": 79242 }, { "epoch": 4.78, "learning_rate": 2.736408492146123e-06, "loss": 0.1909, "step": 79244 }, { "epoch": 4.78, "learning_rate": 2.7348995820318987e-06, "loss": 0.2585, "step": 79246 }, { "epoch": 4.78, "learning_rate": 2.733390671917674e-06, "loss": 0.2351, "step": 79248 }, { "epoch": 4.78, "learning_rate": 2.7318817618034496e-06, "loss": 0.1505, "step": 79250 }, { "epoch": 4.78, "learning_rate": 2.730372851689225e-06, "loss": 0.1023, "step": 79252 }, { "epoch": 4.78, "learning_rate": 2.7288639415750005e-06, "loss": 0.181, "step": 79254 }, { "epoch": 4.78, "learning_rate": 2.7273550314607757e-06, "loss": 0.2551, "step": 79256 }, { "epoch": 4.78, "learning_rate": 2.7258461213465514e-06, "loss": 0.3133, "step": 79258 }, { "epoch": 4.78, "learning_rate": 2.724337211232327e-06, "loss": 0.251, "step": 79260 }, { "epoch": 4.78, "learning_rate": 2.7228283011181027e-06, "loss": 0.2854, "step": 79262 }, { "epoch": 4.78, "learning_rate": 2.7213193910038784e-06, "loss": 0.2224, "step": 79264 }, { "epoch": 4.78, "learning_rate": 2.7198104808896536e-06, "loss": 0.2393, "step": 79266 }, { "epoch": 4.78, "learning_rate": 2.7183015707754293e-06, "loss": 0.1766, "step": 79268 }, { "epoch": 4.78, "learning_rate": 2.7167926606612045e-06, "loss": 0.2001, "step": 79270 }, { "epoch": 4.78, "learning_rate": 2.71528375054698e-06, "loss": 0.2838, "step": 79272 }, { "epoch": 4.78, "learning_rate": 2.7137748404327554e-06, "loss": 0.2585, "step": 79274 }, { "epoch": 4.78, "learning_rate": 2.712265930318531e-06, "loss": 0.2051, "step": 79276 }, { "epoch": 4.79, "learning_rate": 2.7107570202043064e-06, "loss": 0.2676, "step": 79278 }, { "epoch": 4.79, "learning_rate": 2.709248110090082e-06, "loss": 0.1817, "step": 79280 }, { "epoch": 4.79, "learning_rate": 2.7077391999758577e-06, "loss": 0.1316, "step": 79282 }, { "epoch": 4.79, "learning_rate": 2.7062302898616334e-06, "loss": 0.1968, "step": 79284 }, { "epoch": 4.79, "learning_rate": 2.7047213797474086e-06, "loss": 0.297, "step": 79286 }, { "epoch": 4.79, "learning_rate": 2.7032124696331843e-06, "loss": 0.1857, "step": 79288 }, { "epoch": 4.79, "learning_rate": 2.7017035595189595e-06, "loss": 0.245, "step": 79290 }, { "epoch": 4.79, "learning_rate": 2.700194649404735e-06, "loss": 0.2376, "step": 79292 }, { "epoch": 4.79, "learning_rate": 2.6986857392905104e-06, "loss": 0.2369, "step": 79294 }, { "epoch": 4.79, "learning_rate": 2.697176829176286e-06, "loss": 0.2352, "step": 79296 }, { "epoch": 4.79, "learning_rate": 2.6956679190620613e-06, "loss": 0.3333, "step": 79298 }, { "epoch": 4.79, "learning_rate": 2.694159008947837e-06, "loss": 0.1677, "step": 79300 }, { "epoch": 4.79, "learning_rate": 2.6926500988336127e-06, "loss": 0.2629, "step": 79302 }, { "epoch": 4.79, "learning_rate": 2.6911411887193883e-06, "loss": 0.1807, "step": 79304 }, { "epoch": 4.79, "learning_rate": 2.6896322786051636e-06, "loss": 0.1396, "step": 79306 }, { "epoch": 4.79, "learning_rate": 2.6881233684909392e-06, "loss": 0.2377, "step": 79308 }, { "epoch": 4.79, "learning_rate": 2.6866144583767145e-06, "loss": 0.2738, "step": 79310 }, { "epoch": 4.79, "learning_rate": 2.68510554826249e-06, "loss": 0.1644, "step": 79312 }, { "epoch": 4.79, "learning_rate": 2.6835966381482654e-06, "loss": 0.1467, "step": 79314 }, { "epoch": 4.79, "learning_rate": 2.682087728034041e-06, "loss": 0.2155, "step": 79316 }, { "epoch": 4.79, "learning_rate": 2.6805788179198163e-06, "loss": 0.2706, "step": 79318 }, { "epoch": 4.79, "learning_rate": 2.6790699078055924e-06, "loss": 0.3776, "step": 79320 }, { "epoch": 4.79, "learning_rate": 2.6775609976913676e-06, "loss": 0.2229, "step": 79322 }, { "epoch": 4.79, "learning_rate": 2.6760520875771433e-06, "loss": 0.1919, "step": 79324 }, { "epoch": 4.79, "learning_rate": 2.6745431774629185e-06, "loss": 0.1578, "step": 79326 }, { "epoch": 4.79, "learning_rate": 2.673034267348694e-06, "loss": 0.1787, "step": 79328 }, { "epoch": 4.79, "learning_rate": 2.6715253572344695e-06, "loss": 0.2262, "step": 79330 }, { "epoch": 4.79, "learning_rate": 2.670016447120245e-06, "loss": 0.1708, "step": 79332 }, { "epoch": 4.79, "learning_rate": 2.6685075370060208e-06, "loss": 0.2265, "step": 79334 }, { "epoch": 4.79, "learning_rate": 2.666998626891796e-06, "loss": 0.1754, "step": 79336 }, { "epoch": 4.79, "learning_rate": 2.6654897167775717e-06, "loss": 0.1449, "step": 79338 }, { "epoch": 4.79, "learning_rate": 2.6639808066633474e-06, "loss": 0.1895, "step": 79340 }, { "epoch": 4.79, "learning_rate": 2.662471896549123e-06, "loss": 0.224, "step": 79342 }, { "epoch": 4.79, "learning_rate": 2.6609629864348983e-06, "loss": 0.2229, "step": 79344 }, { "epoch": 4.79, "learning_rate": 2.659454076320674e-06, "loss": 0.2774, "step": 79346 }, { "epoch": 4.79, "learning_rate": 2.657945166206449e-06, "loss": 0.1625, "step": 79348 }, { "epoch": 4.79, "learning_rate": 2.656436256092225e-06, "loss": 0.3651, "step": 79350 }, { "epoch": 4.79, "learning_rate": 2.654927345978e-06, "loss": 0.2294, "step": 79352 }, { "epoch": 4.79, "learning_rate": 2.6534184358637758e-06, "loss": 0.2204, "step": 79354 }, { "epoch": 4.79, "learning_rate": 2.651909525749551e-06, "loss": 0.2666, "step": 79356 }, { "epoch": 4.79, "learning_rate": 2.6504006156353267e-06, "loss": 0.2768, "step": 79358 }, { "epoch": 4.79, "learning_rate": 2.6488917055211023e-06, "loss": 0.1568, "step": 79360 }, { "epoch": 4.79, "learning_rate": 2.647382795406878e-06, "loss": 0.1818, "step": 79362 }, { "epoch": 4.79, "learning_rate": 2.6458738852926532e-06, "loss": 0.1621, "step": 79364 }, { "epoch": 4.79, "learning_rate": 2.644364975178429e-06, "loss": 0.2574, "step": 79366 }, { "epoch": 4.79, "learning_rate": 2.642856065064204e-06, "loss": 0.2144, "step": 79368 }, { "epoch": 4.79, "learning_rate": 2.64134715494998e-06, "loss": 0.288, "step": 79370 }, { "epoch": 4.79, "learning_rate": 2.639838244835755e-06, "loss": 0.1692, "step": 79372 }, { "epoch": 4.79, "learning_rate": 2.6383293347215307e-06, "loss": 0.2773, "step": 79374 }, { "epoch": 4.79, "learning_rate": 2.636820424607306e-06, "loss": 0.2641, "step": 79376 }, { "epoch": 4.79, "learning_rate": 2.6353115144930816e-06, "loss": 0.2191, "step": 79378 }, { "epoch": 4.79, "learning_rate": 2.6338026043788573e-06, "loss": 0.1733, "step": 79380 }, { "epoch": 4.79, "learning_rate": 2.632293694264633e-06, "loss": 0.2383, "step": 79382 }, { "epoch": 4.79, "learning_rate": 2.6307847841504082e-06, "loss": 0.2017, "step": 79384 }, { "epoch": 4.79, "learning_rate": 2.629275874036184e-06, "loss": 0.2657, "step": 79386 }, { "epoch": 4.79, "learning_rate": 2.627766963921959e-06, "loss": 0.2549, "step": 79388 }, { "epoch": 4.79, "learning_rate": 2.626258053807735e-06, "loss": 0.2184, "step": 79390 }, { "epoch": 4.79, "learning_rate": 2.62474914369351e-06, "loss": 0.2245, "step": 79392 }, { "epoch": 4.79, "learning_rate": 2.6232402335792857e-06, "loss": 0.3232, "step": 79394 }, { "epoch": 4.79, "learning_rate": 2.6217313234650614e-06, "loss": 0.107, "step": 79396 }, { "epoch": 4.79, "learning_rate": 2.620222413350837e-06, "loss": 0.2592, "step": 79398 }, { "epoch": 4.79, "learning_rate": 2.6187135032366123e-06, "loss": 0.1542, "step": 79400 }, { "epoch": 4.79, "learning_rate": 2.617204593122388e-06, "loss": 0.2632, "step": 79402 }, { "epoch": 4.79, "learning_rate": 2.615695683008163e-06, "loss": 0.2405, "step": 79404 }, { "epoch": 4.79, "learning_rate": 2.614186772893939e-06, "loss": 0.1772, "step": 79406 }, { "epoch": 4.79, "learning_rate": 2.612677862779714e-06, "loss": 0.2837, "step": 79408 }, { "epoch": 4.79, "learning_rate": 2.6111689526654898e-06, "loss": 0.2019, "step": 79410 }, { "epoch": 4.79, "learning_rate": 2.6096600425512654e-06, "loss": 0.1335, "step": 79412 }, { "epoch": 4.79, "learning_rate": 2.6081511324370407e-06, "loss": 0.2354, "step": 79414 }, { "epoch": 4.79, "learning_rate": 2.6066422223228163e-06, "loss": 0.2822, "step": 79416 }, { "epoch": 4.79, "learning_rate": 2.605133312208592e-06, "loss": 0.2252, "step": 79418 }, { "epoch": 4.79, "learning_rate": 2.6036244020943677e-06, "loss": 0.2948, "step": 79420 }, { "epoch": 4.79, "learning_rate": 2.602115491980143e-06, "loss": 0.2798, "step": 79422 }, { "epoch": 4.79, "learning_rate": 2.6006065818659186e-06, "loss": 0.1975, "step": 79424 }, { "epoch": 4.79, "learning_rate": 2.599097671751694e-06, "loss": 0.2164, "step": 79426 }, { "epoch": 4.79, "learning_rate": 2.5975887616374695e-06, "loss": 0.1733, "step": 79428 }, { "epoch": 4.79, "learning_rate": 2.5960798515232447e-06, "loss": 0.182, "step": 79430 }, { "epoch": 4.79, "learning_rate": 2.5945709414090204e-06, "loss": 0.2179, "step": 79432 }, { "epoch": 4.79, "learning_rate": 2.5930620312947956e-06, "loss": 0.2788, "step": 79434 }, { "epoch": 4.79, "learning_rate": 2.5915531211805713e-06, "loss": 0.2933, "step": 79436 }, { "epoch": 4.79, "learning_rate": 2.590044211066347e-06, "loss": 0.2189, "step": 79438 }, { "epoch": 4.79, "learning_rate": 2.5885353009521226e-06, "loss": 0.2985, "step": 79440 }, { "epoch": 4.79, "learning_rate": 2.587026390837898e-06, "loss": 0.1628, "step": 79442 }, { "epoch": 4.8, "learning_rate": 2.5855174807236736e-06, "loss": 0.3382, "step": 79444 }, { "epoch": 4.8, "learning_rate": 2.584008570609449e-06, "loss": 0.1798, "step": 79446 }, { "epoch": 4.8, "learning_rate": 2.5824996604952245e-06, "loss": 0.2276, "step": 79448 }, { "epoch": 4.8, "learning_rate": 2.5809907503809997e-06, "loss": 0.2403, "step": 79450 }, { "epoch": 4.8, "learning_rate": 2.5794818402667754e-06, "loss": 0.1151, "step": 79452 }, { "epoch": 4.8, "learning_rate": 2.577972930152551e-06, "loss": 0.2064, "step": 79454 }, { "epoch": 4.8, "learning_rate": 2.5764640200383267e-06, "loss": 0.2346, "step": 79456 }, { "epoch": 4.8, "learning_rate": 2.574955109924102e-06, "loss": 0.225, "step": 79458 }, { "epoch": 4.8, "learning_rate": 2.5734461998098776e-06, "loss": 0.209, "step": 79460 }, { "epoch": 4.8, "learning_rate": 2.571937289695653e-06, "loss": 0.1417, "step": 79462 }, { "epoch": 4.8, "learning_rate": 2.5704283795814285e-06, "loss": 0.2995, "step": 79464 }, { "epoch": 4.8, "learning_rate": 2.5689194694672038e-06, "loss": 0.2075, "step": 79466 }, { "epoch": 4.8, "learning_rate": 2.5674105593529794e-06, "loss": 0.1573, "step": 79468 }, { "epoch": 4.8, "learning_rate": 2.5659016492387547e-06, "loss": 0.1491, "step": 79470 }, { "epoch": 4.8, "learning_rate": 2.5643927391245303e-06, "loss": 0.1999, "step": 79472 }, { "epoch": 4.8, "learning_rate": 2.562883829010306e-06, "loss": 0.2255, "step": 79474 }, { "epoch": 4.8, "learning_rate": 2.5613749188960817e-06, "loss": 0.1208, "step": 79476 }, { "epoch": 4.8, "learning_rate": 2.559866008781857e-06, "loss": 0.2101, "step": 79478 }, { "epoch": 4.8, "learning_rate": 2.5583570986676326e-06, "loss": 0.1949, "step": 79480 }, { "epoch": 4.8, "learning_rate": 2.556848188553408e-06, "loss": 0.2088, "step": 79482 }, { "epoch": 4.8, "learning_rate": 2.5553392784391835e-06, "loss": 0.1799, "step": 79484 }, { "epoch": 4.8, "learning_rate": 2.5538303683249587e-06, "loss": 0.3453, "step": 79486 }, { "epoch": 4.8, "learning_rate": 2.5523214582107344e-06, "loss": 0.1925, "step": 79488 }, { "epoch": 4.8, "learning_rate": 2.55081254809651e-06, "loss": 0.2736, "step": 79490 }, { "epoch": 4.8, "learning_rate": 2.5493036379822853e-06, "loss": 0.3229, "step": 79492 }, { "epoch": 4.8, "learning_rate": 2.547794727868061e-06, "loss": 0.2043, "step": 79494 }, { "epoch": 4.8, "learning_rate": 2.5462858177538367e-06, "loss": 0.1313, "step": 79496 }, { "epoch": 4.8, "learning_rate": 2.5447769076396123e-06, "loss": 0.3133, "step": 79498 }, { "epoch": 4.8, "learning_rate": 2.5432679975253876e-06, "loss": 0.1882, "step": 79500 }, { "epoch": 4.8, "learning_rate": 2.5417590874111632e-06, "loss": 0.1944, "step": 79502 }, { "epoch": 4.8, "learning_rate": 2.5402501772969385e-06, "loss": 0.2424, "step": 79504 }, { "epoch": 4.8, "learning_rate": 2.538741267182714e-06, "loss": 0.183, "step": 79506 }, { "epoch": 4.8, "learning_rate": 2.5372323570684894e-06, "loss": 0.1925, "step": 79508 }, { "epoch": 4.8, "learning_rate": 2.535723446954265e-06, "loss": 0.1246, "step": 79510 }, { "epoch": 4.8, "learning_rate": 2.5342145368400407e-06, "loss": 0.2316, "step": 79512 }, { "epoch": 4.8, "learning_rate": 2.5327056267258164e-06, "loss": 0.242, "step": 79514 }, { "epoch": 4.8, "learning_rate": 2.5311967166115916e-06, "loss": 0.16, "step": 79516 }, { "epoch": 4.8, "learning_rate": 2.5296878064973673e-06, "loss": 0.2698, "step": 79518 }, { "epoch": 4.8, "learning_rate": 2.5281788963831425e-06, "loss": 0.2193, "step": 79520 }, { "epoch": 4.8, "learning_rate": 2.526669986268918e-06, "loss": 0.2109, "step": 79522 }, { "epoch": 4.8, "learning_rate": 2.5251610761546934e-06, "loss": 0.2162, "step": 79524 }, { "epoch": 4.8, "learning_rate": 2.523652166040469e-06, "loss": 0.2496, "step": 79526 }, { "epoch": 4.8, "learning_rate": 2.5221432559262444e-06, "loss": 0.2104, "step": 79528 }, { "epoch": 4.8, "learning_rate": 2.52063434581202e-06, "loss": 0.2524, "step": 79530 }, { "epoch": 4.8, "learning_rate": 2.5191254356977957e-06, "loss": 0.2396, "step": 79532 }, { "epoch": 4.8, "learning_rate": 2.5176165255835714e-06, "loss": 0.1761, "step": 79534 }, { "epoch": 4.8, "learning_rate": 2.5161076154693466e-06, "loss": 0.2388, "step": 79536 }, { "epoch": 4.8, "learning_rate": 2.5145987053551223e-06, "loss": 0.3467, "step": 79538 }, { "epoch": 4.8, "learning_rate": 2.5130897952408975e-06, "loss": 0.2708, "step": 79540 }, { "epoch": 4.8, "learning_rate": 2.511580885126673e-06, "loss": 0.255, "step": 79542 }, { "epoch": 4.8, "learning_rate": 2.5100719750124484e-06, "loss": 0.2485, "step": 79544 }, { "epoch": 4.8, "learning_rate": 2.508563064898224e-06, "loss": 0.1632, "step": 79546 }, { "epoch": 4.8, "learning_rate": 2.5070541547839993e-06, "loss": 0.228, "step": 79548 }, { "epoch": 4.8, "learning_rate": 2.505545244669775e-06, "loss": 0.2054, "step": 79550 }, { "epoch": 4.8, "learning_rate": 2.5040363345555507e-06, "loss": 0.1731, "step": 79552 }, { "epoch": 4.8, "learning_rate": 2.5025274244413263e-06, "loss": 0.1962, "step": 79554 }, { "epoch": 4.8, "learning_rate": 2.5010185143271016e-06, "loss": 0.1517, "step": 79556 }, { "epoch": 4.8, "learning_rate": 2.4995096042128772e-06, "loss": 0.1675, "step": 79558 }, { "epoch": 4.8, "learning_rate": 2.4980006940986525e-06, "loss": 0.1401, "step": 79560 }, { "epoch": 4.8, "learning_rate": 2.496491783984428e-06, "loss": 0.1944, "step": 79562 }, { "epoch": 4.8, "learning_rate": 2.4949828738702034e-06, "loss": 0.1343, "step": 79564 }, { "epoch": 4.8, "learning_rate": 2.493473963755979e-06, "loss": 0.1932, "step": 79566 }, { "epoch": 4.8, "learning_rate": 2.4919650536417547e-06, "loss": 0.2543, "step": 79568 }, { "epoch": 4.8, "learning_rate": 2.4904561435275304e-06, "loss": 0.1991, "step": 79570 }, { "epoch": 4.8, "learning_rate": 2.488947233413306e-06, "loss": 0.2589, "step": 79572 }, { "epoch": 4.8, "learning_rate": 2.4874383232990813e-06, "loss": 0.2242, "step": 79574 }, { "epoch": 4.8, "learning_rate": 2.485929413184857e-06, "loss": 0.2046, "step": 79576 }, { "epoch": 4.8, "learning_rate": 2.484420503070632e-06, "loss": 0.2656, "step": 79578 }, { "epoch": 4.8, "learning_rate": 2.482911592956408e-06, "loss": 0.3001, "step": 79580 }, { "epoch": 4.8, "learning_rate": 2.481402682842183e-06, "loss": 0.2251, "step": 79582 }, { "epoch": 4.8, "learning_rate": 2.4798937727279588e-06, "loss": 0.3338, "step": 79584 }, { "epoch": 4.8, "learning_rate": 2.478384862613734e-06, "loss": 0.2104, "step": 79586 }, { "epoch": 4.8, "learning_rate": 2.4768759524995097e-06, "loss": 0.2297, "step": 79588 }, { "epoch": 4.8, "learning_rate": 2.4753670423852854e-06, "loss": 0.1723, "step": 79590 }, { "epoch": 4.8, "learning_rate": 2.473858132271061e-06, "loss": 0.2344, "step": 79592 }, { "epoch": 4.8, "learning_rate": 2.4723492221568363e-06, "loss": 0.2382, "step": 79594 }, { "epoch": 4.8, "learning_rate": 2.470840312042612e-06, "loss": 0.25, "step": 79596 }, { "epoch": 4.8, "learning_rate": 2.469331401928387e-06, "loss": 0.2296, "step": 79598 }, { "epoch": 4.8, "learning_rate": 2.467822491814163e-06, "loss": 0.3551, "step": 79600 }, { "epoch": 4.8, "learning_rate": 2.466313581699938e-06, "loss": 0.2687, "step": 79602 }, { "epoch": 4.8, "learning_rate": 2.4648046715857138e-06, "loss": 0.1459, "step": 79604 }, { "epoch": 4.8, "learning_rate": 2.463295761471489e-06, "loss": 0.1387, "step": 79606 }, { "epoch": 4.8, "learning_rate": 2.4617868513572647e-06, "loss": 0.1634, "step": 79608 }, { "epoch": 4.81, "learning_rate": 2.4602779412430403e-06, "loss": 0.199, "step": 79610 }, { "epoch": 4.81, "learning_rate": 2.458769031128816e-06, "loss": 0.2332, "step": 79612 }, { "epoch": 4.81, "learning_rate": 2.4572601210145912e-06, "loss": 0.2203, "step": 79614 }, { "epoch": 4.81, "learning_rate": 2.455751210900367e-06, "loss": 0.1703, "step": 79616 }, { "epoch": 4.81, "learning_rate": 2.454242300786142e-06, "loss": 0.1438, "step": 79618 }, { "epoch": 4.81, "learning_rate": 2.452733390671918e-06, "loss": 0.2418, "step": 79620 }, { "epoch": 4.81, "learning_rate": 2.451224480557693e-06, "loss": 0.2598, "step": 79622 }, { "epoch": 4.81, "learning_rate": 2.4497155704434687e-06, "loss": 0.1848, "step": 79624 }, { "epoch": 4.81, "learning_rate": 2.448206660329244e-06, "loss": 0.1711, "step": 79626 }, { "epoch": 4.81, "learning_rate": 2.44669775021502e-06, "loss": 0.1284, "step": 79628 }, { "epoch": 4.81, "learning_rate": 2.4451888401007953e-06, "loss": 0.177, "step": 79630 }, { "epoch": 4.81, "learning_rate": 2.443679929986571e-06, "loss": 0.1456, "step": 79632 }, { "epoch": 4.81, "learning_rate": 2.442171019872346e-06, "loss": 0.1663, "step": 79634 }, { "epoch": 4.81, "learning_rate": 2.440662109758122e-06, "loss": 0.176, "step": 79636 }, { "epoch": 4.81, "learning_rate": 2.439153199643897e-06, "loss": 0.2534, "step": 79638 }, { "epoch": 4.81, "learning_rate": 2.4376442895296728e-06, "loss": 0.2407, "step": 79640 }, { "epoch": 4.81, "learning_rate": 2.436135379415448e-06, "loss": 0.1693, "step": 79642 }, { "epoch": 4.81, "learning_rate": 2.4346264693012237e-06, "loss": 0.333, "step": 79644 }, { "epoch": 4.81, "learning_rate": 2.4331175591869994e-06, "loss": 0.1395, "step": 79646 }, { "epoch": 4.81, "learning_rate": 2.431608649072775e-06, "loss": 0.1466, "step": 79648 }, { "epoch": 4.81, "learning_rate": 2.4300997389585507e-06, "loss": 0.2072, "step": 79650 }, { "epoch": 4.81, "learning_rate": 2.428590828844326e-06, "loss": 0.2225, "step": 79652 }, { "epoch": 4.81, "learning_rate": 2.4270819187301016e-06, "loss": 0.1629, "step": 79654 }, { "epoch": 4.81, "learning_rate": 2.425573008615877e-06, "loss": 0.2163, "step": 79656 }, { "epoch": 4.81, "learning_rate": 2.4240640985016525e-06, "loss": 0.1268, "step": 79658 }, { "epoch": 4.81, "learning_rate": 2.4225551883874278e-06, "loss": 0.17, "step": 79660 }, { "epoch": 4.81, "learning_rate": 2.4210462782732034e-06, "loss": 0.2429, "step": 79662 }, { "epoch": 4.81, "learning_rate": 2.4195373681589787e-06, "loss": 0.2061, "step": 79664 }, { "epoch": 4.81, "learning_rate": 2.4180284580447543e-06, "loss": 0.1618, "step": 79666 }, { "epoch": 4.81, "learning_rate": 2.41651954793053e-06, "loss": 0.2213, "step": 79668 }, { "epoch": 4.81, "learning_rate": 2.4150106378163057e-06, "loss": 0.1783, "step": 79670 }, { "epoch": 4.81, "learning_rate": 2.413501727702081e-06, "loss": 0.2596, "step": 79672 }, { "epoch": 4.81, "learning_rate": 2.4119928175878566e-06, "loss": 0.1626, "step": 79674 }, { "epoch": 4.81, "learning_rate": 2.410483907473632e-06, "loss": 0.1268, "step": 79676 }, { "epoch": 4.81, "learning_rate": 2.4089749973594075e-06, "loss": 0.2345, "step": 79678 }, { "epoch": 4.81, "learning_rate": 2.4074660872451827e-06, "loss": 0.3247, "step": 79680 }, { "epoch": 4.81, "learning_rate": 2.4059571771309584e-06, "loss": 0.3265, "step": 79682 }, { "epoch": 4.81, "learning_rate": 2.4044482670167336e-06, "loss": 0.1611, "step": 79684 }, { "epoch": 4.81, "learning_rate": 2.4029393569025093e-06, "loss": 0.2121, "step": 79686 }, { "epoch": 4.81, "learning_rate": 2.401430446788285e-06, "loss": 0.1882, "step": 79688 }, { "epoch": 4.81, "learning_rate": 2.3999215366740606e-06, "loss": 0.1886, "step": 79690 }, { "epoch": 4.81, "learning_rate": 2.398412626559836e-06, "loss": 0.1593, "step": 79692 }, { "epoch": 4.81, "learning_rate": 2.3969037164456115e-06, "loss": 0.2233, "step": 79694 }, { "epoch": 4.81, "learning_rate": 2.395394806331387e-06, "loss": 0.187, "step": 79696 }, { "epoch": 4.81, "learning_rate": 2.3938858962171625e-06, "loss": 0.2497, "step": 79698 }, { "epoch": 4.81, "learning_rate": 2.3923769861029377e-06, "loss": 0.1851, "step": 79700 }, { "epoch": 4.81, "learning_rate": 2.3908680759887134e-06, "loss": 0.208, "step": 79702 }, { "epoch": 4.81, "learning_rate": 2.389359165874489e-06, "loss": 0.3537, "step": 79704 }, { "epoch": 4.81, "learning_rate": 2.3878502557602647e-06, "loss": 0.1479, "step": 79706 }, { "epoch": 4.81, "learning_rate": 2.38634134564604e-06, "loss": 0.2268, "step": 79708 }, { "epoch": 4.81, "learning_rate": 2.3848324355318156e-06, "loss": 0.112, "step": 79710 }, { "epoch": 4.81, "learning_rate": 2.383323525417591e-06, "loss": 0.1581, "step": 79712 }, { "epoch": 4.81, "learning_rate": 2.3818146153033665e-06, "loss": 0.1887, "step": 79714 }, { "epoch": 4.81, "learning_rate": 2.3803057051891418e-06, "loss": 0.1628, "step": 79716 }, { "epoch": 4.81, "learning_rate": 2.3787967950749174e-06, "loss": 0.2089, "step": 79718 }, { "epoch": 4.81, "learning_rate": 2.3772878849606927e-06, "loss": 0.2821, "step": 79720 }, { "epoch": 4.81, "learning_rate": 2.3757789748464683e-06, "loss": 0.188, "step": 79722 }, { "epoch": 4.81, "learning_rate": 2.374270064732244e-06, "loss": 0.1954, "step": 79724 }, { "epoch": 4.81, "learning_rate": 2.3727611546180197e-06, "loss": 0.1744, "step": 79726 }, { "epoch": 4.81, "learning_rate": 2.3712522445037953e-06, "loss": 0.2352, "step": 79728 }, { "epoch": 4.81, "learning_rate": 2.3697433343895706e-06, "loss": 0.2522, "step": 79730 }, { "epoch": 4.81, "learning_rate": 2.3682344242753462e-06, "loss": 0.1799, "step": 79732 }, { "epoch": 4.81, "learning_rate": 2.3667255141611215e-06, "loss": 0.2455, "step": 79734 }, { "epoch": 4.81, "learning_rate": 2.365216604046897e-06, "loss": 0.1634, "step": 79736 }, { "epoch": 4.81, "learning_rate": 2.3637076939326724e-06, "loss": 0.1579, "step": 79738 }, { "epoch": 4.81, "learning_rate": 2.362198783818448e-06, "loss": 0.2368, "step": 79740 }, { "epoch": 4.81, "learning_rate": 2.3606898737042233e-06, "loss": 0.2189, "step": 79742 }, { "epoch": 4.81, "learning_rate": 2.359180963589999e-06, "loss": 0.265, "step": 79744 }, { "epoch": 4.81, "learning_rate": 2.3576720534757746e-06, "loss": 0.1937, "step": 79746 }, { "epoch": 4.81, "learning_rate": 2.3561631433615503e-06, "loss": 0.2469, "step": 79748 }, { "epoch": 4.81, "learning_rate": 2.3546542332473256e-06, "loss": 0.2367, "step": 79750 }, { "epoch": 4.81, "learning_rate": 2.3531453231331012e-06, "loss": 0.1882, "step": 79752 }, { "epoch": 4.81, "learning_rate": 2.3516364130188765e-06, "loss": 0.1402, "step": 79754 }, { "epoch": 4.81, "learning_rate": 2.350127502904652e-06, "loss": 0.2387, "step": 79756 }, { "epoch": 4.81, "learning_rate": 2.3486185927904274e-06, "loss": 0.2743, "step": 79758 }, { "epoch": 4.81, "learning_rate": 2.347109682676203e-06, "loss": 0.2358, "step": 79760 }, { "epoch": 4.81, "learning_rate": 2.3456007725619787e-06, "loss": 0.2265, "step": 79762 }, { "epoch": 4.81, "learning_rate": 2.3440918624477544e-06, "loss": 0.2473, "step": 79764 }, { "epoch": 4.81, "learning_rate": 2.3425829523335296e-06, "loss": 0.1587, "step": 79766 }, { "epoch": 4.81, "learning_rate": 2.3410740422193053e-06, "loss": 0.2859, "step": 79768 }, { "epoch": 4.81, "learning_rate": 2.3395651321050805e-06, "loss": 0.259, "step": 79770 }, { "epoch": 4.81, "learning_rate": 2.338056221990856e-06, "loss": 0.2078, "step": 79772 }, { "epoch": 4.81, "learning_rate": 2.3365473118766314e-06, "loss": 0.1726, "step": 79774 }, { "epoch": 4.82, "learning_rate": 2.335038401762407e-06, "loss": 0.226, "step": 79776 }, { "epoch": 4.82, "learning_rate": 2.3335294916481823e-06, "loss": 0.2095, "step": 79778 }, { "epoch": 4.82, "learning_rate": 2.332020581533958e-06, "loss": 0.2724, "step": 79780 }, { "epoch": 4.82, "learning_rate": 2.3305116714197337e-06, "loss": 0.2364, "step": 79782 }, { "epoch": 4.82, "learning_rate": 2.3290027613055093e-06, "loss": 0.2771, "step": 79784 }, { "epoch": 4.82, "learning_rate": 2.3274938511912846e-06, "loss": 0.235, "step": 79786 }, { "epoch": 4.82, "learning_rate": 2.3259849410770603e-06, "loss": 0.2929, "step": 79788 }, { "epoch": 4.82, "learning_rate": 2.3244760309628355e-06, "loss": 0.1854, "step": 79790 }, { "epoch": 4.82, "learning_rate": 2.322967120848611e-06, "loss": 0.2628, "step": 79792 }, { "epoch": 4.82, "learning_rate": 2.3214582107343864e-06, "loss": 0.2715, "step": 79794 }, { "epoch": 4.82, "learning_rate": 2.319949300620162e-06, "loss": 0.1325, "step": 79796 }, { "epoch": 4.82, "learning_rate": 2.3184403905059373e-06, "loss": 0.2023, "step": 79798 }, { "epoch": 4.82, "learning_rate": 2.316931480391713e-06, "loss": 0.2087, "step": 79800 }, { "epoch": 4.82, "learning_rate": 2.3154225702774887e-06, "loss": 0.1228, "step": 79802 }, { "epoch": 4.82, "learning_rate": 2.3139136601632643e-06, "loss": 0.1896, "step": 79804 }, { "epoch": 4.82, "learning_rate": 2.31240475004904e-06, "loss": 0.2345, "step": 79806 }, { "epoch": 4.82, "learning_rate": 2.3108958399348152e-06, "loss": 0.1601, "step": 79808 }, { "epoch": 4.82, "learning_rate": 2.309386929820591e-06, "loss": 0.2812, "step": 79810 }, { "epoch": 4.82, "learning_rate": 2.307878019706366e-06, "loss": 0.2223, "step": 79812 }, { "epoch": 4.82, "learning_rate": 2.306369109592142e-06, "loss": 0.3211, "step": 79814 }, { "epoch": 4.82, "learning_rate": 2.304860199477917e-06, "loss": 0.2164, "step": 79816 }, { "epoch": 4.82, "learning_rate": 2.3033512893636927e-06, "loss": 0.2869, "step": 79818 }, { "epoch": 4.82, "learning_rate": 2.3018423792494684e-06, "loss": 0.3032, "step": 79820 }, { "epoch": 4.82, "learning_rate": 2.300333469135244e-06, "loss": 0.2395, "step": 79822 }, { "epoch": 4.82, "learning_rate": 2.2988245590210193e-06, "loss": 0.294, "step": 79824 }, { "epoch": 4.82, "learning_rate": 2.297315648906795e-06, "loss": 0.3138, "step": 79826 }, { "epoch": 4.82, "learning_rate": 2.29580673879257e-06, "loss": 0.221, "step": 79828 }, { "epoch": 4.82, "learning_rate": 2.294297828678346e-06, "loss": 0.1527, "step": 79830 }, { "epoch": 4.82, "learning_rate": 2.292788918564121e-06, "loss": 0.156, "step": 79832 }, { "epoch": 4.82, "learning_rate": 2.2912800084498968e-06, "loss": 0.2417, "step": 79834 }, { "epoch": 4.82, "learning_rate": 2.289771098335672e-06, "loss": 0.1971, "step": 79836 }, { "epoch": 4.82, "learning_rate": 2.2882621882214477e-06, "loss": 0.2871, "step": 79838 }, { "epoch": 4.82, "learning_rate": 2.2867532781072234e-06, "loss": 0.2166, "step": 79840 }, { "epoch": 4.82, "learning_rate": 2.285244367992999e-06, "loss": 0.2891, "step": 79842 }, { "epoch": 4.82, "learning_rate": 2.2837354578787743e-06, "loss": 0.2146, "step": 79844 }, { "epoch": 4.82, "learning_rate": 2.28222654776455e-06, "loss": 0.1703, "step": 79846 }, { "epoch": 4.82, "learning_rate": 2.280717637650325e-06, "loss": 0.2746, "step": 79848 }, { "epoch": 4.82, "learning_rate": 2.279208727536101e-06, "loss": 0.2141, "step": 79850 }, { "epoch": 4.82, "learning_rate": 2.277699817421876e-06, "loss": 0.2307, "step": 79852 }, { "epoch": 4.82, "learning_rate": 2.2761909073076517e-06, "loss": 0.2247, "step": 79854 }, { "epoch": 4.82, "learning_rate": 2.274681997193427e-06, "loss": 0.3172, "step": 79856 }, { "epoch": 4.82, "learning_rate": 2.2731730870792027e-06, "loss": 0.1758, "step": 79858 }, { "epoch": 4.82, "learning_rate": 2.2716641769649783e-06, "loss": 0.3284, "step": 79860 }, { "epoch": 4.82, "learning_rate": 2.270155266850754e-06, "loss": 0.2116, "step": 79862 }, { "epoch": 4.82, "learning_rate": 2.2686463567365292e-06, "loss": 0.1664, "step": 79864 }, { "epoch": 4.82, "learning_rate": 2.267137446622305e-06, "loss": 0.2374, "step": 79866 }, { "epoch": 4.82, "learning_rate": 2.26562853650808e-06, "loss": 0.1814, "step": 79868 }, { "epoch": 4.82, "learning_rate": 2.264119626393856e-06, "loss": 0.2911, "step": 79870 }, { "epoch": 4.82, "learning_rate": 2.262610716279631e-06, "loss": 0.1848, "step": 79872 }, { "epoch": 4.82, "learning_rate": 2.2611018061654067e-06, "loss": 0.2471, "step": 79874 }, { "epoch": 4.82, "learning_rate": 2.259592896051182e-06, "loss": 0.2147, "step": 79876 }, { "epoch": 4.82, "learning_rate": 2.258083985936958e-06, "loss": 0.1414, "step": 79878 }, { "epoch": 4.82, "learning_rate": 2.2565750758227337e-06, "loss": 0.2311, "step": 79880 }, { "epoch": 4.82, "learning_rate": 2.255066165708509e-06, "loss": 0.1612, "step": 79882 }, { "epoch": 4.82, "learning_rate": 2.2535572555942846e-06, "loss": 0.3019, "step": 79884 }, { "epoch": 4.82, "learning_rate": 2.25204834548006e-06, "loss": 0.2425, "step": 79886 }, { "epoch": 4.82, "learning_rate": 2.2505394353658355e-06, "loss": 0.2469, "step": 79888 }, { "epoch": 4.82, "learning_rate": 2.2490305252516108e-06, "loss": 0.2509, "step": 79890 }, { "epoch": 4.82, "learning_rate": 2.2475216151373864e-06, "loss": 0.2398, "step": 79892 }, { "epoch": 4.82, "learning_rate": 2.2460127050231617e-06, "loss": 0.207, "step": 79894 }, { "epoch": 4.82, "learning_rate": 2.2445037949089374e-06, "loss": 0.2663, "step": 79896 }, { "epoch": 4.82, "learning_rate": 2.242994884794713e-06, "loss": 0.1324, "step": 79898 }, { "epoch": 4.82, "learning_rate": 2.2414859746804887e-06, "loss": 0.2771, "step": 79900 }, { "epoch": 4.82, "learning_rate": 2.239977064566264e-06, "loss": 0.2779, "step": 79902 }, { "epoch": 4.82, "learning_rate": 2.2384681544520396e-06, "loss": 0.2164, "step": 79904 }, { "epoch": 4.82, "learning_rate": 2.236959244337815e-06, "loss": 0.2239, "step": 79906 }, { "epoch": 4.82, "learning_rate": 2.2354503342235905e-06, "loss": 0.1956, "step": 79908 }, { "epoch": 4.82, "learning_rate": 2.2339414241093658e-06, "loss": 0.1425, "step": 79910 }, { "epoch": 4.82, "learning_rate": 2.2324325139951414e-06, "loss": 0.1714, "step": 79912 }, { "epoch": 4.82, "learning_rate": 2.2309236038809167e-06, "loss": 0.1754, "step": 79914 }, { "epoch": 4.82, "learning_rate": 2.2294146937666923e-06, "loss": 0.2472, "step": 79916 }, { "epoch": 4.82, "learning_rate": 2.227905783652468e-06, "loss": 0.229, "step": 79918 }, { "epoch": 4.82, "learning_rate": 2.2263968735382437e-06, "loss": 0.1899, "step": 79920 }, { "epoch": 4.82, "learning_rate": 2.224887963424019e-06, "loss": 0.1697, "step": 79922 }, { "epoch": 4.82, "learning_rate": 2.2233790533097946e-06, "loss": 0.2274, "step": 79924 }, { "epoch": 4.82, "learning_rate": 2.22187014319557e-06, "loss": 0.1898, "step": 79926 }, { "epoch": 4.82, "learning_rate": 2.2203612330813455e-06, "loss": 0.2945, "step": 79928 }, { "epoch": 4.82, "learning_rate": 2.2188523229671207e-06, "loss": 0.1589, "step": 79930 }, { "epoch": 4.82, "learning_rate": 2.2173434128528964e-06, "loss": 0.1949, "step": 79932 }, { "epoch": 4.82, "learning_rate": 2.2158345027386716e-06, "loss": 0.2407, "step": 79934 }, { "epoch": 4.82, "learning_rate": 2.2143255926244473e-06, "loss": 0.182, "step": 79936 }, { "epoch": 4.82, "learning_rate": 2.212816682510223e-06, "loss": 0.1547, "step": 79938 }, { "epoch": 4.82, "learning_rate": 2.2113077723959986e-06, "loss": 0.2424, "step": 79940 }, { "epoch": 4.83, "learning_rate": 2.209798862281774e-06, "loss": 0.2867, "step": 79942 }, { "epoch": 4.83, "learning_rate": 2.2082899521675495e-06, "loss": 0.1651, "step": 79944 }, { "epoch": 4.83, "learning_rate": 2.2067810420533248e-06, "loss": 0.2291, "step": 79946 }, { "epoch": 4.83, "learning_rate": 2.2052721319391005e-06, "loss": 0.249, "step": 79948 }, { "epoch": 4.83, "learning_rate": 2.2037632218248757e-06, "loss": 0.335, "step": 79950 }, { "epoch": 4.83, "learning_rate": 2.2022543117106514e-06, "loss": 0.2851, "step": 79952 }, { "epoch": 4.83, "learning_rate": 2.200745401596427e-06, "loss": 0.2546, "step": 79954 }, { "epoch": 4.83, "learning_rate": 2.1992364914822027e-06, "loss": 0.2104, "step": 79956 }, { "epoch": 4.83, "learning_rate": 2.1977275813679784e-06, "loss": 0.2049, "step": 79958 }, { "epoch": 4.83, "learning_rate": 2.1962186712537536e-06, "loss": 0.1657, "step": 79960 }, { "epoch": 4.83, "learning_rate": 2.1947097611395293e-06, "loss": 0.1236, "step": 79962 }, { "epoch": 4.83, "learning_rate": 2.1932008510253045e-06, "loss": 0.2232, "step": 79964 }, { "epoch": 4.83, "learning_rate": 2.19169194091108e-06, "loss": 0.2582, "step": 79966 }, { "epoch": 4.83, "learning_rate": 2.1901830307968554e-06, "loss": 0.2831, "step": 79968 }, { "epoch": 4.83, "learning_rate": 2.188674120682631e-06, "loss": 0.2121, "step": 79970 }, { "epoch": 4.83, "learning_rate": 2.1871652105684063e-06, "loss": 0.2352, "step": 79972 }, { "epoch": 4.83, "learning_rate": 2.185656300454182e-06, "loss": 0.2884, "step": 79974 }, { "epoch": 4.83, "learning_rate": 2.1841473903399577e-06, "loss": 0.2465, "step": 79976 }, { "epoch": 4.83, "learning_rate": 2.1826384802257333e-06, "loss": 0.1518, "step": 79978 }, { "epoch": 4.83, "learning_rate": 2.1811295701115086e-06, "loss": 0.1623, "step": 79980 }, { "epoch": 4.83, "learning_rate": 2.1796206599972842e-06, "loss": 0.1604, "step": 79982 }, { "epoch": 4.83, "learning_rate": 2.1781117498830595e-06, "loss": 0.1971, "step": 79984 }, { "epoch": 4.83, "learning_rate": 2.176602839768835e-06, "loss": 0.2211, "step": 79986 }, { "epoch": 4.83, "learning_rate": 2.1750939296546104e-06, "loss": 0.2406, "step": 79988 }, { "epoch": 4.83, "learning_rate": 2.173585019540386e-06, "loss": 0.191, "step": 79990 }, { "epoch": 4.83, "learning_rate": 2.1720761094261613e-06, "loss": 0.1455, "step": 79992 }, { "epoch": 4.83, "learning_rate": 2.170567199311937e-06, "loss": 0.2478, "step": 79994 }, { "epoch": 4.83, "learning_rate": 2.1690582891977126e-06, "loss": 0.097, "step": 79996 }, { "epoch": 4.83, "learning_rate": 2.1675493790834883e-06, "loss": 0.2205, "step": 79998 }, { "epoch": 4.83, "learning_rate": 2.1660404689692635e-06, "loss": 0.2457, "step": 80000 }, { "epoch": 4.83, "learning_rate": 2.1645315588550392e-06, "loss": 0.2021, "step": 80002 }, { "epoch": 4.83, "learning_rate": 2.1630226487408145e-06, "loss": 0.2825, "step": 80004 }, { "epoch": 4.83, "learning_rate": 2.16151373862659e-06, "loss": 0.2025, "step": 80006 }, { "epoch": 4.83, "learning_rate": 2.1600048285123654e-06, "loss": 0.1676, "step": 80008 }, { "epoch": 4.83, "learning_rate": 2.158495918398141e-06, "loss": 0.219, "step": 80010 }, { "epoch": 4.83, "learning_rate": 2.1569870082839167e-06, "loss": 0.2935, "step": 80012 }, { "epoch": 4.83, "learning_rate": 2.1554780981696924e-06, "loss": 0.1701, "step": 80014 }, { "epoch": 4.83, "learning_rate": 2.1539691880554676e-06, "loss": 0.2159, "step": 80016 }, { "epoch": 4.83, "learning_rate": 2.1524602779412433e-06, "loss": 0.1802, "step": 80018 }, { "epoch": 4.83, "learning_rate": 2.1509513678270185e-06, "loss": 0.1405, "step": 80020 }, { "epoch": 4.83, "learning_rate": 2.149442457712794e-06, "loss": 0.206, "step": 80022 }, { "epoch": 4.83, "learning_rate": 2.1479335475985694e-06, "loss": 0.2326, "step": 80024 }, { "epoch": 4.83, "learning_rate": 2.146424637484345e-06, "loss": 0.3022, "step": 80026 }, { "epoch": 4.83, "learning_rate": 2.1449157273701203e-06, "loss": 0.1922, "step": 80028 }, { "epoch": 4.83, "learning_rate": 2.143406817255896e-06, "loss": 0.2113, "step": 80030 }, { "epoch": 4.83, "learning_rate": 2.1418979071416717e-06, "loss": 0.1824, "step": 80032 }, { "epoch": 4.83, "learning_rate": 2.1403889970274473e-06, "loss": 0.214, "step": 80034 }, { "epoch": 4.83, "learning_rate": 2.138880086913223e-06, "loss": 0.2797, "step": 80036 }, { "epoch": 4.83, "learning_rate": 2.1373711767989982e-06, "loss": 0.1779, "step": 80038 }, { "epoch": 4.83, "learning_rate": 2.135862266684774e-06, "loss": 0.2307, "step": 80040 }, { "epoch": 4.83, "learning_rate": 2.134353356570549e-06, "loss": 0.2644, "step": 80042 }, { "epoch": 4.83, "learning_rate": 2.132844446456325e-06, "loss": 0.1722, "step": 80044 }, { "epoch": 4.83, "learning_rate": 2.1313355363421e-06, "loss": 0.2103, "step": 80046 }, { "epoch": 4.83, "learning_rate": 2.1298266262278757e-06, "loss": 0.3457, "step": 80048 }, { "epoch": 4.83, "learning_rate": 2.128317716113651e-06, "loss": 0.2088, "step": 80050 }, { "epoch": 4.83, "learning_rate": 2.1268088059994266e-06, "loss": 0.2123, "step": 80052 }, { "epoch": 4.83, "learning_rate": 2.1252998958852023e-06, "loss": 0.2523, "step": 80054 }, { "epoch": 4.83, "learning_rate": 2.123790985770978e-06, "loss": 0.1984, "step": 80056 }, { "epoch": 4.83, "learning_rate": 2.1222820756567532e-06, "loss": 0.1965, "step": 80058 }, { "epoch": 4.83, "learning_rate": 2.120773165542529e-06, "loss": 0.1444, "step": 80060 }, { "epoch": 4.83, "learning_rate": 2.119264255428304e-06, "loss": 0.212, "step": 80062 }, { "epoch": 4.83, "learning_rate": 2.11775534531408e-06, "loss": 0.2362, "step": 80064 }, { "epoch": 4.83, "learning_rate": 2.116246435199855e-06, "loss": 0.2799, "step": 80066 }, { "epoch": 4.83, "learning_rate": 2.1147375250856307e-06, "loss": 0.143, "step": 80068 }, { "epoch": 4.83, "learning_rate": 2.1132286149714064e-06, "loss": 0.2935, "step": 80070 }, { "epoch": 4.83, "learning_rate": 2.111719704857182e-06, "loss": 0.3355, "step": 80072 }, { "epoch": 4.83, "learning_rate": 2.1102107947429573e-06, "loss": 0.216, "step": 80074 }, { "epoch": 4.83, "learning_rate": 2.108701884628733e-06, "loss": 0.2145, "step": 80076 }, { "epoch": 4.83, "learning_rate": 2.107192974514508e-06, "loss": 0.1692, "step": 80078 }, { "epoch": 4.83, "learning_rate": 2.105684064400284e-06, "loss": 0.1489, "step": 80080 }, { "epoch": 4.83, "learning_rate": 2.104175154286059e-06, "loss": 0.0875, "step": 80082 }, { "epoch": 4.83, "learning_rate": 2.1026662441718348e-06, "loss": 0.237, "step": 80084 }, { "epoch": 4.83, "learning_rate": 2.10115733405761e-06, "loss": 0.306, "step": 80086 }, { "epoch": 4.83, "learning_rate": 2.0996484239433857e-06, "loss": 0.2038, "step": 80088 }, { "epoch": 4.83, "learning_rate": 2.0981395138291613e-06, "loss": 0.1853, "step": 80090 }, { "epoch": 4.83, "learning_rate": 2.096630603714937e-06, "loss": 0.2334, "step": 80092 }, { "epoch": 4.83, "learning_rate": 2.0951216936007123e-06, "loss": 0.2091, "step": 80094 }, { "epoch": 4.83, "learning_rate": 2.093612783486488e-06, "loss": 0.226, "step": 80096 }, { "epoch": 4.83, "learning_rate": 2.092103873372263e-06, "loss": 0.2865, "step": 80098 }, { "epoch": 4.83, "learning_rate": 2.090594963258039e-06, "loss": 0.2849, "step": 80100 }, { "epoch": 4.83, "learning_rate": 2.089086053143814e-06, "loss": 0.1357, "step": 80102 }, { "epoch": 4.83, "learning_rate": 2.0875771430295897e-06, "loss": 0.1878, "step": 80104 }, { "epoch": 4.83, "learning_rate": 2.086068232915365e-06, "loss": 0.2023, "step": 80106 }, { "epoch": 4.84, "learning_rate": 2.0845593228011407e-06, "loss": 0.2862, "step": 80108 }, { "epoch": 4.84, "learning_rate": 2.0830504126869163e-06, "loss": 0.1609, "step": 80110 }, { "epoch": 4.84, "learning_rate": 2.081541502572692e-06, "loss": 0.2256, "step": 80112 }, { "epoch": 4.84, "learning_rate": 2.0800325924584677e-06, "loss": 0.4662, "step": 80114 }, { "epoch": 4.84, "learning_rate": 2.078523682344243e-06, "loss": 0.2319, "step": 80116 }, { "epoch": 4.84, "learning_rate": 2.0770147722300186e-06, "loss": 0.1617, "step": 80118 }, { "epoch": 4.84, "learning_rate": 2.075505862115794e-06, "loss": 0.3926, "step": 80120 }, { "epoch": 4.84, "learning_rate": 2.0739969520015695e-06, "loss": 0.383, "step": 80122 }, { "epoch": 4.84, "learning_rate": 2.0724880418873447e-06, "loss": 0.1755, "step": 80124 }, { "epoch": 4.84, "learning_rate": 2.0709791317731204e-06, "loss": 0.2976, "step": 80126 }, { "epoch": 4.84, "learning_rate": 2.069470221658896e-06, "loss": 0.1544, "step": 80128 }, { "epoch": 4.84, "learning_rate": 2.0679613115446717e-06, "loss": 0.235, "step": 80130 }, { "epoch": 4.84, "learning_rate": 2.066452401430447e-06, "loss": 0.1868, "step": 80132 }, { "epoch": 4.84, "learning_rate": 2.0649434913162226e-06, "loss": 0.2213, "step": 80134 }, { "epoch": 4.84, "learning_rate": 2.063434581201998e-06, "loss": 0.2871, "step": 80136 }, { "epoch": 4.84, "learning_rate": 2.0619256710877735e-06, "loss": 0.2312, "step": 80138 }, { "epoch": 4.84, "learning_rate": 2.0604167609735488e-06, "loss": 0.1861, "step": 80140 }, { "epoch": 4.84, "learning_rate": 2.0589078508593244e-06, "loss": 0.2359, "step": 80142 }, { "epoch": 4.84, "learning_rate": 2.0573989407450997e-06, "loss": 0.1704, "step": 80144 }, { "epoch": 4.84, "learning_rate": 2.0558900306308754e-06, "loss": 0.2411, "step": 80146 }, { "epoch": 4.84, "learning_rate": 2.054381120516651e-06, "loss": 0.2232, "step": 80148 }, { "epoch": 4.84, "learning_rate": 2.0528722104024267e-06, "loss": 0.1776, "step": 80150 }, { "epoch": 4.84, "learning_rate": 2.051363300288202e-06, "loss": 0.2429, "step": 80152 }, { "epoch": 4.84, "learning_rate": 2.0498543901739776e-06, "loss": 0.2208, "step": 80154 }, { "epoch": 4.84, "learning_rate": 2.048345480059753e-06, "loss": 0.2402, "step": 80156 }, { "epoch": 4.84, "learning_rate": 2.0468365699455285e-06, "loss": 0.2458, "step": 80158 }, { "epoch": 4.84, "learning_rate": 2.0453276598313037e-06, "loss": 0.2585, "step": 80160 }, { "epoch": 4.84, "learning_rate": 2.0438187497170794e-06, "loss": 0.2111, "step": 80162 }, { "epoch": 4.84, "learning_rate": 2.0423098396028547e-06, "loss": 0.1687, "step": 80164 }, { "epoch": 4.84, "learning_rate": 2.0408009294886303e-06, "loss": 0.1256, "step": 80166 }, { "epoch": 4.84, "learning_rate": 2.039292019374406e-06, "loss": 0.3255, "step": 80168 }, { "epoch": 4.84, "learning_rate": 2.0377831092601817e-06, "loss": 0.2216, "step": 80170 }, { "epoch": 4.84, "learning_rate": 2.036274199145957e-06, "loss": 0.2135, "step": 80172 }, { "epoch": 4.84, "learning_rate": 2.0347652890317326e-06, "loss": 0.1813, "step": 80174 }, { "epoch": 4.84, "learning_rate": 2.033256378917508e-06, "loss": 0.247, "step": 80176 }, { "epoch": 4.84, "learning_rate": 2.0317474688032835e-06, "loss": 0.1616, "step": 80178 }, { "epoch": 4.84, "learning_rate": 2.0302385586890587e-06, "loss": 0.1736, "step": 80180 }, { "epoch": 4.84, "learning_rate": 2.0287296485748344e-06, "loss": 0.1618, "step": 80182 }, { "epoch": 4.84, "learning_rate": 2.0272207384606096e-06, "loss": 0.2174, "step": 80184 }, { "epoch": 4.84, "learning_rate": 2.0257118283463857e-06, "loss": 0.1857, "step": 80186 }, { "epoch": 4.84, "learning_rate": 2.024202918232161e-06, "loss": 0.2071, "step": 80188 }, { "epoch": 4.84, "learning_rate": 2.0226940081179366e-06, "loss": 0.1743, "step": 80190 }, { "epoch": 4.84, "learning_rate": 2.0211850980037123e-06, "loss": 0.1896, "step": 80192 }, { "epoch": 4.84, "learning_rate": 2.0196761878894875e-06, "loss": 0.1915, "step": 80194 }, { "epoch": 4.84, "learning_rate": 2.018167277775263e-06, "loss": 0.2914, "step": 80196 }, { "epoch": 4.84, "learning_rate": 2.0166583676610384e-06, "loss": 0.1993, "step": 80198 }, { "epoch": 4.84, "learning_rate": 2.015149457546814e-06, "loss": 0.3202, "step": 80200 }, { "epoch": 4.84, "learning_rate": 2.0136405474325894e-06, "loss": 0.1684, "step": 80202 }, { "epoch": 4.84, "learning_rate": 2.012131637318365e-06, "loss": 0.2479, "step": 80204 }, { "epoch": 4.84, "learning_rate": 2.0106227272041407e-06, "loss": 0.409, "step": 80206 }, { "epoch": 4.84, "learning_rate": 2.0091138170899164e-06, "loss": 0.2854, "step": 80208 }, { "epoch": 4.84, "learning_rate": 2.0076049069756916e-06, "loss": 0.2874, "step": 80210 }, { "epoch": 4.84, "learning_rate": 2.0060959968614673e-06, "loss": 0.1561, "step": 80212 }, { "epoch": 4.84, "learning_rate": 2.0045870867472425e-06, "loss": 0.1972, "step": 80214 }, { "epoch": 4.84, "learning_rate": 2.003078176633018e-06, "loss": 0.2007, "step": 80216 }, { "epoch": 4.84, "learning_rate": 2.0015692665187934e-06, "loss": 0.2314, "step": 80218 }, { "epoch": 4.84, "learning_rate": 2.000060356404569e-06, "loss": 0.2415, "step": 80220 }, { "epoch": 4.84, "learning_rate": 1.9985514462903443e-06, "loss": 0.2567, "step": 80222 }, { "epoch": 4.84, "learning_rate": 1.99704253617612e-06, "loss": 0.1821, "step": 80224 }, { "epoch": 4.84, "learning_rate": 1.9955336260618957e-06, "loss": 0.2088, "step": 80226 }, { "epoch": 4.84, "learning_rate": 1.9940247159476713e-06, "loss": 0.2476, "step": 80228 }, { "epoch": 4.84, "learning_rate": 1.9925158058334466e-06, "loss": 0.1052, "step": 80230 }, { "epoch": 4.84, "learning_rate": 1.9910068957192222e-06, "loss": 0.187, "step": 80232 }, { "epoch": 4.84, "learning_rate": 1.9894979856049975e-06, "loss": 0.2513, "step": 80234 }, { "epoch": 4.84, "learning_rate": 1.987989075490773e-06, "loss": 0.3215, "step": 80236 }, { "epoch": 4.84, "learning_rate": 1.9864801653765484e-06, "loss": 0.3127, "step": 80238 }, { "epoch": 4.84, "learning_rate": 1.984971255262324e-06, "loss": 0.2595, "step": 80240 }, { "epoch": 4.84, "learning_rate": 1.9834623451480993e-06, "loss": 0.2355, "step": 80242 }, { "epoch": 4.84, "learning_rate": 1.981953435033875e-06, "loss": 0.1617, "step": 80244 }, { "epoch": 4.84, "learning_rate": 1.9804445249196506e-06, "loss": 0.1743, "step": 80246 }, { "epoch": 4.84, "learning_rate": 1.9789356148054263e-06, "loss": 0.1977, "step": 80248 }, { "epoch": 4.84, "learning_rate": 1.9774267046912015e-06, "loss": 0.2019, "step": 80250 }, { "epoch": 4.84, "learning_rate": 1.975917794576977e-06, "loss": 0.2884, "step": 80252 }, { "epoch": 4.84, "learning_rate": 1.9744088844627525e-06, "loss": 0.2829, "step": 80254 }, { "epoch": 4.84, "learning_rate": 1.972899974348528e-06, "loss": 0.1742, "step": 80256 }, { "epoch": 4.84, "learning_rate": 1.9713910642343034e-06, "loss": 0.3049, "step": 80258 }, { "epoch": 4.84, "learning_rate": 1.969882154120079e-06, "loss": 0.3724, "step": 80260 }, { "epoch": 4.84, "learning_rate": 1.9683732440058547e-06, "loss": 0.2465, "step": 80262 }, { "epoch": 4.84, "learning_rate": 1.9668643338916304e-06, "loss": 0.2922, "step": 80264 }, { "epoch": 4.84, "learning_rate": 1.9653554237774056e-06, "loss": 0.2335, "step": 80266 }, { "epoch": 4.84, "learning_rate": 1.9638465136631813e-06, "loss": 0.2421, "step": 80268 }, { "epoch": 4.84, "learning_rate": 1.962337603548957e-06, "loss": 0.2071, "step": 80270 }, { "epoch": 4.85, "learning_rate": 1.960828693434732e-06, "loss": 0.1881, "step": 80272 }, { "epoch": 4.85, "learning_rate": 1.959319783320508e-06, "loss": 0.2578, "step": 80274 }, { "epoch": 4.85, "learning_rate": 1.957810873206283e-06, "loss": 0.2498, "step": 80276 }, { "epoch": 4.85, "learning_rate": 1.9563019630920588e-06, "loss": 0.2013, "step": 80278 }, { "epoch": 4.85, "learning_rate": 1.954793052977834e-06, "loss": 0.2185, "step": 80280 }, { "epoch": 4.85, "learning_rate": 1.9532841428636097e-06, "loss": 0.3181, "step": 80282 }, { "epoch": 4.85, "learning_rate": 1.9517752327493853e-06, "loss": 0.2102, "step": 80284 }, { "epoch": 4.85, "learning_rate": 1.950266322635161e-06, "loss": 0.2275, "step": 80286 }, { "epoch": 4.85, "learning_rate": 1.9487574125209362e-06, "loss": 0.2353, "step": 80288 }, { "epoch": 4.85, "learning_rate": 1.947248502406712e-06, "loss": 0.2141, "step": 80290 }, { "epoch": 4.85, "learning_rate": 1.945739592292487e-06, "loss": 0.1734, "step": 80292 }, { "epoch": 4.85, "learning_rate": 1.944230682178263e-06, "loss": 0.2772, "step": 80294 }, { "epoch": 4.85, "learning_rate": 1.942721772064038e-06, "loss": 0.2337, "step": 80296 }, { "epoch": 4.85, "learning_rate": 1.9412128619498137e-06, "loss": 0.3021, "step": 80298 }, { "epoch": 4.85, "learning_rate": 1.939703951835589e-06, "loss": 0.1488, "step": 80300 }, { "epoch": 4.85, "learning_rate": 1.9381950417213646e-06, "loss": 0.2275, "step": 80302 }, { "epoch": 4.85, "learning_rate": 1.9366861316071403e-06, "loss": 0.1474, "step": 80304 }, { "epoch": 4.85, "learning_rate": 1.935177221492916e-06, "loss": 0.2654, "step": 80306 }, { "epoch": 4.85, "learning_rate": 1.9336683113786912e-06, "loss": 0.1765, "step": 80308 }, { "epoch": 4.85, "learning_rate": 1.932159401264467e-06, "loss": 0.1917, "step": 80310 }, { "epoch": 4.85, "learning_rate": 1.930650491150242e-06, "loss": 0.3808, "step": 80312 }, { "epoch": 4.85, "learning_rate": 1.929141581036018e-06, "loss": 0.1276, "step": 80314 }, { "epoch": 4.85, "learning_rate": 1.927632670921793e-06, "loss": 0.2245, "step": 80316 }, { "epoch": 4.85, "learning_rate": 1.9261237608075687e-06, "loss": 0.151, "step": 80318 }, { "epoch": 4.85, "learning_rate": 1.9246148506933444e-06, "loss": 0.1948, "step": 80320 }, { "epoch": 4.85, "learning_rate": 1.92310594057912e-06, "loss": 0.2294, "step": 80322 }, { "epoch": 4.85, "learning_rate": 1.9215970304648953e-06, "loss": 0.2776, "step": 80324 }, { "epoch": 4.85, "learning_rate": 1.920088120350671e-06, "loss": 0.1958, "step": 80326 }, { "epoch": 4.85, "learning_rate": 1.918579210236446e-06, "loss": 0.2163, "step": 80328 }, { "epoch": 4.85, "learning_rate": 1.917070300122222e-06, "loss": 0.1551, "step": 80330 }, { "epoch": 4.85, "learning_rate": 1.915561390007997e-06, "loss": 0.2692, "step": 80332 }, { "epoch": 4.85, "learning_rate": 1.9140524798937728e-06, "loss": 0.2236, "step": 80334 }, { "epoch": 4.85, "learning_rate": 1.912543569779548e-06, "loss": 0.2901, "step": 80336 }, { "epoch": 4.85, "learning_rate": 1.9110346596653237e-06, "loss": 0.2373, "step": 80338 }, { "epoch": 4.85, "learning_rate": 1.9095257495510993e-06, "loss": 0.1848, "step": 80340 }, { "epoch": 4.85, "learning_rate": 1.908016839436875e-06, "loss": 0.3117, "step": 80342 }, { "epoch": 4.85, "learning_rate": 1.9065079293226503e-06, "loss": 0.2659, "step": 80344 }, { "epoch": 4.85, "learning_rate": 1.904999019208426e-06, "loss": 0.1401, "step": 80346 }, { "epoch": 4.85, "learning_rate": 1.9034901090942016e-06, "loss": 0.211, "step": 80348 }, { "epoch": 4.85, "learning_rate": 1.9019811989799768e-06, "loss": 0.2071, "step": 80350 }, { "epoch": 4.85, "learning_rate": 1.9004722888657525e-06, "loss": 0.1132, "step": 80352 }, { "epoch": 4.85, "learning_rate": 1.8989633787515277e-06, "loss": 0.2873, "step": 80354 }, { "epoch": 4.85, "learning_rate": 1.8974544686373036e-06, "loss": 0.2182, "step": 80356 }, { "epoch": 4.85, "learning_rate": 1.8959455585230789e-06, "loss": 0.2534, "step": 80358 }, { "epoch": 4.85, "learning_rate": 1.8944366484088545e-06, "loss": 0.2801, "step": 80360 }, { "epoch": 4.85, "learning_rate": 1.8929277382946298e-06, "loss": 0.2948, "step": 80362 }, { "epoch": 4.85, "learning_rate": 1.8914188281804054e-06, "loss": 0.1858, "step": 80364 }, { "epoch": 4.85, "learning_rate": 1.8899099180661809e-06, "loss": 0.1602, "step": 80366 }, { "epoch": 4.85, "learning_rate": 1.8884010079519566e-06, "loss": 0.2619, "step": 80368 }, { "epoch": 4.85, "learning_rate": 1.8868920978377318e-06, "loss": 0.2813, "step": 80370 }, { "epoch": 4.85, "learning_rate": 1.8853831877235075e-06, "loss": 0.1569, "step": 80372 }, { "epoch": 4.85, "learning_rate": 1.883874277609283e-06, "loss": 0.2344, "step": 80374 }, { "epoch": 4.85, "learning_rate": 1.8823653674950586e-06, "loss": 0.3888, "step": 80376 }, { "epoch": 4.85, "learning_rate": 1.8808564573808338e-06, "loss": 0.1657, "step": 80378 }, { "epoch": 4.85, "learning_rate": 1.8793475472666095e-06, "loss": 0.267, "step": 80380 }, { "epoch": 4.85, "learning_rate": 1.8778386371523847e-06, "loss": 0.3194, "step": 80382 }, { "epoch": 4.85, "learning_rate": 1.8763297270381606e-06, "loss": 0.2759, "step": 80384 }, { "epoch": 4.85, "learning_rate": 1.8748208169239359e-06, "loss": 0.168, "step": 80386 }, { "epoch": 4.85, "learning_rate": 1.8733119068097115e-06, "loss": 0.2366, "step": 80388 }, { "epoch": 4.85, "learning_rate": 1.8718029966954868e-06, "loss": 0.2729, "step": 80390 }, { "epoch": 4.85, "learning_rate": 1.8702940865812624e-06, "loss": 0.2292, "step": 80392 }, { "epoch": 4.85, "learning_rate": 1.8687851764670379e-06, "loss": 0.185, "step": 80394 }, { "epoch": 4.85, "learning_rate": 1.8672762663528136e-06, "loss": 0.1616, "step": 80396 }, { "epoch": 4.85, "learning_rate": 1.8657673562385888e-06, "loss": 0.3288, "step": 80398 }, { "epoch": 4.85, "learning_rate": 1.8642584461243645e-06, "loss": 0.1889, "step": 80400 }, { "epoch": 4.85, "learning_rate": 1.86274953601014e-06, "loss": 0.3008, "step": 80402 }, { "epoch": 4.85, "learning_rate": 1.8612406258959156e-06, "loss": 0.2006, "step": 80404 }, { "epoch": 4.85, "learning_rate": 1.8597317157816908e-06, "loss": 0.1907, "step": 80406 }, { "epoch": 4.85, "learning_rate": 1.8582228056674665e-06, "loss": 0.2251, "step": 80408 }, { "epoch": 4.85, "learning_rate": 1.8567138955532417e-06, "loss": 0.1928, "step": 80410 }, { "epoch": 4.85, "learning_rate": 1.8552049854390174e-06, "loss": 0.2315, "step": 80412 }, { "epoch": 4.85, "learning_rate": 1.8536960753247929e-06, "loss": 0.1828, "step": 80414 }, { "epoch": 4.85, "learning_rate": 1.8521871652105685e-06, "loss": 0.19, "step": 80416 }, { "epoch": 4.85, "learning_rate": 1.8506782550963438e-06, "loss": 0.1666, "step": 80418 }, { "epoch": 4.85, "learning_rate": 1.8491693449821194e-06, "loss": 0.2454, "step": 80420 }, { "epoch": 4.85, "learning_rate": 1.847660434867895e-06, "loss": 0.2035, "step": 80422 }, { "epoch": 4.85, "learning_rate": 1.8461515247536706e-06, "loss": 0.2358, "step": 80424 }, { "epoch": 4.85, "learning_rate": 1.8446426146394462e-06, "loss": 0.2435, "step": 80426 }, { "epoch": 4.85, "learning_rate": 1.8431337045252215e-06, "loss": 0.172, "step": 80428 }, { "epoch": 4.85, "learning_rate": 1.8416247944109971e-06, "loss": 0.2668, "step": 80430 }, { "epoch": 4.85, "learning_rate": 1.8401158842967726e-06, "loss": 0.1491, "step": 80432 }, { "epoch": 4.85, "learning_rate": 1.8386069741825483e-06, "loss": 0.1994, "step": 80434 }, { "epoch": 4.85, "learning_rate": 1.8370980640683235e-06, "loss": 0.1422, "step": 80436 }, { "epoch": 4.86, "learning_rate": 1.8355891539540992e-06, "loss": 0.3407, "step": 80438 }, { "epoch": 4.86, "learning_rate": 1.8340802438398744e-06, "loss": 0.1731, "step": 80440 }, { "epoch": 4.86, "learning_rate": 1.83257133372565e-06, "loss": 0.1445, "step": 80442 }, { "epoch": 4.86, "learning_rate": 1.8310624236114255e-06, "loss": 0.153, "step": 80444 }, { "epoch": 4.86, "learning_rate": 1.8295535134972012e-06, "loss": 0.2112, "step": 80446 }, { "epoch": 4.86, "learning_rate": 1.8280446033829764e-06, "loss": 0.4033, "step": 80448 }, { "epoch": 4.86, "learning_rate": 1.8265356932687521e-06, "loss": 0.156, "step": 80450 }, { "epoch": 4.86, "learning_rate": 1.8250267831545276e-06, "loss": 0.2727, "step": 80452 }, { "epoch": 4.86, "learning_rate": 1.8235178730403032e-06, "loss": 0.138, "step": 80454 }, { "epoch": 4.86, "learning_rate": 1.8220089629260785e-06, "loss": 0.1532, "step": 80456 }, { "epoch": 4.86, "learning_rate": 1.8205000528118541e-06, "loss": 0.2457, "step": 80458 }, { "epoch": 4.86, "learning_rate": 1.8189911426976296e-06, "loss": 0.1354, "step": 80460 }, { "epoch": 4.86, "learning_rate": 1.8174822325834053e-06, "loss": 0.3542, "step": 80462 }, { "epoch": 4.86, "learning_rate": 1.8159733224691805e-06, "loss": 0.2508, "step": 80464 }, { "epoch": 4.86, "learning_rate": 1.8144644123549562e-06, "loss": 0.2503, "step": 80466 }, { "epoch": 4.86, "learning_rate": 1.8129555022407314e-06, "loss": 0.1929, "step": 80468 }, { "epoch": 4.86, "learning_rate": 1.811446592126507e-06, "loss": 0.1942, "step": 80470 }, { "epoch": 4.86, "learning_rate": 1.8099376820122825e-06, "loss": 0.2496, "step": 80472 }, { "epoch": 4.86, "learning_rate": 1.8084287718980582e-06, "loss": 0.2987, "step": 80474 }, { "epoch": 4.86, "learning_rate": 1.8069198617838334e-06, "loss": 0.283, "step": 80476 }, { "epoch": 4.86, "learning_rate": 1.8054109516696091e-06, "loss": 0.2311, "step": 80478 }, { "epoch": 4.86, "learning_rate": 1.8039020415553846e-06, "loss": 0.1563, "step": 80480 }, { "epoch": 4.86, "learning_rate": 1.8023931314411602e-06, "loss": 0.1696, "step": 80482 }, { "epoch": 4.86, "learning_rate": 1.8008842213269355e-06, "loss": 0.3336, "step": 80484 }, { "epoch": 4.86, "learning_rate": 1.7993753112127111e-06, "loss": 0.2671, "step": 80486 }, { "epoch": 4.86, "learning_rate": 1.7978664010984866e-06, "loss": 0.115, "step": 80488 }, { "epoch": 4.86, "learning_rate": 1.7963574909842623e-06, "loss": 0.1548, "step": 80490 }, { "epoch": 4.86, "learning_rate": 1.7948485808700375e-06, "loss": 0.1924, "step": 80492 }, { "epoch": 4.86, "learning_rate": 1.7933396707558132e-06, "loss": 0.2443, "step": 80494 }, { "epoch": 4.86, "learning_rate": 1.7918307606415884e-06, "loss": 0.2861, "step": 80496 }, { "epoch": 4.86, "learning_rate": 1.790321850527364e-06, "loss": 0.2418, "step": 80498 }, { "epoch": 4.86, "learning_rate": 1.7888129404131398e-06, "loss": 0.2724, "step": 80500 }, { "epoch": 4.86, "learning_rate": 1.7873040302989152e-06, "loss": 0.1925, "step": 80502 }, { "epoch": 4.86, "learning_rate": 1.7857951201846909e-06, "loss": 0.2554, "step": 80504 }, { "epoch": 4.86, "learning_rate": 1.7842862100704661e-06, "loss": 0.1727, "step": 80506 }, { "epoch": 4.86, "learning_rate": 1.7827772999562418e-06, "loss": 0.2373, "step": 80508 }, { "epoch": 4.86, "learning_rate": 1.7812683898420172e-06, "loss": 0.2083, "step": 80510 }, { "epoch": 4.86, "learning_rate": 1.779759479727793e-06, "loss": 0.3161, "step": 80512 }, { "epoch": 4.86, "learning_rate": 1.7782505696135681e-06, "loss": 0.2444, "step": 80514 }, { "epoch": 4.86, "learning_rate": 1.7767416594993438e-06, "loss": 0.2427, "step": 80516 }, { "epoch": 4.86, "learning_rate": 1.7752327493851193e-06, "loss": 0.2083, "step": 80518 }, { "epoch": 4.86, "learning_rate": 1.773723839270895e-06, "loss": 0.3042, "step": 80520 }, { "epoch": 4.86, "learning_rate": 1.7722149291566702e-06, "loss": 0.2091, "step": 80522 }, { "epoch": 4.86, "learning_rate": 1.7707060190424458e-06, "loss": 0.1631, "step": 80524 }, { "epoch": 4.86, "learning_rate": 1.769197108928221e-06, "loss": 0.1873, "step": 80526 }, { "epoch": 4.86, "learning_rate": 1.7676881988139968e-06, "loss": 0.2365, "step": 80528 }, { "epoch": 4.86, "learning_rate": 1.7661792886997722e-06, "loss": 0.228, "step": 80530 }, { "epoch": 4.86, "learning_rate": 1.7646703785855479e-06, "loss": 0.2, "step": 80532 }, { "epoch": 4.86, "learning_rate": 1.7631614684713231e-06, "loss": 0.3891, "step": 80534 }, { "epoch": 4.86, "learning_rate": 1.7616525583570988e-06, "loss": 0.2165, "step": 80536 }, { "epoch": 4.86, "learning_rate": 1.7601436482428742e-06, "loss": 0.2908, "step": 80538 }, { "epoch": 4.86, "learning_rate": 1.75863473812865e-06, "loss": 0.1963, "step": 80540 }, { "epoch": 4.86, "learning_rate": 1.7571258280144251e-06, "loss": 0.1809, "step": 80542 }, { "epoch": 4.86, "learning_rate": 1.7556169179002008e-06, "loss": 0.1515, "step": 80544 }, { "epoch": 4.86, "learning_rate": 1.7541080077859763e-06, "loss": 0.3246, "step": 80546 }, { "epoch": 4.86, "learning_rate": 1.752599097671752e-06, "loss": 0.1104, "step": 80548 }, { "epoch": 4.86, "learning_rate": 1.7510901875575272e-06, "loss": 0.1399, "step": 80550 }, { "epoch": 4.86, "learning_rate": 1.7495812774433028e-06, "loss": 0.2432, "step": 80552 }, { "epoch": 4.86, "learning_rate": 1.748072367329078e-06, "loss": 0.2118, "step": 80554 }, { "epoch": 4.86, "learning_rate": 1.7465634572148538e-06, "loss": 0.1905, "step": 80556 }, { "epoch": 4.86, "learning_rate": 1.7450545471006292e-06, "loss": 0.2465, "step": 80558 }, { "epoch": 4.86, "learning_rate": 1.7435456369864049e-06, "loss": 0.1679, "step": 80560 }, { "epoch": 4.86, "learning_rate": 1.7420367268721801e-06, "loss": 0.116, "step": 80562 }, { "epoch": 4.86, "learning_rate": 1.7405278167579558e-06, "loss": 0.3203, "step": 80564 }, { "epoch": 4.86, "learning_rate": 1.7390189066437312e-06, "loss": 0.2129, "step": 80566 }, { "epoch": 4.86, "learning_rate": 1.737509996529507e-06, "loss": 0.2313, "step": 80568 }, { "epoch": 4.86, "learning_rate": 1.7360010864152822e-06, "loss": 0.1507, "step": 80570 }, { "epoch": 4.86, "learning_rate": 1.7344921763010578e-06, "loss": 0.1545, "step": 80572 }, { "epoch": 4.86, "learning_rate": 1.7329832661868333e-06, "loss": 0.1543, "step": 80574 }, { "epoch": 4.86, "learning_rate": 1.731474356072609e-06, "loss": 0.3138, "step": 80576 }, { "epoch": 4.86, "learning_rate": 1.7299654459583846e-06, "loss": 0.1952, "step": 80578 }, { "epoch": 4.86, "learning_rate": 1.7284565358441598e-06, "loss": 0.3279, "step": 80580 }, { "epoch": 4.86, "learning_rate": 1.7269476257299355e-06, "loss": 0.2692, "step": 80582 }, { "epoch": 4.86, "learning_rate": 1.7254387156157108e-06, "loss": 0.2121, "step": 80584 }, { "epoch": 4.86, "learning_rate": 1.7239298055014864e-06, "loss": 0.2656, "step": 80586 }, { "epoch": 4.86, "learning_rate": 1.7224208953872619e-06, "loss": 0.3195, "step": 80588 }, { "epoch": 4.86, "learning_rate": 1.7209119852730375e-06, "loss": 0.2366, "step": 80590 }, { "epoch": 4.86, "learning_rate": 1.7194030751588128e-06, "loss": 0.2263, "step": 80592 }, { "epoch": 4.86, "learning_rate": 1.7178941650445885e-06, "loss": 0.2341, "step": 80594 }, { "epoch": 4.86, "learning_rate": 1.716385254930364e-06, "loss": 0.2475, "step": 80596 }, { "epoch": 4.86, "learning_rate": 1.7148763448161396e-06, "loss": 0.2124, "step": 80598 }, { "epoch": 4.86, "learning_rate": 1.7133674347019148e-06, "loss": 0.4305, "step": 80600 }, { "epoch": 4.86, "learning_rate": 1.7118585245876905e-06, "loss": 0.188, "step": 80602 }, { "epoch": 4.87, "learning_rate": 1.710349614473466e-06, "loss": 0.264, "step": 80604 }, { "epoch": 4.87, "learning_rate": 1.7088407043592416e-06, "loss": 0.2431, "step": 80606 }, { "epoch": 4.87, "learning_rate": 1.7073317942450169e-06, "loss": 0.2681, "step": 80608 }, { "epoch": 4.87, "learning_rate": 1.7058228841307925e-06, "loss": 0.2572, "step": 80610 }, { "epoch": 4.87, "learning_rate": 1.7043139740165678e-06, "loss": 0.1814, "step": 80612 }, { "epoch": 4.87, "learning_rate": 1.7028050639023434e-06, "loss": 0.2242, "step": 80614 }, { "epoch": 4.87, "learning_rate": 1.7012961537881189e-06, "loss": 0.2187, "step": 80616 }, { "epoch": 4.87, "learning_rate": 1.6997872436738945e-06, "loss": 0.2258, "step": 80618 }, { "epoch": 4.87, "learning_rate": 1.6982783335596698e-06, "loss": 0.2233, "step": 80620 }, { "epoch": 4.87, "learning_rate": 1.6967694234454455e-06, "loss": 0.2209, "step": 80622 }, { "epoch": 4.87, "learning_rate": 1.695260513331221e-06, "loss": 0.2063, "step": 80624 }, { "epoch": 4.87, "learning_rate": 1.6937516032169966e-06, "loss": 0.2401, "step": 80626 }, { "epoch": 4.87, "learning_rate": 1.6922426931027718e-06, "loss": 0.256, "step": 80628 }, { "epoch": 4.87, "learning_rate": 1.6907337829885475e-06, "loss": 0.1788, "step": 80630 }, { "epoch": 4.87, "learning_rate": 1.6892248728743227e-06, "loss": 0.1838, "step": 80632 }, { "epoch": 4.87, "learning_rate": 1.6877159627600986e-06, "loss": 0.2988, "step": 80634 }, { "epoch": 4.87, "learning_rate": 1.6862070526458739e-06, "loss": 0.2269, "step": 80636 }, { "epoch": 4.87, "learning_rate": 1.6846981425316495e-06, "loss": 0.211, "step": 80638 }, { "epoch": 4.87, "learning_rate": 1.6831892324174248e-06, "loss": 0.2461, "step": 80640 }, { "epoch": 4.87, "learning_rate": 1.6816803223032004e-06, "loss": 0.2937, "step": 80642 }, { "epoch": 4.87, "learning_rate": 1.6801714121889759e-06, "loss": 0.1687, "step": 80644 }, { "epoch": 4.87, "learning_rate": 1.6786625020747516e-06, "loss": 0.2239, "step": 80646 }, { "epoch": 4.87, "learning_rate": 1.6771535919605268e-06, "loss": 0.2083, "step": 80648 }, { "epoch": 4.87, "learning_rate": 1.6756446818463025e-06, "loss": 0.2087, "step": 80650 }, { "epoch": 4.87, "learning_rate": 1.674135771732078e-06, "loss": 0.1793, "step": 80652 }, { "epoch": 4.87, "learning_rate": 1.6726268616178536e-06, "loss": 0.2146, "step": 80654 }, { "epoch": 4.87, "learning_rate": 1.6711179515036293e-06, "loss": 0.2729, "step": 80656 }, { "epoch": 4.87, "learning_rate": 1.6696090413894045e-06, "loss": 0.1942, "step": 80658 }, { "epoch": 4.87, "learning_rate": 1.6681001312751802e-06, "loss": 0.2953, "step": 80660 }, { "epoch": 4.87, "learning_rate": 1.6665912211609554e-06, "loss": 0.1957, "step": 80662 }, { "epoch": 4.87, "learning_rate": 1.6650823110467313e-06, "loss": 0.2824, "step": 80664 }, { "epoch": 4.87, "learning_rate": 1.6635734009325065e-06, "loss": 0.2351, "step": 80666 }, { "epoch": 4.87, "learning_rate": 1.6620644908182822e-06, "loss": 0.2062, "step": 80668 }, { "epoch": 4.87, "learning_rate": 1.6605555807040574e-06, "loss": 0.2285, "step": 80670 }, { "epoch": 4.87, "learning_rate": 1.659046670589833e-06, "loss": 0.1988, "step": 80672 }, { "epoch": 4.87, "learning_rate": 1.6575377604756086e-06, "loss": 0.2202, "step": 80674 }, { "epoch": 4.87, "learning_rate": 1.6560288503613842e-06, "loss": 0.2323, "step": 80676 }, { "epoch": 4.87, "learning_rate": 1.6545199402471595e-06, "loss": 0.3537, "step": 80678 }, { "epoch": 4.87, "learning_rate": 1.6530110301329351e-06, "loss": 0.2258, "step": 80680 }, { "epoch": 4.87, "learning_rate": 1.6515021200187106e-06, "loss": 0.2294, "step": 80682 }, { "epoch": 4.87, "learning_rate": 1.6499932099044863e-06, "loss": 0.366, "step": 80684 }, { "epoch": 4.87, "learning_rate": 1.6484842997902615e-06, "loss": 0.1327, "step": 80686 }, { "epoch": 4.87, "learning_rate": 1.6469753896760372e-06, "loss": 0.174, "step": 80688 }, { "epoch": 4.87, "learning_rate": 1.6454664795618124e-06, "loss": 0.212, "step": 80690 }, { "epoch": 4.87, "learning_rate": 1.643957569447588e-06, "loss": 0.2308, "step": 80692 }, { "epoch": 4.87, "learning_rate": 1.6424486593333635e-06, "loss": 0.2291, "step": 80694 }, { "epoch": 4.87, "learning_rate": 1.6409397492191392e-06, "loss": 0.1539, "step": 80696 }, { "epoch": 4.87, "learning_rate": 1.6394308391049144e-06, "loss": 0.2941, "step": 80698 }, { "epoch": 4.87, "learning_rate": 1.63792192899069e-06, "loss": 0.1697, "step": 80700 }, { "epoch": 4.87, "learning_rate": 1.6364130188764656e-06, "loss": 0.1973, "step": 80702 }, { "epoch": 4.87, "learning_rate": 1.6349041087622412e-06, "loss": 0.261, "step": 80704 }, { "epoch": 4.87, "learning_rate": 1.6333951986480165e-06, "loss": 0.2114, "step": 80706 }, { "epoch": 4.87, "learning_rate": 1.6318862885337921e-06, "loss": 0.2908, "step": 80708 }, { "epoch": 4.87, "learning_rate": 1.6303773784195676e-06, "loss": 0.2079, "step": 80710 }, { "epoch": 4.87, "learning_rate": 1.6288684683053433e-06, "loss": 0.1965, "step": 80712 }, { "epoch": 4.87, "learning_rate": 1.6273595581911185e-06, "loss": 0.2822, "step": 80714 }, { "epoch": 4.87, "learning_rate": 1.6258506480768942e-06, "loss": 0.1685, "step": 80716 }, { "epoch": 4.87, "learning_rate": 1.6243417379626694e-06, "loss": 0.2554, "step": 80718 }, { "epoch": 4.87, "learning_rate": 1.622832827848445e-06, "loss": 0.1392, "step": 80720 }, { "epoch": 4.87, "learning_rate": 1.6213239177342205e-06, "loss": 0.2048, "step": 80722 }, { "epoch": 4.87, "learning_rate": 1.6198150076199962e-06, "loss": 0.1814, "step": 80724 }, { "epoch": 4.87, "learning_rate": 1.6183060975057714e-06, "loss": 0.1659, "step": 80726 }, { "epoch": 4.87, "learning_rate": 1.6167971873915471e-06, "loss": 0.1658, "step": 80728 }, { "epoch": 4.87, "learning_rate": 1.6152882772773226e-06, "loss": 0.1528, "step": 80730 }, { "epoch": 4.87, "learning_rate": 1.6137793671630982e-06, "loss": 0.1615, "step": 80732 }, { "epoch": 4.87, "learning_rate": 1.612270457048874e-06, "loss": 0.277, "step": 80734 }, { "epoch": 4.87, "learning_rate": 1.6107615469346491e-06, "loss": 0.1859, "step": 80736 }, { "epoch": 4.87, "learning_rate": 1.6092526368204248e-06, "loss": 0.1355, "step": 80738 }, { "epoch": 4.87, "learning_rate": 1.6077437267062003e-06, "loss": 0.2231, "step": 80740 }, { "epoch": 4.87, "learning_rate": 1.606234816591976e-06, "loss": 0.2361, "step": 80742 }, { "epoch": 4.87, "learning_rate": 1.6047259064777512e-06, "loss": 0.2256, "step": 80744 }, { "epoch": 4.87, "learning_rate": 1.6032169963635268e-06, "loss": 0.2146, "step": 80746 }, { "epoch": 4.87, "learning_rate": 1.601708086249302e-06, "loss": 0.1884, "step": 80748 }, { "epoch": 4.87, "learning_rate": 1.6001991761350777e-06, "loss": 0.1242, "step": 80750 }, { "epoch": 4.87, "learning_rate": 1.5986902660208532e-06, "loss": 0.321, "step": 80752 }, { "epoch": 4.87, "learning_rate": 1.5971813559066289e-06, "loss": 0.2166, "step": 80754 }, { "epoch": 4.87, "learning_rate": 1.5956724457924041e-06, "loss": 0.3976, "step": 80756 }, { "epoch": 4.87, "learning_rate": 1.5941635356781798e-06, "loss": 0.1262, "step": 80758 }, { "epoch": 4.87, "learning_rate": 1.5926546255639552e-06, "loss": 0.1476, "step": 80760 }, { "epoch": 4.87, "learning_rate": 1.591145715449731e-06, "loss": 0.2894, "step": 80762 }, { "epoch": 4.87, "learning_rate": 1.5896368053355061e-06, "loss": 0.2463, "step": 80764 }, { "epoch": 4.87, "learning_rate": 1.5881278952212818e-06, "loss": 0.2103, "step": 80766 }, { "epoch": 4.87, "learning_rate": 1.5866189851070573e-06, "loss": 0.1788, "step": 80768 }, { "epoch": 4.88, "learning_rate": 1.585110074992833e-06, "loss": 0.1945, "step": 80770 }, { "epoch": 4.88, "learning_rate": 1.5836011648786082e-06, "loss": 0.1807, "step": 80772 }, { "epoch": 4.88, "learning_rate": 1.5820922547643838e-06, "loss": 0.3491, "step": 80774 }, { "epoch": 4.88, "learning_rate": 1.580583344650159e-06, "loss": 0.1995, "step": 80776 }, { "epoch": 4.88, "learning_rate": 1.5790744345359347e-06, "loss": 0.2582, "step": 80778 }, { "epoch": 4.88, "learning_rate": 1.5775655244217102e-06, "loss": 0.1328, "step": 80780 }, { "epoch": 4.88, "learning_rate": 1.5760566143074859e-06, "loss": 0.188, "step": 80782 }, { "epoch": 4.88, "learning_rate": 1.5745477041932611e-06, "loss": 0.3437, "step": 80784 }, { "epoch": 4.88, "learning_rate": 1.5730387940790368e-06, "loss": 0.2478, "step": 80786 }, { "epoch": 4.88, "learning_rate": 1.5715298839648122e-06, "loss": 0.1593, "step": 80788 }, { "epoch": 4.88, "learning_rate": 1.570020973850588e-06, "loss": 0.2079, "step": 80790 }, { "epoch": 4.88, "learning_rate": 1.5685120637363631e-06, "loss": 0.1888, "step": 80792 }, { "epoch": 4.88, "learning_rate": 1.5670031536221388e-06, "loss": 0.2901, "step": 80794 }, { "epoch": 4.88, "learning_rate": 1.5654942435079143e-06, "loss": 0.1586, "step": 80796 }, { "epoch": 4.88, "learning_rate": 1.56398533339369e-06, "loss": 0.2724, "step": 80798 }, { "epoch": 4.88, "learning_rate": 1.5624764232794654e-06, "loss": 0.2553, "step": 80800 }, { "epoch": 4.88, "learning_rate": 1.5609675131652408e-06, "loss": 0.2815, "step": 80802 }, { "epoch": 4.88, "learning_rate": 1.5594586030510163e-06, "loss": 0.1973, "step": 80804 }, { "epoch": 4.88, "learning_rate": 1.5579496929367918e-06, "loss": 0.1764, "step": 80806 }, { "epoch": 4.88, "learning_rate": 1.5564407828225674e-06, "loss": 0.2802, "step": 80808 }, { "epoch": 4.88, "learning_rate": 1.5549318727083429e-06, "loss": 0.1922, "step": 80810 }, { "epoch": 4.88, "learning_rate": 1.5534229625941183e-06, "loss": 0.2722, "step": 80812 }, { "epoch": 4.88, "learning_rate": 1.5519140524798938e-06, "loss": 0.2665, "step": 80814 }, { "epoch": 4.88, "learning_rate": 1.5504051423656692e-06, "loss": 0.2784, "step": 80816 }, { "epoch": 4.88, "learning_rate": 1.548896232251445e-06, "loss": 0.2481, "step": 80818 }, { "epoch": 4.88, "learning_rate": 1.5473873221372204e-06, "loss": 0.3224, "step": 80820 }, { "epoch": 4.88, "learning_rate": 1.5458784120229958e-06, "loss": 0.2692, "step": 80822 }, { "epoch": 4.88, "learning_rate": 1.5443695019087713e-06, "loss": 0.1068, "step": 80824 }, { "epoch": 4.88, "learning_rate": 1.542860591794547e-06, "loss": 0.2035, "step": 80826 }, { "epoch": 4.88, "learning_rate": 1.5413516816803224e-06, "loss": 0.2458, "step": 80828 }, { "epoch": 4.88, "learning_rate": 1.5398427715660978e-06, "loss": 0.2836, "step": 80830 }, { "epoch": 4.88, "learning_rate": 1.5383338614518733e-06, "loss": 0.2625, "step": 80832 }, { "epoch": 4.88, "learning_rate": 1.5368249513376488e-06, "loss": 0.2164, "step": 80834 }, { "epoch": 4.88, "learning_rate": 1.5353160412234244e-06, "loss": 0.1043, "step": 80836 }, { "epoch": 4.88, "learning_rate": 1.5338071311091999e-06, "loss": 0.1977, "step": 80838 }, { "epoch": 4.88, "learning_rate": 1.5322982209949753e-06, "loss": 0.1757, "step": 80840 }, { "epoch": 4.88, "learning_rate": 1.5307893108807508e-06, "loss": 0.2038, "step": 80842 }, { "epoch": 4.88, "learning_rate": 1.5292804007665262e-06, "loss": 0.189, "step": 80844 }, { "epoch": 4.88, "learning_rate": 1.527771490652302e-06, "loss": 0.2058, "step": 80846 }, { "epoch": 4.88, "learning_rate": 1.5262625805380774e-06, "loss": 0.1893, "step": 80848 }, { "epoch": 4.88, "learning_rate": 1.524753670423853e-06, "loss": 0.1974, "step": 80850 }, { "epoch": 4.88, "learning_rate": 1.5232447603096285e-06, "loss": 0.2107, "step": 80852 }, { "epoch": 4.88, "learning_rate": 1.521735850195404e-06, "loss": 0.1405, "step": 80854 }, { "epoch": 4.88, "learning_rate": 1.5202269400811796e-06, "loss": 0.2941, "step": 80856 }, { "epoch": 4.88, "learning_rate": 1.518718029966955e-06, "loss": 0.2534, "step": 80858 }, { "epoch": 4.88, "learning_rate": 1.5172091198527305e-06, "loss": 0.2367, "step": 80860 }, { "epoch": 4.88, "learning_rate": 1.515700209738506e-06, "loss": 0.2506, "step": 80862 }, { "epoch": 4.88, "learning_rate": 1.5141912996242814e-06, "loss": 0.2442, "step": 80864 }, { "epoch": 4.88, "learning_rate": 1.512682389510057e-06, "loss": 0.1353, "step": 80866 }, { "epoch": 4.88, "learning_rate": 1.5111734793958325e-06, "loss": 0.1692, "step": 80868 }, { "epoch": 4.88, "learning_rate": 1.509664569281608e-06, "loss": 0.2323, "step": 80870 }, { "epoch": 4.88, "learning_rate": 1.5081556591673835e-06, "loss": 0.1907, "step": 80872 }, { "epoch": 4.88, "learning_rate": 1.506646749053159e-06, "loss": 0.1272, "step": 80874 }, { "epoch": 4.88, "learning_rate": 1.5051378389389346e-06, "loss": 0.2836, "step": 80876 }, { "epoch": 4.88, "learning_rate": 1.50362892882471e-06, "loss": 0.2039, "step": 80878 }, { "epoch": 4.88, "learning_rate": 1.5021200187104855e-06, "loss": 0.1323, "step": 80880 }, { "epoch": 4.88, "learning_rate": 1.500611108596261e-06, "loss": 0.1996, "step": 80882 }, { "epoch": 4.88, "learning_rate": 1.4991021984820366e-06, "loss": 0.1169, "step": 80884 }, { "epoch": 4.88, "learning_rate": 1.497593288367812e-06, "loss": 0.1454, "step": 80886 }, { "epoch": 4.88, "learning_rate": 1.4960843782535875e-06, "loss": 0.1445, "step": 80888 }, { "epoch": 4.88, "learning_rate": 1.494575468139363e-06, "loss": 0.211, "step": 80890 }, { "epoch": 4.88, "learning_rate": 1.4930665580251384e-06, "loss": 0.2946, "step": 80892 }, { "epoch": 4.88, "learning_rate": 1.491557647910914e-06, "loss": 0.129, "step": 80894 }, { "epoch": 4.88, "learning_rate": 1.4900487377966895e-06, "loss": 0.1828, "step": 80896 }, { "epoch": 4.88, "learning_rate": 1.488539827682465e-06, "loss": 0.2686, "step": 80898 }, { "epoch": 4.88, "learning_rate": 1.4870309175682405e-06, "loss": 0.2099, "step": 80900 }, { "epoch": 4.88, "learning_rate": 1.485522007454016e-06, "loss": 0.1887, "step": 80902 }, { "epoch": 4.88, "learning_rate": 1.4840130973397916e-06, "loss": 0.1832, "step": 80904 }, { "epoch": 4.88, "learning_rate": 1.482504187225567e-06, "loss": 0.2173, "step": 80906 }, { "epoch": 4.88, "learning_rate": 1.4809952771113425e-06, "loss": 0.1723, "step": 80908 }, { "epoch": 4.88, "learning_rate": 1.479486366997118e-06, "loss": 0.2039, "step": 80910 }, { "epoch": 4.88, "learning_rate": 1.4779774568828936e-06, "loss": 0.1868, "step": 80912 }, { "epoch": 4.88, "learning_rate": 1.476468546768669e-06, "loss": 0.2073, "step": 80914 }, { "epoch": 4.88, "learning_rate": 1.4749596366544445e-06, "loss": 0.1737, "step": 80916 }, { "epoch": 4.88, "learning_rate": 1.47345072654022e-06, "loss": 0.1791, "step": 80918 }, { "epoch": 4.88, "learning_rate": 1.4719418164259954e-06, "loss": 0.2254, "step": 80920 }, { "epoch": 4.88, "learning_rate": 1.470432906311771e-06, "loss": 0.2275, "step": 80922 }, { "epoch": 4.88, "learning_rate": 1.4689239961975465e-06, "loss": 0.3255, "step": 80924 }, { "epoch": 4.88, "learning_rate": 1.467415086083322e-06, "loss": 0.2126, "step": 80926 }, { "epoch": 4.88, "learning_rate": 1.4659061759690977e-06, "loss": 0.1835, "step": 80928 }, { "epoch": 4.88, "learning_rate": 1.4643972658548731e-06, "loss": 0.2891, "step": 80930 }, { "epoch": 4.88, "learning_rate": 1.4628883557406486e-06, "loss": 0.087, "step": 80932 }, { "epoch": 4.88, "learning_rate": 1.4613794456264242e-06, "loss": 0.2084, "step": 80934 }, { "epoch": 4.89, "learning_rate": 1.4598705355121997e-06, "loss": 0.1831, "step": 80936 }, { "epoch": 4.89, "learning_rate": 1.4583616253979752e-06, "loss": 0.2702, "step": 80938 }, { "epoch": 4.89, "learning_rate": 1.4568527152837506e-06, "loss": 0.3482, "step": 80940 }, { "epoch": 4.89, "learning_rate": 1.4553438051695263e-06, "loss": 0.1726, "step": 80942 }, { "epoch": 4.89, "learning_rate": 1.4538348950553017e-06, "loss": 0.1818, "step": 80944 }, { "epoch": 4.89, "learning_rate": 1.4523259849410772e-06, "loss": 0.2746, "step": 80946 }, { "epoch": 4.89, "learning_rate": 1.4508170748268526e-06, "loss": 0.3142, "step": 80948 }, { "epoch": 4.89, "learning_rate": 1.449308164712628e-06, "loss": 0.1865, "step": 80950 }, { "epoch": 4.89, "learning_rate": 1.4477992545984038e-06, "loss": 0.1359, "step": 80952 }, { "epoch": 4.89, "learning_rate": 1.4462903444841792e-06, "loss": 0.128, "step": 80954 }, { "epoch": 4.89, "learning_rate": 1.4447814343699547e-06, "loss": 0.2401, "step": 80956 }, { "epoch": 4.89, "learning_rate": 1.4432725242557301e-06, "loss": 0.1978, "step": 80958 }, { "epoch": 4.89, "learning_rate": 1.4417636141415056e-06, "loss": 0.2117, "step": 80960 }, { "epoch": 4.89, "learning_rate": 1.4402547040272813e-06, "loss": 0.2047, "step": 80962 }, { "epoch": 4.89, "learning_rate": 1.4387457939130567e-06, "loss": 0.1629, "step": 80964 }, { "epoch": 4.89, "learning_rate": 1.4372368837988322e-06, "loss": 0.1595, "step": 80966 }, { "epoch": 4.89, "learning_rate": 1.4357279736846076e-06, "loss": 0.1276, "step": 80968 }, { "epoch": 4.89, "learning_rate": 1.434219063570383e-06, "loss": 0.2052, "step": 80970 }, { "epoch": 4.89, "learning_rate": 1.4327101534561587e-06, "loss": 0.2482, "step": 80972 }, { "epoch": 4.89, "learning_rate": 1.4312012433419342e-06, "loss": 0.2419, "step": 80974 }, { "epoch": 4.89, "learning_rate": 1.4296923332277096e-06, "loss": 0.3961, "step": 80976 }, { "epoch": 4.89, "learning_rate": 1.428183423113485e-06, "loss": 0.3155, "step": 80978 }, { "epoch": 4.89, "learning_rate": 1.4266745129992608e-06, "loss": 0.2199, "step": 80980 }, { "epoch": 4.89, "learning_rate": 1.4251656028850362e-06, "loss": 0.1932, "step": 80982 }, { "epoch": 4.89, "learning_rate": 1.4236566927708117e-06, "loss": 0.2824, "step": 80984 }, { "epoch": 4.89, "learning_rate": 1.4221477826565871e-06, "loss": 0.2473, "step": 80986 }, { "epoch": 4.89, "learning_rate": 1.4206388725423626e-06, "loss": 0.178, "step": 80988 }, { "epoch": 4.89, "learning_rate": 1.4191299624281383e-06, "loss": 0.1115, "step": 80990 }, { "epoch": 4.89, "learning_rate": 1.4176210523139137e-06, "loss": 0.2159, "step": 80992 }, { "epoch": 4.89, "learning_rate": 1.4161121421996892e-06, "loss": 0.1587, "step": 80994 }, { "epoch": 4.89, "learning_rate": 1.4146032320854646e-06, "loss": 0.2504, "step": 80996 }, { "epoch": 4.89, "learning_rate": 1.41309432197124e-06, "loss": 0.2613, "step": 80998 }, { "epoch": 4.89, "learning_rate": 1.4115854118570157e-06, "loss": 0.1866, "step": 81000 }, { "epoch": 4.89, "learning_rate": 1.4100765017427912e-06, "loss": 0.2193, "step": 81002 }, { "epoch": 4.89, "learning_rate": 1.4085675916285666e-06, "loss": 0.1752, "step": 81004 }, { "epoch": 4.89, "learning_rate": 1.4070586815143423e-06, "loss": 0.173, "step": 81006 }, { "epoch": 4.89, "learning_rate": 1.4055497714001178e-06, "loss": 0.2372, "step": 81008 }, { "epoch": 4.89, "learning_rate": 1.4040408612858934e-06, "loss": 0.1855, "step": 81010 }, { "epoch": 4.89, "learning_rate": 1.4025319511716689e-06, "loss": 0.2946, "step": 81012 }, { "epoch": 4.89, "learning_rate": 1.4010230410574443e-06, "loss": 0.1723, "step": 81014 }, { "epoch": 4.89, "learning_rate": 1.3995141309432198e-06, "loss": 0.3, "step": 81016 }, { "epoch": 4.89, "learning_rate": 1.3980052208289953e-06, "loss": 0.2172, "step": 81018 }, { "epoch": 4.89, "learning_rate": 1.396496310714771e-06, "loss": 0.1131, "step": 81020 }, { "epoch": 4.89, "learning_rate": 1.3949874006005464e-06, "loss": 0.1999, "step": 81022 }, { "epoch": 4.89, "learning_rate": 1.3934784904863218e-06, "loss": 0.1821, "step": 81024 }, { "epoch": 4.89, "learning_rate": 1.3919695803720973e-06, "loss": 0.2911, "step": 81026 }, { "epoch": 4.89, "learning_rate": 1.3904606702578727e-06, "loss": 0.3039, "step": 81028 }, { "epoch": 4.89, "learning_rate": 1.3889517601436484e-06, "loss": 0.2556, "step": 81030 }, { "epoch": 4.89, "learning_rate": 1.3874428500294239e-06, "loss": 0.2594, "step": 81032 }, { "epoch": 4.89, "learning_rate": 1.3859339399151993e-06, "loss": 0.2655, "step": 81034 }, { "epoch": 4.89, "learning_rate": 1.3844250298009748e-06, "loss": 0.2619, "step": 81036 }, { "epoch": 4.89, "learning_rate": 1.3829161196867504e-06, "loss": 0.241, "step": 81038 }, { "epoch": 4.89, "learning_rate": 1.381407209572526e-06, "loss": 0.1929, "step": 81040 }, { "epoch": 4.89, "learning_rate": 1.3798982994583013e-06, "loss": 0.2169, "step": 81042 }, { "epoch": 4.89, "learning_rate": 1.3783893893440768e-06, "loss": 0.1833, "step": 81044 }, { "epoch": 4.89, "learning_rate": 1.3768804792298523e-06, "loss": 0.205, "step": 81046 }, { "epoch": 4.89, "learning_rate": 1.375371569115628e-06, "loss": 0.2173, "step": 81048 }, { "epoch": 4.89, "learning_rate": 1.3738626590014034e-06, "loss": 0.1784, "step": 81050 }, { "epoch": 4.89, "learning_rate": 1.3723537488871788e-06, "loss": 0.1361, "step": 81052 }, { "epoch": 4.89, "learning_rate": 1.3708448387729543e-06, "loss": 0.2907, "step": 81054 }, { "epoch": 4.89, "learning_rate": 1.3693359286587297e-06, "loss": 0.141, "step": 81056 }, { "epoch": 4.89, "learning_rate": 1.3678270185445054e-06, "loss": 0.1713, "step": 81058 }, { "epoch": 4.89, "learning_rate": 1.3663181084302809e-06, "loss": 0.24, "step": 81060 }, { "epoch": 4.89, "learning_rate": 1.3648091983160563e-06, "loss": 0.2961, "step": 81062 }, { "epoch": 4.89, "learning_rate": 1.3633002882018318e-06, "loss": 0.1473, "step": 81064 }, { "epoch": 4.89, "learning_rate": 1.3617913780876074e-06, "loss": 0.1978, "step": 81066 }, { "epoch": 4.89, "learning_rate": 1.360282467973383e-06, "loss": 0.0921, "step": 81068 }, { "epoch": 4.89, "learning_rate": 1.3587735578591584e-06, "loss": 0.2835, "step": 81070 }, { "epoch": 4.89, "learning_rate": 1.3572646477449338e-06, "loss": 0.1843, "step": 81072 }, { "epoch": 4.89, "learning_rate": 1.3557557376307093e-06, "loss": 0.1971, "step": 81074 }, { "epoch": 4.89, "learning_rate": 1.354246827516485e-06, "loss": 0.1319, "step": 81076 }, { "epoch": 4.89, "learning_rate": 1.3527379174022604e-06, "loss": 0.2072, "step": 81078 }, { "epoch": 4.89, "learning_rate": 1.3512290072880358e-06, "loss": 0.2256, "step": 81080 }, { "epoch": 4.89, "learning_rate": 1.3497200971738113e-06, "loss": 0.1767, "step": 81082 }, { "epoch": 4.89, "learning_rate": 1.348211187059587e-06, "loss": 0.21, "step": 81084 }, { "epoch": 4.89, "learning_rate": 1.3467022769453624e-06, "loss": 0.1755, "step": 81086 }, { "epoch": 4.89, "learning_rate": 1.345193366831138e-06, "loss": 0.2658, "step": 81088 }, { "epoch": 4.89, "learning_rate": 1.3436844567169135e-06, "loss": 0.1376, "step": 81090 }, { "epoch": 4.89, "learning_rate": 1.342175546602689e-06, "loss": 0.2416, "step": 81092 }, { "epoch": 4.89, "learning_rate": 1.3406666364884644e-06, "loss": 0.1974, "step": 81094 }, { "epoch": 4.89, "learning_rate": 1.3391577263742401e-06, "loss": 0.1638, "step": 81096 }, { "epoch": 4.89, "learning_rate": 1.3376488162600156e-06, "loss": 0.2081, "step": 81098 }, { "epoch": 4.89, "learning_rate": 1.336139906145791e-06, "loss": 0.2826, "step": 81100 }, { "epoch": 4.9, "learning_rate": 1.3346309960315665e-06, "loss": 0.2487, "step": 81102 }, { "epoch": 4.9, "learning_rate": 1.333122085917342e-06, "loss": 0.1621, "step": 81104 }, { "epoch": 4.9, "learning_rate": 1.3316131758031176e-06, "loss": 0.1703, "step": 81106 }, { "epoch": 4.9, "learning_rate": 1.330104265688893e-06, "loss": 0.1981, "step": 81108 }, { "epoch": 4.9, "learning_rate": 1.3285953555746685e-06, "loss": 0.2226, "step": 81110 }, { "epoch": 4.9, "learning_rate": 1.327086445460444e-06, "loss": 0.2531, "step": 81112 }, { "epoch": 4.9, "learning_rate": 1.3255775353462194e-06, "loss": 0.2531, "step": 81114 }, { "epoch": 4.9, "learning_rate": 1.324068625231995e-06, "loss": 0.1402, "step": 81116 }, { "epoch": 4.9, "learning_rate": 1.3225597151177705e-06, "loss": 0.1992, "step": 81118 }, { "epoch": 4.9, "learning_rate": 1.321050805003546e-06, "loss": 0.1836, "step": 81120 }, { "epoch": 4.9, "learning_rate": 1.3195418948893214e-06, "loss": 0.16, "step": 81122 }, { "epoch": 4.9, "learning_rate": 1.318032984775097e-06, "loss": 0.1894, "step": 81124 }, { "epoch": 4.9, "learning_rate": 1.3165240746608726e-06, "loss": 0.1962, "step": 81126 }, { "epoch": 4.9, "learning_rate": 1.315015164546648e-06, "loss": 0.1409, "step": 81128 }, { "epoch": 4.9, "learning_rate": 1.3135062544324235e-06, "loss": 0.242, "step": 81130 }, { "epoch": 4.9, "learning_rate": 1.311997344318199e-06, "loss": 0.2092, "step": 81132 }, { "epoch": 4.9, "learning_rate": 1.3104884342039746e-06, "loss": 0.3247, "step": 81134 }, { "epoch": 4.9, "learning_rate": 1.30897952408975e-06, "loss": 0.1774, "step": 81136 }, { "epoch": 4.9, "learning_rate": 1.3074706139755255e-06, "loss": 0.2152, "step": 81138 }, { "epoch": 4.9, "learning_rate": 1.305961703861301e-06, "loss": 0.229, "step": 81140 }, { "epoch": 4.9, "learning_rate": 1.3044527937470764e-06, "loss": 0.2191, "step": 81142 }, { "epoch": 4.9, "learning_rate": 1.302943883632852e-06, "loss": 0.2845, "step": 81144 }, { "epoch": 4.9, "learning_rate": 1.3014349735186275e-06, "loss": 0.3612, "step": 81146 }, { "epoch": 4.9, "learning_rate": 1.299926063404403e-06, "loss": 0.2156, "step": 81148 }, { "epoch": 4.9, "learning_rate": 1.2984171532901785e-06, "loss": 0.2999, "step": 81150 }, { "epoch": 4.9, "learning_rate": 1.296908243175954e-06, "loss": 0.2066, "step": 81152 }, { "epoch": 4.9, "learning_rate": 1.2953993330617296e-06, "loss": 0.2043, "step": 81154 }, { "epoch": 4.9, "learning_rate": 1.293890422947505e-06, "loss": 0.1942, "step": 81156 }, { "epoch": 4.9, "learning_rate": 1.2923815128332805e-06, "loss": 0.2115, "step": 81158 }, { "epoch": 4.9, "learning_rate": 1.290872602719056e-06, "loss": 0.2569, "step": 81160 }, { "epoch": 4.9, "learning_rate": 1.2893636926048316e-06, "loss": 0.2807, "step": 81162 }, { "epoch": 4.9, "learning_rate": 1.2878547824906073e-06, "loss": 0.1506, "step": 81164 }, { "epoch": 4.9, "learning_rate": 1.2863458723763827e-06, "loss": 0.1749, "step": 81166 }, { "epoch": 4.9, "learning_rate": 1.2848369622621582e-06, "loss": 0.2125, "step": 81168 }, { "epoch": 4.9, "learning_rate": 1.2833280521479336e-06, "loss": 0.2288, "step": 81170 }, { "epoch": 4.9, "learning_rate": 1.281819142033709e-06, "loss": 0.1855, "step": 81172 }, { "epoch": 4.9, "learning_rate": 1.2803102319194848e-06, "loss": 0.2061, "step": 81174 }, { "epoch": 4.9, "learning_rate": 1.2788013218052602e-06, "loss": 0.3132, "step": 81176 }, { "epoch": 4.9, "learning_rate": 1.2772924116910357e-06, "loss": 0.1887, "step": 81178 }, { "epoch": 4.9, "learning_rate": 1.2757835015768111e-06, "loss": 0.175, "step": 81180 }, { "epoch": 4.9, "learning_rate": 1.2742745914625866e-06, "loss": 0.1992, "step": 81182 }, { "epoch": 4.9, "learning_rate": 1.2727656813483622e-06, "loss": 0.2315, "step": 81184 }, { "epoch": 4.9, "learning_rate": 1.2712567712341377e-06, "loss": 0.151, "step": 81186 }, { "epoch": 4.9, "learning_rate": 1.2697478611199132e-06, "loss": 0.239, "step": 81188 }, { "epoch": 4.9, "learning_rate": 1.2682389510056886e-06, "loss": 0.1137, "step": 81190 }, { "epoch": 4.9, "learning_rate": 1.2667300408914643e-06, "loss": 0.1773, "step": 81192 }, { "epoch": 4.9, "learning_rate": 1.2652211307772397e-06, "loss": 0.2902, "step": 81194 }, { "epoch": 4.9, "learning_rate": 1.2637122206630152e-06, "loss": 0.1601, "step": 81196 }, { "epoch": 4.9, "learning_rate": 1.2622033105487906e-06, "loss": 0.1401, "step": 81198 }, { "epoch": 4.9, "learning_rate": 1.260694400434566e-06, "loss": 0.1954, "step": 81200 }, { "epoch": 4.9, "learning_rate": 1.2591854903203418e-06, "loss": 0.2206, "step": 81202 }, { "epoch": 4.9, "learning_rate": 1.2576765802061172e-06, "loss": 0.1949, "step": 81204 }, { "epoch": 4.9, "learning_rate": 1.2561676700918927e-06, "loss": 0.2051, "step": 81206 }, { "epoch": 4.9, "learning_rate": 1.2546587599776681e-06, "loss": 0.1694, "step": 81208 }, { "epoch": 4.9, "learning_rate": 1.2531498498634436e-06, "loss": 0.2294, "step": 81210 }, { "epoch": 4.9, "learning_rate": 1.2516409397492192e-06, "loss": 0.2187, "step": 81212 }, { "epoch": 4.9, "learning_rate": 1.2501320296349947e-06, "loss": 0.3218, "step": 81214 }, { "epoch": 4.9, "learning_rate": 1.2486231195207702e-06, "loss": 0.3339, "step": 81216 }, { "epoch": 4.9, "learning_rate": 1.2471142094065456e-06, "loss": 0.2845, "step": 81218 }, { "epoch": 4.9, "learning_rate": 1.2456052992923213e-06, "loss": 0.3946, "step": 81220 }, { "epoch": 4.9, "learning_rate": 1.2440963891780967e-06, "loss": 0.1965, "step": 81222 }, { "epoch": 4.9, "learning_rate": 1.2425874790638722e-06, "loss": 0.1819, "step": 81224 }, { "epoch": 4.9, "learning_rate": 1.2410785689496476e-06, "loss": 0.1448, "step": 81226 }, { "epoch": 4.9, "learning_rate": 1.239569658835423e-06, "loss": 0.1964, "step": 81228 }, { "epoch": 4.9, "learning_rate": 1.2380607487211988e-06, "loss": 0.1643, "step": 81230 }, { "epoch": 4.9, "learning_rate": 1.2365518386069742e-06, "loss": 0.1895, "step": 81232 }, { "epoch": 4.9, "learning_rate": 1.2350429284927497e-06, "loss": 0.1897, "step": 81234 }, { "epoch": 4.9, "learning_rate": 1.2335340183785251e-06, "loss": 0.1499, "step": 81236 }, { "epoch": 4.9, "learning_rate": 1.2320251082643006e-06, "loss": 0.2434, "step": 81238 }, { "epoch": 4.9, "learning_rate": 1.2305161981500762e-06, "loss": 0.2383, "step": 81240 }, { "epoch": 4.9, "learning_rate": 1.229007288035852e-06, "loss": 0.1593, "step": 81242 }, { "epoch": 4.9, "learning_rate": 1.2274983779216274e-06, "loss": 0.2139, "step": 81244 }, { "epoch": 4.9, "learning_rate": 1.2259894678074028e-06, "loss": 0.2108, "step": 81246 }, { "epoch": 4.9, "learning_rate": 1.2244805576931783e-06, "loss": 0.1421, "step": 81248 }, { "epoch": 4.9, "learning_rate": 1.222971647578954e-06, "loss": 0.1802, "step": 81250 }, { "epoch": 4.9, "learning_rate": 1.2214627374647294e-06, "loss": 0.2352, "step": 81252 }, { "epoch": 4.9, "learning_rate": 1.2199538273505049e-06, "loss": 0.162, "step": 81254 }, { "epoch": 4.9, "learning_rate": 1.2184449172362803e-06, "loss": 0.2176, "step": 81256 }, { "epoch": 4.9, "learning_rate": 1.2169360071220558e-06, "loss": 0.3405, "step": 81258 }, { "epoch": 4.9, "learning_rate": 1.2154270970078314e-06, "loss": 0.2431, "step": 81260 }, { "epoch": 4.9, "learning_rate": 1.2139181868936069e-06, "loss": 0.1432, "step": 81262 }, { "epoch": 4.9, "learning_rate": 1.2124092767793823e-06, "loss": 0.1797, "step": 81264 }, { "epoch": 4.9, "learning_rate": 1.2109003666651578e-06, "loss": 0.1651, "step": 81266 }, { "epoch": 4.91, "learning_rate": 1.2093914565509333e-06, "loss": 0.1366, "step": 81268 }, { "epoch": 4.91, "learning_rate": 1.207882546436709e-06, "loss": 0.1581, "step": 81270 }, { "epoch": 4.91, "learning_rate": 1.2063736363224844e-06, "loss": 0.2029, "step": 81272 }, { "epoch": 4.91, "learning_rate": 1.2048647262082598e-06, "loss": 0.2219, "step": 81274 }, { "epoch": 4.91, "learning_rate": 1.2033558160940353e-06, "loss": 0.1971, "step": 81276 }, { "epoch": 4.91, "learning_rate": 1.2018469059798107e-06, "loss": 0.2244, "step": 81278 }, { "epoch": 4.91, "learning_rate": 1.2003379958655864e-06, "loss": 0.1649, "step": 81280 }, { "epoch": 4.91, "learning_rate": 1.1988290857513619e-06, "loss": 0.2517, "step": 81282 }, { "epoch": 4.91, "learning_rate": 1.1973201756371373e-06, "loss": 0.1671, "step": 81284 }, { "epoch": 4.91, "learning_rate": 1.1958112655229128e-06, "loss": 0.1865, "step": 81286 }, { "epoch": 4.91, "learning_rate": 1.1943023554086884e-06, "loss": 0.2214, "step": 81288 }, { "epoch": 4.91, "learning_rate": 1.1927934452944639e-06, "loss": 0.1844, "step": 81290 }, { "epoch": 4.91, "learning_rate": 1.1912845351802393e-06, "loss": 0.3315, "step": 81292 }, { "epoch": 4.91, "learning_rate": 1.1897756250660148e-06, "loss": 0.1939, "step": 81294 }, { "epoch": 4.91, "learning_rate": 1.1882667149517903e-06, "loss": 0.1258, "step": 81296 }, { "epoch": 4.91, "learning_rate": 1.186757804837566e-06, "loss": 0.2042, "step": 81298 }, { "epoch": 4.91, "learning_rate": 1.1852488947233414e-06, "loss": 0.209, "step": 81300 }, { "epoch": 4.91, "learning_rate": 1.1837399846091168e-06, "loss": 0.2063, "step": 81302 }, { "epoch": 4.91, "learning_rate": 1.1822310744948923e-06, "loss": 0.3496, "step": 81304 }, { "epoch": 4.91, "learning_rate": 1.1807221643806677e-06, "loss": 0.2991, "step": 81306 }, { "epoch": 4.91, "learning_rate": 1.1792132542664434e-06, "loss": 0.2583, "step": 81308 }, { "epoch": 4.91, "learning_rate": 1.1777043441522189e-06, "loss": 0.2893, "step": 81310 }, { "epoch": 4.91, "learning_rate": 1.1761954340379943e-06, "loss": 0.2439, "step": 81312 }, { "epoch": 4.91, "learning_rate": 1.1746865239237698e-06, "loss": 0.1763, "step": 81314 }, { "epoch": 4.91, "learning_rate": 1.1731776138095454e-06, "loss": 0.2324, "step": 81316 }, { "epoch": 4.91, "learning_rate": 1.171668703695321e-06, "loss": 0.1289, "step": 81318 }, { "epoch": 4.91, "learning_rate": 1.1701597935810966e-06, "loss": 0.2859, "step": 81320 }, { "epoch": 4.91, "learning_rate": 1.168650883466872e-06, "loss": 0.1916, "step": 81322 }, { "epoch": 4.91, "learning_rate": 1.1671419733526475e-06, "loss": 0.2056, "step": 81324 }, { "epoch": 4.91, "learning_rate": 1.165633063238423e-06, "loss": 0.2842, "step": 81326 }, { "epoch": 4.91, "learning_rate": 1.1641241531241986e-06, "loss": 0.2033, "step": 81328 }, { "epoch": 4.91, "learning_rate": 1.162615243009974e-06, "loss": 0.2697, "step": 81330 }, { "epoch": 4.91, "learning_rate": 1.1611063328957495e-06, "loss": 0.1277, "step": 81332 }, { "epoch": 4.91, "learning_rate": 1.159597422781525e-06, "loss": 0.2462, "step": 81334 }, { "epoch": 4.91, "learning_rate": 1.1580885126673004e-06, "loss": 0.1749, "step": 81336 }, { "epoch": 4.91, "learning_rate": 1.156579602553076e-06, "loss": 0.2286, "step": 81338 }, { "epoch": 4.91, "learning_rate": 1.1550706924388515e-06, "loss": 0.1718, "step": 81340 }, { "epoch": 4.91, "learning_rate": 1.153561782324627e-06, "loss": 0.2269, "step": 81342 }, { "epoch": 4.91, "learning_rate": 1.1520528722104024e-06, "loss": 0.234, "step": 81344 }, { "epoch": 4.91, "learning_rate": 1.1505439620961781e-06, "loss": 0.195, "step": 81346 }, { "epoch": 4.91, "learning_rate": 1.1490350519819536e-06, "loss": 0.1016, "step": 81348 }, { "epoch": 4.91, "learning_rate": 1.147526141867729e-06, "loss": 0.177, "step": 81350 }, { "epoch": 4.91, "learning_rate": 1.1460172317535045e-06, "loss": 0.1312, "step": 81352 }, { "epoch": 4.91, "learning_rate": 1.14450832163928e-06, "loss": 0.2222, "step": 81354 }, { "epoch": 4.91, "learning_rate": 1.1429994115250556e-06, "loss": 0.2974, "step": 81356 }, { "epoch": 4.91, "learning_rate": 1.141490501410831e-06, "loss": 0.3233, "step": 81358 }, { "epoch": 4.91, "learning_rate": 1.1399815912966065e-06, "loss": 0.2486, "step": 81360 }, { "epoch": 4.91, "learning_rate": 1.138472681182382e-06, "loss": 0.1783, "step": 81362 }, { "epoch": 4.91, "learning_rate": 1.1369637710681574e-06, "loss": 0.2649, "step": 81364 }, { "epoch": 4.91, "learning_rate": 1.135454860953933e-06, "loss": 0.2293, "step": 81366 }, { "epoch": 4.91, "learning_rate": 1.1339459508397085e-06, "loss": 0.2065, "step": 81368 }, { "epoch": 4.91, "learning_rate": 1.132437040725484e-06, "loss": 0.2025, "step": 81370 }, { "epoch": 4.91, "learning_rate": 1.1309281306112594e-06, "loss": 0.2879, "step": 81372 }, { "epoch": 4.91, "learning_rate": 1.1294192204970351e-06, "loss": 0.2074, "step": 81374 }, { "epoch": 4.91, "learning_rate": 1.1279103103828106e-06, "loss": 0.1555, "step": 81376 }, { "epoch": 4.91, "learning_rate": 1.126401400268586e-06, "loss": 0.1884, "step": 81378 }, { "epoch": 4.91, "learning_rate": 1.1248924901543615e-06, "loss": 0.1648, "step": 81380 }, { "epoch": 4.91, "learning_rate": 1.123383580040137e-06, "loss": 0.165, "step": 81382 }, { "epoch": 4.91, "learning_rate": 1.1218746699259126e-06, "loss": 0.2966, "step": 81384 }, { "epoch": 4.91, "learning_rate": 1.120365759811688e-06, "loss": 0.1697, "step": 81386 }, { "epoch": 4.91, "learning_rate": 1.1188568496974635e-06, "loss": 0.2156, "step": 81388 }, { "epoch": 4.91, "learning_rate": 1.117347939583239e-06, "loss": 0.152, "step": 81390 }, { "epoch": 4.91, "learning_rate": 1.1158390294690144e-06, "loss": 0.2136, "step": 81392 }, { "epoch": 4.91, "learning_rate": 1.11433011935479e-06, "loss": 0.2506, "step": 81394 }, { "epoch": 4.91, "learning_rate": 1.1128212092405657e-06, "loss": 0.2463, "step": 81396 }, { "epoch": 4.91, "learning_rate": 1.1113122991263412e-06, "loss": 0.1905, "step": 81398 }, { "epoch": 4.91, "learning_rate": 1.1098033890121167e-06, "loss": 0.2815, "step": 81400 }, { "epoch": 4.91, "learning_rate": 1.1082944788978921e-06, "loss": 0.2069, "step": 81402 }, { "epoch": 4.91, "learning_rate": 1.1067855687836678e-06, "loss": 0.1847, "step": 81404 }, { "epoch": 4.91, "learning_rate": 1.1052766586694432e-06, "loss": 0.2856, "step": 81406 }, { "epoch": 4.91, "learning_rate": 1.1037677485552187e-06, "loss": 0.3321, "step": 81408 }, { "epoch": 4.91, "learning_rate": 1.1022588384409941e-06, "loss": 0.1862, "step": 81410 }, { "epoch": 4.91, "learning_rate": 1.1007499283267696e-06, "loss": 0.1686, "step": 81412 }, { "epoch": 4.91, "learning_rate": 1.0992410182125453e-06, "loss": 0.1491, "step": 81414 }, { "epoch": 4.91, "learning_rate": 1.0977321080983207e-06, "loss": 0.1676, "step": 81416 }, { "epoch": 4.91, "learning_rate": 1.0962231979840962e-06, "loss": 0.16, "step": 81418 }, { "epoch": 4.91, "learning_rate": 1.0947142878698716e-06, "loss": 0.1971, "step": 81420 }, { "epoch": 4.91, "learning_rate": 1.093205377755647e-06, "loss": 0.1313, "step": 81422 }, { "epoch": 4.91, "learning_rate": 1.0916964676414228e-06, "loss": 0.2547, "step": 81424 }, { "epoch": 4.91, "learning_rate": 1.0901875575271982e-06, "loss": 0.3657, "step": 81426 }, { "epoch": 4.91, "learning_rate": 1.0886786474129737e-06, "loss": 0.1563, "step": 81428 }, { "epoch": 4.91, "learning_rate": 1.0871697372987491e-06, "loss": 0.2153, "step": 81430 }, { "epoch": 4.92, "learning_rate": 1.0856608271845246e-06, "loss": 0.3058, "step": 81432 }, { "epoch": 4.92, "learning_rate": 1.0841519170703002e-06, "loss": 0.1999, "step": 81434 }, { "epoch": 4.92, "learning_rate": 1.0826430069560757e-06, "loss": 0.265, "step": 81436 }, { "epoch": 4.92, "learning_rate": 1.0811340968418511e-06, "loss": 0.2012, "step": 81438 }, { "epoch": 4.92, "learning_rate": 1.0796251867276266e-06, "loss": 0.3428, "step": 81440 }, { "epoch": 4.92, "learning_rate": 1.0781162766134023e-06, "loss": 0.2122, "step": 81442 }, { "epoch": 4.92, "learning_rate": 1.0766073664991777e-06, "loss": 0.2507, "step": 81444 }, { "epoch": 4.92, "learning_rate": 1.0750984563849532e-06, "loss": 0.1829, "step": 81446 }, { "epoch": 4.92, "learning_rate": 1.0735895462707286e-06, "loss": 0.185, "step": 81448 }, { "epoch": 4.92, "learning_rate": 1.072080636156504e-06, "loss": 0.1677, "step": 81450 }, { "epoch": 4.92, "learning_rate": 1.0705717260422798e-06, "loss": 0.1473, "step": 81452 }, { "epoch": 4.92, "learning_rate": 1.0690628159280552e-06, "loss": 0.2141, "step": 81454 }, { "epoch": 4.92, "learning_rate": 1.0675539058138307e-06, "loss": 0.3071, "step": 81456 }, { "epoch": 4.92, "learning_rate": 1.0660449956996061e-06, "loss": 0.192, "step": 81458 }, { "epoch": 4.92, "learning_rate": 1.0645360855853816e-06, "loss": 0.2314, "step": 81460 }, { "epoch": 4.92, "learning_rate": 1.0630271754711572e-06, "loss": 0.1931, "step": 81462 }, { "epoch": 4.92, "learning_rate": 1.0615182653569327e-06, "loss": 0.2102, "step": 81464 }, { "epoch": 4.92, "learning_rate": 1.0600093552427081e-06, "loss": 0.1431, "step": 81466 }, { "epoch": 4.92, "learning_rate": 1.0585004451284836e-06, "loss": 0.1579, "step": 81468 }, { "epoch": 4.92, "learning_rate": 1.0569915350142593e-06, "loss": 0.259, "step": 81470 }, { "epoch": 4.92, "learning_rate": 1.055482624900035e-06, "loss": 0.2574, "step": 81472 }, { "epoch": 4.92, "learning_rate": 1.0539737147858104e-06, "loss": 0.2432, "step": 81474 }, { "epoch": 4.92, "learning_rate": 1.0524648046715858e-06, "loss": 0.1622, "step": 81476 }, { "epoch": 4.92, "learning_rate": 1.0509558945573613e-06, "loss": 0.192, "step": 81478 }, { "epoch": 4.92, "learning_rate": 1.0494469844431368e-06, "loss": 0.1758, "step": 81480 }, { "epoch": 4.92, "learning_rate": 1.0479380743289124e-06, "loss": 0.3338, "step": 81482 }, { "epoch": 4.92, "learning_rate": 1.0464291642146879e-06, "loss": 0.2324, "step": 81484 }, { "epoch": 4.92, "learning_rate": 1.0449202541004633e-06, "loss": 0.171, "step": 81486 }, { "epoch": 4.92, "learning_rate": 1.0434113439862388e-06, "loss": 0.2753, "step": 81488 }, { "epoch": 4.92, "learning_rate": 1.0419024338720142e-06, "loss": 0.2092, "step": 81490 }, { "epoch": 4.92, "learning_rate": 1.04039352375779e-06, "loss": 0.2056, "step": 81492 }, { "epoch": 4.92, "learning_rate": 1.0388846136435654e-06, "loss": 0.2004, "step": 81494 }, { "epoch": 4.92, "learning_rate": 1.0373757035293408e-06, "loss": 0.1763, "step": 81496 }, { "epoch": 4.92, "learning_rate": 1.0358667934151163e-06, "loss": 0.1948, "step": 81498 }, { "epoch": 4.92, "learning_rate": 1.034357883300892e-06, "loss": 0.2033, "step": 81500 }, { "epoch": 4.92, "learning_rate": 1.0328489731866674e-06, "loss": 0.213, "step": 81502 }, { "epoch": 4.92, "learning_rate": 1.0313400630724428e-06, "loss": 0.2626, "step": 81504 }, { "epoch": 4.92, "learning_rate": 1.0298311529582183e-06, "loss": 0.1504, "step": 81506 }, { "epoch": 4.92, "learning_rate": 1.0283222428439938e-06, "loss": 0.1808, "step": 81508 }, { "epoch": 4.92, "learning_rate": 1.0268133327297694e-06, "loss": 0.2665, "step": 81510 }, { "epoch": 4.92, "learning_rate": 1.0253044226155449e-06, "loss": 0.2401, "step": 81512 }, { "epoch": 4.92, "learning_rate": 1.0237955125013203e-06, "loss": 0.2435, "step": 81514 }, { "epoch": 4.92, "learning_rate": 1.0222866023870958e-06, "loss": 0.1987, "step": 81516 }, { "epoch": 4.92, "learning_rate": 1.0207776922728712e-06, "loss": 0.2834, "step": 81518 }, { "epoch": 4.92, "learning_rate": 1.019268782158647e-06, "loss": 0.1786, "step": 81520 }, { "epoch": 4.92, "learning_rate": 1.0177598720444224e-06, "loss": 0.2404, "step": 81522 }, { "epoch": 4.92, "learning_rate": 1.0162509619301978e-06, "loss": 0.2447, "step": 81524 }, { "epoch": 4.92, "learning_rate": 1.0147420518159733e-06, "loss": 0.2733, "step": 81526 }, { "epoch": 4.92, "learning_rate": 1.0132331417017487e-06, "loss": 0.3287, "step": 81528 }, { "epoch": 4.92, "learning_rate": 1.0117242315875244e-06, "loss": 0.1817, "step": 81530 }, { "epoch": 4.92, "learning_rate": 1.0102153214732999e-06, "loss": 0.2031, "step": 81532 }, { "epoch": 4.92, "learning_rate": 1.0087064113590753e-06, "loss": 0.1832, "step": 81534 }, { "epoch": 4.92, "learning_rate": 1.0071975012448508e-06, "loss": 0.3034, "step": 81536 }, { "epoch": 4.92, "learning_rate": 1.0056885911306264e-06, "loss": 0.1221, "step": 81538 }, { "epoch": 4.92, "learning_rate": 1.0041796810164019e-06, "loss": 0.1866, "step": 81540 }, { "epoch": 4.92, "learning_rate": 1.0026707709021773e-06, "loss": 0.2198, "step": 81542 }, { "epoch": 4.92, "learning_rate": 1.0011618607879528e-06, "loss": 0.2192, "step": 81544 }, { "epoch": 4.92, "learning_rate": 9.996529506737282e-07, "loss": 0.2355, "step": 81546 }, { "epoch": 4.92, "learning_rate": 9.98144040559504e-07, "loss": 0.2664, "step": 81548 }, { "epoch": 4.92, "learning_rate": 9.966351304452796e-07, "loss": 0.1719, "step": 81550 }, { "epoch": 4.92, "learning_rate": 9.95126220331055e-07, "loss": 0.2878, "step": 81552 }, { "epoch": 4.92, "learning_rate": 9.936173102168305e-07, "loss": 0.2686, "step": 81554 }, { "epoch": 4.92, "learning_rate": 9.92108400102606e-07, "loss": 0.3019, "step": 81556 }, { "epoch": 4.92, "learning_rate": 9.905994899883816e-07, "loss": 0.142, "step": 81558 }, { "epoch": 4.92, "learning_rate": 9.89090579874157e-07, "loss": 0.2311, "step": 81560 }, { "epoch": 4.92, "learning_rate": 9.875816697599325e-07, "loss": 0.2343, "step": 81562 }, { "epoch": 4.92, "learning_rate": 9.86072759645708e-07, "loss": 0.2336, "step": 81564 }, { "epoch": 4.92, "learning_rate": 9.845638495314834e-07, "loss": 0.247, "step": 81566 }, { "epoch": 4.92, "learning_rate": 9.83054939417259e-07, "loss": 0.1838, "step": 81568 }, { "epoch": 4.92, "learning_rate": 9.815460293030346e-07, "loss": 0.2463, "step": 81570 }, { "epoch": 4.92, "learning_rate": 9.8003711918881e-07, "loss": 0.1783, "step": 81572 }, { "epoch": 4.92, "learning_rate": 9.785282090745855e-07, "loss": 0.1094, "step": 81574 }, { "epoch": 4.92, "learning_rate": 9.77019298960361e-07, "loss": 0.2789, "step": 81576 }, { "epoch": 4.92, "learning_rate": 9.755103888461366e-07, "loss": 0.1845, "step": 81578 }, { "epoch": 4.92, "learning_rate": 9.74001478731912e-07, "loss": 0.1455, "step": 81580 }, { "epoch": 4.92, "learning_rate": 9.724925686176875e-07, "loss": 0.2568, "step": 81582 }, { "epoch": 4.92, "learning_rate": 9.70983658503463e-07, "loss": 0.3022, "step": 81584 }, { "epoch": 4.92, "learning_rate": 9.694747483892384e-07, "loss": 0.226, "step": 81586 }, { "epoch": 4.92, "learning_rate": 9.67965838275014e-07, "loss": 0.2499, "step": 81588 }, { "epoch": 4.92, "learning_rate": 9.664569281607895e-07, "loss": 0.2181, "step": 81590 }, { "epoch": 4.92, "learning_rate": 9.64948018046565e-07, "loss": 0.1798, "step": 81592 }, { "epoch": 4.92, "learning_rate": 9.634391079323404e-07, "loss": 0.2867, "step": 81594 }, { "epoch": 4.92, "learning_rate": 9.61930197818116e-07, "loss": 0.1439, "step": 81596 }, { "epoch": 4.93, "learning_rate": 9.604212877038916e-07, "loss": 0.2374, "step": 81598 }, { "epoch": 4.93, "learning_rate": 9.58912377589667e-07, "loss": 0.1824, "step": 81600 }, { "epoch": 4.93, "learning_rate": 9.574034674754425e-07, "loss": 0.2557, "step": 81602 }, { "epoch": 4.93, "learning_rate": 9.55894557361218e-07, "loss": 0.2676, "step": 81604 }, { "epoch": 4.93, "learning_rate": 9.543856472469936e-07, "loss": 0.1705, "step": 81606 }, { "epoch": 4.93, "learning_rate": 9.52876737132769e-07, "loss": 0.1511, "step": 81608 }, { "epoch": 4.93, "learning_rate": 9.513678270185445e-07, "loss": 0.2737, "step": 81610 }, { "epoch": 4.93, "learning_rate": 9.4985891690432e-07, "loss": 0.1897, "step": 81612 }, { "epoch": 4.93, "learning_rate": 9.483500067900955e-07, "loss": 0.1325, "step": 81614 }, { "epoch": 4.93, "learning_rate": 9.46841096675871e-07, "loss": 0.2091, "step": 81616 }, { "epoch": 4.93, "learning_rate": 9.453321865616465e-07, "loss": 0.1597, "step": 81618 }, { "epoch": 4.93, "learning_rate": 9.43823276447422e-07, "loss": 0.3077, "step": 81620 }, { "epoch": 4.93, "learning_rate": 9.423143663331975e-07, "loss": 0.2861, "step": 81622 }, { "epoch": 4.93, "learning_rate": 9.40805456218973e-07, "loss": 0.335, "step": 81624 }, { "epoch": 4.93, "learning_rate": 9.392965461047485e-07, "loss": 0.205, "step": 81626 }, { "epoch": 4.93, "learning_rate": 9.377876359905242e-07, "loss": 0.1813, "step": 81628 }, { "epoch": 4.93, "learning_rate": 9.362787258762997e-07, "loss": 0.1389, "step": 81630 }, { "epoch": 4.93, "learning_rate": 9.347698157620751e-07, "loss": 0.1373, "step": 81632 }, { "epoch": 4.93, "learning_rate": 9.332609056478507e-07, "loss": 0.2788, "step": 81634 }, { "epoch": 4.93, "learning_rate": 9.317519955336262e-07, "loss": 0.2705, "step": 81636 }, { "epoch": 4.93, "learning_rate": 9.302430854194017e-07, "loss": 0.1419, "step": 81638 }, { "epoch": 4.93, "learning_rate": 9.287341753051772e-07, "loss": 0.3641, "step": 81640 }, { "epoch": 4.93, "learning_rate": 9.272252651909526e-07, "loss": 0.1843, "step": 81642 }, { "epoch": 4.93, "learning_rate": 9.257163550767282e-07, "loss": 0.1894, "step": 81644 }, { "epoch": 4.93, "learning_rate": 9.242074449625036e-07, "loss": 0.2934, "step": 81646 }, { "epoch": 4.93, "learning_rate": 9.226985348482792e-07, "loss": 0.2016, "step": 81648 }, { "epoch": 4.93, "learning_rate": 9.211896247340547e-07, "loss": 0.1753, "step": 81650 }, { "epoch": 4.93, "learning_rate": 9.196807146198302e-07, "loss": 0.3258, "step": 81652 }, { "epoch": 4.93, "learning_rate": 9.181718045056057e-07, "loss": 0.0982, "step": 81654 }, { "epoch": 4.93, "learning_rate": 9.166628943913811e-07, "loss": 0.2978, "step": 81656 }, { "epoch": 4.93, "learning_rate": 9.151539842771567e-07, "loss": 0.2089, "step": 81658 }, { "epoch": 4.93, "learning_rate": 9.136450741629321e-07, "loss": 0.3133, "step": 81660 }, { "epoch": 4.93, "learning_rate": 9.121361640487077e-07, "loss": 0.2119, "step": 81662 }, { "epoch": 4.93, "learning_rate": 9.106272539344832e-07, "loss": 0.15, "step": 81664 }, { "epoch": 4.93, "learning_rate": 9.091183438202587e-07, "loss": 0.1638, "step": 81666 }, { "epoch": 4.93, "learning_rate": 9.076094337060342e-07, "loss": 0.3207, "step": 81668 }, { "epoch": 4.93, "learning_rate": 9.061005235918096e-07, "loss": 0.1235, "step": 81670 }, { "epoch": 4.93, "learning_rate": 9.045916134775852e-07, "loss": 0.1776, "step": 81672 }, { "epoch": 4.93, "learning_rate": 9.030827033633606e-07, "loss": 0.2377, "step": 81674 }, { "epoch": 4.93, "learning_rate": 9.015737932491362e-07, "loss": 0.1748, "step": 81676 }, { "epoch": 4.93, "learning_rate": 9.000648831349117e-07, "loss": 0.221, "step": 81678 }, { "epoch": 4.93, "learning_rate": 8.985559730206872e-07, "loss": 0.1754, "step": 81680 }, { "epoch": 4.93, "learning_rate": 8.970470629064627e-07, "loss": 0.2015, "step": 81682 }, { "epoch": 4.93, "learning_rate": 8.955381527922381e-07, "loss": 0.2706, "step": 81684 }, { "epoch": 4.93, "learning_rate": 8.940292426780137e-07, "loss": 0.2686, "step": 81686 }, { "epoch": 4.93, "learning_rate": 8.925203325637891e-07, "loss": 0.2011, "step": 81688 }, { "epoch": 4.93, "learning_rate": 8.910114224495647e-07, "loss": 0.2909, "step": 81690 }, { "epoch": 4.93, "learning_rate": 8.895025123353402e-07, "loss": 0.1451, "step": 81692 }, { "epoch": 4.93, "learning_rate": 8.879936022211157e-07, "loss": 0.1635, "step": 81694 }, { "epoch": 4.93, "learning_rate": 8.864846921068912e-07, "loss": 0.2234, "step": 81696 }, { "epoch": 4.93, "learning_rate": 8.849757819926666e-07, "loss": 0.2111, "step": 81698 }, { "epoch": 4.93, "learning_rate": 8.834668718784422e-07, "loss": 0.1691, "step": 81700 }, { "epoch": 4.93, "learning_rate": 8.819579617642176e-07, "loss": 0.2187, "step": 81702 }, { "epoch": 4.93, "learning_rate": 8.804490516499932e-07, "loss": 0.1896, "step": 81704 }, { "epoch": 4.93, "learning_rate": 8.789401415357689e-07, "loss": 0.2368, "step": 81706 }, { "epoch": 4.93, "learning_rate": 8.774312314215443e-07, "loss": 0.2113, "step": 81708 }, { "epoch": 4.93, "learning_rate": 8.759223213073199e-07, "loss": 0.2987, "step": 81710 }, { "epoch": 4.93, "learning_rate": 8.744134111930953e-07, "loss": 0.1816, "step": 81712 }, { "epoch": 4.93, "learning_rate": 8.729045010788708e-07, "loss": 0.1476, "step": 81714 }, { "epoch": 4.93, "learning_rate": 8.713955909646464e-07, "loss": 0.1548, "step": 81716 }, { "epoch": 4.93, "learning_rate": 8.698866808504218e-07, "loss": 0.2289, "step": 81718 }, { "epoch": 4.93, "learning_rate": 8.683777707361974e-07, "loss": 0.1797, "step": 81720 }, { "epoch": 4.93, "learning_rate": 8.668688606219728e-07, "loss": 0.1562, "step": 81722 }, { "epoch": 4.93, "learning_rate": 8.653599505077484e-07, "loss": 0.218, "step": 81724 }, { "epoch": 4.93, "learning_rate": 8.638510403935238e-07, "loss": 0.2238, "step": 81726 }, { "epoch": 4.93, "learning_rate": 8.623421302792993e-07, "loss": 0.2824, "step": 81728 }, { "epoch": 4.93, "learning_rate": 8.608332201650749e-07, "loss": 0.1884, "step": 81730 }, { "epoch": 4.93, "learning_rate": 8.593243100508503e-07, "loss": 0.3514, "step": 81732 }, { "epoch": 4.93, "learning_rate": 8.578153999366259e-07, "loss": 0.18, "step": 81734 }, { "epoch": 4.93, "learning_rate": 8.563064898224013e-07, "loss": 0.1803, "step": 81736 }, { "epoch": 4.93, "learning_rate": 8.547975797081769e-07, "loss": 0.2729, "step": 81738 }, { "epoch": 4.93, "learning_rate": 8.532886695939523e-07, "loss": 0.1458, "step": 81740 }, { "epoch": 4.93, "learning_rate": 8.517797594797278e-07, "loss": 0.212, "step": 81742 }, { "epoch": 4.93, "learning_rate": 8.502708493655034e-07, "loss": 0.2146, "step": 81744 }, { "epoch": 4.93, "learning_rate": 8.487619392512788e-07, "loss": 0.2816, "step": 81746 }, { "epoch": 4.93, "learning_rate": 8.472530291370544e-07, "loss": 0.2118, "step": 81748 }, { "epoch": 4.93, "learning_rate": 8.457441190228298e-07, "loss": 0.3667, "step": 81750 }, { "epoch": 4.93, "learning_rate": 8.442352089086054e-07, "loss": 0.193, "step": 81752 }, { "epoch": 4.93, "learning_rate": 8.427262987943808e-07, "loss": 0.1862, "step": 81754 }, { "epoch": 4.93, "learning_rate": 8.412173886801563e-07, "loss": 0.1364, "step": 81756 }, { "epoch": 4.93, "learning_rate": 8.397084785659319e-07, "loss": 0.1266, "step": 81758 }, { "epoch": 4.93, "learning_rate": 8.381995684517073e-07, "loss": 0.2108, "step": 81760 }, { "epoch": 4.93, "learning_rate": 8.366906583374829e-07, "loss": 0.1835, "step": 81762 }, { "epoch": 4.94, "learning_rate": 8.351817482232583e-07, "loss": 0.2635, "step": 81764 }, { "epoch": 4.94, "learning_rate": 8.336728381090338e-07, "loss": 0.2347, "step": 81766 }, { "epoch": 4.94, "learning_rate": 8.321639279948093e-07, "loss": 0.1593, "step": 81768 }, { "epoch": 4.94, "learning_rate": 8.306550178805848e-07, "loss": 0.2606, "step": 81770 }, { "epoch": 4.94, "learning_rate": 8.291461077663604e-07, "loss": 0.1364, "step": 81772 }, { "epoch": 4.94, "learning_rate": 8.276371976521358e-07, "loss": 0.3004, "step": 81774 }, { "epoch": 4.94, "learning_rate": 8.261282875379114e-07, "loss": 0.2215, "step": 81776 }, { "epoch": 4.94, "learning_rate": 8.246193774236868e-07, "loss": 0.2984, "step": 81778 }, { "epoch": 4.94, "learning_rate": 8.231104673094623e-07, "loss": 0.2429, "step": 81780 }, { "epoch": 4.94, "learning_rate": 8.216015571952378e-07, "loss": 0.2265, "step": 81782 }, { "epoch": 4.94, "learning_rate": 8.200926470810135e-07, "loss": 0.3472, "step": 81784 }, { "epoch": 4.94, "learning_rate": 8.18583736966789e-07, "loss": 0.1605, "step": 81786 }, { "epoch": 4.94, "learning_rate": 8.170748268525645e-07, "loss": 0.1406, "step": 81788 }, { "epoch": 4.94, "learning_rate": 8.1556591673834e-07, "loss": 0.1099, "step": 81790 }, { "epoch": 4.94, "learning_rate": 8.140570066241155e-07, "loss": 0.1957, "step": 81792 }, { "epoch": 4.94, "learning_rate": 8.12548096509891e-07, "loss": 0.2317, "step": 81794 }, { "epoch": 4.94, "learning_rate": 8.110391863956665e-07, "loss": 0.1603, "step": 81796 }, { "epoch": 4.94, "learning_rate": 8.09530276281442e-07, "loss": 0.332, "step": 81798 }, { "epoch": 4.94, "learning_rate": 8.080213661672175e-07, "loss": 0.2031, "step": 81800 }, { "epoch": 4.94, "learning_rate": 8.06512456052993e-07, "loss": 0.1841, "step": 81802 }, { "epoch": 4.94, "learning_rate": 8.050035459387685e-07, "loss": 0.1375, "step": 81804 }, { "epoch": 4.94, "learning_rate": 8.03494635824544e-07, "loss": 0.1707, "step": 81806 }, { "epoch": 4.94, "learning_rate": 8.019857257103195e-07, "loss": 0.2639, "step": 81808 }, { "epoch": 4.94, "learning_rate": 8.00476815596095e-07, "loss": 0.1586, "step": 81810 }, { "epoch": 4.94, "learning_rate": 7.989679054818705e-07, "loss": 0.1456, "step": 81812 }, { "epoch": 4.94, "learning_rate": 7.97458995367646e-07, "loss": 0.2997, "step": 81814 }, { "epoch": 4.94, "learning_rate": 7.959500852534215e-07, "loss": 0.1377, "step": 81816 }, { "epoch": 4.94, "learning_rate": 7.94441175139197e-07, "loss": 0.2142, "step": 81818 }, { "epoch": 4.94, "learning_rate": 7.929322650249725e-07, "loss": 0.2354, "step": 81820 }, { "epoch": 4.94, "learning_rate": 7.91423354910748e-07, "loss": 0.1562, "step": 81822 }, { "epoch": 4.94, "learning_rate": 7.899144447965235e-07, "loss": 0.2109, "step": 81824 }, { "epoch": 4.94, "learning_rate": 7.88405534682299e-07, "loss": 0.2117, "step": 81826 }, { "epoch": 4.94, "learning_rate": 7.868966245680745e-07, "loss": 0.2278, "step": 81828 }, { "epoch": 4.94, "learning_rate": 7.8538771445385e-07, "loss": 0.174, "step": 81830 }, { "epoch": 4.94, "learning_rate": 7.838788043396255e-07, "loss": 0.1819, "step": 81832 }, { "epoch": 4.94, "learning_rate": 7.82369894225401e-07, "loss": 0.2251, "step": 81834 }, { "epoch": 4.94, "learning_rate": 7.808609841111765e-07, "loss": 0.2809, "step": 81836 }, { "epoch": 4.94, "learning_rate": 7.79352073996952e-07, "loss": 0.1176, "step": 81838 }, { "epoch": 4.94, "learning_rate": 7.778431638827275e-07, "loss": 0.2224, "step": 81840 }, { "epoch": 4.94, "learning_rate": 7.763342537685031e-07, "loss": 0.2587, "step": 81842 }, { "epoch": 4.94, "learning_rate": 7.748253436542786e-07, "loss": 0.1429, "step": 81844 }, { "epoch": 4.94, "learning_rate": 7.733164335400541e-07, "loss": 0.1947, "step": 81846 }, { "epoch": 4.94, "learning_rate": 7.718075234258296e-07, "loss": 0.372, "step": 81848 }, { "epoch": 4.94, "learning_rate": 7.702986133116051e-07, "loss": 0.1884, "step": 81850 }, { "epoch": 4.94, "learning_rate": 7.687897031973806e-07, "loss": 0.1931, "step": 81852 }, { "epoch": 4.94, "learning_rate": 7.672807930831561e-07, "loss": 0.1418, "step": 81854 }, { "epoch": 4.94, "learning_rate": 7.657718829689316e-07, "loss": 0.1827, "step": 81856 }, { "epoch": 4.94, "learning_rate": 7.64262972854707e-07, "loss": 0.2225, "step": 81858 }, { "epoch": 4.94, "learning_rate": 7.627540627404826e-07, "loss": 0.341, "step": 81860 }, { "epoch": 4.94, "learning_rate": 7.61245152626258e-07, "loss": 0.2517, "step": 81862 }, { "epoch": 4.94, "learning_rate": 7.597362425120336e-07, "loss": 0.208, "step": 81864 }, { "epoch": 4.94, "learning_rate": 7.582273323978091e-07, "loss": 0.2383, "step": 81866 }, { "epoch": 4.94, "learning_rate": 7.567184222835846e-07, "loss": 0.18, "step": 81868 }, { "epoch": 4.94, "learning_rate": 7.552095121693601e-07, "loss": 0.2174, "step": 81870 }, { "epoch": 4.94, "learning_rate": 7.537006020551355e-07, "loss": 0.2502, "step": 81872 }, { "epoch": 4.94, "learning_rate": 7.521916919409111e-07, "loss": 0.1921, "step": 81874 }, { "epoch": 4.94, "learning_rate": 7.506827818266866e-07, "loss": 0.2017, "step": 81876 }, { "epoch": 4.94, "learning_rate": 7.491738717124621e-07, "loss": 0.1247, "step": 81878 }, { "epoch": 4.94, "learning_rate": 7.476649615982377e-07, "loss": 0.1632, "step": 81880 }, { "epoch": 4.94, "learning_rate": 7.461560514840131e-07, "loss": 0.2228, "step": 81882 }, { "epoch": 4.94, "learning_rate": 7.446471413697887e-07, "loss": 0.2083, "step": 81884 }, { "epoch": 4.94, "learning_rate": 7.431382312555641e-07, "loss": 0.284, "step": 81886 }, { "epoch": 4.94, "learning_rate": 7.416293211413397e-07, "loss": 0.2162, "step": 81888 }, { "epoch": 4.94, "learning_rate": 7.401204110271152e-07, "loss": 0.3138, "step": 81890 }, { "epoch": 4.94, "learning_rate": 7.386115009128907e-07, "loss": 0.1196, "step": 81892 }, { "epoch": 4.94, "learning_rate": 7.371025907986662e-07, "loss": 0.1967, "step": 81894 }, { "epoch": 4.94, "learning_rate": 7.355936806844416e-07, "loss": 0.1529, "step": 81896 }, { "epoch": 4.94, "learning_rate": 7.340847705702172e-07, "loss": 0.1446, "step": 81898 }, { "epoch": 4.94, "learning_rate": 7.325758604559926e-07, "loss": 0.2481, "step": 81900 }, { "epoch": 4.94, "learning_rate": 7.310669503417682e-07, "loss": 0.2801, "step": 81902 }, { "epoch": 4.94, "learning_rate": 7.295580402275437e-07, "loss": 0.2624, "step": 81904 }, { "epoch": 4.94, "learning_rate": 7.280491301133191e-07, "loss": 0.2044, "step": 81906 }, { "epoch": 4.94, "learning_rate": 7.265402199990947e-07, "loss": 0.2747, "step": 81908 }, { "epoch": 4.94, "learning_rate": 7.250313098848701e-07, "loss": 0.188, "step": 81910 }, { "epoch": 4.94, "learning_rate": 7.235223997706457e-07, "loss": 0.1839, "step": 81912 }, { "epoch": 4.94, "learning_rate": 7.220134896564211e-07, "loss": 0.2267, "step": 81914 }, { "epoch": 4.94, "learning_rate": 7.205045795421967e-07, "loss": 0.1432, "step": 81916 }, { "epoch": 4.94, "learning_rate": 7.189956694279722e-07, "loss": 0.2405, "step": 81918 }, { "epoch": 4.94, "learning_rate": 7.174867593137477e-07, "loss": 0.2027, "step": 81920 }, { "epoch": 4.94, "learning_rate": 7.159778491995233e-07, "loss": 0.258, "step": 81922 }, { "epoch": 4.94, "learning_rate": 7.144689390852987e-07, "loss": 0.1786, "step": 81924 }, { "epoch": 4.94, "learning_rate": 7.129600289710743e-07, "loss": 0.2307, "step": 81926 }, { "epoch": 4.94, "learning_rate": 7.114511188568498e-07, "loss": 0.193, "step": 81928 }, { "epoch": 4.95, "learning_rate": 7.099422087426252e-07, "loss": 0.1618, "step": 81930 }, { "epoch": 4.95, "learning_rate": 7.084332986284008e-07, "loss": 0.2465, "step": 81932 }, { "epoch": 4.95, "learning_rate": 7.069243885141762e-07, "loss": 0.3109, "step": 81934 }, { "epoch": 4.95, "learning_rate": 7.054154783999518e-07, "loss": 0.2387, "step": 81936 }, { "epoch": 4.95, "learning_rate": 7.039065682857272e-07, "loss": 0.2131, "step": 81938 }, { "epoch": 4.95, "learning_rate": 7.023976581715028e-07, "loss": 0.3065, "step": 81940 }, { "epoch": 4.95, "learning_rate": 7.008887480572783e-07, "loss": 0.3152, "step": 81942 }, { "epoch": 4.95, "learning_rate": 6.993798379430537e-07, "loss": 0.1384, "step": 81944 }, { "epoch": 4.95, "learning_rate": 6.978709278288293e-07, "loss": 0.2439, "step": 81946 }, { "epoch": 4.95, "learning_rate": 6.963620177146047e-07, "loss": 0.2224, "step": 81948 }, { "epoch": 4.95, "learning_rate": 6.948531076003803e-07, "loss": 0.2746, "step": 81950 }, { "epoch": 4.95, "learning_rate": 6.933441974861557e-07, "loss": 0.3167, "step": 81952 }, { "epoch": 4.95, "learning_rate": 6.918352873719313e-07, "loss": 0.3151, "step": 81954 }, { "epoch": 4.95, "learning_rate": 6.903263772577068e-07, "loss": 0.3373, "step": 81956 }, { "epoch": 4.95, "learning_rate": 6.888174671434823e-07, "loss": 0.2412, "step": 81958 }, { "epoch": 4.95, "learning_rate": 6.873085570292579e-07, "loss": 0.2506, "step": 81960 }, { "epoch": 4.95, "learning_rate": 6.857996469150333e-07, "loss": 0.2047, "step": 81962 }, { "epoch": 4.95, "learning_rate": 6.842907368008088e-07, "loss": 0.3109, "step": 81964 }, { "epoch": 4.95, "learning_rate": 6.827818266865843e-07, "loss": 0.2461, "step": 81966 }, { "epoch": 4.95, "learning_rate": 6.812729165723598e-07, "loss": 0.2068, "step": 81968 }, { "epoch": 4.95, "learning_rate": 6.797640064581354e-07, "loss": 0.271, "step": 81970 }, { "epoch": 4.95, "learning_rate": 6.782550963439108e-07, "loss": 0.1889, "step": 81972 }, { "epoch": 4.95, "learning_rate": 6.767461862296864e-07, "loss": 0.5184, "step": 81974 }, { "epoch": 4.95, "learning_rate": 6.752372761154618e-07, "loss": 0.1777, "step": 81976 }, { "epoch": 4.95, "learning_rate": 6.737283660012373e-07, "loss": 0.2016, "step": 81978 }, { "epoch": 4.95, "learning_rate": 6.722194558870129e-07, "loss": 0.1566, "step": 81980 }, { "epoch": 4.95, "learning_rate": 6.707105457727883e-07, "loss": 0.2836, "step": 81982 }, { "epoch": 4.95, "learning_rate": 6.692016356585639e-07, "loss": 0.2345, "step": 81984 }, { "epoch": 4.95, "learning_rate": 6.676927255443393e-07, "loss": 0.1226, "step": 81986 }, { "epoch": 4.95, "learning_rate": 6.661838154301149e-07, "loss": 0.1957, "step": 81988 }, { "epoch": 4.95, "learning_rate": 6.646749053158903e-07, "loss": 0.1139, "step": 81990 }, { "epoch": 4.95, "learning_rate": 6.631659952016658e-07, "loss": 0.2343, "step": 81992 }, { "epoch": 4.95, "learning_rate": 6.616570850874414e-07, "loss": 0.1924, "step": 81994 }, { "epoch": 4.95, "learning_rate": 6.601481749732168e-07, "loss": 0.2748, "step": 81996 }, { "epoch": 4.95, "learning_rate": 6.586392648589925e-07, "loss": 0.2859, "step": 81998 }, { "epoch": 4.95, "learning_rate": 6.571303547447679e-07, "loss": 0.1733, "step": 82000 }, { "epoch": 4.95, "learning_rate": 6.556214446305434e-07, "loss": 0.1198, "step": 82002 }, { "epoch": 4.95, "learning_rate": 6.541125345163189e-07, "loss": 0.2854, "step": 82004 }, { "epoch": 4.95, "learning_rate": 6.526036244020944e-07, "loss": 0.1867, "step": 82006 }, { "epoch": 4.95, "learning_rate": 6.5109471428787e-07, "loss": 0.324, "step": 82008 }, { "epoch": 4.95, "learning_rate": 6.495858041736454e-07, "loss": 0.1903, "step": 82010 }, { "epoch": 4.95, "learning_rate": 6.480768940594209e-07, "loss": 0.2296, "step": 82012 }, { "epoch": 4.95, "learning_rate": 6.465679839451964e-07, "loss": 0.17, "step": 82014 }, { "epoch": 4.95, "learning_rate": 6.450590738309719e-07, "loss": 0.1759, "step": 82016 }, { "epoch": 4.95, "learning_rate": 6.435501637167474e-07, "loss": 0.3565, "step": 82018 }, { "epoch": 4.95, "learning_rate": 6.420412536025229e-07, "loss": 0.1818, "step": 82020 }, { "epoch": 4.95, "learning_rate": 6.405323434882985e-07, "loss": 0.2655, "step": 82022 }, { "epoch": 4.95, "learning_rate": 6.390234333740739e-07, "loss": 0.1055, "step": 82024 }, { "epoch": 4.95, "learning_rate": 6.375145232598494e-07, "loss": 0.1613, "step": 82026 }, { "epoch": 4.95, "learning_rate": 6.360056131456249e-07, "loss": 0.2085, "step": 82028 }, { "epoch": 4.95, "learning_rate": 6.344967030314004e-07, "loss": 0.1882, "step": 82030 }, { "epoch": 4.95, "learning_rate": 6.32987792917176e-07, "loss": 0.2232, "step": 82032 }, { "epoch": 4.95, "learning_rate": 6.314788828029514e-07, "loss": 0.1723, "step": 82034 }, { "epoch": 4.95, "learning_rate": 6.29969972688727e-07, "loss": 0.2026, "step": 82036 }, { "epoch": 4.95, "learning_rate": 6.284610625745025e-07, "loss": 0.1647, "step": 82038 }, { "epoch": 4.95, "learning_rate": 6.26952152460278e-07, "loss": 0.2354, "step": 82040 }, { "epoch": 4.95, "learning_rate": 6.254432423460535e-07, "loss": 0.1866, "step": 82042 }, { "epoch": 4.95, "learning_rate": 6.23934332231829e-07, "loss": 0.2581, "step": 82044 }, { "epoch": 4.95, "learning_rate": 6.224254221176046e-07, "loss": 0.1477, "step": 82046 }, { "epoch": 4.95, "learning_rate": 6.2091651200338e-07, "loss": 0.1857, "step": 82048 }, { "epoch": 4.95, "learning_rate": 6.194076018891555e-07, "loss": 0.2525, "step": 82050 }, { "epoch": 4.95, "learning_rate": 6.17898691774931e-07, "loss": 0.2086, "step": 82052 }, { "epoch": 4.95, "learning_rate": 6.163897816607065e-07, "loss": 0.304, "step": 82054 }, { "epoch": 4.95, "learning_rate": 6.14880871546482e-07, "loss": 0.1768, "step": 82056 }, { "epoch": 4.95, "learning_rate": 6.133719614322575e-07, "loss": 0.2585, "step": 82058 }, { "epoch": 4.95, "learning_rate": 6.11863051318033e-07, "loss": 0.121, "step": 82060 }, { "epoch": 4.95, "learning_rate": 6.103541412038085e-07, "loss": 0.1573, "step": 82062 }, { "epoch": 4.95, "learning_rate": 6.08845231089584e-07, "loss": 0.2265, "step": 82064 }, { "epoch": 4.95, "learning_rate": 6.073363209753595e-07, "loss": 0.1957, "step": 82066 }, { "epoch": 4.95, "learning_rate": 6.05827410861135e-07, "loss": 0.1683, "step": 82068 }, { "epoch": 4.95, "learning_rate": 6.043185007469105e-07, "loss": 0.1867, "step": 82070 }, { "epoch": 4.95, "learning_rate": 6.02809590632686e-07, "loss": 0.1553, "step": 82072 }, { "epoch": 4.95, "learning_rate": 6.013006805184615e-07, "loss": 0.1817, "step": 82074 }, { "epoch": 4.95, "learning_rate": 5.997917704042371e-07, "loss": 0.2466, "step": 82076 }, { "epoch": 4.95, "learning_rate": 5.982828602900126e-07, "loss": 0.2078, "step": 82078 }, { "epoch": 4.95, "learning_rate": 5.967739501757881e-07, "loss": 0.1904, "step": 82080 }, { "epoch": 4.95, "learning_rate": 5.952650400615636e-07, "loss": 0.2794, "step": 82082 }, { "epoch": 4.95, "learning_rate": 5.93756129947339e-07, "loss": 0.2052, "step": 82084 }, { "epoch": 4.95, "learning_rate": 5.922472198331146e-07, "loss": 0.2395, "step": 82086 }, { "epoch": 4.95, "learning_rate": 5.907383097188901e-07, "loss": 0.1137, "step": 82088 }, { "epoch": 4.95, "learning_rate": 5.892293996046656e-07, "loss": 0.1663, "step": 82090 }, { "epoch": 4.95, "learning_rate": 5.877204894904411e-07, "loss": 0.1909, "step": 82092 }, { "epoch": 4.95, "learning_rate": 5.862115793762166e-07, "loss": 0.1431, "step": 82094 }, { "epoch": 4.96, "learning_rate": 5.847026692619921e-07, "loss": 0.21, "step": 82096 }, { "epoch": 4.96, "learning_rate": 5.831937591477675e-07, "loss": 0.234, "step": 82098 }, { "epoch": 4.96, "learning_rate": 5.816848490335431e-07, "loss": 0.2608, "step": 82100 }, { "epoch": 4.96, "learning_rate": 5.801759389193186e-07, "loss": 0.1412, "step": 82102 }, { "epoch": 4.96, "learning_rate": 5.786670288050941e-07, "loss": 0.1807, "step": 82104 }, { "epoch": 4.96, "learning_rate": 5.771581186908696e-07, "loss": 0.2783, "step": 82106 }, { "epoch": 4.96, "learning_rate": 5.756492085766451e-07, "loss": 0.1949, "step": 82108 }, { "epoch": 4.96, "learning_rate": 5.741402984624206e-07, "loss": 0.2065, "step": 82110 }, { "epoch": 4.96, "learning_rate": 5.72631388348196e-07, "loss": 0.2297, "step": 82112 }, { "epoch": 4.96, "learning_rate": 5.711224782339717e-07, "loss": 0.2568, "step": 82114 }, { "epoch": 4.96, "learning_rate": 5.696135681197472e-07, "loss": 0.207, "step": 82116 }, { "epoch": 4.96, "learning_rate": 5.681046580055226e-07, "loss": 0.2214, "step": 82118 }, { "epoch": 4.96, "learning_rate": 5.665957478912982e-07, "loss": 0.1923, "step": 82120 }, { "epoch": 4.96, "learning_rate": 5.650868377770736e-07, "loss": 0.1725, "step": 82122 }, { "epoch": 4.96, "learning_rate": 5.635779276628492e-07, "loss": 0.1712, "step": 82124 }, { "epoch": 4.96, "learning_rate": 5.620690175486247e-07, "loss": 0.276, "step": 82126 }, { "epoch": 4.96, "learning_rate": 5.605601074344002e-07, "loss": 0.2186, "step": 82128 }, { "epoch": 4.96, "learning_rate": 5.590511973201757e-07, "loss": 0.2535, "step": 82130 }, { "epoch": 4.96, "learning_rate": 5.575422872059511e-07, "loss": 0.1937, "step": 82132 }, { "epoch": 4.96, "learning_rate": 5.560333770917267e-07, "loss": 0.2394, "step": 82134 }, { "epoch": 4.96, "learning_rate": 5.545244669775021e-07, "loss": 0.1154, "step": 82136 }, { "epoch": 4.96, "learning_rate": 5.530155568632777e-07, "loss": 0.3194, "step": 82138 }, { "epoch": 4.96, "learning_rate": 5.515066467490532e-07, "loss": 0.1297, "step": 82140 }, { "epoch": 4.96, "learning_rate": 5.499977366348287e-07, "loss": 0.161, "step": 82142 }, { "epoch": 4.96, "learning_rate": 5.484888265206042e-07, "loss": 0.2098, "step": 82144 }, { "epoch": 4.96, "learning_rate": 5.469799164063796e-07, "loss": 0.2893, "step": 82146 }, { "epoch": 4.96, "learning_rate": 5.454710062921552e-07, "loss": 0.2325, "step": 82148 }, { "epoch": 4.96, "learning_rate": 5.439620961779306e-07, "loss": 0.275, "step": 82150 }, { "epoch": 4.96, "learning_rate": 5.424531860637062e-07, "loss": 0.179, "step": 82152 }, { "epoch": 4.96, "learning_rate": 5.409442759494818e-07, "loss": 0.1988, "step": 82154 }, { "epoch": 4.96, "learning_rate": 5.394353658352572e-07, "loss": 0.1995, "step": 82156 }, { "epoch": 4.96, "learning_rate": 5.379264557210328e-07, "loss": 0.166, "step": 82158 }, { "epoch": 4.96, "learning_rate": 5.364175456068082e-07, "loss": 0.269, "step": 82160 }, { "epoch": 4.96, "learning_rate": 5.349086354925838e-07, "loss": 0.1372, "step": 82162 }, { "epoch": 4.96, "learning_rate": 5.333997253783592e-07, "loss": 0.1674, "step": 82164 }, { "epoch": 4.96, "learning_rate": 5.318908152641347e-07, "loss": 0.2823, "step": 82166 }, { "epoch": 4.96, "learning_rate": 5.303819051499103e-07, "loss": 0.2006, "step": 82168 }, { "epoch": 4.96, "learning_rate": 5.288729950356857e-07, "loss": 0.2172, "step": 82170 }, { "epoch": 4.96, "learning_rate": 5.273640849214613e-07, "loss": 0.2415, "step": 82172 }, { "epoch": 4.96, "learning_rate": 5.258551748072367e-07, "loss": 0.2169, "step": 82174 }, { "epoch": 4.96, "learning_rate": 5.243462646930123e-07, "loss": 0.2888, "step": 82176 }, { "epoch": 4.96, "learning_rate": 5.228373545787877e-07, "loss": 0.1623, "step": 82178 }, { "epoch": 4.96, "learning_rate": 5.213284444645632e-07, "loss": 0.169, "step": 82180 }, { "epoch": 4.96, "learning_rate": 5.198195343503388e-07, "loss": 0.1124, "step": 82182 }, { "epoch": 4.96, "learning_rate": 5.183106242361142e-07, "loss": 0.2094, "step": 82184 }, { "epoch": 4.96, "learning_rate": 5.168017141218898e-07, "loss": 0.1564, "step": 82186 }, { "epoch": 4.96, "learning_rate": 5.152928040076652e-07, "loss": 0.266, "step": 82188 }, { "epoch": 4.96, "learning_rate": 5.137838938934408e-07, "loss": 0.328, "step": 82190 }, { "epoch": 4.96, "learning_rate": 5.122749837792164e-07, "loss": 0.2082, "step": 82192 }, { "epoch": 4.96, "learning_rate": 5.107660736649918e-07, "loss": 0.1728, "step": 82194 }, { "epoch": 4.96, "learning_rate": 5.092571635507674e-07, "loss": 0.1599, "step": 82196 }, { "epoch": 4.96, "learning_rate": 5.077482534365428e-07, "loss": 0.2497, "step": 82198 }, { "epoch": 4.96, "learning_rate": 5.062393433223184e-07, "loss": 0.2, "step": 82200 }, { "epoch": 4.96, "learning_rate": 5.047304332080938e-07, "loss": 0.1944, "step": 82202 }, { "epoch": 4.96, "learning_rate": 5.032215230938693e-07, "loss": 0.2101, "step": 82204 }, { "epoch": 4.96, "learning_rate": 5.017126129796449e-07, "loss": 0.3072, "step": 82206 }, { "epoch": 4.96, "learning_rate": 5.002037028654203e-07, "loss": 0.2941, "step": 82208 }, { "epoch": 4.96, "learning_rate": 4.986947927511959e-07, "loss": 0.1695, "step": 82210 }, { "epoch": 4.96, "learning_rate": 4.971858826369713e-07, "loss": 0.1616, "step": 82212 }, { "epoch": 4.96, "learning_rate": 4.956769725227468e-07, "loss": 0.2305, "step": 82214 }, { "epoch": 4.96, "learning_rate": 4.941680624085223e-07, "loss": 0.179, "step": 82216 }, { "epoch": 4.96, "learning_rate": 4.926591522942978e-07, "loss": 0.1972, "step": 82218 }, { "epoch": 4.96, "learning_rate": 4.911502421800734e-07, "loss": 0.1703, "step": 82220 }, { "epoch": 4.96, "learning_rate": 4.896413320658488e-07, "loss": 0.3104, "step": 82222 }, { "epoch": 4.96, "learning_rate": 4.881324219516244e-07, "loss": 0.1734, "step": 82224 }, { "epoch": 4.96, "learning_rate": 4.866235118373998e-07, "loss": 0.1957, "step": 82226 }, { "epoch": 4.96, "learning_rate": 4.851146017231753e-07, "loss": 0.2255, "step": 82228 }, { "epoch": 4.96, "learning_rate": 4.83605691608951e-07, "loss": 0.1776, "step": 82230 }, { "epoch": 4.96, "learning_rate": 4.820967814947264e-07, "loss": 0.1728, "step": 82232 }, { "epoch": 4.96, "learning_rate": 4.80587871380502e-07, "loss": 0.2221, "step": 82234 }, { "epoch": 4.96, "learning_rate": 4.790789612662774e-07, "loss": 0.2265, "step": 82236 }, { "epoch": 4.96, "learning_rate": 4.775700511520529e-07, "loss": 0.2121, "step": 82238 }, { "epoch": 4.96, "learning_rate": 4.7606114103782844e-07, "loss": 0.239, "step": 82240 }, { "epoch": 4.96, "learning_rate": 4.7455223092360395e-07, "loss": 0.222, "step": 82242 }, { "epoch": 4.96, "learning_rate": 4.730433208093794e-07, "loss": 0.1756, "step": 82244 }, { "epoch": 4.96, "learning_rate": 4.715344106951549e-07, "loss": 0.277, "step": 82246 }, { "epoch": 4.96, "learning_rate": 4.700255005809304e-07, "loss": 0.192, "step": 82248 }, { "epoch": 4.96, "learning_rate": 4.685165904667059e-07, "loss": 0.1656, "step": 82250 }, { "epoch": 4.96, "learning_rate": 4.6700768035248143e-07, "loss": 0.1897, "step": 82252 }, { "epoch": 4.96, "learning_rate": 4.6549877023825694e-07, "loss": 0.1717, "step": 82254 }, { "epoch": 4.96, "learning_rate": 4.639898601240324e-07, "loss": 0.2054, "step": 82256 }, { "epoch": 4.96, "learning_rate": 4.624809500098079e-07, "loss": 0.2754, "step": 82258 }, { "epoch": 4.96, "learning_rate": 4.609720398955834e-07, "loss": 0.1305, "step": 82260 }, { "epoch": 4.97, "learning_rate": 4.594631297813589e-07, "loss": 0.2489, "step": 82262 }, { "epoch": 4.97, "learning_rate": 4.579542196671344e-07, "loss": 0.1982, "step": 82264 }, { "epoch": 4.97, "learning_rate": 4.5644530955290993e-07, "loss": 0.2461, "step": 82266 }, { "epoch": 4.97, "learning_rate": 4.5493639943868544e-07, "loss": 0.2386, "step": 82268 }, { "epoch": 4.97, "learning_rate": 4.53427489324461e-07, "loss": 0.1941, "step": 82270 }, { "epoch": 4.97, "learning_rate": 4.519185792102365e-07, "loss": 0.2464, "step": 82272 }, { "epoch": 4.97, "learning_rate": 4.50409669096012e-07, "loss": 0.3023, "step": 82274 }, { "epoch": 4.97, "learning_rate": 4.489007589817875e-07, "loss": 0.2349, "step": 82276 }, { "epoch": 4.97, "learning_rate": 4.47391848867563e-07, "loss": 0.2451, "step": 82278 }, { "epoch": 4.97, "learning_rate": 4.458829387533385e-07, "loss": 0.2086, "step": 82280 }, { "epoch": 4.97, "learning_rate": 4.44374028639114e-07, "loss": 0.1843, "step": 82282 }, { "epoch": 4.97, "learning_rate": 4.428651185248895e-07, "loss": 0.1573, "step": 82284 }, { "epoch": 4.97, "learning_rate": 4.41356208410665e-07, "loss": 0.2074, "step": 82286 }, { "epoch": 4.97, "learning_rate": 4.398472982964405e-07, "loss": 0.2852, "step": 82288 }, { "epoch": 4.97, "learning_rate": 4.38338388182216e-07, "loss": 0.2408, "step": 82290 }, { "epoch": 4.97, "learning_rate": 4.368294780679915e-07, "loss": 0.1806, "step": 82292 }, { "epoch": 4.97, "learning_rate": 4.35320567953767e-07, "loss": 0.2422, "step": 82294 }, { "epoch": 4.97, "learning_rate": 4.338116578395425e-07, "loss": 0.3029, "step": 82296 }, { "epoch": 4.97, "learning_rate": 4.32302747725318e-07, "loss": 0.3123, "step": 82298 }, { "epoch": 4.97, "learning_rate": 4.307938376110935e-07, "loss": 0.1447, "step": 82300 }, { "epoch": 4.97, "learning_rate": 4.29284927496869e-07, "loss": 0.2207, "step": 82302 }, { "epoch": 4.97, "learning_rate": 4.277760173826445e-07, "loss": 0.2831, "step": 82304 }, { "epoch": 4.97, "learning_rate": 4.2626710726842e-07, "loss": 0.2192, "step": 82306 }, { "epoch": 4.97, "learning_rate": 4.247581971541956e-07, "loss": 0.2995, "step": 82308 }, { "epoch": 4.97, "learning_rate": 4.232492870399711e-07, "loss": 0.3592, "step": 82310 }, { "epoch": 4.97, "learning_rate": 4.217403769257466e-07, "loss": 0.181, "step": 82312 }, { "epoch": 4.97, "learning_rate": 4.2023146681152207e-07, "loss": 0.2546, "step": 82314 }, { "epoch": 4.97, "learning_rate": 4.187225566972976e-07, "loss": 0.1855, "step": 82316 }, { "epoch": 4.97, "learning_rate": 4.172136465830731e-07, "loss": 0.2473, "step": 82318 }, { "epoch": 4.97, "learning_rate": 4.157047364688486e-07, "loss": 0.1854, "step": 82320 }, { "epoch": 4.97, "learning_rate": 4.141958263546241e-07, "loss": 0.209, "step": 82322 }, { "epoch": 4.97, "learning_rate": 4.126869162403996e-07, "loss": 0.2127, "step": 82324 }, { "epoch": 4.97, "learning_rate": 4.1117800612617506e-07, "loss": 0.1902, "step": 82326 }, { "epoch": 4.97, "learning_rate": 4.0966909601195057e-07, "loss": 0.1926, "step": 82328 }, { "epoch": 4.97, "learning_rate": 4.081601858977261e-07, "loss": 0.2072, "step": 82330 }, { "epoch": 4.97, "learning_rate": 4.066512757835016e-07, "loss": 0.1667, "step": 82332 }, { "epoch": 4.97, "learning_rate": 4.051423656692771e-07, "loss": 0.2705, "step": 82334 }, { "epoch": 4.97, "learning_rate": 4.036334555550526e-07, "loss": 0.2212, "step": 82336 }, { "epoch": 4.97, "learning_rate": 4.021245454408281e-07, "loss": 0.2128, "step": 82338 }, { "epoch": 4.97, "learning_rate": 4.0061563532660356e-07, "loss": 0.1926, "step": 82340 }, { "epoch": 4.97, "learning_rate": 3.9910672521237907e-07, "loss": 0.2361, "step": 82342 }, { "epoch": 4.97, "learning_rate": 3.975978150981546e-07, "loss": 0.1747, "step": 82344 }, { "epoch": 4.97, "learning_rate": 3.960889049839301e-07, "loss": 0.2531, "step": 82346 }, { "epoch": 4.97, "learning_rate": 3.9457999486970565e-07, "loss": 0.2164, "step": 82348 }, { "epoch": 4.97, "learning_rate": 3.9307108475548115e-07, "loss": 0.17, "step": 82350 }, { "epoch": 4.97, "learning_rate": 3.9156217464125666e-07, "loss": 0.2482, "step": 82352 }, { "epoch": 4.97, "learning_rate": 3.900532645270321e-07, "loss": 0.182, "step": 82354 }, { "epoch": 4.97, "learning_rate": 3.885443544128077e-07, "loss": 0.1584, "step": 82356 }, { "epoch": 4.97, "learning_rate": 3.870354442985832e-07, "loss": 0.2354, "step": 82358 }, { "epoch": 4.97, "learning_rate": 3.855265341843587e-07, "loss": 0.2453, "step": 82360 }, { "epoch": 4.97, "learning_rate": 3.8401762407013415e-07, "loss": 0.2054, "step": 82362 }, { "epoch": 4.97, "learning_rate": 3.8250871395590965e-07, "loss": 0.2559, "step": 82364 }, { "epoch": 4.97, "learning_rate": 3.8099980384168516e-07, "loss": 0.2198, "step": 82366 }, { "epoch": 4.97, "learning_rate": 3.7949089372746067e-07, "loss": 0.2172, "step": 82368 }, { "epoch": 4.97, "learning_rate": 3.779819836132362e-07, "loss": 0.2266, "step": 82370 }, { "epoch": 4.97, "learning_rate": 3.764730734990117e-07, "loss": 0.2143, "step": 82372 }, { "epoch": 4.97, "learning_rate": 3.749641633847872e-07, "loss": 0.1783, "step": 82374 }, { "epoch": 4.97, "learning_rate": 3.734552532705627e-07, "loss": 0.2582, "step": 82376 }, { "epoch": 4.97, "learning_rate": 3.719463431563382e-07, "loss": 0.1469, "step": 82378 }, { "epoch": 4.97, "learning_rate": 3.704374330421137e-07, "loss": 0.2739, "step": 82380 }, { "epoch": 4.97, "learning_rate": 3.689285229278892e-07, "loss": 0.2305, "step": 82382 }, { "epoch": 4.97, "learning_rate": 3.6741961281366473e-07, "loss": 0.273, "step": 82384 }, { "epoch": 4.97, "learning_rate": 3.659107026994402e-07, "loss": 0.1509, "step": 82386 }, { "epoch": 4.97, "learning_rate": 3.644017925852157e-07, "loss": 0.0827, "step": 82388 }, { "epoch": 4.97, "learning_rate": 3.628928824709912e-07, "loss": 0.197, "step": 82390 }, { "epoch": 4.97, "learning_rate": 3.613839723567667e-07, "loss": 0.1534, "step": 82392 }, { "epoch": 4.97, "learning_rate": 3.598750622425422e-07, "loss": 0.248, "step": 82394 }, { "epoch": 4.97, "learning_rate": 3.583661521283178e-07, "loss": 0.23, "step": 82396 }, { "epoch": 4.97, "learning_rate": 3.5685724201409323e-07, "loss": 0.337, "step": 82398 }, { "epoch": 4.97, "learning_rate": 3.5534833189986874e-07, "loss": 0.1378, "step": 82400 }, { "epoch": 4.97, "learning_rate": 3.5383942178564425e-07, "loss": 0.2254, "step": 82402 }, { "epoch": 4.97, "learning_rate": 3.5233051167141976e-07, "loss": 0.1517, "step": 82404 }, { "epoch": 4.97, "learning_rate": 3.5082160155719526e-07, "loss": 0.18, "step": 82406 }, { "epoch": 4.97, "learning_rate": 3.4931269144297077e-07, "loss": 0.2316, "step": 82408 }, { "epoch": 4.97, "learning_rate": 3.4780378132874623e-07, "loss": 0.2318, "step": 82410 }, { "epoch": 4.97, "learning_rate": 3.4629487121452173e-07, "loss": 0.2941, "step": 82412 }, { "epoch": 4.97, "learning_rate": 3.4478596110029724e-07, "loss": 0.1891, "step": 82414 }, { "epoch": 4.97, "learning_rate": 3.432770509860728e-07, "loss": 0.2268, "step": 82416 }, { "epoch": 4.97, "learning_rate": 3.417681408718483e-07, "loss": 0.1522, "step": 82418 }, { "epoch": 4.97, "learning_rate": 3.402592307576238e-07, "loss": 0.2439, "step": 82420 }, { "epoch": 4.97, "learning_rate": 3.387503206433993e-07, "loss": 0.2246, "step": 82422 }, { "epoch": 4.97, "learning_rate": 3.372414105291748e-07, "loss": 0.1556, "step": 82424 }, { "epoch": 4.98, "learning_rate": 3.357325004149503e-07, "loss": 0.2242, "step": 82426 }, { "epoch": 4.98, "learning_rate": 3.342235903007258e-07, "loss": 0.1862, "step": 82428 }, { "epoch": 4.98, "learning_rate": 3.327146801865013e-07, "loss": 0.3099, "step": 82430 }, { "epoch": 4.98, "learning_rate": 3.312057700722768e-07, "loss": 0.281, "step": 82432 }, { "epoch": 4.98, "learning_rate": 3.296968599580523e-07, "loss": 0.1367, "step": 82434 }, { "epoch": 4.98, "learning_rate": 3.2818794984382783e-07, "loss": 0.1839, "step": 82436 }, { "epoch": 4.98, "learning_rate": 3.2667903972960334e-07, "loss": 0.2494, "step": 82438 }, { "epoch": 4.98, "learning_rate": 3.2517012961537884e-07, "loss": 0.2192, "step": 82440 }, { "epoch": 4.98, "learning_rate": 3.2366121950115435e-07, "loss": 0.1926, "step": 82442 }, { "epoch": 4.98, "learning_rate": 3.2215230938692986e-07, "loss": 0.2839, "step": 82444 }, { "epoch": 4.98, "learning_rate": 3.206433992727053e-07, "loss": 0.2611, "step": 82446 }, { "epoch": 4.98, "learning_rate": 3.191344891584808e-07, "loss": 0.1881, "step": 82448 }, { "epoch": 4.98, "learning_rate": 3.1762557904425633e-07, "loss": 0.2517, "step": 82450 }, { "epoch": 4.98, "learning_rate": 3.1611666893003184e-07, "loss": 0.1522, "step": 82452 }, { "epoch": 4.98, "learning_rate": 3.146077588158074e-07, "loss": 0.2071, "step": 82454 }, { "epoch": 4.98, "learning_rate": 3.1309884870158285e-07, "loss": 0.2343, "step": 82456 }, { "epoch": 4.98, "learning_rate": 3.1158993858735836e-07, "loss": 0.1781, "step": 82458 }, { "epoch": 4.98, "learning_rate": 3.1008102847313387e-07, "loss": 0.2776, "step": 82460 }, { "epoch": 4.98, "learning_rate": 3.085721183589094e-07, "loss": 0.1379, "step": 82462 }, { "epoch": 4.98, "learning_rate": 3.070632082446849e-07, "loss": 0.1608, "step": 82464 }, { "epoch": 4.98, "learning_rate": 3.055542981304604e-07, "loss": 0.1514, "step": 82466 }, { "epoch": 4.98, "learning_rate": 3.040453880162359e-07, "loss": 0.1779, "step": 82468 }, { "epoch": 4.98, "learning_rate": 3.0253647790201135e-07, "loss": 0.2347, "step": 82470 }, { "epoch": 4.98, "learning_rate": 3.0102756778778686e-07, "loss": 0.1624, "step": 82472 }, { "epoch": 4.98, "learning_rate": 2.995186576735624e-07, "loss": 0.2222, "step": 82474 }, { "epoch": 4.98, "learning_rate": 2.9800974755933793e-07, "loss": 0.2021, "step": 82476 }, { "epoch": 4.98, "learning_rate": 2.9650083744511344e-07, "loss": 0.2285, "step": 82478 }, { "epoch": 4.98, "learning_rate": 2.949919273308889e-07, "loss": 0.1871, "step": 82480 }, { "epoch": 4.98, "learning_rate": 2.934830172166644e-07, "loss": 0.2169, "step": 82482 }, { "epoch": 4.98, "learning_rate": 2.919741071024399e-07, "loss": 0.2331, "step": 82484 }, { "epoch": 4.98, "learning_rate": 2.904651969882154e-07, "loss": 0.2824, "step": 82486 }, { "epoch": 4.98, "learning_rate": 2.889562868739909e-07, "loss": 0.2801, "step": 82488 }, { "epoch": 4.98, "learning_rate": 2.8744737675976643e-07, "loss": 0.1349, "step": 82490 }, { "epoch": 4.98, "learning_rate": 2.8593846664554194e-07, "loss": 0.2333, "step": 82492 }, { "epoch": 4.98, "learning_rate": 2.8442955653131745e-07, "loss": 0.2941, "step": 82494 }, { "epoch": 4.98, "learning_rate": 2.8292064641709296e-07, "loss": 0.2698, "step": 82496 }, { "epoch": 4.98, "learning_rate": 2.8141173630286846e-07, "loss": 0.2357, "step": 82498 }, { "epoch": 4.98, "learning_rate": 2.7990282618864397e-07, "loss": 0.3234, "step": 82500 }, { "epoch": 4.98, "learning_rate": 2.783939160744195e-07, "loss": 0.2094, "step": 82502 }, { "epoch": 4.98, "learning_rate": 2.7688500596019493e-07, "loss": 0.1711, "step": 82504 }, { "epoch": 4.98, "learning_rate": 2.7537609584597044e-07, "loss": 0.1865, "step": 82506 }, { "epoch": 4.98, "learning_rate": 2.7386718573174595e-07, "loss": 0.1361, "step": 82508 }, { "epoch": 4.98, "learning_rate": 2.7235827561752146e-07, "loss": 0.2418, "step": 82510 }, { "epoch": 4.98, "learning_rate": 2.70849365503297e-07, "loss": 0.226, "step": 82512 }, { "epoch": 4.98, "learning_rate": 2.693404553890725e-07, "loss": 0.2296, "step": 82514 }, { "epoch": 4.98, "learning_rate": 2.67831545274848e-07, "loss": 0.2235, "step": 82516 }, { "epoch": 4.98, "learning_rate": 2.663226351606235e-07, "loss": 0.1852, "step": 82518 }, { "epoch": 4.98, "learning_rate": 2.64813725046399e-07, "loss": 0.1036, "step": 82520 }, { "epoch": 4.98, "learning_rate": 2.633048149321745e-07, "loss": 0.2033, "step": 82522 }, { "epoch": 4.98, "learning_rate": 2.6179590481795e-07, "loss": 0.2892, "step": 82524 }, { "epoch": 4.98, "learning_rate": 2.602869947037255e-07, "loss": 0.2361, "step": 82526 }, { "epoch": 4.98, "learning_rate": 2.5877808458950103e-07, "loss": 0.1715, "step": 82528 }, { "epoch": 4.98, "learning_rate": 2.572691744752765e-07, "loss": 0.137, "step": 82530 }, { "epoch": 4.98, "learning_rate": 2.5576026436105204e-07, "loss": 0.2622, "step": 82532 }, { "epoch": 4.98, "learning_rate": 2.5425135424682755e-07, "loss": 0.228, "step": 82534 }, { "epoch": 4.98, "learning_rate": 2.5274244413260306e-07, "loss": 0.1256, "step": 82536 }, { "epoch": 4.98, "learning_rate": 2.5123353401837857e-07, "loss": 0.1566, "step": 82538 }, { "epoch": 4.98, "learning_rate": 2.49724623904154e-07, "loss": 0.1509, "step": 82540 }, { "epoch": 4.98, "learning_rate": 2.4821571378992953e-07, "loss": 0.2009, "step": 82542 }, { "epoch": 4.98, "learning_rate": 2.4670680367570504e-07, "loss": 0.1615, "step": 82544 }, { "epoch": 4.98, "learning_rate": 2.4519789356148054e-07, "loss": 0.3153, "step": 82546 }, { "epoch": 4.98, "learning_rate": 2.4368898344725605e-07, "loss": 0.2313, "step": 82548 }, { "epoch": 4.98, "learning_rate": 2.4218007333303156e-07, "loss": 0.1495, "step": 82550 }, { "epoch": 4.98, "learning_rate": 2.4067116321880707e-07, "loss": 0.2252, "step": 82552 }, { "epoch": 4.98, "learning_rate": 2.391622531045826e-07, "loss": 0.229, "step": 82554 }, { "epoch": 4.98, "learning_rate": 2.3765334299035808e-07, "loss": 0.1603, "step": 82556 }, { "epoch": 4.98, "learning_rate": 2.361444328761336e-07, "loss": 0.2133, "step": 82558 }, { "epoch": 4.98, "learning_rate": 2.3463552276190907e-07, "loss": 0.1576, "step": 82560 }, { "epoch": 4.98, "learning_rate": 2.3312661264768458e-07, "loss": 0.2103, "step": 82562 }, { "epoch": 4.98, "learning_rate": 2.316177025334601e-07, "loss": 0.1335, "step": 82564 }, { "epoch": 4.98, "learning_rate": 2.301087924192356e-07, "loss": 0.1362, "step": 82566 }, { "epoch": 4.98, "learning_rate": 2.2859988230501108e-07, "loss": 0.2286, "step": 82568 }, { "epoch": 4.98, "learning_rate": 2.2709097219078664e-07, "loss": 0.1204, "step": 82570 }, { "epoch": 4.98, "learning_rate": 2.2558206207656212e-07, "loss": 0.2943, "step": 82572 }, { "epoch": 4.98, "learning_rate": 2.2407315196233763e-07, "loss": 0.2928, "step": 82574 }, { "epoch": 4.98, "learning_rate": 2.2256424184811313e-07, "loss": 0.1772, "step": 82576 }, { "epoch": 4.98, "learning_rate": 2.2105533173388861e-07, "loss": 0.2437, "step": 82578 }, { "epoch": 4.98, "learning_rate": 2.1954642161966412e-07, "loss": 0.1942, "step": 82580 }, { "epoch": 4.98, "learning_rate": 2.1803751150543963e-07, "loss": 0.2087, "step": 82582 }, { "epoch": 4.98, "learning_rate": 2.1652860139121514e-07, "loss": 0.1704, "step": 82584 }, { "epoch": 4.98, "learning_rate": 2.1501969127699062e-07, "loss": 0.2743, "step": 82586 }, { "epoch": 4.98, "learning_rate": 2.1351078116276613e-07, "loss": 0.1944, "step": 82588 }, { "epoch": 4.98, "learning_rate": 2.1200187104854166e-07, "loss": 0.2609, "step": 82590 }, { "epoch": 4.99, "learning_rate": 2.1049296093431717e-07, "loss": 0.2453, "step": 82592 }, { "epoch": 4.99, "learning_rate": 2.0898405082009268e-07, "loss": 0.2717, "step": 82594 }, { "epoch": 4.99, "learning_rate": 2.0747514070586816e-07, "loss": 0.1475, "step": 82596 }, { "epoch": 4.99, "learning_rate": 2.0596623059164367e-07, "loss": 0.2089, "step": 82598 }, { "epoch": 4.99, "learning_rate": 2.0445732047741917e-07, "loss": 0.2096, "step": 82600 }, { "epoch": 4.99, "learning_rate": 2.0294841036319466e-07, "loss": 0.2799, "step": 82602 }, { "epoch": 4.99, "learning_rate": 2.0143950024897016e-07, "loss": 0.1812, "step": 82604 }, { "epoch": 4.99, "learning_rate": 1.9993059013474567e-07, "loss": 0.2064, "step": 82606 }, { "epoch": 4.99, "learning_rate": 1.9842168002052118e-07, "loss": 0.2133, "step": 82608 }, { "epoch": 4.99, "learning_rate": 1.969127699062967e-07, "loss": 0.1272, "step": 82610 }, { "epoch": 4.99, "learning_rate": 1.9540385979207222e-07, "loss": 0.2527, "step": 82612 }, { "epoch": 4.99, "learning_rate": 1.938949496778477e-07, "loss": 0.2173, "step": 82614 }, { "epoch": 4.99, "learning_rate": 1.923860395636232e-07, "loss": 0.2441, "step": 82616 }, { "epoch": 4.99, "learning_rate": 1.9087712944939872e-07, "loss": 0.1696, "step": 82618 }, { "epoch": 4.99, "learning_rate": 1.893682193351742e-07, "loss": 0.151, "step": 82620 }, { "epoch": 4.99, "learning_rate": 1.878593092209497e-07, "loss": 0.1697, "step": 82622 }, { "epoch": 4.99, "learning_rate": 1.8635039910672524e-07, "loss": 0.192, "step": 82624 }, { "epoch": 4.99, "learning_rate": 1.8484148899250072e-07, "loss": 0.1922, "step": 82626 }, { "epoch": 4.99, "learning_rate": 1.8333257887827623e-07, "loss": 0.1447, "step": 82628 }, { "epoch": 4.99, "learning_rate": 1.8182366876405174e-07, "loss": 0.3002, "step": 82630 }, { "epoch": 4.99, "learning_rate": 1.8031475864982722e-07, "loss": 0.1472, "step": 82632 }, { "epoch": 4.99, "learning_rate": 1.7880584853560275e-07, "loss": 0.171, "step": 82634 }, { "epoch": 4.99, "learning_rate": 1.7729693842137826e-07, "loss": 0.1725, "step": 82636 }, { "epoch": 4.99, "learning_rate": 1.7578802830715374e-07, "loss": 0.2423, "step": 82638 }, { "epoch": 4.99, "learning_rate": 1.7427911819292925e-07, "loss": 0.1269, "step": 82640 }, { "epoch": 4.99, "learning_rate": 1.7277020807870476e-07, "loss": 0.2466, "step": 82642 }, { "epoch": 4.99, "learning_rate": 1.7126129796448027e-07, "loss": 0.1095, "step": 82644 }, { "epoch": 4.99, "learning_rate": 1.6975238785025577e-07, "loss": 0.1406, "step": 82646 }, { "epoch": 4.99, "learning_rate": 1.6824347773603128e-07, "loss": 0.1825, "step": 82648 }, { "epoch": 4.99, "learning_rate": 1.6673456762180676e-07, "loss": 0.1748, "step": 82650 }, { "epoch": 4.99, "learning_rate": 1.6522565750758227e-07, "loss": 0.1266, "step": 82652 }, { "epoch": 4.99, "learning_rate": 1.637167473933578e-07, "loss": 0.2617, "step": 82654 }, { "epoch": 4.99, "learning_rate": 1.6220783727913329e-07, "loss": 0.2317, "step": 82656 }, { "epoch": 4.99, "learning_rate": 1.606989271649088e-07, "loss": 0.1777, "step": 82658 }, { "epoch": 4.99, "learning_rate": 1.591900170506843e-07, "loss": 0.1506, "step": 82660 }, { "epoch": 4.99, "learning_rate": 1.5768110693645978e-07, "loss": 0.1753, "step": 82662 }, { "epoch": 4.99, "learning_rate": 1.5617219682223532e-07, "loss": 0.2687, "step": 82664 }, { "epoch": 4.99, "learning_rate": 1.5466328670801082e-07, "loss": 0.1826, "step": 82666 }, { "epoch": 4.99, "learning_rate": 1.531543765937863e-07, "loss": 0.2265, "step": 82668 }, { "epoch": 4.99, "learning_rate": 1.5164546647956181e-07, "loss": 0.2677, "step": 82670 }, { "epoch": 4.99, "learning_rate": 1.5013655636533732e-07, "loss": 0.2364, "step": 82672 }, { "epoch": 4.99, "learning_rate": 1.4862764625111283e-07, "loss": 0.221, "step": 82674 }, { "epoch": 4.99, "learning_rate": 1.4711873613688834e-07, "loss": 0.1568, "step": 82676 }, { "epoch": 4.99, "learning_rate": 1.4560982602266384e-07, "loss": 0.1467, "step": 82678 }, { "epoch": 4.99, "learning_rate": 1.4410091590843933e-07, "loss": 0.2207, "step": 82680 }, { "epoch": 4.99, "learning_rate": 1.4259200579421486e-07, "loss": 0.2241, "step": 82682 }, { "epoch": 4.99, "learning_rate": 1.4108309567999037e-07, "loss": 0.2862, "step": 82684 }, { "epoch": 4.99, "learning_rate": 1.3957418556576585e-07, "loss": 0.1763, "step": 82686 }, { "epoch": 4.99, "learning_rate": 1.3806527545154136e-07, "loss": 0.25, "step": 82688 }, { "epoch": 4.99, "learning_rate": 1.3655636533731686e-07, "loss": 0.2464, "step": 82690 }, { "epoch": 4.99, "learning_rate": 1.3504745522309237e-07, "loss": 0.1866, "step": 82692 }, { "epoch": 4.99, "learning_rate": 1.3353854510886788e-07, "loss": 0.3244, "step": 82694 }, { "epoch": 4.99, "learning_rate": 1.320296349946434e-07, "loss": 0.2665, "step": 82696 }, { "epoch": 4.99, "learning_rate": 1.3052072488041887e-07, "loss": 0.2765, "step": 82698 }, { "epoch": 4.99, "learning_rate": 1.2901181476619438e-07, "loss": 0.2179, "step": 82700 }, { "epoch": 4.99, "learning_rate": 1.2750290465196988e-07, "loss": 0.1981, "step": 82702 }, { "epoch": 4.99, "learning_rate": 1.259939945377454e-07, "loss": 0.2285, "step": 82704 }, { "epoch": 4.99, "learning_rate": 1.244850844235209e-07, "loss": 0.2201, "step": 82706 }, { "epoch": 4.99, "learning_rate": 1.229761743092964e-07, "loss": 0.1563, "step": 82708 }, { "epoch": 4.99, "learning_rate": 1.214672641950719e-07, "loss": 0.2448, "step": 82710 }, { "epoch": 4.99, "learning_rate": 1.1995835408084742e-07, "loss": 0.3154, "step": 82712 }, { "epoch": 4.99, "learning_rate": 1.1844944396662292e-07, "loss": 0.2623, "step": 82714 }, { "epoch": 4.99, "learning_rate": 1.1694053385239841e-07, "loss": 0.1909, "step": 82716 }, { "epoch": 4.99, "learning_rate": 1.1543162373817392e-07, "loss": 0.2205, "step": 82718 }, { "epoch": 4.99, "learning_rate": 1.1392271362394942e-07, "loss": 0.2343, "step": 82720 }, { "epoch": 4.99, "learning_rate": 1.1241380350972494e-07, "loss": 0.2088, "step": 82722 }, { "epoch": 4.99, "learning_rate": 1.1090489339550044e-07, "loss": 0.2058, "step": 82724 }, { "epoch": 4.99, "learning_rate": 1.0939598328127594e-07, "loss": 0.256, "step": 82726 }, { "epoch": 4.99, "learning_rate": 1.0788707316705143e-07, "loss": 0.1904, "step": 82728 }, { "epoch": 4.99, "learning_rate": 1.0637816305282694e-07, "loss": 0.2286, "step": 82730 }, { "epoch": 4.99, "learning_rate": 1.0486925293860246e-07, "loss": 0.149, "step": 82732 }, { "epoch": 4.99, "learning_rate": 1.0336034282437796e-07, "loss": 0.1763, "step": 82734 }, { "epoch": 4.99, "learning_rate": 1.0185143271015346e-07, "loss": 0.1414, "step": 82736 }, { "epoch": 4.99, "learning_rate": 1.0034252259592896e-07, "loss": 0.1716, "step": 82738 }, { "epoch": 4.99, "learning_rate": 9.883361248170448e-08, "loss": 0.2612, "step": 82740 }, { "epoch": 4.99, "learning_rate": 9.732470236747997e-08, "loss": 0.1409, "step": 82742 }, { "epoch": 4.99, "learning_rate": 9.581579225325548e-08, "loss": 0.2412, "step": 82744 }, { "epoch": 4.99, "learning_rate": 9.430688213903098e-08, "loss": 0.2235, "step": 82746 }, { "epoch": 4.99, "learning_rate": 9.279797202480648e-08, "loss": 0.2183, "step": 82748 }, { "epoch": 4.99, "learning_rate": 9.128906191058199e-08, "loss": 0.1906, "step": 82750 }, { "epoch": 4.99, "learning_rate": 8.97801517963575e-08, "loss": 0.2613, "step": 82752 }, { "epoch": 4.99, "learning_rate": 8.8271241682133e-08, "loss": 0.1858, "step": 82754 }, { "epoch": 4.99, "learning_rate": 8.67623315679085e-08, "loss": 0.2453, "step": 82756 }, { "epoch": 5.0, "learning_rate": 8.525342145368401e-08, "loss": 0.2167, "step": 82758 }, { "epoch": 5.0, "learning_rate": 8.37445113394595e-08, "loss": 0.1782, "step": 82760 }, { "epoch": 5.0, "learning_rate": 8.223560122523503e-08, "loss": 0.3303, "step": 82762 }, { "epoch": 5.0, "learning_rate": 8.072669111101052e-08, "loss": 0.2597, "step": 82764 }, { "epoch": 5.0, "learning_rate": 7.921778099678603e-08, "loss": 0.2425, "step": 82766 }, { "epoch": 5.0, "learning_rate": 7.770887088256154e-08, "loss": 0.1771, "step": 82768 }, { "epoch": 5.0, "learning_rate": 7.619996076833703e-08, "loss": 0.1914, "step": 82770 }, { "epoch": 5.0, "learning_rate": 7.469105065411254e-08, "loss": 0.2095, "step": 82772 }, { "epoch": 5.0, "learning_rate": 7.318214053988805e-08, "loss": 0.1496, "step": 82774 }, { "epoch": 5.0, "learning_rate": 7.167323042566354e-08, "loss": 0.1751, "step": 82776 }, { "epoch": 5.0, "learning_rate": 7.016432031143905e-08, "loss": 0.342, "step": 82778 }, { "epoch": 5.0, "learning_rate": 6.865541019721456e-08, "loss": 0.2802, "step": 82780 }, { "epoch": 5.0, "learning_rate": 6.714650008299006e-08, "loss": 0.3301, "step": 82782 }, { "epoch": 5.0, "learning_rate": 6.563758996876556e-08, "loss": 0.1693, "step": 82784 }, { "epoch": 5.0, "learning_rate": 6.412867985454108e-08, "loss": 0.2247, "step": 82786 }, { "epoch": 5.0, "learning_rate": 6.261976974031657e-08, "loss": 0.1662, "step": 82788 }, { "epoch": 5.0, "learning_rate": 6.111085962609207e-08, "loss": 0.2599, "step": 82790 }, { "epoch": 5.0, "learning_rate": 5.960194951186758e-08, "loss": 0.174, "step": 82792 }, { "epoch": 5.0, "learning_rate": 5.809303939764308e-08, "loss": 0.2319, "step": 82794 }, { "epoch": 5.0, "learning_rate": 5.658412928341859e-08, "loss": 0.2604, "step": 82796 }, { "epoch": 5.0, "learning_rate": 5.507521916919409e-08, "loss": 0.2419, "step": 82798 }, { "epoch": 5.0, "learning_rate": 5.3566309054969593e-08, "loss": 0.2258, "step": 82800 }, { "epoch": 5.0, "learning_rate": 5.205739894074511e-08, "loss": 0.1788, "step": 82802 }, { "epoch": 5.0, "learning_rate": 5.05484888265206e-08, "loss": 0.1185, "step": 82804 }, { "epoch": 5.0, "learning_rate": 4.9039578712296117e-08, "loss": 0.201, "step": 82806 }, { "epoch": 5.0, "learning_rate": 4.753066859807162e-08, "loss": 0.2353, "step": 82808 }, { "epoch": 5.0, "learning_rate": 4.602175848384712e-08, "loss": 0.1576, "step": 82810 }, { "epoch": 5.0, "learning_rate": 4.451284836962263e-08, "loss": 0.2724, "step": 82812 }, { "epoch": 5.0, "learning_rate": 4.300393825539813e-08, "loss": 0.213, "step": 82814 }, { "epoch": 5.0, "learning_rate": 4.149502814117363e-08, "loss": 0.2755, "step": 82816 }, { "epoch": 5.0, "learning_rate": 3.998611802694914e-08, "loss": 0.2453, "step": 82818 }, { "epoch": 5.0, "learning_rate": 3.847720791272464e-08, "loss": 0.2058, "step": 82820 }, { "epoch": 5.0, "learning_rate": 3.6968297798500146e-08, "loss": 0.1552, "step": 82822 }, { "epoch": 5.0, "learning_rate": 3.5459387684275653e-08, "loss": 0.1853, "step": 82824 }, { "epoch": 5.0, "learning_rate": 3.395047757005115e-08, "loss": 0.1516, "step": 82826 }, { "epoch": 5.0, "learning_rate": 3.2441567455826656e-08, "loss": 0.1675, "step": 82828 }, { "epoch": 5.0, "learning_rate": 3.0932657341602164e-08, "loss": 0.1586, "step": 82830 }, { "epoch": 5.0, "learning_rate": 2.9423747227377668e-08, "loss": 0.187, "step": 82832 }, { "epoch": 5.0, "learning_rate": 2.7914837113153172e-08, "loss": 0.2575, "step": 82834 }, { "epoch": 5.0, "eval_cer": 0.07275090531382321, "eval_loss": 0.4541075527667999, "eval_runtime": 9398.6705, "eval_samples_per_second": 1.763, "eval_steps_per_second": 0.22, "step": 82835 } ], "max_steps": 82840, "num_train_epochs": 5, "total_flos": 8.903740709279682e+20, "trial_name": null, "trial_params": null }